The Neuroanatomical Organization Intrinsic Brain Activity Measured By

The Neuroanatomical Organization of Intrinsic Brain Activity Measured by fMRI in the Human Visual Cortex

Nicolás Gravel

This research and thesis was kindly supported by: The National Commission for Science and Technology of Chile, the Research School of Behavioural and Cognitive Neurosciences of the University of Groningen, the Laboratory of Experimental Ophthalmology of the University Medical Center Groningen, the Graduate School of Medical Sciences of the University of Groningen and the Professor Mulder Stichting.

The Neuroanatomical Organization of Intrinsic Brain Activity Measured by fMRI in the Human Visual Cortex ISBN digital version: 978-94-034-0592-6 ISBN printed version: 978-94-034-0593-3 Publisher: University of Groningen Cover art: Nicolás Gravel, ”Resting state fMRI activity depicted on a cortical surface reconstruction of the Calcarine sulcus”. The Neuroanatomical Organization of Intrinsic Brain Activity Measured by fMRI in the Human Visual Cortex

Proefschrift

ter verkrijging van de graad van doctor aan de Rijksuniversiteit Groningen op gezag van de rector magniﬁcus prof. dr. E. Sterken en volgens het besluit van het College voor Promoties.

De openbare verdediging zal plaatsvinden op

woensdag 18 april 2018 om 12:45 uur

door

Nicolás Gaspar Gravel Araneda

geboren op 6 mei 1985 te Santiago, Chile Promotores Prof. dr. F. W. Cornelissen Prof. dr. N. M. Jansonius

Copromotor Dr. R. Renken

Beoordelingscommissie Prof. dr. B. van de Berg Prof. dr. C. F. Beckmann Prof. dr. J. B.T.M. Roerdink Contents

List of Figures vi

List of Tables viii

List of Acronyms xi

1 General introduction 1 1.1 Outline ...... 2 1.2 Background ...... 3 1.2.1 Visual pathways ...... 3 1.2.2 Retinotopic organization of the visual cortex . . . . . 3 1.2.3 Hierarchical organization of the visual cortex . . . . . 4 1.2.4 Neurovascular coupling ...... 4 1.2.5 Functional magnetic resonance imaging ...... 5 1.2.6 Retinotopic mapping using population receptive ﬁeld (pRF) and connective ﬁeld (CF) modeling ...... 5 1.2.7 Resting state ...... 6 1.3 References ...... 7

2 Connective field estimates from resting state fMRI activity 13 2.1 Introduction ...... 13 2.2 Materials and Methods ...... 15 2.2.1 Subjects ...... 15 2.2.2 Stimulus ...... 15 2.2.3 Resting state ...... 15 2.2.4 Data acquisition ...... 15 2.2.5 Preprocessing ...... 16 2.2.6 Analysis ...... 16 2.3 Results ...... 19 2.3.1 Deriving connective field models based on resting state fMRI data ...... 19 2.3.2 Spatial aspects of resting state connective field map estimation ...... 20 2.4 Discussion ...... 25

i ii CONTENTS

2.4.1 Connective field models can be estimated based on resting state data ...... 25 2.4.2 Agreement between resting state and visual field mapping based connective field parameters ...... 25 2.4.3 Spatial changes: possible mechanisms ...... 26 2.4.4 Limitations and future directions ...... 26 2.4.5 Concluding remarks ...... 27 2.5 Supplementary Material ...... 27 2.6 References ...... 27

3 Synchronization clusters derived from fMRI activity in the early visual cortex 33 3.1 Introduction ...... 33 3.2 Methods ...... 35 3.2.1 Data ...... 35 3.2.2 Analysis ...... 36 3.2.3 Reproducibility of the synchronization clusters . . . . 38 3.2.4 Visuotopic organization of phase synchronization clusters 39 3.2.5 Spatial extent of phase synchronization ...... 39 3.2.6 Intra and inter-hemispheric cluster connectivity . . . 40 3.3 Results ...... 41 3.3.1 Synchronization clusters derived from resting state and visual field mapping are similar ...... 41 3.3.2 Visuotopic organization of the synchronization clusters 41 3.3.3 Comparison of the spatial extent of phase synchronization between RS and VFM ...... 42 3.3.4 Homotopic anatomical connectivity of cluster synchronization ...... 44 3.4 Discussion ...... 46 3.4.1 Phase-synchronization-based parcellation of RS-fMRI signals reveals topographically organized clusters in early visual cortex ...... 47 3.4.2 Similar location and shape, but different spatial extent of phase synchronization clusters for resting state and visual field mapping ...... 47 3.4.3 Spatial extent of phase synchrony resting: possible mechanisms ...... 49 3.4.4 Limitations and future directions ...... 50 3.5 Conclusion ...... 51 3.6 Supplementary Material ...... 51 3.7 References ...... 52

4 On the feasibility of assessing connective ﬁeld maps and synchronization clusters using 3T fMRI 59 4.1 Introduction ...... 59 4.2 Materials and Methods ...... 60 4.2.1 Participants ...... 60 4.2.2 Stimulus ...... 61 CONTENTS iii

4.2.3 Resting state ...... 61 4.2.4 Data acquisition ...... 61 4.2.5 Preprocessing ...... 62 4.2.6 Analysis ...... 62 4.3 Results ...... 64 4.3.1 Deriving connective ﬁeld estimates from VFM and RS 3T fMRI activity ...... 64 4.3.2 Synchronization clusters derived from VFM and RS 3T fMRI activity ...... 66 4.3.3 Limitations ...... 67 4.4 Conclusion ...... 69 4.5 References ...... 69

5 Propagation of BOLD activity across human visual cortex 73 5.1 Introduction ...... 73 5.2 Methods ...... 75 5.2.1 Data ...... 75 5.2.2 Selection of regions of interest ...... 76 5.2.3 Effective connectivity model for BOLD propagation . 77 5.3 Results ...... 81 5.3.1 Propagation of BOLD activity across early visual cortex measured with a noise-diffusion network model of EC ...... 81 5.3.2 Common underlying structures in EC ...... 82 5.3.3 Differences in EC and Σ between RS and VFM . . . 85 5.4 Discussion ...... 88 5.4.1 Recurrent connectivity and its role in visual processing 88 5.4.2 Cortical excitability: possible mechanisms ...... 90 5.4.3 Relation of the BOLD autocovariance decay constant to behavioral condition ...... 90 5.4.4 Limitations and interpretability of the model . . . . . 91 5.5 Concluding remarks ...... 94 5.6 Supplementary Material ...... 95 5.7 References ...... 95

6 General discussion 103 6.1 Summary of findings ...... 103 6.1.1 Visuotopic maps from resting state ...... 103 6.1.2 Similar synchronization clusters in resting state and visual field mapping ...... 104 6.1.3 It is feasible to obtain connective field maps and synchronization clusters from 3T fMRI activity . . . . . 104 6.1.4 Propagation of BOLD activity reveals task-dependent changes in effective connectivity ...... 105 6.2 Discussion ...... 105 6.2.1 Evolution, development and homeostasis ...... 105 6.2.2 Evoked versus intrinsic activity ...... 106 6.2.3 Predictive coding and enaction ...... 107 iv CONTENTS

6.3 Future research and applications ...... 107 6.4 Conclusions ...... 108 6.5 References ...... 109

Appendices 113

Thesis summary 113

Dutch summary 117

Acknowledgements 119

List of Figures

2.1 Distributions of explained variance ...... 20 2.2 Overall modeling performance characteristics ...... 21 2.3 Visualization of connective field maps ...... 22 2.4 Position scatter for V1-referred connective fields ...... 23 2.5 V1-referred connective field size during visual field mapping and resting state scans grouped over participants ...... 24 2.6 Relation between eccentricity and V1-referred connective field size in visual areas V2 and V3 grouped over participants ...... 24

3.1 Synchronization clusters obtained from VFM and RS ...... 42 3.2 Spatial aspects of the synchronization clusters ...... 43 3.3 Phase synchronization as a function of cortical distance in areas V1, V2 and V3 ...... 44 3.4 Phase synchronization as a function of visuotopic distance . . . . 45 3.5 Voxels in clusters with similar visual ﬁeld position selectivity have higher PLV across visual ﬁeld maps and hemispheres ...... 46

4.1 Visualization of pRF and connective field maps obtained from VFM and RS 3T f-MRI data ...... 65 4.2 Connective field size as a function of pRF eccentricity in V2 and V3 67 4.3 Synchronization clusters obtained from VFM and RS fMRI data 68 4.4 Cluster membership probability as a function of visual field distance in RS and VFM ...... 68

5.1 Apparent propagation of BOLD activity during RS depicted in the flattened cortical surface reconstruction ...... 82 5.2 Modeling the propagation of BOLD activity across visual field maps V1, V2 and V3 ...... 83 5.3 Common underlying structure of EC across visual cortical areas V1, V2 and V3 ...... 84 5.4 Common structure in EC for RS and VFM illustrated in visual field space ...... 86 5.5 Differences in EC and Σ between RS and VFM suggest a re-configuration of feedforward and feedback interactions ...... 87

List of Tables

4.1 Correlations between visual field maps derived using population receptive field and connective field modeling ...... 64 4.2 Correlations between CF maps derived from VFM and RS data . 66

5.1 Goodness of ﬁt between modeled and empirical spatiotemporal covariances ...... 85

viii

List of Acronyms

BOLD Blood Oxygen Level Dependent CF Connective Field DC Direct Current DCM Dynamic Causal Modeling DCT Discrete Cosine Transform EC Eﬀective Connectivity FC Functional Connectivity fMRI Functional Magnetic Resonance Imaging FWHM Full Width at Half Maximum LGN Lateral Geniculate Nucleus LO Lyapunov Optimization MVAR Multi-Variate Auto-Regressive NMI Normalized Mutual Information PLV Phase Locking Values PS Phase Synchronization pRF Population Receptive Field RGC Retinal Ganglion Cells ROI Region Of Interest RS Resting State SC Synchronization Cluster SNR Signal to Noise Ratio T1 Longitudinal relaxation time of functional scan T2 Transverse relaxation time of anatomical scan TE Time to Echo TR Time to Repetition VE Variance Explained VFM Visual Field Mapping V1/V2/V3 Human visual cortical areas 1,2 and 3 3T/7T 3/7 Tesla (units of magnetic ﬂux density)

General introduction 1

Invasive electrophysiological recordings of neuronal activity from the visual cortex of cats and other animals have revealed that spontaneous neuronal activity reflects the underlying neuroanatomical organization [30, 70, 40]. However, for non-invasive and indirect functional magnetic resonance imaging (fMRI) recordings of neuronal activity from visual cortical areas, this relationship between intrinsic activity (often referred to as “resting-state”) and cortical neuroanatomical organization is not immediately obvious [42, 36, 62, 11, 27]. When measured using fMRI, intrinsic fluctuations in blood-oxygen level dependent (BOLD) activity are correlated between distant brain regions that are anatomically connected, such as homologous areas in the two hemispheres [6, 73]. For this reason, resting state (RS) fMRI has been widely used to study whole-brain interactions in health and disease [63, 72]. However, interpreting patterns of RS-fMRI activity at a more local scale (e.g., that of the visual cortex) remains challenging as activity in nearby sites can be correlated as a result of either neuroanatomical connections, or metabolic and vascular relationships [12, 47, 45, 34, 58, 46]. Currently, this uncertainty limits the use of RS-fMRI for characterizing visual cortical activity in health and disease. A better understanding of the relationships between neuronal activity, brain anatomy and hemodynamics would help to develop RS-fMRI as a valuable tool for fundamental and applied —non-invasive— research in humans. Aiming to contribute to the understanding of these relationships, this thesis focuses on the following question: What can we learn about the neuroanatomical organization of the human visual cortex from RS-fMRI recordings? With this central question in mind, I hypothesize that RS-fMRI activity in the human visual cortex reflects its underlying neuroanatomical organization. Using advanced anatomical MRI techniques and novel fMRI analysis methods, I will characterize and interpret the spatial and temporal organization of in-

1 . G

trinsic fluctuations in BOLD activity across striate (V1) and extrastriate visual cortex (V2 and V3). Moreover, I will discuss how these fluctuations may be shaped by neuroanatomical, physiological and vascular factors. In the following sections, I outline the questions and main findings of each experimental chapter, while in the background section, I describe the main anatomical features of the human visual system, summarize the experimental meth- odologies used in this thesis and briefly describe the emergence of the field of resting state research. The latter section reviews some of the previous studies that my work builds on and helps to remind that the brain is not only concerned with the demands imposed by the environment but also with internally generated dynamics. As we will see in the experimental chapters of my thesis, the neuroanatomical organization of the visual system forms a complex recurrent network of feedforward, feedback and lateral connections that cannot be fully grasped from the classical “sandwich” perspective (perception-cognition- action) of brain function [21] but perhaps —as I will discuss later (in chapter 6)— by a more constructivist and enactive view of brain function that accounts for recurrent neuronal processing and circular causality [71, 38].

1.1 Outline This thesis comprises four experimental sections. In chapter 2 I ask whether visuotopic organization 1 can be derived from 7T RS-fMRI activity. Based on the hypothesis that intrinsic fluctuations in BOLD activity reflect underlying neuroanatomical organization, I show that it is possible to map —based on RS- fMRI recordings— cortico-cortical neuronal interactions between V1, V2 and V3 using connective field modeling [32, 27]. Based on the hypothesis that both intrinsic and stimulus-evoked BOLD fluctuations are anchored by the same neuroanatomical connections, in chapter 3 I ask whether patterns of synchronized fMRI activity in RS and visual field mapping (VFM) are comparable (at 7T). By examining local patterns of phase covariation, I find synchronization clusters that are similar, regardless of whether they are derived from RS or VFM data. However, in activity obtained during VFM, phase synchronization is spatially more extensive than in RS derived activity, reflecting stimulus driven interactions between local responses. Nevertheless, the resemblance between RS and VFM-derived synchronization clusters suggests that they share a common neuroanatomical origin [28]. Methods to analyse and interpret RS-fMRI activity can become valuable tools for the study of human neuronal activity in vivo, in particular if they can be generalized to different magnetic field strengths. To verify this for connective field modeling and cluster synchronization analysis of RS data, in chapter 4 I study the feasibility of reproducing the main findings of chapters 2 & 3 using data acquired with a 3T rather than a 7T scanner. Despite the lower resolution and signal-to-noise ratio of the 3T data, I find that the results obtained with it are in fair agreement to those obtained previously with 7T data.

1When the image’s spatial relationships formed in the retina are preserved in the visual cortex in the form of visual ﬁeld maps: nearby neurons respond to nearby locations in the image.

2 1.2. Background

In chapter 5 I ask whether the propagation of BOLD activity within and between visual ﬁeld maps relates to structured neuronal activity. To explore this question, I implement a modeling approach aimed at disentangling the contributions of local activity and directed interactions in shaping BOLD activity propagation [26]. Applying this approach to 7T fMRI data reveals changes in cortical excitability and directed interactions in RS and VFM, pointing to a task-dependent reconﬁguration of local, feedforward and feedback interactions across the visual system [29].

1.2 Background

In this section, I provide a general overview for those interested in studying the human visual cortex using RS-fMRI. I brieﬂy describe the anatomical organization of the human visual system and provide some background on the ﬁeld of fMRI and resting state measurements. For more detailed accounts on the neuronal and anatomical organization of the human visual cortex see e.g. [37, 74].

1.2.1 Visual pathways Light, in the form of structured energy distributions, is transduced by photore- ceptor cells in the retina (rods and cones) into electrical membrane potentials [24]. These potentials are then shaped by a recurrent cellular network consisting of bipolar, horizontal, and amacrine cells. Together they form a mosaic of functional subunits that contribute to inhibitory and sensitivity-adjustment mechanisms and communicate with the retinal ganglion cells (RGC). The ax- onal projections from the RGC exit the eye to form the optic nerve. These fiber projections from the two eyes decussate in the optic chiasm, with fibers from the left and right half of each retina going to the left and right hemispheres, respectively. Optic nerve fibers project primarily to the lateral geniculate nucleus (LGN), which relays to occipital visual cortical areas –the topic of the research in this thesis.

1.2.2 Retinotopic organization of the visual cortex Shaped by evolutionary and developmental constraints, the anatomical circuitry of the human sensory cortices follows the topographic organization of their corresponding sensory surfaces. For the visual system, the result is that the image’s spatial relationships formed in the retina are preserved in many regions of the cortex in the form of visual field maps: nearby neurons respond to nearby locations in the image [37]. These retino-cortical maps are said to be retinotopically organized, yet their actual location and shape is determined by different constraints from those that shape the eye [64, 55]. The topographic organization of the primary visual cortex (V1), located along the calcarine sulcus, accom- modates an hemifield representation of the retinal image in which the foveal region is greatly magnified –a phenomenon called cortical magnification [65].

3 . G

The nearby visual field maps V2 and V3 cluster around V1, sharing parallel eccentricity representations. However, the visual field maps in V2 and V3 are discontinuous. They split into dorsal and ventral parts, with angular representations alternating in visual field sign along the horizontal meridian. As a result, V2 and V3 are organized into two quarter hemifields. Other extrastriate visual cortical areas are organized similarly [74]. The multiplicity of visual field maps in the visual cortex appears to reflect the fact that our perceptual machinery is attuned to stable structural features in our habitat. As neurons process different aspects of the image, cortical circuitry is organized into receptive fields that preserve its spatial organization. As a result, cortical areas subserving different functions still preserve this level of organization [74]. In the experimental chapters that will follow, I relate task-dependent changes in fMRI activity to the retinotopic organization of V1, V2 and V3.

1.2.3 Hierarchical organization of the visual cortex Traditionally, neuroanatomical connections among visual cortical areas have been thought to follow a primarily parallel-hierarchical bottom-up architecture [19]. In this view, parallel neuronal pathways, originating from anatomically and functionally distinct cell types in the retina [14], connect to thalamic and cortical structures following a precise retinotopic ordering. The visual processing of retinal signals unfolds then through a series of stages in which low- level stimulus features are processed ﬁrst by ’early’ visual cortices such as V1, and increasingly complex features are processed sequentially by ’higher’ extrastriate visual cortices [50]. However, feedforward connections along the visual hierarchy areas are typically complemented by reciprocal feedback connections [23, 10, 41]. For example, extra-retinal inputs to V1 may arrive from extrastriate visual areas such as V2 and V3 [54, 53, 61], but also indirectly from more distant corical areas [66, 56, 20, 7, 15]. Moreover, reciprocal connections between V1 and visual thalamic structures have long been described, yet, their functional relevance is not yet fully understood [31, 35, 3, 69]. The integration of feedforward and feedback interactions along the visual hierarchy enables the modulation of ongoing cortical dynamics by incoming sensory input [25, 61]. In chapter 5,I will implement a generative model of directed interactions to examine the presence of state-dependent changes in the feedback and feedforward interactions between V1, V2 and V3.

1.2.4 Neurovascular coupling In most of the visual cortex, blood is supplied by the calcarine and basilar ar- teries. Flowing through their ramiﬁcations, blood arrives in a network of small arterioles and capillary beds. After irrigating these capillary beds, which are arranged perpendicularly to the cortical surface, blood is drained through small venous channels into the dural venous sinuses. Neuroglia and astrocytes [59], the brain’s housekeepers, regulate blood ﬂow at the level of the capillary beds. These cells form part of the blood-brain barrier. They coordinate local blood

4 1.2. Background supply, nutrient transport and neurotransmitter release, and thus provide a crucial contribution to neuronal activity [12]. Thanks to this coupling between blood supply and neuronal activity, it has become possible to —indirectly— estimate neuronal population activity using non-invasive instruments like magnetic resonance scanners. In the next sections, I will show how fMRI has allowed to map the topographical and neuroanatomical organization of visual ﬁeld maps in the human visual cortex.

1.2.5 Functional magnetic resonance imaging In order to estimate neuronal activity patterns across early visual cortical areas I used fMRI. This technique measures changes in blood oxygenation, which has been shown to be indirectly related to neuronal activity [45]. Local neuronal activity induces changes in the ratio between oxyhemoglobin and deoxyhemoglobin, which can be detected due to their differential magnetic susceptibility [57]. This effect is named the blood-oxygen level dependent (BOLD) effect, which is the basis for fMRI. In my studies, I have used primarily 7T fMRI data. However, in chapter 4, I also examine 3T data. Higher magnetic field strengths allow for a better signal-to-noise ratio and spatiotemporal resolution. However, the temporal resolution of fMRI is limited by the hemodynamic response to neuronal activity, not by the magnetic field strength.

1.2.6 Retinotopic mapping using population receptive field (pRF) and connective field (CF) modeling The development of functional magnetic resonance imaging (fMRI) has allowed to map the topographical and neuroanatomical organization of visual field maps in the human visual cortex [18]. Modeling the hemodynamic responses to hypothetical neuronal activity elicited by a stimulus allows the mapping of population receptive field properties [68]. The population receptive field (pRF) method re- lies on this approach [17]. The method uses a parameterized forward model of the neuronal population responses, a description of the hemodynamic response function (a two-gamma HRF model) [8], and the stimulus (here I used drifting checkerboard bars) to predict evoked BOLD activity. The population receptive field model used in this thesis corresponds to a circular Gaussian characterized by three parameters: x and y (positions in the stimuli screen), and size (σ). To find the most likely retinotopic map, a set of candidate population receptive field models are combined with the stimulus aperture to generate predictions of the neuronal responses each candidate pRF would produce. Subsequent con- volution of this predicted neuronal response with the HRF gives a set of candidate BOLD responses for each combination of pRF parameters. The pRF parameters associated with the best fitting candidate BOLD responses are then chosen to summarize the response of each fMRI recording site. Somewhat similar to the way in which the visual field is mapped on the surface of the cortex using pRFs, CF modeling describes the neuronal interactions between different cortical visual areas in terms of spatial integration and cortical selectivity maps. CF modeling enables the characterization of a target record-

5 . G

ing site (e.g. V2 and V3) in terms of the aggregate BOLD activity in a source brain area (e.g. V1), thus providing a description of the preferred locations on the cortical surface to which these target sites respond [32]. Locations in the primary visual cortex are associated with visual field positions obtained during visual field mapping (VFM). Therefore, visual field coordinates can be inferred for the target recording sites from the preferred locations in the source region, allowing the reconstruction of visuotopic maps even in the absence of a stimulus (e.g., RS). The technique is prominently used in chapters 2 & 4.

1.2.7 Resting state Resting state (RS) research has become very popular in the last decades. De- parting from the idea of the brain as a primarily reflexive device driven by the momentary demands of the environment [60, 67], RS studies suggest that the brain’s operations are mainly intrinsic, involving the integration of slow internal and rapid sensory dynamics [75, 9, 44, 16]. From this perspective, sensory input modulates rather than ‘forms’ brain function [13]. Early evidence on the role of intrinsic brain activity came from studies of animal physiology and anatomy. Using decorticated cats, Brown [9] showed that rhythmic behaviours reminiscent of walking and running were still possible, even in the absence of sensory and cortical input. Intrinsic rhythms in brain activity were further found in humans. By recording scalp electrical potentials, Berger [5] found large-amplitude rhythmic fluctuations that appeared only when subjects had their eyes closed. These fluctuations were reported to occur at a rate of approximately 10 oscillations per second, what is now called the alpha frequency band [2]. Likewise, studying spontaneous electrical activity in the rabbit olfactory bulb, Adrian [1] found that the persistent activity of cells in the bulb unfolded as waves of electrical activity, even in the absence of sensory stimulation. Despite these early findings, attempts to characterize and quantify intrinsic brain activity only gained momentum in the last two decades. Using multi- electrode recordings of retinal ganglion cells, Galli and Maffei [22] and Meister et al. [49] showed spontaneous wave-like propagation of neuronal firing prior to visual experience. These patterns were hypothesized to play a role inthe development and formation of retinotopically ordered maps in the visual system [39, 48]. Later studies using large-scale neuronal recording approaches showed that both stimulus-evoked and spontaneous neuronal activity closely reflected the functional and anatomical organization of the visual cortex [4, 70, 40]. Parallel investigations in humans and non-human primates using fMRI recordings from the somatosensory cortex also indicated a link between the brain’s intrinsic activity and its neuroanatomical organization [6, 73]. Moreover, using whole-brain fMRI recordings, it has been recently shown that the propagation of BOLD activity across distant brain areas reflects the transition between different behavioral states [51, 52]. This suggests a meaningful relationship between BOLD activity propagation and neuronal interactions between distant brain regions. Therefore, in this thesis, I ask whether task-dependent differences in BOLD activity between RS and VFM can also reveal relevant aspects of brain organization and activity, yet at a much more local scale. Thereby, I focus on

6 1.3. References early cortical visual ﬁeld maps, which are richly interconnected and for which regional variation in the hemodynamic response is less pronounced [33, 43].

1.3 References [1] E. D. Adrian. The electrical activity of the mammalian olfactory bulb. Electroencephalography and Clinical Neurophysiology, 2(1):377–388, Janu- ary 1950. . [2] ED Adrian and BHC Matthews. The Berger Rhythm: Potential changes from the occipital lobes in man. Brain, 57(4):355–385, December 1934. . [3] Ian M Andolina, Helen E Jones, Wei Wang, and Adam M Sillito. Cor- ticothalamic feedback enhances stimulus response precision in the visual system. Proc. Natl. Acad. Sci. U. S. A., 104(5):1685–1690, 30 January 2007. [4] A. Arieli, D. Shoham, R. Hildesheim, and A. Grinvald. Coherent spatiotemporal patterns of ongoing activity revealed by real-time optical imaging coupled with single-unit recording in the cat visual cortex. Journal of Neurophysiology, 73(5):2072–2093, May 1995.

[5] Hans Berger. Über das elektrenkephalogramm des menschen. Archiv f. Psychiatrie, 87(1):527–570, December 1929. . [6] Bharat Biswal, F Zerrin Yetkin, Victor M Haughton, and James S Hyde. Functional connectivity in the motor cortex of resting human brain using echo-planar mri. Magn. Reson. Med., 34(4):537–541, 1995. [7] Sander E Bosch, Janneke F M Jehee, Guillén Fernández, and Christian F Doeller. Reinstatement of associative memories in early visual cortex is signaled by the hippocampus. J. Neurosci., 34(22):7493–7500, 28 May 2014. [8] Geoffrey M. Boynton, Stephen A. Engel, Gary H. Glover, and David J. Heeger. Linear Systems Analysis of Functional Magnetic Resonance Ima- ging in Human V1. Journal of Neuroscience, 16(13):4207–4221, July 1996. [9] T. Graham Brown. On the nature of the fundamental activity of the nervous centres; together with an analysis of the conditioning of rhythmic activity in progression, and a theory of the evolution of function in the nervous system. The Journal of Physiology, 48(1):18–46, March 1914. . [10] J Bullier, J M Hupé, A C James, and P Girard. The role of feedback connections in shaping the responses of visual cortical neurons. Prog. Brain Res., 134:193–204, 2001. [11] Omar H Butt, Noah C Benson, Ritobrato Datta, and Geoffrey K Aguirre. The fine-scale functional correlation of striate cortex in sighted and blind people. J. Neurosci., 33(41):16209–16219, 9 October 2013.

7 . G

[12] György Buzsáki, Buzsáki György, Kaila Kai, and Raichle Marcus. Inhib- ition and brain work. Neuron, 56(5):771–783, 2007. [13] Felicity Callard and Daniel S Margulies. The Subject at Rest: Novel conceptualizations of self and brain from cognitive neuroscience’s study of the resting state฀. Subjectivity, 4(3):227–257, September 2011. . [14] Edward M. Callaway. Structure and function of parallel pathways in the primate early visual system. The Journal of Physiology, 566(1). . [15] Russell W Chan, Alex T L Leong, Leon C Ho, Patrick P Gao, Ed- die C Wong, Celia M Dong, Xunda Wang, Jufang He, Ying-Shing Chan, Lee Wei Lim, and Ed X Wu. Low-frequency hippocampal-cortical activity drives brain-wide resting-state functional MRI connectivity. Proc. Natl. Acad. Sci. U. S. A., 114(33):E6972–E6981, 15 August 2017. [16] Gustavo Deco, Morten L. Kringelbach, Viktor K. Jirsa, and Petra Ritter. The dynamics of resting fluctuations in the brain: metastability andits dynamical cortical core. Scientific Reports, 7(1):3095, June 2017. . [17] Serge O Dumoulin and Brian A Wandell. Population receptive field estimates in human visual cortex. Neuroimage, 39(2):647–660, 15 January 2008. [18] Stephen A Engel, David E Rumelhart, Brian A Wandell, Adrian T Lee, Gary H Glover, Chichilnisky Eduardo-Jose, and Michael N Shadlen. fMRI of human visual cortex. Nature, 369(6481):525–525, 1994. [19] D. J. Felleman and D. C. Van Essen. Distributed hierarchical processing in the primate cerebral cortex. Cerebral Cortex (New York, N.Y.: 1991), 1 (1). [20] Stephanie J. Forkel, Michel Thiebaut de Schotten, Jamie M. Kawadler, Flavio Dell’Acqua, Adrian Danek, and Marco Catani. The anatomy of fronto-occipital connections from early blunt dissections to contemporary tractography. Cortex, 56(Supplement C):73–84, July 2014. . [21] Karl J. Friston. Precision Psychiatry. Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, 2(8):640–643, November 2017. . [22] Lucia Galli and Lamberto Maffei. Spontaneous impulse activity of rat retinal ganglion cells in prenatal life, October 1988. [23] Ralf A. W. Galuske, Kerstin E. Schmidt, Rainer Goebel, Stephen G. Lomber, and Bertram R. Payne. The role of feedback in shaping neural representations in cat visual cortex. Proc Natl Acad Sci U S A, 99(26): 17083–17088, December 2002. . [24] J. J. Gibson. Ecological optics. Vision Research, 1(3):253–262, October 1961. . [25] Charles D Gilbert and Mariano Sigman. Brain states: top-down influences in sensory processing. Neuron, 54(5):677–696, 7 June 2007.

8 1.3. References

[26] Matthieu Gilson, Ruben Moreno-Bote, Adrián Ponce-Alvarez, Petra Ritter, and Gustavo Deco. Estimation of directed eﬀective connectivity from fMRI functional connectivity hints at asymmetries of cortical connectome. PLoS Comput. Biol., 12(3):e1004762, March 2016. [27] Nicolás Gravel, Ben Harvey, Barbara Nordhjem, Koen V Haak, Serge O Dumoulin, Remco Renken, Branislava Curčić-Blake, and Frans W Cor- nelissen. Cortical connective ﬁeld estimates from resting state fMRI activity. Front. Neurosci., 8:339, 31 October 2014. [28] Nicolás Gravel, Ben M. Harvey, Remco J. Renken, Serge O. Dumoulin, and Frans W. Cornelissen. Phase-synchronization-based parcellation of resting state fMRI signals reveals topographically organized clusters in early visual cortex. NeuroImage, September 2017. . [29] Nicolás Gravel, Remco J. Renken, Ben M. Harvey, Gustavo Deco, Frans W. Cornelissen, and Matthieu Gilson. Propagation of BOLD activity reveals directed interactions across human visual cortex. bioRxiv, page 172452, October 2017. .

[30] Amiram Grinvald, Edmund Lieke, Ron D. Frostig, Charles D. Gilbert, and Torsten N. Wiesel. Functional architecture of cortex revealed by optical imaging of intrinsic signals. Nature, 324(6095):361, November 1986. .

[31] R W Guillery and S Murray Sherman. Thalamic relay functions and their role in corticocortical communication: generalizations from the visual system. Neuron, 33(2):163–175, 17 January 2002. [32] Koen V Haak, Jonathan Winawer, Ben M Harvey, Remco Renken, Serge O Dumoulin, Brian A Wandell, and Frans W Cornelissen. Con- nective field modeling. Neuroimage, 66:376–384, 1 February 2013. [33] Daniel A. Handwerker, John M. Ollinger, and Mark D’Esposito. Vari- ation of BOLD hemodynamic responses across subjects and brain regions and their effects on statistical analyses. NeuroImage, 21(4):1639–1651, April 2004. . [34] Robert V Harrison, Noam Harel, Jaswinder Panesar, and Richard J Mount. Blood capillary distribution correlates with hemodynamic-based functional imaging in cerebral cortex. Cereb. Cortex, 12(3):225–233, March 2002. [35] J Michael Hasse and Farran Briggs. Corticogeniculate feedback sharpens the temporal precision and spatial resolution of visual signals in the ferret. Proc. Natl. Acad. Sci. U. S. A., 114(30):E6222–E6230, 25 July 2017. [36] Jakob Heinzle, Thorsten Kahnt, and John-Dylan Haynes. Topographic- ally specific functional connectivity between visual field maps in the human brain. Neuroimage, 56(3):1426–1436, 1 June 2011.

9 . G

[37] David H. Hubel. The Visual Cortex of the Brain. Scientiﬁc American, 209 (5), 1963. . [38] Friston Karl. A Free Energy Principle for Biological Systems. Entropy, 14(11):2100–2121, October 2012. . [39] Lawrence C. Katz and Justin C. Crowley. Development of cortical circuits: lessons from ocular dominance columns. Nat. Rev. Neurosci., 3(1):34–42, January 2002. . [40] Tal Kenet, Dmitri Bibitchkov, Misha Tsodyks, Amiram Grinvald, and Amos Arieli. Spontaneously emerging cortical representations of visual attributes. Nature, 425(6961):954–956, 30 October 2003. [41] Bart Krekelberg. Primate motion perception. In How Animals See the World. Oxford University Press, 2012. [42] Christopher M Lewis, Conrado A Bosman, Thilo Womelsdorf, and Pas- cal Fries. Stimulus-induced visual cortical networks are recapitulated by spontaneous local and interareal synchronization. Proc. Natl. Acad. Sci. U. S. A., 113(5):E606–15, 2 February 2016. [43] Fa-Hsuan Lin, Jonathan R Polimeni, Jo-Fu Lotus Lin, Kevin W-K Tsai, Ying-Hua Chu, Pu-Yeh Wu, Yi-Tien Li, Yi-Cheng Hsu, Shang-Yueh Tsai, and Wen-Jui Kuo. Relative latency and temporal variability of hemodynamic responses at the human primary visual cortex. Neuroimage, 22 January 2017. [44] R. R. Llinas. The intrinsic electrophysiological properties of mammalian neurons: insights into central nervous system function. Science, 242(4886): 1654–1664, December 1988. . [45] Nikos K Logothetis and Brian A Wandell. Interpreting the BOLD signal. Annu. Rev. Physiol., 66:735–769, 2004. [46] Celine Mateo, Per M. Knutsen, Philbert S. Tsai, Andy Y. Shih, and David Kleinfeld. Entrainment of Arteriole Vasomotor Fluctuations by Neural Activity Is a Basis of Blood-Oxygenation-Level-Dependent ฀Resting- State Connectivity. Neuron, October 2017. . [47] Teppei Matsui, Tomonari Murakami, and Kenichi Ohki. Transient neuronal coactivations embedded in globally propagating waves underlie resting-state functional connectivity. Proc. Natl. Acad. Sci. U. S. A., 16 May 2016. [48] Todd McLaughlin, Christine L. Torborg, Marla B. Feller, and Dennis D. M. O’Leary. Retinotopic map reﬁnement requires spontaneous retinal waves during a brief critical period of development. Neuron, 40(6):1147– 1160, December 2003. [49] Markus Meister, Rachel O. L. Wong, Denis A. Baylor, and Carla J. Shatz. Synchronous bursts of action potentials in ganglion cells of the developing mammalian retina, May 1991.

10 1.3. References

[50] Mortimer Mishkin, Leslie G Ungerleider, and Kathleen A Macko. Ob- ject vision and spatial vision: two cortical pathways. Trends Neurosci., 6: 414–417, 1983. [51] Anish Mitra, Abraham Z Snyder, Enzo Tagliazucchi, Helmut Laufs, and Marcus E Raichle. Propagated infra-slow intrinsic brain activity reorgan- izes across wake and slow wave sleep. Elife, 4, 9 November 2015. [52] Anish Mitra, Abraham Z. Snyder, Carl D. Hacker, Mrinal Pahwa, Enzo Tagliazucchi, Helmut Laufs, Eric C. Leuthardt, and Marcus E. Raichle. Human cortical hippocampal dialogue in wake and slow-wave sleep. PNAS, 113(44):E6868–E6876, November 2016. . [53] Lars Muckli and Lucy S Petro. Network interactions: non-geniculate input to V1. Curr. Opin. Neurobiol., 23(2):195–201, April 2013. [54] S Murray Sherman and R W Guillery. Functional Connections of Cortical Areas: A New View from the Thalamus. MIT Press, 9 August 2013. [55] Ian Nauhaus and Kristina J. Nielsen. Building maps from maps in primary visual cortex. Curr. Opin. Neurobiol., 24(1):1–6, February 2014. . [56] Behrad Noudoost and Tirin Moore. Control of visual cortical signals by prefrontal dopamine. Nature, 474(7351):372–375, 15 May 2011. [57] S Ogawa, T M Lee, A R Kay, and D W Tank. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc Natl Acad Sci U S A, 87(24):9868–9872, December 1990. [58] Pratik Y O’Herron, Philip andand Chhatbar, Levy Manuel, Shen Zhim- ing, Adrien E Schramm, Lu Zhongyang, and Kara Prakash. Neural correlates of single-vessel haemodynamic responses in vivo. Nature, 2016. [59] J C Pang, P A Robinson, K M Aquino, and N Vasan. Eﬀects of as- trocytic dynamics on spatiotemporal hemodynamics: Modeling and en- hanced data analysis. Neuroimage, 14 October 2016. [60] I. P. Pavlov and For Mem R. S. Croonian Lecture. Certain problems in the physiology of the cerebral hemispheres. Proc. R. Soc. Lond. B, 103 (722):97–110, June 1928. . [61] Lucy S Petro, Luca Vizioli, and Lars Muckli. Contributions of cortical feedback to sensory processing in primary visual cortex. Front. Psychol., 5: 1223, 6 November 2014. [62] Mathijs Raemaekers, Raemaekers Mathijs, Schellekens Wouter, Richard J A van Wezel, Petridou Natalia, Kristo Gert, and Nick F Ramsey. Pat- terns of resting state connectivity in human primary visual cortical areas: A 7T fMRI study. Neuroimage, 84:911–921, 2014. [63] M E Raichle. The restless brain: how intrinsic activity organizes brain function. Philos. Trans. R. Soc. Lond. B Biol. Sci., 370(1668):20140172– 20140172, 2015.

11 . G

[64] Lars Reichl, Dominik Heide, Siegrid Löwel, Justin C. Crowley, Matthias Kaschube, and Fred Wolf. Coordinated Optimization of Visual Cortical Maps (I) Symmetry-based Analysis. PLoS Computational Biology, 8(11), 2012. . [65] Mark M Schira, Christopher W Tyler, Michael Breakspear, and Branka Spehar. The foveal confluence in human visual cortex. J. Neurosci., 29(28): 9050–9058, 15 July 2009. [66] Jeremy D. Schmahmann and Deepak N. Pandya. The Complex History of the Fronto-Occipital Fasciculus. Journal of the History of the Neurosciences, 16(4):362–377, October 2007. . [67] Charles Sherrington. Ferrier Lecture: Some Functional Problems At- taching to Convergence. Proceedings of the Royal Society of London. Series B, Containing Papers of a Biological Character, 105(737):332–362, 1929. [68] A T Smith, K D Singh, A L Williams, and M W Greenlee. Estimating receptive field size from fMRI data in human striate and extrastriate visual cortex. Cereb. Cortex, 11(12):1182–1190, December 2001. [69] M. Steriade. Corticothalamic resonance, states of vigilance and mentation. Neuroscience, 101(2):243–276, 2000. [70] M. Tsodyks, T. Kenet, A. Grinvald, and A. Arieli. Linking Spontan- eous Activity of Single Cortical Neurons and the Underlying Functional Architecture. Science, 286(5446):1943–1946, December 1999. . [71] Francisco J Varela. Autopoiesis and a Biology of Intentionality. 1991. [72] Alberto L Vazquez, Matthew C Murphy, and Seong-Gi Kim. Neuronal and physiological correlation to hemodynamic resting-state fluctuations in health and disease. Brain Connect., 4(9):727–740, November 2014. [73] J L Vincent, G H Patel, M D Fox, A Z Snyder, J T Baker, D C Van Es- sen, J M Zempel, L H Snyder, M Corbetta, and M E Raichle. Intrinsic functional architecture in the anaesthetized monkey brain. Nature, 447 (7140):83–86, 3 May 2007. [74] Brian A Wandell, Serge O Dumoulin, and Alyssa A Brewer. Visual field maps in human cortex. Neuron, 56(2):366–383, 25 October 2007. [75] James William. The Principles of Psychology. Philosophy and Phenomeno- logical Research, 44(1):124–126, 1983.

12 Cortical connective ﬁeld estimates from resting state fMRI activity 2

One way to study connectivity in visual cortical areas is by examining spontaneous neuronal activity. In the absence of visual input, such activity remains shaped by the underlying neuronal architecture and, presumably, may still reflect visuotopic organization. Here, we applied population connective field modeling (CF) to estimate the spatial profile of functional connectivity in the early visual cortex during resting state functional magnetic resonance imaging (RS-fMRI). This model-based analysis estimates the spatial integration between blood-oxygen level dependent (BOLD) signals in distinct cortical visual field maps using fMRI. Just as population receptive field (pRF) mapping predicts the collective neuronal activity in a voxel as a function of response selectivity to stimulus position in visual space, connective field modeling predicts the activity of voxels in one visual area as a function of the aggregate activity in voxels in another visual area. In combination with pRF mapping, CF locations on the cortical surface can be interpreted in visual space, thus enabling reconstruction of visuotopic maps from resting state data. We demonstrate that V1→V2 and V1→V3 connective field maps estimated from resting state fMRI data show visuotopic organization. Therefore, we conclude that, despite some variability in CF estimates between RS scans, neuronal properties such as CF maps and CF size can be derived from resting state data.

2.1 Introduction

The human visual cortex is a highly complex and interconnected system operating at various temporal and spatial scales, and as such, non-invasive assessment of the neuronal correlates of human visual processing are of great importance.

13 . C MRI

A significant contribution toward understanding human visual processing can be made by studying cortico-cortical interactions between different visual areas [22, 36, 18]. One way to study these neuronal correlates is by examining spontaneous BOLD co-fluctuations during resting state [22, 36]. Given that resting state BOLD fluctuations are partly shaped by the underlying functional and neuroanatomical organization [4, 37, 30, 5, 11, 25, 49], analysis of resting state activity offers a possibility to examine intrinsic functional connectivity of the visual system as well as the extent of variability of these processes. Although functional magnetic resonance imaging (fMRI) indirectly measures neuronal activity, accurate methods to map neuronal response selectivity in the early visual cortex from the blood-oxygen level dependent (BOLD) signal have been developed [43, 14, 13]. With these methods, the unifying concept of classical receptive field [23] has found its place in fMRI, under the definition of population receptive field (pRF). The term pRF was first used to describe population encoding in macaque early visual areas [47]. Used in fMRI, the term describes the aggregate responses of fMRI recording sites (voxels) to presented stimuli, in terms of the position and size of the visual field area to which each recording site responds. The parametric modeling approach of the pRF technique has allowed non-invasive investigation of neuronal response selectivity, its cortical organization, and the computational properties of the visual system. A recent complementary method, called connective field (CF) modeling [18], extends this type of analysis to model cortico-cortical interactions in terms of spatially localized patterns of functional connectivity. Specifically, this method enables characterization of a recording site in terms of aggregate cortical activity in another brain area, thus extending the concept of receptive field from a description of preferred locations in visual (stimulus) space to preferred locations on the cortical surface. Connective field modeling was originally conceived as a method to analyze responses evoked by visual field mapping stimuli, though the analysis does not use a description of the stimulus. As such, it could in principle be applied to explore cortico-cortical connectivity profiles during different experimental conditions as well as resting state. To realize this potential, a number of questions must be addressed. In this paper, we try to provide answers to at least four of them. First, how do we measure CF models in the presence of substantial physiological measurement noise? Second, how much scan time is sufficient to achieve accurate discrimination of CF models obtained from resting state data? Third, how do CF parameters obtained from resting state compare to those obtained from stimulus-evoked activity? Four, to what extent do CF parameters vary between resting state scans? While previous studies have examined cortico-cortical interactions in the early visual cortex during resting state [22, 36], our current study focuses on the application of the CF method. These previous studies used model-free approaches whereas the CFmethod is a model-based approach. To the extent that the model adequately describes the underlying neuronal activity, model-based approaches provide summary descriptions of aggregate neuronal activity, which is another reason to examine the application of the CF method to analyze resting state fMRI data.

14 2.2. Materials and Methods

2.2 Materials and Methods

2.2.1 Subjects We recruited four subjects with normal visual acuity (age: S1=26, S2=30, S3= 31, S4=40 years old). Experimental procedures were approved by the medical ethics committee of the University Medical Center Utrecht.

2.2.2 Stimulus Visual stimuli were presented by back-projection onto a 15.0x7.9 cm gamma- corrected screen inside the MRI bore. The subject viewed the display through prisms and mirrors, and the total distance from the subject’s eyes (in the scanner) to the display screen was 36 cm. Visible display resolution was 1024x538 pixels. The stimuli were generated in Matlab (Mathworks, Natick, MA,USA) using the PsychToolbox [8, 35]. The mapping paradigm consisted of drifting bar apertures at various orientations, which exposed a 100% contrast checkerboard moving parallel to the bar orientation. After each horizontal or vertical bar orientation pass, 30 s of mean-luminance stimulus were displayed. Sub- jects ﬁxated a dot in the center of the visual stimulus. The dot changed colors between red and green at random intervals. To ensure attention was maintained, subjects pressed a button on a response box every time the color changed (detailed procedures can be found in [13, 20]). The radius of the stimulation area covered 6.25 degrees of visual angle from the ﬁxation point.

2.2.3 Resting state During the resting state scans, the stimulus was replaced with a black screen and subjects closed their eyes. We chose this so that there was no visual input; neither from outside the stimulus area (hence eyes closed) nor from light com- ing through the eyelids (hence the black screen). The lights in the scanning room were oﬀ and blackout blinds removed light from outside the room. The room was in complete darkness.

2.2.4 Data acquisition Functional T2*-weighted 2D echo planar images were acquired on a 7 Tesla scanner (Philips, Best, Netherlands) using a 32 channel head coil at a voxel resolution of 1.98x1.98x2.00 mm, with a field of view of 190x190x50 mm. TR was 1500 ms, TE was 25 ms, and flip angle was 80 degrees. The volume orientation differs between subjects, though in all cases it was approximately perpendicular to the calcarine sulcus. High resolution T1-weighted structural images acquired at 7T using a 32 channel head coil at a resolution of 0.49x0.49x0.80 mm, with a field of view of 252x252x190 mm. TR was 7 ms, TE was 2.84 ms, and flip angle was 8 degrees. We compensated for intensity gradients across the image using an MP2RAGE sequence, dividing the T1 by a co-acquired proton density scan of the same resolution, with a TR of 5.8 ms, TE was 2.84 ms and flip angle

15 . C MRI

was 1 degree. In total, eight 240-volumes functional scans were acquired; comprising 5 resting state scans (RS) and 3 interleaved visual ﬁeld mapping scans (VFM). The ﬁrst scan was a RS scan. Physiological data were not collected.

2.2.5 Preprocessing First, the T1-weighted structural volumes were resampled to 1 mm isotropic voxel resolution. Gray and white matter were automatically segmented using Freesurfer and hand edited in ITKGray to minimize segmentation errors [45]. The cortical surface was reconstructed at the white/gray matter boundary and rendered as a smoothed 3D mesh [48]. Motion correction within and between scans was applied for the VFM and the RS scans [34]. To clean the resting scan signals from DC baseline drift and reduce high frequency nuisance from physiological variation, time courses were band pass filtered with a high-pass discrete cosine transform filter (DCT) with cut-off frequency of 0.01 Hz and a low-pass 4th order Butterworth filter with cutoff frequency of 0.1 Hz. Finally, functional data were aligned to the anatomical scans [34] and interpolated to the anatomical segmentation space.

2.2.6 Analysis 2.2.6.1 Population receptive field mapping Early visual areas V1, V2 and V3 were mapped using the population receptive field (pRF) method [13]. The method uses a parameterized forward model of the underlying neuronal population, a description of the hemodynamic response (a two-gamma HRF model) [7], and the stimulus aperture. The model we chose corresponds to a circular Gaussian characterized by three parameters: x and y (positions), and size (σ). A set of candidate population receptive field models are combined with the stimulus aperture to generate predictions of the neuronal responses each candidate pRF would produce. Subsequent convolu- tion of this predicted neuronal response time course with the HRF give a set of candidate predicted fMRI response time courses for each combination of pRF parameters. The best fitting predicted fMRI time courses and their associated pRF parameters are then chosen to summarize the response of each recording site [13].

2.2.6.2 Connective field mapping Connective field (CF) model parameters were estimated for both the VFM and RS scans using the connective field modeling method described by Haak et al. (2013). CF models summarize the activity of each recording site in a target region of interest (ROI) in terms of the aggregate activity contributed by a set of recording sites in a source ROI [18]. Specifically, the BOLD activity over a particular part of a source region (the CF) is integrated (summed) to yield the BOLD activity at a target recording site, whose neuronal response we are trying to describe. As we aim to determine the source CF for all target recording sites within an ROI simultaneously, we describe a target visual field map ROI (e.g.

16 2.2. Materials and Methods

V2 or V3). As candidate source CFs are limited to a particular visual field map, this is described as the source ROI (here, always V1). First, a discrete parameter space of 2-dimensional Gaussians of different candidate sizes (σ) is generated for each candidate location (each recording site inside the source ROI, V1), giving a set of candidate V1-referred CF models. In the next step, similarly to the pRF approach, a candidate predicted time course is generated for each candidate CF model by calculating the Gaussian weighted sum of the measured signals from the candidate CF (including the preferred recording site and its neighbors). These candidate time courses predictions are compared to the measured time course of each recording site in the target ROI (V2 and V3), and the best fitting prediction and its associate V1-referred CF parameters are chosen for each target recording site. Furthermore, because CF preferred locations in V1 cortical surface are associated with preferred visual field positions during pRF mapping, coordinates in visual space can be inferred for target recording sites. This allows the reconstruction of visuotopic maps even in the absence of stimuli. Note that the size of a CF represents the Gaussian spread along the cortical surface (mm) and is defined as the shortest path distance between pairs of vertices in the 3D mesh associated with the gray/white matter border. The location and size of the ROIs are defined during pRF mapping. These parameters (location and size of the source ROI) may restrict CF position but not CF size. By emphasizing the spatial profile of functional connectivity, a CF allows to examine spatially localized connectivity patterns among brain areas. As with most functional connectivity measures, CF models do not infer the temporal order of the responses in target and source recording sites.

2.2.6.3 Discriminability criterion By emphasizing local over long-range functional connectivity, biologically inspired models like pRF and CF are generally robust to global effects (e.g. physiological noise). Nevertheless, evaluation of model significance can be frustrated by the noisy and non-stationary nature of the time series obtained from resting state. To overcome this issue and assess the statistical significance of connective field models estimated from the RS, we apply a strategy based on surrogate data testing. First, we distinguish the contribution of topographically organized BOLD co-fluctuations from spatially uncorrelated random BOLD fluctuations. This distinction allows defining a criterion in terms of model discriminability. In this context, we define discriminability as the distinction between topographically organized BOLD co-fluctuations and spatially uncorrelated random BOLD fluctuations. To determine model discriminability, we estimated null distributions from the variance explained (VE) of connective field models obtained from surrogate V1 BOLD time courses. To generate these surrogate BOLD signals, artificial time courses were produced with the iterative amplitude adjusted Fourier transform (iAAFT) method [41, 46]. This method randomizes the phase of the original signal, but preserves its autocorrelation, linear structure and amplitude distribution. The spatial correlation between BOLD time courses in the source region is lost but their fundamental statistical properties are preserved. Each CF model estimation was accompanied of an estimation based on surrogate time courses. For the present analysis, the

17 . C MRI

null distributions obtained from 240 volumes (each RS scan) are comparable across subjects and target ROIs (V2 and V3); therefore, we combined all estimates into one null distribution and used the 5th percentile as discrimination threshold. Second, we estimated the amount of data that is sufficient to discriminate RS-based CF models by examining the dependence of discrimination accuracy on data quantity. First, CF models were calculated for different amounts of RS data (both for original and for surrogate data). Segments of 40, 80, 120, 160, 200 and 240 volumes starting from the beginning of each RS scan were used. Next, VE estimates (adjusted for the degrees of freedom in each number of volumes) were grouped according to their corresponding seg- ment length, obtaining original and null VE distributions for each amount of volumes. These distributions allow the application of a receiver-operator characteristic (ROC) analysis. By assessing the performance of a binary classifier as its discrimination threshold is varied, ROC analysis provides quantitative measures of model discrimination performance. To discriminate CF models attributed to genuine BOLD co-fluctuations from those attributed to random BOLD activity, the corresponding VE cutoff threshold is moved from 0 to 1 across the original and the null distributions, producing a contingency matrix of true positives (hits), false positives (false alarms), true negatives (correct rejec- tions) and false negatives (miss). Using the contingency matrix, values of true positive rate (sensitivity) and false positive rate (1-specificity) are computed and plotted as ROC curves. In ROC space, a diagonal line corresponds to random discrimination. The area under the ROC curve (AUC) is commonly used to quantify classifier discriminability, with a value of 0.5 corresponding to random, and a value of 1 to perfect, classification. We choose informedness as our discriminability index, which corresponds to twice the area between the curve and the diagonal: 2*AUC-1 [19, 15]. It has the advantage that 0 represents random, and 1 perfect, classification. Finally, we estimated the dependence of discrimination accuracy on the VE cutoff threshold by calculating F1 score for each amount of volumes.

2.2.6.4 Spatial analysis In the spatial domain, we estimate CF size change and position scatter during RS using VFM-based size and position as reference. First, to assess CF position variability in the RS, we assume that connective ﬁelds are topographically organized. This implies that neuronal activity in neighboring cortical locations in the target ROI may correlate with neuronal activity in neighboring cortical locations inside the source ROI that represent the same portions of visual space, as shown by VFM. This assumption allows us to estimate position variability as position scatter of V1-referred CFs by calculating their displacement on the V1 cortical surface with respect to their VFM-based reference positions. We proceeded as follows: for each recording site in the target ROI, position scatter was calculated as the shortest distance along the cortical manifold between the VFM-based center position and the RS-based position. This distance was computed in millimeters using Dijkstra’s algorithm [12]. Estimates whose associated models scored a VE above discrimination threshold (VEthreshold = 0.35) were retained. To quantify the variability in position scatter for each subject

18 2.3. Results and each RS scan, the median (to assess tendency) and the median absolute deviation (MAD; to assess dispersion) were calculated for each RS scan and subject. To assess RS scan-to-scan variability, we also calculated these values for all RS scan pairs. In order to determine a possible influence of cortical distance (e.g. shared vasculature, spatial blurring), we compared position scatter to the distance between CF centers and their associated recording sites in the target area. We then compared position scatter as a function of VFM-based reference eccentricity. Finally, agreement in eccentricity estimates was quantified by calculating linear correlation coefficient for VFM- and RS-based eccentricities. Second, we examined differences in size for V1→V2 and V1→V3 models between RS- and VFM-based estimates. RS-based size estimates for V1→V2 and V1→V3 from all participants were grouped by map combination and compared to those obtained based on VFM using a two-sample Kolmogorov-Smirnov test (KS-test). Subsequently, we examined the relation of RS-based CF size as a function of VFM-reference eccentricity by binning eccentricity in bins of 1 degree and calculating linear fits over the mean with bootstrapped confidence intervals (1000 iterations).

2.3 Results

2.3.1 Deriving connective field models based on resting state fMRI data Our first analysis concerned two questions: whether CF models could be obtained in presence of substantial physiological measurement noise; and, if the models obtained could be discriminated based on the contribution of genuine spontaneous BOLD co-fluctuations. Figure 2.1 shows the distributions of variance explained (VE) for actual (blue) and surrogate (black) data. We used the VE of CFs obtained from surrogate data as null-distribution (240 volumes, TR: 1.5 s). The VE cutoff threshold was estimated based on the 5th percentile ofthe null-distributions and lies around ∼ 0.35 for all subjects. The majority of the models have a VE that exceeds this cutoff threshold. Importantly, this analysis demonstrates that the estimation of CF models based on genuine spontaneous BOLD co-fluctuations is possible even in presence of substantial physiological measurement noise. Nevertheless, we cannot determine the effect that these confounds exert in the estimation of CF parameters. In addition, we examined the dependence of discrimination accuracy on the amount of volumes included in the analysis. To do so, we calculated VE (adjusted for degrees of freedom) for actual and surrogate data for various amounts of volumes and applied a ROC analysis. Figure 2.2 summarizes the results of the analysis for a single subject (Subject 3). First, it shows the VE distributions for actual (black) and surrogate data (red) as a function of the amount of volumes included in the analysis. VE drops with the number of volumes, but drops more sharply for the surrogate data (Figure 2.2A). The resulting ROC curves are shown in Figure 2.2B; they show detection probability as a function of false alarm probability for each amount of volumes. Detection probability in-

19 . C MRI

Figure 2.1: Distributions of variance explained. Histogram of relative frequency of recording sites in the target area as a function of the variance explained by their respective connective field models in source obtained during resting state. The black distribution represents the hypothesis of non-discriminability (noisy and spatially uncorrelated signals obtained with the IAAFT method) and was generated by fitting connective field models to surrogate BOLD signals. The blue distribution illustrates a typical outcome for an actual resting state scan.

creases with the amount of volumes (Figure 2.2C). Figure 2.2D shows discrimination accuracy (F1 score) as a function of the VE threshold for each amount of volumes analyzed. This analysis also indicates that CF modeling could be based on even shorter scan periods with retaining reasonable discrimination accuracy. However, fewer models are expected to lie above threshold. Finally, it must be noted that, even though this analysis provides a strategy to optimize modeling accuracy by adjusting the VE cutoﬀ threshold, in the remaining analysis we use a threshold of VEthreshold = 0.35, which corresponds to the 5th percentile of the null-distribution obtained after grouping the VE of surrogate RS-based models from all scans and subjects.

2.3.2 Spatial aspects of resting state connective ﬁeld map estimation The next question we address is whether the topographical maps basedonRS data have similar characteristics as the one based on VFM data (our current reference). Also, how variable are the results between RS scans? To provide an impression of this variability, Figure 2.3 shows both VFM and RS derived CF maps for a single participant (maps for other participants are shown in supplementary materials). V2 and V3 CF parameter maps (V1-referred) are plotted on a smoothed 3D mesh representing gray matter along the cortical surface. Ec- centricity, polar angle and size (σ) are plotted in three columns. In top row of panels, CF parameters estimated based on VFM data are shown. These maps serve as our reference. In the lower rows of panels, these same parameters are plotted for all RS scans. As shown previously [18], the VFM derived maps show a clear visuotopic organization (note that in the context of CF modeling,

20 2.3. Results

Figure 2.2: Overall modeling performance characteristics for a single subject. (A) De- pendence of explained variance on the amount of volumes. The whisker box-plots illustrates the distributions of explained variance adjusted by degrees of freedom. Red distributions were obtained from RS data and black distributions from surrogate RS data. The central mark is the median, the edges of the box are the 25th and 75th percentiles, and the whiskers indicate the most extreme data points with an interquartile range of 1.5 (Tukey box-plot). (B) Receiver operating characteristic (ROC) curves corresponding to each amount of volumes. (C) Discrim- inability increases as a function of the amount of volumes (we choose discriminability index in the form of informedness: 2*AUC-1). (D) Discrimination accuracy (F1 score) as a function of the adjusted VE cut-oﬀ threshold for each amount of volumes (colors as in B). Data are for → V1 V2 CF models from subject 3.

eccentricity and polar angle maps are inferred from a pRF mapping and associated to each recording site in the source region, in this case V1). In some RS scans eccentricity and polar angles maps resemble the VFM-based reference, although some variability can be observed (Figure 2.3, RS4, RS5). To quantify the variability of the individual maps, the median position displacement in CF cortical location (relative to the VFM reference and between all RS scan pairs; in mm) and the MAD were calculated for RS1 to RS5 (values are reported in the legend of Figure 2.3). These values conﬁrm the visual impression that RS4 and RS5 most clearly resemble the visuotopic organization observed in the VFM-based maps (results are shown for participant 3, those for the other participants are shown in the supplementary material). Figure 2.4A plots the change in V1-referred CF center position between RS

21 . C MRI

Figure 2.3: Visualization of connective field maps for a single subject. From left to right: eccentricity, polar angle and size. Top panel corresponds to visual field mapping (VFM) based estimates. Lower panels show parameter estimates for each resting state (RS) scan. → For V1 V2 CF models, the position displacement in CF cortical location (in mm) between VFM and RS-based estimates for RS1 to RS5 is: median (MAD) = 10.0 (5.4); 8.5 (5); 5.8 (3.7); 3.8 (3.4); and 4.1 (3.0), respectively (total = 5.4 (3.9). Corresponding position displacement values between RS4 and RS5 (the RS scans with lowest displacement: 4.1 (3.1); between RS1 and RS2 (the RS scans with highest displacement): 8.5 (5.8); between RS1 and RS4: → 10.5 (6.6); when grouping results for all RS scan pairs: 8.6 (5.9). For V1 V3 CF models, the corresponding values are: 13.6 (6.3); 14.4 (6.8); 7.9 (5.4); 6.7 (5.5); and 7.1 (4.2) (total = 8.7 (5.5)). Eccentricity and polar angle are inferred from V1 pRF mapping (see methods for de- → → tails). Data are for V1 V2 and V1 V3 models estimated for subject 3 (data for other subjects included in supplementary materials). A threshold of VEthreshold = 0.35 was applied. Median cortical displacements reflect the agreement between RS and VFM maps and between different RS maps.

22 2.3. Results

Figure 2.4: Position scatter for V1-referred CFs for a single subject. (A) Joint histogram of cortical displacement in V1-referred CF centers as a function of adjusted explained variance. The goodness of fit tends to decrease with larger displacements (colorbar depicts frequency of voxels after grouping data from all RS scans; the number of voxels that entered the analysis is: 1622 for → → V1 V2 and 1467 for V1 V3). (B) Position scatter as a function of the distance from the target → → voxel. A cortical distance effect can be seen in V1 V2 (R=0.90, p<0.0001) but not in V1 V3 (R=0.11, p<0.0001 (C) No systematic deviations from the median distance are observed for eccentricity (data was binned in eccentricity bins of 0.25 degrees). Points represent the median of each bin and error-bars the median absolute deviation for the corresponding bin. (D) There → is good agreement between RS-based eccentricity and VFM reference eccentricity (V1 V2: R → = 0.97, p<0.0001; V1 V3: R=0.70, p<0.0001) (data was binned in eccentricity bins of 0.25 deg. Points represent the median of each bin and error-bars the median absolute deviation for the corresponding bin). Data are from subject 3. A cutoff threshold of VEthreshold = 0.5 (F1 ∼ 0.85) was applied in (B), (C) and (D). and VFM-based reference position as a function of VE (of the RS model). CFs with higher VE show smaller cortical displacements. The majority of CFs (as indicated by the heat map) have a high VE and show relatively small displacements. Figure 2.4B shows a distance effect for V1→V2 (R=0.90, p<0.0001) but not for V1→V3 (R=0.11, p<0.0001). Figure 2.4C shows that there are no systematic deviations from the median cortical displacement as a function of eccentricity. Figure 2.4D shows a good agreement between RS- and VFM-based eccentricities (V1→V2: R = 0.97, p<0.0001; V1→V3: R=0.70, p<0.0001). Figure 2.5 shows VFM and RS-based V1-referred CF size distributions for V2 and V3 (data grouped over all scans and participants, N=4). RS-based CF size tend to be smaller than those estimated based on VFM data (V1→V2: p<0.0001, KS-test = 0.240; V1→V3: p<0.0001, KS-test = 0.0001). Moreover, we cannot confirm a difference in RS-based CF size estimates for V1→V2 or V1→V3 (p =0.0065, KS-test = 0.015). Figure 2.6 plots the relationship between CF size and eccentricity for VFM- and RS-based estimates. The left panel shows that VFM-based CF size estimates for V1→V2 do not increase significantly with eccentricity (black line),

23 . C MRI

Figure 2.5: V1-referred connective field size during visual field mapping (VFM) and resting state (RS) scans grouped over participants (N=4). Resting state based CF size → is generally smaller than their visual field mapping based CF size (V1 V2: p<0.0001, KS- → test = 0.240; V1 V3: p<0.0001, KS-test = 0.0001). CF size does not increase in the visual hierarchy when measured during resting state (p = 0.0065, KS-test = 0.015). A cutoff threshold of VEthreshold = 0.35 was applied.

Figure 2.6: Relation between eccentricity and V1-referred connective field size in visual areas V2 (black) and V3 (yellow) grouped over participants (N=4). Resting state based CF size estimates do not increase with eccentricity. Eccentricity was binned in intervals of 1 deg. Dots indicate the mean of VE weighted CF size for each bin. Linear fits were calculated for these means. Dashed lines correspond to the 95% bootstrap confidence interval of the linear fit (1000 iterations). A cutoff threshold VEthreshold = 0.35 was applied.

24 2.4. Discussion whereas those for V1→V3 do (yellow line). The right panel shows that RS- based CF size for V2 (black line) and V3 (yellow line) do not increase signi- ﬁcantly with eccentricity. Together, the analyses shown in Figures 2.5 and 2.6 show that RS-based CF size estimates are smaller than those estimated based on VFM. In RS, CF size does not appear to increase with eccentricity, neither within the visual hierarchy.

2.4 Discussion

2.4.1 Connective field models can be estimated based on resting state data We have shown that connective field (CF) modeling can be based on resting state data. This indicates that spontaneous BOLD co-fluctuations in the early visual cortex state preserve fine-grained topographic connectivity structure. While this preservation of topographic connectivity corroborates results of previous studies [22, 36] our study goes beyond these by examining both the topography and the spatial properties of the functional connections. In order to assess the statistical significance of our CF estimates, we determined a VE cutoff threshold taking into account the VE of CF models based on surrogate data (Figure 2.1). Rather than using an arbitrary threshold or selecting a comparison region that is not functionally connected —as is common practice— we created ”null” distributions to quantitatively assess model discrimination. This involves disrupting the phase correlations across recording sites in the source ROI in order to destroy the local structure of BOLD co-fluctuations. Further- more, we examined the dependence of discrimination accuracy on the amount of data and found six minutes of scanning (240 volumes using a TR of 1.5 s at 7T) to be more than sufficient to achieve good discrimination (Figure 2.2).

2.4.2 Agreement between resting state and visual field mapping based connective field parameters Although data obtained during resting state (RS) provide different information than data obtained during stimulation, a comparison of the maps estimated from RS to those estimated based on VFM reveals a fairly close agreement between the two (Figure 2.3). Some RS maps show patterns of visuotopic organization that agree well with their VFM reference (Figure 2.3, RS4, RS5). Nevertheless, we observed substantial variability in CF model parameters for different RS scans. We quantified the degree of agreement by measuring CF position scatter as the cortical displacement between RS- and VFM-based CF cortical positions and show that the median cortical displacement reflects the agreement observed in Figure 2.3 (data for other subjects are shown in supplementary materials). Besides the observed variability in visuotopic organization, CF size estimates obtained for RS scans were generally smaller than those obtained for VFM (Figure 2.5). Moreover, contrary to estimates based on VFM, RS-based CF size did not increase with eccentricity neither throughout the visual hierarchy (Figure 2.6).

25 . C MRI

2.4.3 Spatial changes: possible mechanisms In the absence of visual input, changes in CF size and variability in CF position may reflect a reduction in the amount of spatial integration and selectivity, respectively. Possible mechanisms underlying these changes in CFs may involve temporal restructuring of corticothalamic network activity in a state-dependent way [33, 52, 2, 9], as well as intracortical processing mediated by horizontal connections and feedback signals from higher cortical stages [38, 44, 29, 6, 40]. Decreased corticothalamic feedback and cortical lateral inhibition in the absence of visual input likely plays a role in the shrinkage of CFs, as well as in the reduced visuotopic organization observed of the higher-scatter CF maps. These input changes might adjust the balance between excitation and inhibition in cortical neuronal populations that eventually shapes cortico-cortical connectivity as a function of stimulation, behavioral context and physiological state [27, 28, 44, 38, 32, 42, 51, 16, 17]. During resting state, a variety of ongoing processes may modulate connectivity between visual areas. In particular, the transitional period from wakefulness to sleep leads to a progressive inhibition of synaptic transmission through thalamic relay neurons [44, 29], which is another possible cause to the changes observed. Another reason to speculate that there may be differences between the RS and VFM results is related to the origin of the BOLD signal. Given that the majority of the brain’s energy budget is devoted to ongoing intrinsic activity (e.g. resting state), the metabolic costs of the adjustment between excitation and inhibition may reflect in the BOLD signal. The relative contribution of excitatory and inhibitory signals to the BOLD signal changes between the RS and VFM scans. Inhibitory functions, which may be supported more by oxidative mechanisms than excitatory signaling, may contribute less to the measured BOLD signal [10]). As a consequence, resting state BOLD co-fluctuations may provide a different picture of the neuronal connections.

2.4.4 Limitations and future directions The current study assesses connective field properties in four healthy participants. Even though the results are consistent between participants, further studies involving more participants are advised. Moreover, the CF models were estimated based on entire RS scans. As such, they only estimate average CF properties and do not capture potential temporal variations in these. To establish the possible neuronal mechanisms underlying the observed changes in CF properties, further research is still necessary. In its current implementation, the present method cannot determine the precise factors that contribute to this variability. Large-scale network interactions, physiological processes and measurement noise might all influence the variability observed. Important to note is, however, that biologically inspired methods like pRF and CF modeling that emphasize local connectivity are generally robust to global effects like physiological noise. In future studies, extending the present analysis with dynamic functional connectivity metrics [39, 26, 1, 24], might help to disclose relevant temporal and spatial repertories in various experimental conditions allowing to study phenomena that unfold over time, such as attention, contextual modulation and object

26 2.5. Supplementary Material recognition. Adding independent measures of neuronal activity like electroencephalography [53] or other neurophysiological recordings seems a promising path to capture relevant temporal variations in neuronal activity. Future analyses could also take into account simultaneously recorded physiological data and draining veins in the preprocessing of the data, as these are known to inﬂu- ence resting state functional connectivity estimates [3, 18, 22, 31, 50]. Lastly, it should be noted that some of the possible mechanisms underlying changes in CF properties are based on animal models [52, 44, 21, 29, 2, 51]. Because certain experimental manipulations are not possible in human subjects, com- parative approaches between humans and animal models are needed to bridge the gap in resting state fMRI investigations Hutchison2012-ko,Mantini2012- yx. Examining the correspondence of functional and anatomical connectivity in homologous brain architectures will help to further elucidate the mechanisms underlying neuronal activity.

2.4.5 Concluding remarks We have shown that CF estimates can be obtained based on RS data. We observed good agreement can be observed between RS- and VFM-based maps, and between diﬀerent RS-based maps 1. This implies that local functional connectivity in visual cortical areas during resting state, as measured with CF modeling, may reﬂect the underlying neuronal architecture. However, we found that CF estimates may vary between RS scans even for high VE scans. The present study cannot determine to what extent this variability is explained by genuine changes in the neuronal properties of the visual system or by various external sources of noise. Nevertheless, we show that neuronal properties such as CF maps and CF size can be derived from resting state data.

2.5 Supplementary Material Please refer to the online version of this article (doi.org/10.3389/fnins.2014.00339) for the supplementary material.

2.6 References [1] Elena A Allen, Eswar Damaraju, Sergey M Plis, Erik B Erhardt, Tom Eichele, and Vince D Calhoun. Tracking whole-brain connectivity dynamics in the resting state. Cereb. Cortex, 24(3):663–676, March 2014. [2] Ian M Andolina, Helen E Jones, Wei Wang, and Adam M Sillito. Cor- ticothalamic feedback enhances stimulus response precision in the visual system. Proc. Natl. Acad. Sci. U. S. A., 104(5):1685–1690, 30 January 2007. [3] R M Birn, Z S Saad, and P A Bandettini. Spatial heterogeneity of the nonlinear dynamics in the fMRI BOLD response. Neuroimage, 14(4): 817–826, October 2001.

1However, only part of the RS data supported the derivation of retinotopically organized maps.

27 . C MRI

[4] B B Biswal, J Van Kylen, and J S Hyde. Simultaneous assessment of flow and BOLD signals in resting-state functional connectivity maps. NMR Biomed., 10(4-5):165–170, June 1997. [5] M Boly, E Balteau, C Schnakers, C Degueldre, G Moonen, A Luxen, C Phillips, P Peigneux, P Maquet, and S Laureys. Baseline brain activity fluctuations predict somatosensory perception in humans. Proc. Natl. Acad. Sci. U. S. A., 104(29):12187–12192, 17 July 2007. [6] Eliã P Botelho, Cecília Ceriatte, Juliana G M Soares, Ricardo Gattass, and Mario Fiorani. Quantification of early stages of cortical reorganiza- tion of the topographic map of V1 following retinal lesions in monkeys. Cereb. Cortex, 24(1):1–16, January 2014. [7] Geoffrey M. Boynton, Stephen A. Engel, Gary H. Glover, and David J. Heeger. Linear Systems Analysis of Functional Magnetic Resonance Ima- ging in Human V1. Journal of Neuroscience, 16(13):4207–4221, July 1996. [8] D H Brainard. The psychophysics toolbox. Spat. Vis., 10(4):433–436, 1997. [9] Juliane Britz and Christoph M Michel. State-dependent visual processing. Front. Psychol., 2:370, 16 December 2011. [10] György Buzsáki, Kai Kaila, and Marcus Raichle. Inhibition and brain work. Neuron, 56(5):771–783, 6 December 2007. [11] Gustavo Deco, Viktor K Jirsa, and Anthony R McIntosh. Emerging concepts for the dynamical organization of resting-state activity in the brain. Nat. Rev. Neurosci., 12(1):43–56, January 2011. [12] E W Dijkstra. A note on two problems in connexion with graphs. Numer. Math., 1(1):269–271, 1959. [13] Serge O Dumoulin and Brian A Wandell. Population receptive field estimates in human visual cortex. Neuroimage, 39(2):647–660, 15 January 2008. [14] S A Engel, G H Glover, and B A Wandell. Retinotopic organization in human visual cortex and the spatial precision of functional MRI. Cereb. Cortex, 7(2):181–192, March 1997. [15] Tom Fawcett. An introduction to ROC analysis. Pattern Recognit. Lett., 27(8):861–874, 2006. [16] Adam S Greenberg, Timothy Verstynen, Yu-Chin Chiu, Steven Yantis, Walter Schneider, and Marlene Behrmann. Visuotopic cortical connectivity underlying attention revealed with white-matter tractography. J. Neur- osci., 32(8):2773–2782, 22 February 2012. [17] Koen V Haak, Frans W Cornelissen, and Antony B Morland. Population receptive field dynamics in human visual cortex. PLoS One, 7(5):e37686, 23 May 2012.

28 2.6. References

[18] Koen V Haak, Jonathan Winawer, Ben M Harvey, Remco Renken, Serge O Dumoulin, Brian A Wandell, and Frans W Cornelissen. Con- nective field modeling. Neuroimage, 66:376–384, 1 February 2013. [19] J A Hanley and B J McNeil. The meaning and use of the area undera receiver operating characteristic (ROC) curve. Radiology, 143(1):29–36, April 1982. [20] Ben M Harvey and Serge O Dumoulin. The relationship between cortical magnification factor and population receptive field size in human visual cortex: constancies in cortical architecture. J. Neurosci., 31(38):13604– 13612, 21 September 2011. [21] S Shuichi Haupt, S Shuichi Haupt, Friederike Spengler, Robert Huse- mann, and Hubert R Dinse. Receptive field scatter, topography and map variability in different layers of the hindpaw representation of rat somatosensory cortex. Exp. Brain Res., 155(4):485–499, 2004. [22] Jakob Heinzle, Thorsten Kahnt, and John-Dylan Haynes. Topographic- ally specific functional connectivity between visual field maps in the human brain. Neuroimage, 56(3):1426–1436, 1 June 2011. [23] D H Hubel and T N Wiesel. Receptive fields, binocular interaction and functional architecture in the cat’s visual cortex. J. Physiol., 160(1):106– 154, 1962. [24] R Matthew Hutchison, Thilo Womelsdorf, Elena A Allen, Peter ABan- dettini, Vince D Calhoun, Maurizio Corbetta, Stefania Della Penna, Jeff H Duyn, Gary H Glover, Javier Gonzalez-Castillo, Daniel A Handwerker, Shella Keilholz, Vesa Kiviniemi, David A Leopold, Francesco de Pasquale, Olaf Sporns, Martin Walter, and Catie Chang. Dynamic functional connectivity: promise, issues, and interpretations. Neuroimage, 80:360–378, 15 October 2013. [25] R Matthew Hutchison, Thilo Womelsdorf, Joseph S Gati, Stefan Ever- ling, and Ravi S Menon. Resting-state networks show dynamic functional connectivity in awake humans and anesthetized macaques. Hum. Brain Mapp., 34(9):2154–2177, September 2013. [26] Vesa Kiviniemi, Tapani Vire, Jukka Remes, Ahmed Abou Elseoud, Tuomo Starck, Osmo Tervonen, and Juha Nikkinen. A sliding time- window ICA reveals spatial variability of the default mode network in time. Brain Connect., 1(4):339–347, 2011. [27] S M Kosslyn, W L Thompson, I J Kim, and N M Alpert. Topographical representations of mental images in primary visual cortex. Nature, 378 (6556):496–498, 30 November 1995. [28] D Lehmann, W K Strik, B Henggeler, T Koenig, and M Koukkou. Brain electric microstates and momentary conscious mind states as building blocks of spontaneous thinking: I. visual imagery and abstract thoughts. Int. J. Psychophysiol., 29(1):1–11, June 1998.

29 . C MRI

[29] Rodolfo R Llinás and Mircea Steriade. Bursting of thalamic neurons and states of vigilance. J. Neurophysiol., 95(6):3297–3308, June 2006. [30] Nikos Logothetis. Object vision and visual awareness. Curr. Opin. Neuro- biol., 8(4):536–544, 1998. [31] Nikos K Logothetis, Yusuke Murayama, Mark Augath, Theodor Steffen, Joachim Werner, and Axel Oeltermann. How not to study spontaneous activity. Neuroimage, 45(4):1080–1089, 2009. [32] Julio C Martinez-Trujillo and Stefan Treue. Feature-based attention increases the selectivity of population responses in primate visual cortex. Curr. Biol., 14(9):744–751, 4 May 2004. [33] D N Mastronarde. Correlated firing of retinal ganglion cells. Trends Neurosci., 12(2):75–80, February 1989. [34] O Nestares and D J Heeger. Robust multiresolution alignment of MRI brain volumes. Magn. Reson. Med., 43(5):705–715, May 2000. [35] D G Pelli. The VideoToolbox software for visual psychophysics: trans- forming numbers into movies. Spat. Vis., 10(4):437–442, 1997. [36] Mathijs Raemaekers, Wouter Schellekens, Richard J A van Wezel, Nat- alia Petridou, Gert Kristo, and Nick F Ramsey. Patterns of resting state connectivity in human primary visual cortical areas: a 7T fMRI study. Neuroimage, 84:911–921, 1 January 2014. [37] M E Raichle, A M MacLeod, A Z Snyder, W J Powers, D A Gusnard, and G L Shulman. A default mode of brain function. Proceedings of the National Academy of Sciences, 98(2):676–682, 2001. [38] R P Rao and D H Ballard. Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. Nat. Neurosci., 2(1):79–87, January 1999. [39] Unal Sakoğlu, Godfrey D Pearlson, Kent A Kiehl, Y Michelle Wang, Andrew M Michael, and Vince D Calhoun. A method for evaluating dynamic functional network connectivity and task-modulation: application to schizophrenia. MAGMA, 23(5-6):351–366, December 2010. [40] Michael C Schmid and Georgios A Keliris. Filling-in versus filling-out: patterns of cortical short-term plasticity. Trends Cogn. Sci., 18(7):342–344, July 2014. [41] T Schreiber and A Schmitz. Improved surrogate data for nonlinearity tests. Phys. Rev. Lett., 77(4):635–638, 22 July 1996. [42] Scott D Slotnick, William L Thompson, and Stephen M Kosslyn. Visual mental imagery induces retinotopically organized activation of early visual areas. Cereb. Cortex, 15(10):1570–1583, October 2005.

30 2.6. References

[43] A T Smith, K D Singh, A L Williams, and M W Greenlee. Estimating receptive field size from fMRI data in human striate and extrastriate visual cortex. Cereb. Cortex, 11(12):1182–1190, December 2001. [44] M Steriade. Corticothalamic resonance, states of vigilance and mentation. Neuroscience, 101(2):243–276, 2000. [45] P C Teo, G Sapiro, and B A Wandell. Creating connected representations of cortical gray matter for functional MRI visualization. IEEE Trans. Med. Imaging, 16(6):852–863, 1997. [46] Victor Venema, Steffen Meyer, Sebastián Gimeno García, Anke Kniffka, Clemens Simmer, Suzanne Crewell, Ulrich Löhnert, Thomas Trautmann, and Andreas Macke. Surrogate cloud fields generated with the iterative amplitude adapted fourier transform algorithm. Tellus A, 2006. [47] J D Victor, K Purpura, E Katz, and B Mao. Population encoding of spatial frequency, orientation, and color in macaque V1. J. Neurophysiol., 72(5): 2151–2166, November 1994. [48] Brian A Wandell, Serge O Dumoulin, and Alyssa A Brewer. Visual field maps in human cortex. Neuron, 56(2):366–383, 25 October 2007. [49] Zheng Wang, Li Min Chen, László Négyessy, Robert M Friedman, Ar- abinda Mishra, John C Gore, and Anna W Roe. The relationship of anatomical and functional connectivity to resting-state connectivity in primate somatosensory cortex. Neuron, 78(6):1116–1126, 19 June 2013. [50] Jonathan Winawer, Hiroshi Horiguchi, Rory A Sayres, Kaoru Amano, and Brian A Wandell. Mapping hv4 and ventral occipital cortex: the venous eclipse. J. Vis., 10(5):1, 1 May 2010. [51] Thilo Womelsdorf, Katharina Anton-Erxleben, and Stefan Treue. Re- ceptive field shift and shrinkage in macaque middle temporal area through attentional gain modulation. J. Neurosci., 28(36):8934–8944, 3 September 2008. [52] F Wörgötter, K Suder, Y Zhao, N Kerscher, U T Eysel, and K Funke. State-dependent receptive-field restructuring in the visual cortex. Nature, 396(6707):165–168, 12 November 1998. [53] Han Yuan, Vadim Zotev, Raquel Phillips, Wayne C Drevets, and Jerzy Bodurka. Spatiotemporal dynamics of the brain at rest–exploring EEG microstates as electrophysiological signatures of BOLD resting state networks. Neuroimage, 60(4):2062–2072, 1 May 2012.

Phase-synchronization-based parcellation of resting state fMRI signals reveals topographically organized clusters in early visual cortex 3

Resting-state fMRI is widely used to study brain function and connectivity. However, interpreting patterns of resting state (RS) fMRI activity remains challenging as they may arise from different neuronal mechanisms than those triggered by exogenous events. Currently, this limits the use of RS-fMRI for understanding cortical function in health and disease. Here, we examine the phase synchronization (PS) properties of blood-oxygen level dependent (BOLD) signals obtained during visual field mapping (VFM) and RS with 7T fMRI. This data- driven approach exploits spatiotemporal covariations in the phase of BOLD recordings to establish the presence of clusters of synchronized activity. We find that, in both VFM and RS data, selecting the most synchronized neighboring recording sites identifies spatially localized PS clusters that follow the topographic organization of the visual cortex. However, in activity obtained during VFM, PS is spatially more extensive than in RS activity, likely reflecting stimulus-driven interactions between local responses. Nevertheless, the similarity of the PS clusters obtained for RS and stimulus-driven fMRI suggests that they share a common neuroanatomical origin. Our finding justifies and facilitates direct comparison of RS and stimulus-evoked activity.

3.1 Introduction Resting state (RS) fMRI is a popular method to examine brain function and connectivity in health and disease. Blood-oxygen level dependent (BOLD) ﬂuc- tuations exhibit extensive spatial structure during RS, making RS-fMRI a valuable measure of brain function and metabolism [55, 71, 76]. It is also becom-

33 . S MRI

ing an increasingly important measure in clinical diagnosis [18, 36]. However, the physiological mechanisms underlying RS-fMRI activity are still poorly understood. Responses at diﬀerent recording sites can be correlated as a result of neuroanatomical connections, or metabolic and hemodynamic relationships [12, 42, 41, 66]. Therefore, interpretation of function and connectivity patterns estimated from RS remains challenging. Furthermore, high-resolution fMRI and methods to analyze patterns of neuronal responses on the cortical surface are becoming increasingly common, particularly in vision research [1, 15, 61, 21]. This creates a need for methods that can adequately describe the interactions between structural connections, neuronal metabolism and hemodynamics at the cortical surface level.

Previous studies have examined the structure of occipital BOLD fluctuations in RS and shown that they are influenced by several factors: local spatial fluctuations within visual field maps that may reflect aggregate population activity [10, 25, 48], widespread iso-eccentric fluctuations within and between visual areas [2, 76, 11] that may partly reflect the transition between foveal and peripheral regions [6], and visuotopically organized cofluctuations between visual areas [54, 20, 6, 25], which may reflect links between neuronal populations sharing the same visual field selectivity within and across areas. However, the neuronal correlates of these components, their functional relevance, and their interactions are not yet fully understood. For this reason, demonstrating that patterns of BOLD responses derived from both stimulus-driven and spontaneous brain activity can be parcelled into similar, spatially specific motifs would provide valuable insights into the neuronal correlates underlying RS-fMRI activity.

Here, we take a new approach by investigating the spatial structure that arises from local covariations in the phase of low-frequency BOLD fluctuations within early visual cortex. First, we ask whether clustering based on such phase synchronization (PS) analysis establishes modular spatial structure. Next, we ask whether the synchronization clusters derived from signals recorded during RS and VFM are similar. Finally, we went on to examine whether synchronization between clusters with similar visual field position selectivity across visual areas reflected the underlying layout of neuroanatomical connections, which may provide clues to their origin.

To establish and quantify the spatial structure of PS we take a modeling approach in which we analyze how PS and clustering change with cortical and visuotopic distance. Together, this lets us characterize the spatial extent of local synchrony and parcelate recording sites into topographically localized clusters.

To preview our main ﬁnding, we ﬁnd that clustering the phase covariations in both RS and VFM results in discrete —topographically organized— synchronization clusters that are stable across sessions and conditions (VFM and RS). This suggests that the synchronization clusters obtained from both RSand VFM activity are subserved by common neuroanatomical underpinnings.

34 3.2. Methods

3.2 Methods

3.2.1 Data The empirical data used in the main body of this paper has previously beende- scribed in Gravel et al. [20]. The main results were replicated using a second data set that has previously been described in Raemaekers et al. [54]. The methods and results for this second dataset are presented in the supplementary information.

3.2.1.1 Participants Data was acquired for four participants with normal visual acuity (2 females, 2 males, age 26-40). Experimental procedures were approved by the medical ethics committee of the University Medical Center Utrecht.

3.2.1.2 Visual field mapping stimulus Visual stimuli were presented by back-projection onto a 15.0 × 7.9 cm gamma- corrected screen inside the MRI bore. The subject viewed the display through prisms and mirrors, and the total distance from the subject’s eyes (in the scanner) to the display screen was 36 cm. Visible display resolution was 1024 × 538 pixels. The stimuli were generated in Matlab (Mathworks, Natick, MA,USA) using the PsychToolbox [9, 49] . The visual field mapping paradigm consisted of drifting bar apertures at various orientations, which exposed a 100% contrast checkerboard (switching contrast at 5 Hz) moving parallel to the bar orientation. After each horizontal or vertical bar orientation pass, 30 s of mean-luminance stimulus were displayed. Throughout the VFM, subjects fixated a dot inthe center of the visual stimulus. The dot changed color between red and green at random intervals. To ensure attention was maintained, subjects pressed a button on a response box every time the color changed. Detailed procedures can be found in Dumoulin and Wandell [15] and Harvey and Dumoulin [24]. The radius of the stimulation area covered 6.25 deg (eccentricity) of visual angle from the fixation point.

3.2.1.3 Resting state During the resting state scans, the stimulus was replaced with a black screen and subjects closed their eyes. The lights in the scanning room were oﬀ and blackout blinds removed light from outside the room. The room was in complete darkness. Thus, visual stimulation was minimized. The subjects were instructed to think of nothing in particular and not to fall asleep.

3.2.1.4 MRI acquisition Functional T2*-weighted 2D echo planar images were acquired on a 7 Tesla scanner (Philips, Best, Netherlands) using a 32 channel head coil at a voxel resolution of 1.98 × 1.98 × 2.00 mm, with a ﬁeld of view of 190 × 190 × 50 mm .

35 . S MRI

◦ TR was 1500 ms, TE was 25 ms, and flip angle was set to 80 . The volume orientation was approximately perpendicular to the calcarine sulcus. In total, eight 240 volumes functional scans were acquired, comprising 5 resting state scans (RS) interleaved with 3 VFM scans (first was an RS scan). High resolution T1- weighted structural images were acquired at a resolution of 0.49 × 0.49 × 0.80 252 × 252 × 190 mm, with a field of view◦ of mm. TR was 7 ms, TE was 2.84 ms, and flip angle was 8 . We compensated for intensity gradients across the image using an MP2RAGE sequence, dividing the T1 by a co-acquired proton density scan of the◦ same resolution, with a TR of 5.8 ms, TE was 2.84 ms, and flip angle was 1 . Physiological recordings were not collected.

3.2.1.5 Preprocessing First, the T1-weighted structural volumes were resampled to 1mm isotropic voxel resolution. Gray and white matter were automatically labeled using Free- surfer and labels were manually edited in ITKGray to minimize segmentation errors [64]. The cortical surface was reconstructed at the white/gray matter boundary and rendered as a smoothed 3D mesh [73]. Motion correction within and between scans was applied for the VFM and the RS scans [45]. Sub- sequently, data were aligned to the anatomical scans and interpolated to the anatomical segmentation space [45]. Instrumental drift was removed by detrending with a discrete cosine transform (DCT) filter with cutoff frequency of 0.01 Hz. In order to reduce nuisance from high frequency physiological variation, the detrended signals were filtered with a low-pass 4th order Butter- worth filter with cutoff frequency of 0.1 Hz. The resulting signals were used for population receptive field (pRF) modeling (Section 2.6.1). Additionally, in order to analyze narrow band low-frequency synchronization patterns (Section 2.6.2), detrended signals were band-pass filtered using two 4th order bandpass Butterworth filters with cutoff 0.04 Hz and 0.07 Hz [19].

3.2.2 Analysis 3.2.2.1 Visual field mapping Visual field maps V1, V2, and V3 were mapped using the pRF method [15]. This summarizes the visual field position to which each recording site responds as a circular Gaussian in visual space. An isotropic 2D Gaussian was chosen as pRF shape characterized by three parameters: x and y (position), and size (sigma). Here, a large set of candidate pRF models are combined with the stimulus aperture to generate predictions of the neuronal responses each candidate pRF would produce. This predicted neuronal response time course is convolved with the hemodynamic response function (a two-gamma HRF model) [8] to give a set of candidate predicted fMRI response time courses for each combination of pRF parameters. The best fitting predicted fMRI time course and its associated pRF parameters are then chosen to summarize the response of each recording site, together with the model variance explained to summarize the model goodness of fit [15]. Recording sites were excluded from subsequent

36 3.2. Methods analyses if their best ﬁtting pRF models explained less than 30% of response variance, or had visual ﬁeld eccentricities beyond 6 deg.

3.2.2.2 Estimating phase synchronization To establish the synchronization of low-frequency BOLD fluctuations, we estimated the phase locking values (PLV) of the BOLD signals of all pairs of recording sites within a visual field map. The PLV is a measure of phase synchronization widely described in the literature [63, 31] and it has been used in fMRI [19, 34, 52]. We first obtained estimates of the instantaneous phase of the band-pass filtered BOLD signal by computing the Hilbert transform and then taking the angle of the resulting analytical signal [33]. Data were band- pass filtered because instantaneous phase estimates thus obtained only admit clear physical interpretation for a narrow frequency band [19]. After obtaining instantaneous phase estimates for each recording site at each time point, we discarded the 5 first and 5 last time points (the first and last 7.5 seconds of recording) to avoid edge effects inherent to the Hilbert transform. Afterwards, we computed the PLV matrix by using the following equation: T 1 i∆ϕp,q (t) PLVp,q = e (3.1) T ∑ t=1

where PLVp,q represents the time averaged symmetric phase locking values matrix between the recording sites pairs p,q and ∆φp,q(t) is the instantaneous phase diﬀerence between any two recording sites at time t and T is the total duration of the recording.This matrix summarizes the average synchronization tendency of low-frequency BOLD ﬂuctuations over the scanning session.

3.2.2.3 Parcellation into synchronization clusters To identify the spatial structure that arises from local covariations in the phase of low-frequency BOLD fluctuations, we searched for groups of neighboring recording sites (contiguous along the grid defined by the cortical surface reconstruction) that were, on average, highly synchronous during the entire duration of a scan. This was achieved by selecting the 5% strongest entries in the PLV matrix. Afterwards, we pruned the 5% strongest entries in the PLV matrix by setting the synchronization values of recording sites beyond 3 mm distance to zero. The cortical distances between recording sites were estimated asthe shortest distance along the cortical surface manifold using Dijkstra’s algorithm [14] (Dijkstra, 1959). We choose a 3 mm distance threshold to emphasize short range (local, contiguous along the cortical surface) over long range (large-scale, between distant cortical folds) interactions. Pruning the PLV matrix before clustering helped minimize the contribution of long-range interactions while maintaining local interaction structure. This implies that two distant voxels could be assigned to the same cluster only if they were chained together through synchronized neighbors. Next, we clustered the pruned PLV matrix using the Louvain clustering algorithm [4]. The Louvain algorithm finds the optimal modular partition ofnon-

37 . S MRI

overlapping clusters by maximizing the number of within-cluster connections while minimizing the number of between-cluster connections. The algorithm depends on a scaling parameter that determines the scale of the modular partition, which we set to 1, the default value defined for classic modularity [59]. Due to the nondeterministic nature of clustering algorithms, slightly different solutions are obtained each time the algorithm is run. Therefore we applied an iterative approach. First, we run the Louvain clustering algorithm 10 times and then calculated an agreement matrix for all iterations. The entries in this matrix indicate the probability that a link was assigned to the same partition across all iterations. Next, we run the clustering algorithm another 10 times on the most probable entries of the agreement matrix (agreement >0.9). This approach converges to the most probable modular partition [3] for each VFM and RS scan. Lastly, we computed a consensus partition from the resulting modular partitions of the individual VFM and RS scans. For illustrative pur- poses, we render the consensus partitions obtained from all VFM and RS scans on inflated reconstructions of the cortical surface and in visual space (using the pRF positions of the recording sites in each cluster).

3.2.3 Reproducibility of the synchronization clusters To determine whether identified PS clusters were similar across scanning sessions, we computed the normalized mutual information (NMI) between modular partitions derived from different scans. First, we computed NMI values between all scan pairs of the same category (3 VFM and 5 RS). Second, we computed the NMI between the consensus RS and VFM partitions. To demonstrate that NMI values reflected genuine features of the data (similar spatial covariations) rather than methodological artifacts, we constructed a surrogate NMI distribution under the null hypothesis that high NMI values could be obtained from data in which there are no meaningful spatial correlations (e.g. spatially randomized data). This allowed us to compare empirical NMI estimates toa baseline. To construct the surrogate NMI distribution, we first generated 30 surrogate data sets for each visual area by permuting the indices of the grid defined by the cortical surface reconstruction. This removed the spatial structure of the data while keeping the autocovariance and temporal covariations intact. We choose to shuffle grid and not voxel space because shuffling before interpolating to grid- space can remove some extremes in the data, effectively smoothing it out while bringing everything towards the mean [22]. Second, we computed modular partitions for each surrogate data set as we did with the empirical data. Third, we computed the NMI between these surrogate partitions and the empirical partitions, obtaining a surrogate distribution of NMI values. Finally, we compared this distribution to the empirical NMI distributions using a Mann-Whitney test. This gave the probability of observing each empirical NMI distribution by chance. Additionally, we examined whether the size of the clusters derived from RS and VFM were similar by computing the cortical surface area of each cluster. Only recording sites directly on the white-gray matter boundary were considered for area computation. Cluster areas were then averaged for each visual field map

38 3.2. Methods and hemisphere, giving a total of 24 averages for VFM and RS. We used the Pearson correlation coeﬃcient between average cluster areas derived from RS and VFM data to compare their similarity.

3.2.4 Visuotopic organization of phase synchronization clusters To examine the visuotopic organization of clusters, we investigated how the probability of any two recording sites sharing a cluster changed as a function of the visual field distance between their population receptive fields (pRF). For all visual field maps, hemispheres and subjects, we computed the binomial probability of sharing a cluster for the RS-derived consensus partition for bins of 0.25 deg of visual field distance. The visual field distances between all recording pairs were estimated in the radial and angular directions as the difference in their corresponding pRF eccentricities and arc distance, respectively. The arc distance was computed by multiplying the pRF polar angle difference with the average pRF eccentricity of any two recording sites. To estimate how the range of cluster membership probability changes with radial and arc distance we fitted binomial distributions with the radial and arc distances as independent variables, and computed the decay factor of these probabilities. To ensure probabilities were computed between truly adjacent locations, only arc distances between iso-eccentric and radial distances between iso-angular locations were considered (a tolerance of 2 deg was used).

3.2.5 Spatial extent of phase synchronization We also determined the spatial extent of phase synchronization (PS) by estimating how it decreases with cortical and visuotopic distance. First, for each visual field map, subject and scan, we computed the average PLV for every 1mm increase in cortical distance. We then quantified the spatial extent of PS by drawing on a framework developed in geophysics for the empirical evaluation of spatial dependencies: the variogram [75, 48]. The variogram provides a measure of how much two samples vary depending on their distance. In order to adapt the method to describe the dependency of the PLV with cortical distance, we proceeded as follows. First, we defined an exponential model by centering an exponential function at zero distance and keeping its maximum amplitude at 1, the maximum possible value for the PLV. We then varied the decay factor (in mm of cortical distance), and the baseline PLV of this exponential model to fit the distance-binned PLV data. We choose the exponential model because the relation between functional connectivity and cortical distance is well described by an exponential decay [10] and because the model’s parameters have a direct physical interpretation, in millimeters of cortical distance and PS magnitude (PLV). This allowed us to summarize the spatial extent of the aggregate synchronous activity contributed by a set of neighboring recording sites (synchronization range) and their long range baseline interactions (baseline). One important difference between our approach and variogram analysis —as applied in geophysics— is that here we use the decay factor as a measure of the synchronization range. In a standard variogram analysis, the range is evaluated, which typically corresponds to 3 times the decay factor and roughly de-

39 . S MRI

scribes the region in which spatial correlations dissipate into randomness (here baseline). Synchronization range (the decay factor) and baseline were estimated for each subject, scan and visual area separately. Additionally, we examined how PS decreases with visuotopic distance. Sim- ilarly as previously described, for each visual ﬁeld map combination (V1-V1, V2-V2, V3-V3, V1-V2, V1-V3 and V2-V3), subject and scan, we computed the average PLV for every 1 deg increase in visuotopic distance. Visuotopic distances were computed using the following equation:

2 2 2 2 Dp,q = √rq + rq − 2rprq cos(θp − θq) (3.2)

where D(p,q) represents the visuotopic distance between the recording sites p,q. As explained previously, we modeled the range and baseline of PS in visual space by using an exponential decay function. To test for diﬀerences in range and baseline between VFM and RS derived estimates, we computed two sample t-tests. To test whether there was a correlation between synchronization range and the cortical areas (in mm) of the synchronization clusters in diﬀerent scans, we computed the Pearson correlation between the synchronization range and cluster cortical areas for data grouped across subjects, visual areas and scans (5 for RS and 3 for VFM).

3.2.6 Intra and inter-hemispheric cluster connectivity We examined whether clusters sharing similar visual field selectivity have higher PLV across areas and hemispheres. To answer this, we proceeded as follows. Clusters were first grouped over foveal and peripheral quadrants using the averaged eccentricity and polar angle pRF preferences of their cortical grid points. The transition between fovea and periphery was set to 2.2 deg. The grouping process resulted in a matrix of 24 ROIs, 4 for each area (V1, V2 and V3) in each hemisphere. Signals for each ROI were obtained by averaging the minimally preprocessed signals (after detrending) and filtering in the in the 0.04 - 0.07 Hz. From these signals, the PLV was computed resulting in 24 × 24 PLV matrices for each subject, scan and condition (12× hemisphere). These matrices were then averaged over scans and the resulting grand averages classified as intra- or inter-hemispheric functional connections. Subsequently, these connections were z-scored (separately) and evaluated for significance across subjects using permutations corrected for multiple comparisons. We then asked if the the resulting PLV-based functional connectivity matrices matched homotopic anatomical connections. For this purpose, we created a binary matrix with ones indicating homotopic connections between foveal and peripheral quadrants in different areas and hemispheres and used the Spearman correlation between the PLV-based connectivity matrices and the homotopic anatomical connectivity matrix.

40 3.3. Results

3.3 Results

3.3.1 Synchronization clusters derived from resting state and visual ﬁeld mapping are similar

We first asked two questions: 1) whether spatially localized patterns of functional connectivity within visual areas could be derived from resting state (RS) data; and 2) whether these patterns were similar to those evoked by visual field mapping (VFM) stimuli. Both in RS and VFM, clustering the 5% most synchronized neighbours revealed a modular network structure that maps to discrete cortical regions that group recording sites with similar eccentricity and polar angle preference. Figure 3.1 shows the parcellation based on the consensus clustering across individual scans in both RS and VFM for one participant. PS clusters obtained from one RS scan were significantly more similar to those from another RS scan than they were to clusters from surrogate RS data with recording sites randomly permuted (average V1 normalized mutual information (to empirical RS data was 0.744, while to surrogate RS data was 0.635. p <0.0001. All subjects show this result in V1, V2 and V3, as shown in supplementary table 1). Importantly, consensus cluster partitions obtained from VFM data were significantly more similar to clusters from RS data (= 0.745) than they were to clusters from surrogate RS data with recording site locations randomly permuted (= 0.623) (p <0.001) (Sup- plementary table 2). These results indicate that PS clusters were stable in space both across scanning runs and different conditions. The average cortical surface area of clusters derived from VFM and RS was strongly correlated (Pearson correlation coefficient r = 0.94, p <10 -11) but clusters were significantly larger in VFM than in RS (p = 0.0045) (Figure 3.2A).

3.3.2 Visuotopic organization of the synchronization clusters

Next, we examined the visuotopic organization of the synchronization clusters to test whether the clustering reflects neuronal response preferences of the recording sites. We quantified how the probability of any two recording sites sharing a cluster changed as a function of the radial and arc distances between recording sites. In both radial and angular directions, recording sites are more likely to share a cluster if their pRFs are close in visual space (Figures 3.2B & 3.2C). However, the probability of sharing a cluster declines faster with visual field distance in the angular than the radial direction. Therefore, the visual field extent of clusters is elongated along the radial direction both in RS (decay factor: 1.93 deg for the radial direction and 0.3 deg for the angular direction) and in VFM (decay factor: 1.65 deg for the radial direction and 0.25 deg for the angular direction). This systematic asymmetry in the relationship between visual position preferences and clustering cannot be straightforwardly explained by relationships to cortical location.

41 . S MRI

Figure 3.1: Synchronization clusters obtained from VFM and RS. Consensus synchronization clusters partitions across all VFM and RS scans of a single subject (results for other subjects are included in supplementary materials). Modules depicted in the cortical surface reconstruction and in visual space (their perimeter) using each recording site’s pRF position. Different clusters are shown in different colors, which correspond between cortical surface and visual field representations (recording sites within 0-6 deg of eccentricity). The number of clusters and colors may not match between RS and VFM. Data for other subjects are included in the supplementary material (Figure 3.1S).

3.3.3 Comparison of the spatial extent of phase synchronization between RS and VFM We next asked whether the spatial extent of phase synchronization differed between VFM and RS scans. To test this, we fit the relationship between the PLV and cortical distance as an exponential function. This help us quantifies the synchronization range and baseline (long-range) magnitude of the PLV as the decay factor and offset, respectively (Figure 3.3). Across subjects and scans, in V1, V2 and V3, the synchronization range spread over a larger cortical distance in VFM (mean (SD): V1 = 7.72 mm (0.75); V2 = 7.81 mm (1.13) and V3 = 7.92 mm (1.79)) than in RS (mean (SD): V1 = 3.19 mm (1.17); V2 = 2.97 mm (1.41) and V3 = 3.11 mm (0.86)). In VFM, baseline synchronization estimates were smaller (mean (SD): V1 = 0.24 (0.022); V2 = 0.37 (0.07) and V3 = 0.36 (0.09)) than those for RS (mean (SD): V1 = 0.46 (0.069); V2 = 0.45 (0.068) and V3 = 0.45 (0.11)). Differences between VFM and RS derived synchronization range and baseline were significant in all cases (two sample t-test: p <10-5 for V1; p = 0.0252 for V2; p = 0.0307 for V3). Next, we asked how phase synchronization changed between voxels, within and between areas, with close visuotopic selectivity in VFM and RS.Similarly to what we previously described, we fit an

42 3.3. Results

Figure 3.2: Spatial aspects of the synchronization clusters. (A) Correlation between cortical surface areas of VFM and RS derived clusters. Average areas of cluster were obtained for each visual field map in each subject hemispheres. The average cortical surface area of clusters was significantly correlated (r = 0.94, p <10-11) but nevertheless were significantly larger for estimates based on VFM or RS data (p = 0.0045, Wilcoxon signed rank test). As most data points are distributed above the unity line, this indicates that the synchronization clusters were slightly larger when derived from the VFM data. (B) In RS, shared cluster membership probability decreases with visual field distance. The probability of two recording sites sharing a synchronization cluster decreases with the visuotopic distance between the recording sites’ pRFs for both the radial (blue) and the angular direction ( red). The spread of shared cluster membership was greater along the radial direction (decay factor = 1.93 deg for RS and 1.65 deg for VFM) compared to the angular direction (decay factor = 0.3 deg for RS and 0.25 deg for VFM). Error bars correspond to the binomial probability and its confidence interval for bins of 0.25 deg of visuotopic distance. The continuous trace corresponds to the binomial probability fit of alldata within 0-6 deg of visuotopic distance. The binomial fits were obtained after grouping the data over areas, scans, subjects and conditions. (C) Same as in (B) but for the VFM data.

exponential function to the relationship between the PLV and visuotopic distance instead (Figure 3.4). Across subjects and scans, within V1, V2, V3 and between V1-V2, V1-V3 and V2-V3, the synchronization range spread over a larger visuotopic distance in VFM (mean (SD): V1 = 0.99 deg (0.14); V2 = 1 deg (0.13); V3 = 0.95 deg (0.19); V1-V2 = 0.97 deg (0.14); V1-V3 = 0.89 deg (0.15) and V2-V3 = 0.94 deg (0.14)) than in RS (mean (SD): V1 = 0.45 deg (0.17); V2 = 0.51 deg (0.16); V3 = 0.44 deg (0.22); V1-V2 = 0.42 deg (0.18); V1-V3 = 0.32 deg (0.18) and V2-V3 = 0.41 deg (0.21)). In VFM, baseline synchronization estimates were smaller (mean (SD): V1 = 0.21 (0.013); V2 = 0.22 (0.017); V3 = 0.24 (0.024); V1-V2 = 0.22 (0.013); V1-V3 = 0.23 (0.015) and V2-V3 = 0.23 (0.011)) than those for RS (mean (SD): V1 = 0.45 (0.08); V2 = 0.39 (0.08); V3 = 0.43 (0.11); V1-V2 = 0.39 (0.073); V1-V3 = 0.38 (0.074) and V2-V3 = 0.4 (0.095))). Differences between VFM and RS derived synchronization range and baseline were significant in all cases (two sample t-test: p <10-5 for all cases). Finally, we asked whether there was a correlation between synchronization range and the areas of the synchronization clusters in different RS scans. To this end, we computed the Pearson correlation between the synchronization range and the mean cluster areas for data grouped across subject, resting state scans and visual areas. For cortical distances, we found no significant cor-

43 . S MRI

Figure 3.3: Phase synchronization as a function of cortical distance in areas V1, V2 and V3. In VFM, phase synchronization range spread over a larger cortical distance than in RS scans, likely reflecting stimulus induced interactions. In contrast, in RS, baseline synchronization is higher, possibly indicating increased long range interactions or the effect of global signals. To quantify range and baseline, we fit an exponential function (dashed lines) to the average PLV binned by cortical distance (colored points). Error bars show the standard error of the mean corrected for upsampling the functional data to the reconstructed cortical surface grid (upsampling factor = 8). Across scans, in V1, V2 and V3, the synchronization range spread over a larger cortical distance in VFM (mean (SD): V1 = 7.66 mm (0.40); V2 = 9.31 mm (0.49) and V3 = 10.66 mm (1.34)) than in RS (mean (SD): V1 = 3.91 mm (1.55); V2 = 4.00 mm (2.35) and V3 = 3.45 mm (0.83)). In VFM, baseline synchronization estimates were smaller (mean (SD): V1 = 0.23 (0.007); V2 = 0.30 (0.026) and V3 = 0.23 (0.039)) than those for RS (mean (SD): V1 = 0.46 (0.045); V2 = 0.44 (0.095) and V3 = 0.48 (0.18)). Differences between VFM and RS derived synchronization range and baseline estimates were significant in all cases (p <0.05, two sample t-test). Global average PLV floor, computed by permuting voxel indexes across subjects and scans, was: 0.33 for VFM scans and 0.42 for RS scans. Data for a single subject. Data for other subjects are included in the supplementary material (Figure 3.3S).

relation between synchronization range and mean cluster areas ( r = -0.044, p = 0.737, n = 60). For visuotopic distances, we found a weak correlation ( r = 0.31, p = 0.0128, n = 60).

3.3.4 Homotopic anatomical connectivity of cluster synchronization We also asked whether clusters sharing similar visual ﬁeld selectivity had higher PLV across visual areas and hemispheres. To answer this, we grouped the data in each hemisphere across upper/lower foveal and peripheral quadrants and computed signiﬁcant PLV-based functional (intra- and inter-hemispheric)connections for RS and VFM data (p <0.001, permutations corrected for multiple comparisons. See Figure 3.4). The resulting binary matrices were compared to a binary matrix consisting of connections between ROIs with similar visuotopic selectivity using Spearman correlation (intra-hemispheric: 0.45 for RS and 0.56 for VFM. Inter-hemispheric: 0.46 for RS and 0.57 for VFM. p <0.0001 for all cases). The match between PLV-based functional connections and anatomical connections increased in VFM.

44 3.3. Results

Figure 3.4: Phase synchronization as a function of visuotopic distance for areas V1, V2 and V3. Phase synchronization, expressed as PLV, was grouped from 0-10 deg in bins of 1deg. Points represents the voxel averages across bins for each scan. To quantify differences between VFM and RS we fit an exponential function to the average PLV binned by visuotopic distance and compute the range and baseline. Across scans, within V1, V2, V3 and between V1-V2, V1-V3 and V2-V3, the synchronization range spread over a larger visuotopic distance in VFM (mean (SD): V1 = 1.02 deg (0.1); V2 = 0.98 deg (0.05); V3 = 1.04 deg (0.2); V1-V2 = 0.97 deg (0.09); V1-V3 = 0.94 deg (0.12) and V2-V3 = 0.98 deg (0.07)) than in RS (mean (SD): V1 = 0.55 deg (0.15); V2 = 0.47 deg (0.088); V3 = 0.38 deg (0.099); V1-V2 = 0.43 deg (0.22); V1-V3 = 0.36 deg (0.13) and V2-V3 = 0.38 deg (0.075)). In VFM, baseline synchronization estimates were smaller (mean (SD): V1 = 0.21 (0.016); V2 = 0.21 (0.016); V3 = 0.22 (0.04); V1-V2 = 0.20 (0.012); V1-V3 = 0.21 (0.014) and V2-V3 = 0.22 (0.089)) than those for RS (mean (SD): V1 = 0.44 (0.053); V2 = 0.43 (0.13); V3 = 0.48 (0.16); V1-V2 = 0.37 (0.088); V1-V3 = 0.36 (0.1) and V2-V3 = 0.44 (0.14))). Differences between VFM and RS derived synchronization range and baseline were significant in all cases (two sample t-test: p <10-5 for all cases). Data for a single subject. Data for other subjects are included in the supplementary material (Figure 3.4S).

45 . S MRI

Figure 3.5: Voxels in clusters with similar visual field position selectivity have higher PLV across visual field maps and hemispheres. For each subject, clusters were grouped over upper/lower foveal (below 2.2 deg eccentricity) and peripheral (above 2.2 deg eccentricity) quadrants and PLV matrices for intra and inter -hemispheric PLV-based functional connections were computed. Diagonal and off-diagonal quadrants in each matrix represent within- and between-hemisphere PLV across visual cortical areas (grouped by the colors), respectively. Inside each colored box, quadrants are grouped in the following order (from left to to right): upper fovea, upper periphery, lower fovea and lower periphery. The resulting PLV matrices were averaged across scans and z-scored. Subsequently, a significant PLV structure across subjects (black squares for intra-hemispheric and gray squares for inter-hemispheric connections) was established using permutations (p <0.001. Corrected for multiple comparisons). Furthermore, we estimated the degree of homotopy in the resulting PLV-structures by using the Spearman correlation between these and binary matrices (for within-hemisphere and between-hemisphere connections) with ones indicating homotopy (see methods section 3.2.6.). Correlations increase during VFM (intra-hemispheric: 0.45 for RS and 0.56 for VFM. Inter-hemispheric: 0.46 for RS and 0.57 for VFM. p <0.0001 for all cases).

3.4 Discussion

Using covariations in local phase synchrony to parcelate the fMRI signal, we revealed a stable spatial structure in BOLD fluctuations recorded during resting state (RS). This spatial structure, which we identified as phase synchronization (PS) clusters, was also observed in response to visual field mapping (VFM) stimuli. Interestingly, the magnitude and spatial extent of PS varied in different RS scans yet the shape, elongated along eccentricity, and location of the PS clusters remained stable. Importantly, the topographical organization of synchronized hemodynamic activity across areas strongly reflected the layout of homotopic anatomical connections. The similar shape and location of the PS clusters obtained for RS and stimulus-driven fMRI, together with the high synchrony between clusters sharing similar visual field position selectivity across visual cortical areas and hemispheres, suggest that they share a common neuroanatomical origin. Therefore, our findings justify and facilitate direct comparison of RSand stimulus-evoked activity.

46 3.4. Discussion

3.4.1 Phase-synchronization-based parcellation of RS-fMRI signals reveals topographically organized clusters in early visual cortex Analyzing the phase synchronization (PS) of low-frequency BOLD fluctuations within early visual cortex, we revealed synchronization clusters during resting state (RS). The spatial structure was consistent among different RS scans, indicating that within-area local interactions are preserved during RS. This within- area stability is in contrast to the variability in between-area interactions observed in previous studies [25, 54, 10, 20]. Whereas between-area interactions are less stable during RS, during VFM they become more stable [20, 2, 25, 58], giving rise to a synchronization-based structure that is consistently found across subjects (as shown by Figure 3.4). Both in RS and VFM, the PLV structure resembled the layout of homotopic anatomical connections (cortical connections between visual areas sharing similar visual field position selectivity) within and between hemispheres (Figure 3.5). The finding that across-area cluster synchronization follows the underlying layout of homotopic anatomical connections points to a structural anchoring in spontaneous fMRI activity.

3.4.2 Similar location and shape, but different spatial extent of phase synchronization clusters for resting state and visual field mapping Similar cluster structure was observed in BOLD activity during RS and in response to visual field mapping (VFM) stimuli. However, the PS clusters found in VFM were slightly larger than those found in RS (Figure 3.2A). This mirrors our result that synchronization range spreads over a larger cortical extent in VFM, but local variations in the magnitude and spatial extent of PS do not affect the overall location and shape of the PS clusters. The fact that PS between neighboring recording sites (e.g. synchronization range) decreases more gradually with cortical and visual distance in VFM than in RS likely reflects a stimulus-evoked increase in spatial interactions [32, 60]. The narrower synchronization range in RS data is consistent with recent findings demonstrating that within- and between-area interactions in primary and extrastriate visual areas have a narrower spatial footprint during RS than during VFM [25, 54, 10, 20]. However, long range (baseline) PLV interactions were higher during RS, possibly indicating increased long-range interactions or global activity. In some RS scans, between ROI interactions in PS resembled those seen in visual field mapping. Interestingly, in a previous study of the first dataset, connective field models computed from those scans revealed well-ordered topographic maps [20]. An interesting feature of the cluster shape is its radial-tangential anisotropy. PS clusters derived from both RS and VFM are radially elongated in the visual field maps (Figures 3.2B & 3.2C). One possible factor for this elongation is a certain non-uniformity introduced by the shape of the ROIs. However, the ROIs are not always elongated (while V2 and V3 quarter fields sometimes are, V1 quarter fields generally not). Moreover, clusters are too small to be affected

47 . S MRI

by ROI shape. Another possibility is that, due to the log-scaling between visual field and cortical coordinates [62], circular connectivity on the cortical surface may look radially elongated when depicted in visual space [44, 65]. To explore these possibilities, we smoothed the time series using a 2D Gaussian kernel (2 and 15 mm FWHM, Figure 3.2S in supplementary material) defined onto the reconstructed cortical surface (grid). Anisotropy was preserved but its range widened with smoothing. We also modeled the effect of circular connectivity (without using the time series) on anisotropy by clustering toy models of circular connectivity on the cortical surface. Two toy cases were considered: 1) a matrix of weights constructed by assigning a 2 mm FWHM Gaussian connectivity profile to each location; and, 2) a binary network of nearest neighbors chains (3 mm and 6 mm threshold). Anisotropic clusters with similar spatial ranges were recovered in both cases. Clusters obtained from a shuffled grid (were 2D spatial relations are destroyed) did not show anisotropy (Figure 3.2S in supplementary materials). In sum, all lines of evidence seem to support the second option: circular connectivity on the cortical surface looks elongated in visual space. This cluster elongation in the eccentricity direction may have functional consequences. One hypothetical functional structure that has a similar radially elongated shape is the so-called “integration field” identified by crowding research [30, 65, 7, 50, 44]. Crowding is the breakdown of object recognition in peripheral vision when similar objects are too closely spaced, which increases perceptual uncertainty. It has been speculated that spacing effects on perceptual uncertainty at the behavioral level are a result of signal correlations in neuronal populations [69].

Besides differences in cluster size and synchronization range, RS and VFM data also differ in the decay of the phase locking values (PLV) with cortical distance (Figure 3.3). For VFM, the PLV was very stable over scans whereas in RS it was rather variable. Searching for the origin of this variability, we reana- lyzed the data after having applied global signal regression. This suppressed the long-range variability in particular in the RS data, suggesting that large-scale interactions across areas and global systemic variations affect the baseline synchronization level in each scan. We conclude that these are an important com- ponent of the observed variability in occipital RS recordings. Another cause of variability, at the behavioral level, could be the relationship between resting state BOLD fluctuations and oculo-motor behaviour that occurs subliminally and spontaneously during the scanning sessions [56].

Hence, while we derive similar synchronization clusters from RS and VFM in terms of their location and shape, the magnitude and spatial extent of PS differs between RS and VFM data. Despite local variations in the magnitude and spatial extent of PS during different scans and conditions, the location and shape of PS clusters is stable, demonstrating that PS clusters reflect a stable ar- chitectural property. In addition, we show that the topographical organization of synchronized hemodynamic activity across areas strongly reflects the layout of homotopic anatomical connections. Taken together, these findings support the hypothesis that spontaneous fMRI activity reflects the organization of the underlying anatomical connections [71, 29, 68, 26].

48 3.4. Discussion

3.4.3 Spatial extent of phase synchrony resting: possible mechanisms But which mechanisms could underlie the stability and spatial organization of PS? The observed spatial specificity of the spontaneous BOLD fluctuations can only emerge if the intrinsic neuronal activity within and across visual cortical areas is topographically organized. However, the neuronal basis of these BOLD patterns is not yet fully understood. It is debated on whether spontaneous fMRI activity reflects the consequences of population spiking activity, sub-threshold neuronal activity [39], or metabolic relationships between neurons and astrocytes (e.g. neuro-vascular coupling) [46, 47]. On the one hand, BOLD phase synchronization patterns may reflect the consequences of intrinsic switching of spiking input activity to aggregate neuronal assemblies sharing similar visual field position selectivity and tuning characteristics [29, 5, 37, 72]. Alternatively, those patterns may be the footprint of slow subthreshold fluctuations in local field potentials, which can be visuotopically organized and good predictors of the BOLD signal [40, 13]. A recent study by Matsui and colleagues [42] using neuronal calcium signals and simultaneous hemodynamic recordings unifies these contrasting findings by showing that both global fluctuations, in the form of propagating waves, and transient local coactivations are necessary for setting the spatial structure of hemodynamic functional connectivity [42, 51]. Also important is the role of astrocytes, the main link between neuronal metabolism and blood flow. They certainly play a role setting the pace of visuotopically organized hemodynamic fluctuations, as it is has been shown they modulate the delay between neuronal activity and its hemodynamic response [47]. Together, these findings suggest that multiple physiological factors may influence spontaneous hemodynamic activity in a way that gives rise to visuotopically organized fluctuations. An anatomical factor that may influence the spatial extent of phase synchrony, and therefore the size and location of the clusters, is the density and distribution of capillary beds. Shaped by the intricate branching patterns of the supporting vascular network, inhomogeneity in the density and distribution of capillary beds may lead to varying blood perfusion and deoxyhemoglobin washout rates [53, 1, 46]. In addition to the effects exerted by the supporting vascular network, the energy expense of different populations of neurons and glial cells may covary with vessel density, structuring hemodynamic fluctuations in locally and functionally segregated neighborhoods [67, 12] whose responses are -on average- biased together [28]. This structured hemodynamic variability may lead to inhomogeneity in the coupling between the hemodynamic response of a recording site and the neuronal activity in the neighborhood of the site, and a concomitant bias among neighboring sites [27]. This inhomogeneity hypothesis is consistent with evidence from animal models pointing to the limits of hemodynamic measurements in neuroimaging studies. [23] demonstrated that the distribution of blood capillaries correlates with hemodynamic functional connectivity at the millimeter scale, suggesting a direct relationship between a neuronal population’s metabolic demand and the density of its supporting capillary networks. In another study, [70] delineated functional clusters of slow metabolic fluctuations across the mouse cortex, con-

49 . S MRI

firming a relationship between metabolic demand and capillary network density. Furthermore, simultaneous measurements of neuronal spiking, metabolic demand and vessel responses demonstrate that vessel dilatory responses are effectively coupled to spiking and metabolic activity [46]. However, they also show vascular responses to stimuli that elicit little to no neuronal activity in the surrounding tissue, revealing limitations of the link between hemodynamic and neuronal responses. Large draining veins also affect the spatial structure of the BOLD signal. They are known to modulate the phase of nearby metabolic fluctuations [74] with little effect on signal amplitude and reliability. On the other hand, par- enchymal draining veins inside recording sites have a large modulatory effect on the amplitude of local BOLD signals, and are sometimes an order of magnitude larger than those induced by neuronal metabolism. Pial veins, which are adjacent to the cortical surface and predominantly parallel to the sulci, similarly reduce spatial specificity and introduce noise [74]. As pial branches drain oxygen from large patches of cortex, they may play a substantial role in determining the location and shape of the synchronization clusters. Together, these findings suggest that links between the neuronal metabolism and the hemodynamic signals have limited spatial resolution [17, 35, 43, 46]. At finer scales, vascular network structures play an important role in shaping the spatial structure of the BOLD signal. We conclude that synchronization clusters most likely reflect a combination of 1) shared metabolic demands of local neuronal populations with similar visuotopic selectivity, 2) linked neuronal activity of neuronal populations with similar visuotopic selectivity across areas, and 3) the density of supporting capillary networks, regardless of whether responses are stimulus driven or endogenous. Our findings indicate that, despite limitations in the resolution of metabolism-sensitive measurements such as fMRI to determine the contribution of neuronal activity to hemodynamic signals, it is still possible to study the neuronal properties of aggregate neuronal populations.

3.4.4 Limitations and future directions We computed phase locking values (PLV) to quantify the degree of synchronization of low-frequency BOLD fluctuations despite the Pearson correlation being the most used functional connectivity measure in RS BOLD analysis. However, this is justified by two facts. Correlation values come with 2 drawbacks: First, there is an ongoing debate◦ on how to deal with negative correlations. Second, if two signals are 90 out of phase, they will result in a Pearson correlation of 0, even though these signals are highly locked. PLV does not suffer from these drawbacks. Nonetheless, we have provided a direct comparison between PLV measures and Pearson correlation measures on the effect of measured anisotropy in the supplementary material (Figure 3.2S). This shows that, for the current data under investigation, results are highly comparable. Given that the PLV measure is convenient to interpret (e.g. are the signals phase locked), we opted to keep this measure. We use the same frequency band (0.04 - 0.07 Hz) for RS and VFM. How- ever, during VFM responses are locked to the stimuli period, the peak frequency

50 3.5. Conclusion band is lower than in RS. Generally in RS data, most power is below 0.03 Hz, however, the peak power might not be always the product of neuronal activity. Some scans can show peaks around 0.1 Hz or even higher. As shown by Glerean and colleagues, 0.04-0.07 Hz is the the most reliable frequency band from which to extract resting state networks [19]. In VFM, we use the same frequency band as a in RS to be able to compare the results. We deliberately remove some of the possible useful information in the low frequency side because we are not pulling out the VFM response per-se but focus on spontaneous fluctuations and try to avoid contamination by low frequency artifacts (e.g. respiration). The current study analyses stationary synchronization patterns only. As such, they provide a time averaged picture and do not capture potentially relevant transient dynamics [38]. Time-resolved methods such as the analysis of propagation and transient coactivation patterns might help disclose relevant dynamics such as visuotopically organized cofluctuations, patterns of diffusion and possibly waves [38, 42, 1]. Further research is also required to establish the possible neuronal mechanisms underlying visuotopically organized covariations in the phase of fMRI signals. Adding independent measures of neurophysiological and visceral activity, like electroencephalography and electrogastrography, seems a promising path to study the relation of neuronal activity to hemostatic and visceral activity during resting state [57, 77]. Future analyses could also take into account measures of blood flow, as blood arrival times are known to affect resting state functional connectivity [16].

3.5 Conclusion

We have demonstrated that BOLD signals fluctuations across early visual cortical areas can be segmented into highly reproducible synchronization clusters irrespective whether they are derived from RS-fMRI or visual field mapping data. Importantly, we show that the spatial footprint of these synchronization clusters is independent of the magnitude and spatial extent of PS in different RS and VFM scans and reflect the layout of homotopic anatomical connections along the visual hierarchy. This indicates that both intrinsic and stimulus- evoked fMRI fluctuations are anchored by the same neuroanatomical connections. The fact that the PS clusters for RS and VFM are similarly elongated in the eccentricity direction also points to a shared neuroanatomical basis. Our findings thus justify and facilitate direct comparison of RS and stimulus-evoked activity.

3.6 Supplementary Material

Please refer to the online version of this article (doi.org/10.1016/j.neuroimage.2017.08.063) for the supplementary material.

51 . S MRI

3.7 References [1] Kevin M Aquino, Mark M Schira, P A Robinson, Peter M Drysdale, and Breakspear Michael. Hemodynamic traveling waves in human visual cortex. PLoS Comput. Biol., 8(3):e1002435, 2012. [2] Michael J Arcaro, Christopher J Honey, Ryan E B Mruczek, Sabine Kast- ner, and Uri Hasson. Widespread correlation patterns of fMRI signal across visual cortex reflect eccentricity organization. Elife, 4, 19 February 2015. [3] Richard F Betzel, Griffa Alessandra, Avena-Koenigsberger Andrea, Goñi Joaquín, Thiran Jean-Philippe, Hagmann Patric, and Sporns Olaf. Multi- scale community organization of the human structural connectome and its relationship with resting-state functional connectivity. Network Science, 1 (03):353–373, 2013. [4] Vincent D Blondel, Jean-Loup Guillaume, Renaud Lambiotte, and Etienne Lefebvre. Fast unfolding of communities in large networks. J. Stat. Mech., 2008(10):P10008, 9 October 2008. [5] Barak Blumenfeld, Dmitri Bibitchkov, and Misha Tsodyks. Neural network model of the primary visual cortex: from functional architecture to lateral connectivity and back. J. Comput. Neurosci., 20(2):219–241, April 2006. [6] Andrew S Bock, Paola Binda, Noah C Benson, Holly Bridge, Kate E Watkins, and Ione Fine. Resting-State retinotopic organization in the absence of retinal input and visual experience. J. Neurosci., 35(36):12366– 12382, 9 September 2015. [7] H Bouma. Interaction effects in parafoveal letter recognition. Nature, 226 (5241):177–178, 11 April 1970. [8] Geoffrey M. Boynton, Stephen A. Engel, Gary H. Glover, and David J. Heeger. Linear Systems Analysis of Functional Magnetic Resonance Ima- ging in Human V1. Journal of Neuroscience, 16(13):4207–4221, July 1996. [9] D H Brainard. The psychophysics toolbox. Spat. Vis., 10(4):433–436, 1997. [10] Omar H Butt, Noah C Benson, Ritobrato Datta, and Geoffrey K Aguirre. The fine-scale functional correlation of striate cortex in sighted and blind people. J. Neurosci., 33(41):16209–16219, 9 October 2013. [11] Omar H Butt, Noah C Benson, Ritobrato Datta, and Geoffrey K Aguirre. Hierarchical and homotopic correlations of spontaneous neural activity within the visual cortex of the sighted and blind. Front. Hum. Neurosci., 9: 25, 10 February 2015. [12] György Buzsáki, Buzsáki György, Kaila Kai, and Raichle Marcus. Inhib- ition and brain work. Neuron, 56(5):771–783, 2007.

52 3.7. References

[13] Matteo Carandini, Daisuke Shimaoka, L Federico Rossi, Tatsuo K Sato, Andrea Benucci, and Thomas Knöpfel. Imaging the awake visual cortex with a genetically encoded voltage indicator. J. Neurosci., 35(1):53–63, 7 January 2015. [14] E W Dijkstra. A note on two problems in connexion with graphs. Numer. Math., 1(1):269–271, 1959. [15] Serge O Dumoulin and Brian A Wandell. Population receptive field estimates in human visual cortex. Neuroimage, 39(2):647–660, 15 January 2008. [16] Sinem B Erdoğan, Yunjie Tong, Lia M Hocke, Kimberly P Lindsey, and Blaise deB Frederick. Correcting for blood arrival time in global mean regression enhances functional connectivity analysis of resting state fMRI- BOLD signals. Front. Hum. Neurosci., 10:311, 28 June 2016. [17] Elia Formisano, Formisano Elia, and Kriegeskorte Nikolaus. Seeing patterns through the hemodynamic veil — the future of pattern-information fMRI. Neuroimage, 62(2):1249–1256, 2012. [18] Michael D Fox. Clinical applications of resting state functional connectivity. Front. Syst. Neurosci., 2010. [19] Enrico Glerean, Juha Salmi, Juha M Lahnakoski, Iiro P Jääskeläinen, and Mikko Sams. Functional magnetic resonance imaging phase synchronization as a measure of dynamic functional connectivity. Brain Connect., 2 (2):91–101, 11 June 2012. [20] Nicolás Gravel, Ben Harvey, Barbara Nordhjem, Koen V Haak, Serge O Dumoulin, Remco Renken, Branislava Curčić-Blake, and Frans W Cor- nelissen. Cortical connective field estimates from resting state fMRI activity. Front. Neurosci., 8:339, 31 October 2014. [21] Koen V Haak, Jonathan Winawer, Ben M Harvey, Remco Renken, Serge O Dumoulin, Brian A Wandell, and Frans W Cornelissen. Con- nective field modeling. Neuroimage, 66:376–384, 1 February 2013. [22] Donald J Hagler, Jr, Ayse Pinar Saygin, and Martin I Sereno. Smoothing and cluster thresholding for cortical surface-based group analysis of fMRI data. Neuroimage, 33(4):1093–1103, December 2006. [23] Robert V Harrison, Noam Harel, Jaswinder Panesar, and Richard J Mount. Blood capillary distribution correlates with hemodynamic-based functional imaging in cerebral cortex. Cereb. Cortex, 12(3):225–233, March 2002. [24] Ben M Harvey and Serge O Dumoulin. The relationship between cortical magnification factor and population receptive field size in human visual cortex: constancies in cortical architecture. J. Neurosci., 31(38):13604– 13612, 21 September 2011.

53 . S MRI

[25] Jakob Heinzle, Thorsten Kahnt, and John-Dylan Haynes. Topographic- ally speciﬁc functional connectivity between visual ﬁeld maps in the human brain. Neuroimage, 56(3):1426–1436, 1 June 2011.

[26] C J Honey, O Sporns, L Cammoun, X Gigandet, J P Thiran, R Meuli, and P Hagmann. Predicting human resting-state functional connectivity from structural connectivity. Proc. Natl. Acad. Sci. U. S. A., 106(6):2035– 2040, 10 February 2009.

[27] Yukiyasu Kamitani and Frank Tong. Decoding the visual and subjective contents of the human brain. Nat. Neurosci., 8(5):679–685, May 2005.

[28] S Kastner, M A Pinsk, P De Weerd, R Desimone, and L G Ungerleider. Increased activity in human visual cortex during directed attention in the absence of visual stimulation. Neuron, 22(4):751–761, April 1999.

[29] Tal Kenet, Dmitri Bibitchkov, Misha Tsodyks, Amiram Grinvald, and Amos Arieli. Spontaneously emerging cortical representations of visual attributes. Nature, 425(6961):954–956, 30 October 2003.

[30] F L Kooi, A Toet, S P Tripathy, and D M Levi. The eﬀect of similarity and duration on spatial interaction in peripheral vision. Spat. Vis., 8(2): 255–279, 1994.

[31] J P Lachaux, E Rodriguez, J Martinerie, and F J Varela. Measuring phase synchrony in brain signals. Hum. Brain Mapp., 8(4):194–208, 1999.

[32] T C Lacy, K M Aquino, P A Robinson, and M M Schira. Shock-like haemodynamic responses induced in the primary visual cortex by moving visual stimuli. J. R. Soc. Interface, 13(125), December 2016.

[33] A Laird, J Carew, and M E Meyerand. Analysis of the instantaneous phase signal of a fMRI time series via the hilbert transform. In Con- ference Record of Thirty-Fifth Asilomar Conference on Signals, Systems and Computers (Cat.No.01CH37256), 2001.

[34] Angela R Laird, Baxter P Rogers, John D Carew, Konstantinos Arfanakis, Chad H Moritz, and M Elizabeth Meyerand. Characterizing instantaneous phase relationships in whole-brain fMRI activation data. Hum. Brain Mapp., 16(2):71–80, June 2002.

[35] A T Lee, G H Glover, and C H Meyer. Discrimination of large venous vessels in time-course spiral blood-oxygen-level-dependent magnetic- resonance functional neuroimaging. Magn. Reson. Med., 33(6):745–754, June 1995.

[36] M H Lee, C D Smyser, and J S Shimony. Resting-state fMRI: a review of methods and clinical applications. AJNR Am. J. Neuroradiol., 34(10): 1866–1872, October 2013.

54 3.7. References

[37] Christopher M Lewis, Conrado A Bosman, Thilo Womelsdorf, and Pas- cal Fries. Stimulus-induced visual cortical networks are recapitulated by spontaneous local and interareal synchronization. Proc. Natl. Acad. Sci. U. S. A., 113(5):E606–15, 2 February 2016. [38] Xiao Liu and Jeff H Duyn. Time-varying functional network information extracted from brief instances of spontaneous brain activity. Proc. Natl. Acad. Sci. U. S. A., 110(11):4392–4397, 12 March 2013. [39] N K Logothetis, J Pauls, M Augath, T Trinath, and A Oeltermann. Neurophysiological investigation of the basis of the fMRI signal. Nature, 412(6843):150–157, 12 July 2001. [40] Nikos K Logothetis and Brian A Wandell. Interpreting the BOLD signal. Annu. Rev. Physiol., 66:735–769, 2004. [41] Nikos K Logothetis, Murayama Yusuke, Augath Mark, Steffen Theodor, Werner Joachim, and Oeltermann Axel. How not to study spontaneous activity. Neuroimage, 45(4):1080–1089, 2009. [42] Teppei Matsui, Tomonari Murakami, and Kenichi Ohki. Transient neuronal coactivations embedded in globally propagating waves underlie resting-state functional connectivity. Proc. Natl. Acad. Sci. U. S. A., 16 May 2016. [43] R S Menon and S G Kim. Spatial and temporal limits in cognitive neuroimaging with fMRI. Trends Cogn. Sci., 3(6):207–216, June 1999. [44] Anirvan S Nandy and Bosco S Tjan. Saccade-confounded image statist- ics explain visual crowding. Nat. Neurosci., 15(3):463–9, S1–2, 8 January 2012. [45] O Nestares and D J Heeger. Robust multiresolution alignment of MRI brain volumes. Magn. Reson. Med., 43(5):705–715, May 2000. [46] Philip O’Herron, O’herron Philip, Pratik Y Chhatbar, Levy Manuel, Shen Zhiming, Adrien E Schramm, Lu Zhongyang, and Kara Prakash. Neural correlates of single-vessel haemodynamic responses in vivo. Nature, 2016. [47] J C Pang, P A Robinson, K M Aquino, and N Vasan. Effects of as- trocytic dynamics on spatiotemporal hemodynamics: Modeling and en- hanced data analysis. Neuroimage, 14 October 2016. [48] Andrew Parker, Parker Andrew, Bridge Holly, and Coullon Gaelle. Spa- tial scale of correlated signals in 7T BOLD imaging. In Proceedings of the 9th EAI International Conference on Bio-inspired Information and Commu- nications Technologies (formerly BIONETICS), 2016. [49] D G Pelli. The VideoToolbox software for visual psychophysics: trans- forming numbers into movies. Spat. Vis., 10(4):437–442, 1997.

55 . S MRI

[50] Denis G Pelli. Crowding: a cortical constraint on object recognition. Curr. Opin. Neurobiol., 18(4):445–451, August 2008. [51] M Andrea Pisauro, Neel T Dhruv, Matteo Carandini, and Andrea Be- nucci. Fast hemodynamic responses in the visual cortex of the awake mouse. J. Neurosci., 33(46):18343–18351, 13 November 2013. [52] Adrián Ponce-Alvarez, Gustavo Deco, Patric Hagmann, Gian Luca Ro- mani, Dante Mantini, and Maurizio Corbetta. Resting-state temporal synchronization networks emerge from connectivity topology and heterogeneity. PLoS Comput. Biol., 11(2):e1004100, February 2015. [53] Dmitry E Postnov, Olga V Sosnovtseva, and Mosekilde Erik. Oscillator clustering in a resource distribution chain. Chaos: An Interdisciplinary Journal of Nonlinear Science, 15(1):013704, 2005. [54] Mathijs Raemaekers, Raemaekers Mathijs, Schellekens Wouter, Richard J A van Wezel, Petridou Natalia, Kristo Gert, and Nick F Ramsey. Pat- terns of resting state connectivity in human primary visual cortical areas: A 7T fMRI study. Neuroimage, 84:911–921, 2014. [55] M E Raichle. The restless brain: how intrinsic activity organizes brain function. Philos. Trans. R. Soc. Lond. B Biol. Sci., 370(1668):20140172– 20140172, 2015. [56] Michal Ramot, Meytal Wilf, Hagar Goldberg, Tali Weiss, Leon Y Deouell, and Rafael Malach. Coupling between spontaneous (resting state) fMRI fluctuations and human oculo-motor activity. Neuroimage, 58(1):213–225, 1 September 2011. [57] Craig G Richter, Mariana Babo-Rebelo, Denis Schwartz, and Cather- ine Tallon-Baudry. Phase-amplitude coupling at the organism level: The amplitude of spontaneous alpha rhythm fluctuations varies with the phase of the infra-slow gastric basal rhythm. Neuroimage, 21 August 2016. [58] Leor Roseman, Martin I Sereno, Robert Leech, Mendel Kaelen, Csaba Orban, John McGonigle, Amanda Feilding, David J Nutt, and Robin L Carhart-Harris. LSD alters eyes-closed functional connectivity within the early visual cortex in a retinotopic fashion. Hum. Brain Mapp., 37(8): 3031–3040, August 2016. [59] Mikail Rubinov and Olaf Sporns. Complex network measures of brain connectivity: uses and interpretations. Neuroimage, 52(3):1059–1069, September 2010. [60] Wouter Schellekens, Richard J A Van Wezel, Natalia Petridou, Nick F Ramsey, and Mathijs Raemaekers. Integration of motion responses underlying directional motion anisotropy in human early visual cortical areas. PLoS One, 8(6):e67468, 28 June 2013. [61] Mark M Schira, Christopher W Tyler, Michael Breakspear, and Branka Spehar. The foveal confluence in human visual cortex. J. Neurosci., 29(28): 9050–9058, 15 July 2009.

56 3.7. References

[62] Mark M Schira, Christopher W Tyler, Branka Spehar, and Michael Breakspear. Modeling magnification and anisotropy in the primate foveal confluence. PLoS Comput. Biol., 6(1):e1000651, 29 January 2010. [63] P Tass, M G Rosenblum, J Weule, J Kurths, A Pikovsky, J Volkmann, A Schnitzler, and Freund H.-J. Detection of n : m phase locking from noisy data: Application to magnetoencephalography. Phys. Rev. Lett., 81 (15):3291–3294, 1998. [64] P C Teo, G Sapiro, and B A Wandell. Creating connected representations of cortical gray matter for functional MRI visualization. IEEE Trans. Med. Imaging, 16(6):852–863, December 1997. [65] A Toet and D M Levi. The two-dimensional shape of spatial interaction zones in the parafovea. Vision Res., 32(7):1349–1357, July 1992. [66] Yunjie Tong, Lia M Hocke, Xiaoying Fan, Amy C Janes, and Blaise Deb Frederick. Can apparent resting state connectivity arise from systemic fluctuations? Front. Hum. Neurosci., 9:285, 15 May 2015. [67] Yunjie Tong, Tong Yunjie, Lia M Hocke, Kimberly P Lindsey, Sinem B Erdoğan, Vitaliano Gordana, Carolyn E Caine, and Blaise Deb Frederick. Systemic Low-Frequency oscillations in BOLD signal vary with tissue type. Front. Neurosci., 10, 2016. [68] M Tsodyks, T Kenet, A Grinvald, and A Arieli. Linking spontaneous activity of single cortical neurons and the underlying functional architecture. Science, 286(5446):1943–1946, 3 December 1999. [69] Ronald van den Berg, Addie Johnson, Angela Martinez Anton, Anne L Schepers, and Frans W Cornelissen. Comparing crowding in human and ideal observers. J. Vis., 12(6):13, 12 June 2012. [70] Alberto L Vazquez, Matthew C Murphy, and Seong-Gi Kim. Neuronal and physiological correlation to hemodynamic resting-state fluctuations in health and disease. Brain Connect., 4(9):727–740, November 2014. [71] J L Vincent, G H Patel, M D Fox, A Z Snyder, J T Baker, D C Van Es- sen, J M Zempel, L H Snyder, M Corbetta, and M E Raichle. Intrinsic functional architecture in the anaesthetized monkey brain. Nature, 447 (7140):83–86, 3 May 2007. [72] Martin Vinck and Conrado A Bosman. More gamma more predictions: Gamma-Synchronization as a key mechanism for efficient integration of classical receptive field inputs with surround predictions. Front. Syst. Neur- osci., 10:35, 25 April 2016. [73] B A Wandell, S Chial, and B T Backus. Visualization and measurement of the cortical surface. J. Cogn. Neurosci., 12(5):739–752, September 2000. [74] Jonathan Winawer, Hiroshi Horiguchi, Rory A Sayres, Kaoru Amano, and Brian A Wandell. Mapping hv4 and ventral occipital cortex: the venous eclipse. J. Vis., 10(5):1, 1 May 2010.

57 . S MRI

[75] Jun Ye, Nicole A Lazar, and Yehua Li. Nonparametric variogram modeling with hole eﬀect structure in analyzing the spatial characteristics of fMRI data. J. Neurosci. Methods, 240:101–115, 30 January 2015. [76] B T Thomas Yeo, Fenna M Krienen, Jorge Sepulcre, Mert R Sabuncu, Danial Lashkari, Marisa Hollinshead, Joshua L Roﬀman, Jordan W Smoller, Lilla Zöllei, Jonathan R Polimeni, Bruce Fischl, Hesheng Liu, and Randy L Buckner. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J. Neurophysiol., 106(3): 1125–1165, September 2011. [77] Han Yuan, Yuan Han, Zotev Vadim, Phillips Raquel, Wayne C Drevets, and Bodurka Jerzy. Spatiotemporal dynamics of the brain at rest — exploring EEG microstates as electrophysiological signatures of BOLD resting state networks. Neuroimage, 60(4):2062–2072, 2012.

58 On the feasibility of assessing connective ﬁeld maps and synchronization clusters using 3T fMRI 4

Biologically inspired models of fMRI activity like connective field modeling are potential tools for the study of human neuronal activity in vivo. However, for the summary descriptions they provide to become useful tools in brain research, results should generalize to different magnetic field strengths. To achieve this, here we try to reproduce results previously obtained at 7T using 3T data. First, we compute connective field models for V1→V2 and V1→V3. Then, we determine the spatial structure of synchronized activity by detecting clusters of synchronized fMRI activity. Finally, we compare the results obtained with different magnetic field strengths. Despite the lower resolution and signal-to- noise ratio of the 3T data. We find that the results obtained with it are in fair agreement to those obtained previously with 7T data. Our findings justify and facilitate the direct comparison of RS and stimulus-evoked activity acquired at 3T.

4.1 Introduction

Thanks to indirect forms of neuronal recordings such as magnetic resonance imaging, accurate methods to characterize blood oxygen dependent fluctuations (BOLD) across the human visual cortex have been developed [1, 7, 12, 13]. The methods implemented in chapter 2 (to estimate cortical connective fields) and chapter 3 (to detect synchronization clusters) are examples of this. However, when a method has been shown to reveal relevant aspects of brain activity using a high-resolution acquisition approach (e.g. 7 Tesla fMRI), it is also important to assess its performance using a lower-resolution approach (at 3 Tesla fMRI). Higher magnetic fields increase the signal-to-noise (SNR) ratio, the tissue spe-

59 . O T MRI

cificity and the spatial resolution of fMRI recordings. However, 3T scanners are much more abundant than 7T. Therefore, it is important to determine the limitations and applicability of those methods in providing meaningful markers of neuronal activity at different magnetic field strengths. To generalize results obtained at a higher magnetic field, key findings should be reproducible at lower magnetic fields. In chapter 2 [10], we have used blood-oxygen level dependent (BOLD) fluctuations recorded using 7T fMRI during resting-state (RS) to study intrinsic functional connectivity across visual cortical areas using the connective field (CF) modeling method [12] and shown that spontaneous fluctuations in BOLD activity are retinotopically organized. Based on the hypothesis that BOLD cofluctuations between topographically connected maps reflect the underlying neuroanatomical organization, CF modeling enables the characterization of a target recording site (e.g. V2 and V3) in terms of the aggregate BOLD activity in a source brain area (e.g. V1), thus providing a description of the preferred locations on the cortical surface to which these target sites respond. Locations in primary visual cortex are associated with visual field positions obtained during visual field mapping (VFM). Therefore, visual field coordinates can be inferred for the target recording sites from the preferred locations in the source region, allowing the reconstruction of visuotopic maps even in the absence of a stimulus (i.e RS). Here, we ask whether RS data recorded at 3T can also reveal meaningful CF maps. In addition, we ask whether the phase synchronization based parcellation approach implemented in chapter 3 reveals similar patterns when applied to 3T data. To assess these questions, we apply the two methods to RS and VFM BOLD data recorded at 3T fMRI and qualitatively compare the results to those obtained in our previous studies using 7T fMRI [10, 11] (See chapters 2 & 3). To preview our results, we find that CF parameters determined based on 3T fMRI data are noisier yet still fairly comparable to those based on 7T data. Phase synchronization clusters are also comparable. In addition to the lower spatial resolution and SNR of 3T, we identify as sources of noise and variability the lower quality of the gray/white matter segmentations, motion related artifacts and difficulties in the alignment of the RS functional data with the anatomical volumes. We will show that 3T fMRI data can be used to derive biologically plausible CF models and synchronization clusters from RS data, yet care must be taken when assessing and interpreting the results, given various sources of variability that are more prominently present at 3T. In future studies, improvements to the analysis, the measurement and the preprocessing approaches could help overcome such limitations.

4.2 Materials and Methods

4.2.1 Participants Data was acquired for four female right-handed participants with normal visual acuity (age 20-30). Experimental procedures were approved by the medical

60 4.2. Materials and Methods ethics committee of the University Medical Center Groningen.

4.2.2 Stimulus Visual stimuli were presented on an MR compatible display screen (BOLD- screen 24 LCD; Cambridge Research Systems, Cambridge, UK). The screen was located at the head-end of the MRI scanner. Participants viewed the screen through a tilted mirror attached to the 32-channel SENSE head mounted coil. Distance from the eyes to the screen (measured through the mirror) was approximately 75 cm. Screen size was 36×23 degrees of visual angle. The stimuli were generated in Matlab (Mathworks, Natick, MA, USA), using the Psychtoolbox [5, 16]. The visual field mapping paradigm consisted on a drifting bar aperture defined by high-contrast flickering texture [8]. The bar aperture moved in8 different directions (four bar orientations: horizontal, vertical and the two diagonal orientations), with for each orientation two opposite directions). The bar moved across the screen in 16 equally spaced steps each lasting 1 TR. After each pass and a half, 12 s of a blank stimulus at mean luminance was presented full screen. To ensure stable fixation, participants were instructed to focus on a colored dot in the center of the screen and press a button as soon as the dot changed color.

4.2.3 Resting state During the resting state scans, the stimulus was replaced with a black screen and participants closed their eyes. The lights in the scanning room were oﬀ. The room was in complete darkness. Participants were instructed to let their mind roam freely (e.g. not focus on one speciﬁc thought) , and not to fall asleep.

4.2.4 Data acquisition High-resolution T1-weighted structural images were acquired on a 3 Tesla scanner (Philips, Best, Netherlands) using a 32-channel head coil at a resolution of 1 × 1 × 1 mm, with a field of view of 256 × 256 × 170 mm. TR was 9 ms, TE was 3.54 ms. Functional, T2*-weighted, 2D echo planar images were acquired at a voxel resolution of 2.5 × 2.5 × 2.5, with a field of view of 190 × 190 × 50 mm. TR was 1500 ms, TE was 30 ms. The slice orientation was approximately parallel to the calcarine sulcus. The VFM scan lasted 210 s (e.g. 132 acquisi- tions) per run. Eight volumes were discarded from the beginning of each scan to ensure the signal had reached a steady state. The RS scan lasted 370 s per run and a total of 340 volumes functional scans were acquired. In total, 3 VFM and 2 RS runs were acquired. For VFM, the acquired volume covered occipital and frontal regions but not dorsal regions. A short anatomical scan with the same field of view of the VFM functional scans was acquired and used to obtain a good alignment between the VFM functional data and the anatomical volume. For RS, the whole brain was recorded. For the alignment of the RS data, the mean functional data was used, which generally resulted in a slightly less accurate alignment.

61 . O T MRI

4.2.5 Preprocessing Gray and white matter were automatically segmented using Freesurfer and hand edited in ITKGray to minimize segmentation errors [18]. The cortical surface was reconstructed at the white/gray matter boundary and rendered as a smoothed 3D mesh [19]. Motion correction within and between scans was applied for the VFM and the RS scans [15]. To clean the resting state signals from DC drift and reduce high frequency nuisance, time courses were band pass filtered with a high-pass discrete cosine transform filter (DCT) with cutoff frequency of 0.01 Hz and a low-pass 4th order Butterworth filter with cutoff frequency of 0.1 Hz. Finally, functional data were aligned to the anatomical scans [15] and interpolated to the anatomical segmentation space.

4.2.6 Analysis Since the focus of the present study was to compare methods results obtained at 3T with those previously obtained at 7 T (in chapter 2 & 3), here we just summarize key steps in the analysis. Further details about the methods can be found in chapter 2 & 3.

4.2.6.1 Population receptive field mapping As described in previous chapters, visual field maps V1, V2 and V3 were mapped using the population receptive field (pRF) method [8]. Briefly, the method consists in combining 2D Gaussian models of the underlying neuronal population response, a two-gamma model of the hemodynamic response [4], and the stimulus aperture, to generate predictions of the VFM evoked BOLD responses. The pRF parameters associated with the best fMRI time series predictions are then chosen to summarize the response of each recording point [8].

4.2.6.2 Connective field mapping Connective field (CF) model parameters were estimated for both the VFM and RS scans in the same manner as described in chapter 2. Briefly, the method consists of summarizing the activity of the recording sites of a target region of interest (ROI) in terms of the aggregate activity contributed by a set of recording sites in a source ROI [12]. The aggregate activity of these sites is obtained by calculating the Gaussian weighted sum of the measured signals (thus including the preferred recording site and its neighbours). The resulting candidate times series predictions of the aggregate activity are compared to the measured time series of each recording point in the target ROI (V2 and/or V3), and the best fitting prediction and its associated CF parameters (position and size) are then selected. Furthermore, because CF preferred locations on the cortical surface are associated with preferred visual field positions during pRF mapping [12], coordinates in visual space can be inferred for target recording sites, allowing the reconstruction of visuotopic maps using RS data (See chapter 2 [10]). CF models were estimated for each participant, scan (VFM and RS), and visual field map combination (V1→V2 and V1→V3). The source and target ROIs used

62 4.2. Materials and Methods to compute CF models were deﬁned in the ﬁrst layer of the cortical surface reconstruction. To compare the CF map scatter between 3T and 7T- derived CF estimates, we computed the VFM-referred CF position displacement (see section 2.2.6.4). To assess the topographic organization of CF maps, eccentricity and polar angle CF maps were compared to pRF maps (and also between VFM- and RS-derived CF maps) using Pearson and circular correlation respectively. Finally, we examined the relationship between CF size and pRF eccentricity.

4.2.6.3 Parcellation into synchronization clusters

First, we established the degree of synchronization of low-frequency BOLD fluctuations by computing the phase locking values (PLV) between the instantaneous phase of the band-pass filtered BOLD signas of all pairs [9, 14, 17, 11] of recording sites within a visual field map. Estimates of the instantaneous phase were obtained by using the Hilbert transform. After obtaining instantaneous phase estimates for each recording site, we discarded the 5 first and 5 last time points and computed PLV matrices for each participant, scan and visual field map. This matrix summarizes the average synchronization tendency of low-frequency BOLD fluctuations over the scanning session [17, 11]. Second, we identify clusters of highly synchronized neighbouring recording sites. To this end, we first select the 5% strongest entries in the PLV matrices, and next, pruned these by setting the phase locking values of recording sites beyond 3 mm of cortical distance to zero. Cortical distance was computed using Dijkstra’s algorithm [6]. Pruning the PLV matrices before clustering helped minimize the contribution of long-range interactions while maintaining local interaction structure. subsequently, we clustered the pruned PLV matrices using an iterative implementation of the Louvain clustering algorithm [3]. Start- ing from the PLV matrices from each individual RS and VFM scans, this approach converges —for each condition— to the most probable phase synchronization (PS) cluster partition [2] (Further details about the implementation of this approach can be found in chapter 3 [11]). PS clusters were estimated for each condition, participant and visual field map. To aid visualization, we render the resulting clusters in visual space (using the pRF positions of the recording sites in each cluster). Finally, we examined the shape of the clusters in terms of their visuotopic organization. To this end, we computed the binomial probability of any two recording sites sharing a cluster (cluster membership probability) as a function of the visual field distance between their pRFs (for bins of 0.25 deg of visual field distance). Visual field distances between all recording pairs were estimated for the radial and arc directions as the difference in their corresponding pRF eccentricities and arc distance, respectively. To estimate how the cluster membership probability changed with radial and arc distance, we fitted binomial distributions and estimated the decay factor of these probabilities. Fits were computed after grouping the probabilities over areas, participants and conditions (See chapter 3 for details [11]).

63 . O T MRI

Table 4.1: Correlations between visual field maps derived using population receptive field and connective field modeling. Correlations were computed to estimate the agreement of the eccentricity (r, Pearson correlations) and polar angle (ϑ, circular correlations) maps in V2 and V3 derived using pRF modeling to those derived from visual field mapping and resting state data using the connective field method. Grouped data from 5 participants (six hemispheres).

VFM RS1 RS2 r ϑr ϑr ϑ → V1 → V2 0.90 0.93 0.43 0.71 0.47 0.76 V1 V3 0.85 0.83 0.19 0.41 0.25 0.43

4.3 Results An important requisite for the implementation of CF modeling is the success- ful mapping of visual field position selectivity for the cortical regions of interest, particularly the source region. Here we assessed this by using population receptive field (pRF) mapping [8]. However, pRF mapping depends on the adequate segmentation of gray and white matter. To illustrate the influence of the segmentation on the results, here we show results for a brain with a satisfactorily and an unsatisfactorily (pRF) mapped hemisphere (see top panels in Figure 4.1) in which V1 and dorsal V2 and V3 could not be satisfactorily mapped. Adequate segmentations could be obtained for 6 hemispheres out of the 5 avail- able brains only. As a consequence, we report CF properties based on those 6 hemispheres.

4.3.1 Deriving connective field estimates from VFM and RS 3T fMRI activity Figure 4.1 (bottom panels) illustrates CF maps derived from VFM and RS data obtained using 3T fMRI. With a good segmentation provided, VFM-derived CF maps obtained using 3T are of a similar quality to those obtained when using 7T fMRI (compare left hemisphere in Figure 4.1 to top panel in Figure 2.3) [12, 10]. Similar to what was previously observed at 7T, RS-derived CF maps partly reflect the functional topographic organization revealed with pRF mapping. However, VFM-referred displacements in CF position (CF-scatter) was greater in RS-CF maps derived at 3T than those obtained at 7T. Median CF-scatter for RS1-derived V1→V2 and V1→V3 CF maps was 21.5 mm and 19.5 mm respectively whereas for RS was 19 mm, 17.4 mm. These values approximately doubled those obtained at 7T (see Figures 2.3 & 2.4A). We then quantified the visuotopic organization of the resulting CF maps by correlating the CF-derived eccentricity and polar angle to their corresponding pRF-derived counterparts. Table 4.1 illustrates these results. The best agreement was found for V1→V2 CF models. These values are comparable to those

64 4.3. Results

Figure 4.1: Visualization of pRF and CF maps obtained from VFM and RS 3T f-MRI data for a single participant. From left to right: eccentricity, polar angle, and pRF/CF size. → → The top two panels correspond to pRF and CF maps (for V1 V2 and V1 V3) derived from VFM estimates. The lower panels show parameter estimates for two resting state runs. Me- → → dian CF-scatter for RS1-derived V1 V2 and V1 V3 CF maps was 20 mm and 24.3 mm respectively whereas for RS was 16.1 mm, 19.6 mm. A threshold 0.35% variance explained was applied for visualization and computation of CF-scatter. To illustrate the inﬂuence of the segmentation, we show results for a satisfactorily (left) and an unsatisfactorily (pRF) mapped (right) hemisphere. Data are for participant 3.

65 . O T MRI

Table 4.2: Correlations between CF maps derived from VFM and RS data. Correlations were computed to estimate the agreement of the eccentricity (r, Pearson correlations) and polar angle (ϑ, circular correlations) maps in V2 and V3 derived using CF modeling. RS-derived maps were more similar to the VFM-derived map than between them. Grouped data from 5 participants (six hemispheres).

VFM-RS1 VFM-RS2 RS1-RS2 r ϑr ϑr ϑ → V1 → V2 0.3 0.71 0.36 0.74 0.3 0.61 V1 V3 0.23 0.4 0.3 0.38 0.15 0.25

shown in [12] (see Table 1 in Haak et al. [12]). Additionally, we computed correlations between CF maps derived from VFM and RS. Table 4.2 shows that RS-derived CF maps were more similar to VFM-derived CF maps than between them. Finally, we examined the relationship between CF size and pRF eccentricity. Figure 4.2 shows a weak but signiﬁcant correlation of CF size with eccentricity in VFM but not for RS. The same trend was observed in results obtained at 7T. However, comparing Figure 4.2 to Figure 2.6 (chapter 2) shows that the average CF size values for VFM data obtained at 3T ( 2 mm) were smaller than those obtained at 7T ( 3 mm).

4.3.2 Synchronization clusters derived from VFM and RS 3T fMRI activity Both in VFM and RS, clustering the most synchronized neighbouring recording sites revealed a modular structure that grouped locations with similar visual field position selectivity. Using each recording sites pRF position, Figure 4.3 illustrates synchronization clusters in visual field space (for the same participant shown in Figure 4.1). Note that the perimeter of some clusters in the left visual hemifield of V1 (left column in each condition) extend into the right hemifield. This apparent elongation is the consequence of poor pRF mapping (for V1and dorsal V2/V3 in the right hemisphere, see top panel in Figure 4.1). Finally, we examined how the probability of any two recording sites sharing a cluster decreased with the visuotopic distance between them (in the radial and arc directions). The range of shared cluster membership was greater along the radial◦ direction compared to the arc◦ direction both in VFM (decay factor: 1.48 for the◦ radial direction and 0.24 for the◦ arc direction) and in RS (decay factor: 2.35 for the radial direction and 0.39 for the arc direction). However, decay factors for both directions were slightly larger in◦ RS. A similar pattern was observed◦ in 7T data (decay factors for◦ VFM: 1.65 for the radial direction◦ and 0.25 for the arc direction; for RS: 1.93 for the radial direction and 0.3 for the arc direction).

66 4.3. Results

Figure 4.2: Connective field size as a function of pRF eccentricity in V2 and V3. A weak but significant correlation was found in VFM but not for RS. Gray circles are the raw data between 0-6 deg of eccentricity. Red dots indicate the VE-weighted mean CF size mean of each 1 deg eccentricity bin. Linear fits were calculated for these means. Dashed lines correspond to the 95% bootstrap confidence interval of the linear fit (1000 iterations). The Pear- son correlation coefficient for the raw data between 0-6 deg of eccentricity is reported together with its p-value. Grouped data from 5 participants (six hemispheres).

4.3.3 Limitations Since an accurate identification of visual field maps could not always be ac- complished with the current 3T data set, the present implementation of the assessed methods and interpretability of the results remains somewhat limited. We categorized the potential limitations affecting the obtention of meaningful CF models in two categories: 1) Factors relating to the acquisition and preprocessing of the data (i.e artifacts resulting from excessive movement, poor segmentation and misaligned volumes that may hinder an accurate pRF mapping. 2) Factors relating to the nature of the measurements (lower resolution, tissue specificity and SNR). We attribute the differences between 3T- and 7T-derived results mostly to the second category. Problems derived from the acquisition and preprocessing of the data can be identified, controlled and avoided, but the nature of the measurements cannot be easily changed. For example, achieving a good alignment and segmentation is crucial for obtaining good results –regardless of the magnetic field strength. Procedures such as using a short anatomical scan with the same field of view than the functional volumes to guide the alignment greatly

67 . O T MRI

Figure 4.3: Synchronization clusters obtained from VFM and RS fMRI data for a single participant. The clusters are depicted as a perimeter in visual ﬁeld space (hemiﬁelds) using each recording sites pRF positions (only recording sites within 0-6 degrees of eccentricity were used). Data are for participant 3 (same as in Figure 4.1).

Figure 4.4: Cluster membership probability as a function of visual field distance in RS and VFM. The probability of any two recording sites sharing a cluster decreases with the visuotopic distance between them for both in the radial (blue) and arc (red) directions. To quantify these decreases,◦ we fit an exponential decay function (colored lines) to the binomial probabilities of each 0.25 bin. The error bars correspond to confidence interval ◦of the bins. The continuous trace corresponds to the binomial probability fit of all data within 0-6 of visuotopic distance. The range of cluster membership probability◦ was greater along the radial direction◦ compared to the arc direction both in VFM (decay factor:◦ 1.48 for the radial direction◦ and 0.24 for the arc direction) and in RS (decay factor: 2.35 for the radial direction and 0.39 for the arc direction). The spread was greater along the radial direction compared to the arc direction. Data arefor5 participants.

68 4.4. Conclusion improves its quality. Unfortunately, such a short anatomical scan was not acquired for the RS data. Therefore, the alignment of the RS data was achieved manually using the mean functional volumes. Unfortunately, this procedure is less accurate and more prone to human error than using a short anatomical volume to guide the alignment. This resulted in the RS data being less accurately matched to the surface reconstruction and therefore —potentially— in higher levels of measurement error. It is important to note that the skills necessary to obtain a good alignment using the mean functional data are difficult to achieve, yet crucial. In future studies, extra care in this step is advised. Whole brain functional volumes typically suffer from signal drop-off and inhomogen- eous intensity. These artifacts depend on the coil used and can be minimized. The take home message is: always acquire a short anatomical scan, regardless of your protocol, and ideally refine the segmentation and alignment by hand. Another limitation is that here we compare results obtained from different magnetic strength fields and participants. A comparison of 3T- and 7T-derived results obtained from the same participants would be desirable. Moreover, we show CF maps and PS clusters for a participant in which the right hemisphere was not satisfactorily segmented. While this can be seen as a limitation, it helped to emphasize the influence of segmentation quality. By illustrating sat- isfactory and an unsatisfactory results in the same brain, Figure 4.1 stress that the specificity and accuracy of pRF and CF modeling is crucially dependent on the quality of the data (alignment and segmentation). Nevertheless, we believe that the present study suffices to demonstrate that meaningful CF models and PS clusters can be obtained from 3T data. Visuo- topically organized CF maps obtained at 3T were not as accurate as those based in 7T data, but still fairly meaningful to aid the analysis and interpretation of 3T data, and perhaps serve as a tool to explore cortico-cortical interactions in other experimental conditions and cortical areas.

4.4 Conclusion We have shown that two measures of local fMRI activity, CF estimates and phase synchronization clusters can be obtained based on 3T fMRI data. For CF models, a fair agreement can be observed between VFM/RS-based CF maps and pRF maps, and good agreement between diﬀerent RS-based based CF maps. PS clusters obtained from VFM and RS were also in good agreement. Both CF models and PS clusters results were qualitatively similar to those previously observed for 7T data. This corroborates the view that local functional connectivity reﬂects the underlying neuroanatomical architecture and endorses 3T fMRI as a suitable tool to study the extent of variability of these processes.

4.5 References [1] Kevin M Aquino, Mark M Schira, P A Robinson, Peter M Drysdale, and Breakspear Michael. Hemodynamic traveling waves in human visual cortex. PLoS Comput. Biol., 8(3):e1002435, 2012.

69 . O T MRI

[2] Richard F Betzel, Griﬀa Alessandra, Avena-Koenigsberger Andrea, Goñi Joaquín, Thiran Jean-Philippe, Hagmann Patric, and Sporns Olaf. Multi- scale community organization of the human structural connectome and its relationship with resting-state functional connectivity. Network Science, 1 (03):353–373, 2013.

[3] Vincent D Blondel, Jean-Loup Guillaume, Renaud Lambiotte, and Etienne Lefebvre. Fast unfolding of communities in large networks. J. Stat. Mech., 2008(10):P10008, 9 October 2008.

[4] Geoﬀrey M. Boynton, Stephen A. Engel, Gary H. Glover, and David J. Heeger. Linear Systems Analysis of Functional Magnetic Resonance Ima- ging in Human V1. Journal of Neuroscience, 16(13):4207–4221, July 1996.

[5] D H Brainard. The psychophysics toolbox. Spat. Vis., 10(4):433–436, 1997.

[6] E W Dijkstra. A note on two problems in connexion with graphs. Numer. Math., 1(1):269–271, 1959.

[7] Serge O Dumoulin and Brian A Wandell. Population receptive ﬁeld estimates in human visual cortex. Neuroimage, 39(2):647–660, 15 January 2008.

[8] Serge O Dumoulin and Brian A Wandell. Population receptive ﬁeld estimates in human visual cortex. Neuroimage, 39(2):647–660, 15 January 2008.

[9] Enrico Glerean, Juha Salmi, Juha M Lahnakoski, Iiro P Jääskeläinen, and Mikko Sams. Functional magnetic resonance imaging phase synchronization as a measure of dynamic functional connectivity. Brain Connect., 2 (2):91–101, 11 June 2012.

[10] Nicolás Gravel, Ben Harvey, Barbara Nordhjem, Koen V Haak, Serge O Dumoulin, Remco Renken, Branislava Curčić-Blake, and Frans W Cor- nelissen. Cortical connective ﬁeld estimates from resting state fMRI activity. Front. Neurosci., 8:339, 31 October 2014.

[11] Nicolás Gravel, Ben M. Harvey, Remco J. Renken, Serge O. Dumoulin, and Frans W. Cornelissen. Phase-synchronization-based parcellation of resting state fMRI signals reveals topographically organized clusters in early visual cortex. NeuroImage, September 2017. .

[12] Koen V Haak, Jonathan Winawer, Ben M Harvey, Remco Renken, Serge O Dumoulin, Brian A Wandell, and Frans W Cornelissen. Con- nective ﬁeld modeling. Neuroimage, 66:376–384, 1 February 2013.

[13] Koen V. Haak, Andre F. Marquand, and Christian F. Beckmann. Con- nectopic mapping with resting-state fMRI. NeuroImage, June 2017. .

70 4.5. References

[14] Angela R Laird, Baxter P Rogers, John D Carew, Konstantinos Arfanakis, Chad H Moritz, and M Elizabeth Meyerand. Characterizing instantaneous phase relationships in whole-brain fMRI activation data. Hum. Brain Mapp., 16(2):71–80, June 2002. [15] O Nestares and D J Heeger. Robust multiresolution alignment of MRI brain volumes. Magn. Reson. Med., 43(5):705–715, May 2000. [16] D G Pelli. The VideoToolbox software for visual psychophysics: trans- forming numbers into movies. Spat. Vis., 10(4):437–442, 1997. [17] Adrián Ponce-Alvarez, Gustavo Deco, Patric Hagmann, Gian Luca Ro- mani, Dante Mantini, and Maurizio Corbetta. Resting-state temporal synchronization networks emerge from connectivity topology and heterogeneity. PLoS Comput. Biol., 11(2):e1004100, February 2015. [18] P C Teo, G Sapiro, and B A Wandell. Creating connected representations of cortical gray matter for functional MRI visualization. IEEE Trans. Med. Imaging, 16(6):852–863, 1997. [19] B A Wandell, S Chial, and B T Backus. Visualization and measurement of the cortical surface. J. Cogn. Neurosci., 12(5):739–752, September 2000.

Propagation of BOLD activity reveals directed interactions across human visual cortex 5

It has recently been shown that large-scale propagation of blood-oxygen level dependent (BOLD) activity is constrained by anatomical connections and reflects transitions between behavioral states. It remains to be seen, however, if the propagation of BOLD activity can also relate to the brain anatomical structure at a more local scale. Here, we hypothesized that BOLD propagation reflects structured neuronal activity across early visual field maps. To explore this hypothesis, we characterize the propagation of BOLD activity across V1, V2 and V3 using a modeling approach that aims to disentangle the contributions of local activity and directed interactions in shaping BOLD propagation. It does so by estimating the effective connectivity (EC) and the excitability of a noise-diffusion network to reproduce the spatiotemporal covariance structure of the data. We apply our approach to 7T fMRI recordings acquired during resting state (RS) and visual field mapping (VFM). Our results reveal different EC interactions and changes in cortical excitability in RS and VFM, and point to a reconfiguration of feedforward and feedback interactions across the visual system. We conclude that the propagation of BOLD activity has functional relevance, as it reveals directed interactions and changes in cortical excitability in a task-dependent manner.

5.1 Introduction

Neuronal connections among cortical areas can be observed at a variety of scales, from laminar circuits to cortico-thalamic and cortico-cortical connections. To- gether they form a complex and intricate set of connections that serve as pathways for signal transmission and processing. The anatomy and function of

73 . P BOLD

such connections has been a focus of much research during the last decades [6, 84]. Thanks to non-invasive forms of neuronal recordings, like functional magnetic resonance imaging fMRI (see Raichle [68] for a detailed review), the link between structural and functional connectivity in the human brain has be- gun to be unravelled in-vivo. However, due to the multiple physiological mechanisms contributing to the BOLD signal (e.g. metabolic, vascular, neuronal, etc), its limited temporal resolution, and high noise level in its measurements [81, 13], it is still difficult to quantify and interpret this relationship link using fMRI. Several strategies have been proposed to address these issues and infer the efficacy with which anatomical connections modulate interactions between brain regions –referred to as “effective connectivity” (EC) [18, 24, 76, 22]. How- ever, most such strategies rely on the assumption of temporal precedence –the temporal resolution of the BOLD signal and its measurement must be sufficient to capture the time scale of modulatory influences. Crucially, the observed responses should reflect temporal dependencies within the system under scrutiny, which may not be the case if hemodynamic response delays differ between regions [2, 4, 31, 28]. Nevertheless, for whole-brain fMRI recordings, it has been recently shown that the propagation of BOLD activity reflects the transition between different behavioral states and may partly map to anatomical paths [57, 58]. This suggests a meaningful relationship between BOLD activity propagation and the modulation of communication between distant brain regions. Therefore, we ask if the propagation of BOLD activity can also reveal relevant aspects of brain activity at a more local scale, such as that of early cortical visual field maps, which are richly interconnected and where regional variation in the hemodynamic response is less pronounced [28, 49]. In the present study, we hypothesize that the propagation of BOLD activity across early cortical visual field maps V1, V2 and V3 reflects structured neuronal activity (e.g. modulation of EC weights) within and between these maps. To examine this hypothesis, we implement a data-driven model of EC informed by the empirical temporal autocovariance of the BOLD data. This model reproduces the spatiotemporal covariance structure (the zero-lag covariance and time-lag covariance) of the empirical data and the local cortical excitability. Our model thus accounts for the spatiotemporal stat- istics of BOLD activity and thereby the propagation of BOLD activity across different brain locations [22, 23]. We apply this approach to resting state (RS) and visual field mapping (VFM) fMRI recordings of the early cortical visual field maps V1, V2 and V3 in healthy human participants. In both RS and VFM data, we find a common structure underlying the EC of all participants, regardless of inter-participant variation in EC estimates. The common structure in EC links homotopic regions with similar visual field position selectivity both within and between the early cortical visual field maps, e.g. across both their topography and hierarchy. Furthermore, the resulting ECs capture different interaction regimes in RS and VFM. Within-area interactions are increased in RS, whereas in VFM between-area interactions are increased, particularly feedback interactions from V3 to V1. Moreover, local cortical excitability in V1 is greatly increased during RS but, during VFM, decreases to levels that are comparable to that of other visual areas. These differences point to a change of input to V1, and appear to reflect a re-configuration of feedfor-

74 5.2. Methods ward, lateral and feedback interactions. Finally, we interpret our results under the framework of predictive coding, emphasizing the role of recurrent cortical feedback during visual processing. Taken together, our results demonstrate that the propagation of BOLD activity through early visual cortex, as assessed by the present analysis, has functional relevance.

5.2 Methods

5.2.1 Data The empirical data used in this study stem from the dataset presented inGravel et al. [25]. It comprise visual ﬁeld mapping (VFM) and resting state (RS) 7T fMRI data from four healthy subjects with normal visual acuity. Experimental procedures were approved by the medical ethics committee of the University Medical Center Utrecht.

5.2.1.1 Visual field mapping Visual stimuli were presented by back-projection onto a 15.0 × 7.9 cm gamma- corrected screen inside the MRI bore. Participants viewed the display through prisms and mirrors, and the total distance from the participant’s eyes (in the scanner) to the display screen was 36 cm. Visible display resolution was 1024× 538 pixels. The stimuli were generated in Matlab (Mathworks, Natick, MA, USA) using the PsychToolbox [10, 66]. The visual field mapping paradigm consisted of drifting bar apertures at various orientations, which exposed a 100% contrast checkerboard moving parallel to the bar orientation. After each horizontal or vertical bar orientation pass, 30 s of mean-luminance stimulus were displayed. Throughout the VFM, participants fixated a dot in the center ofthe visual stimulus. The dot changed color between red and green at random intervals. To ensure attention was maintained, participants pressed a button on a response box every time the color changed. Detailed procedures can be found in Dumoulin and Wandell [16] and Harvey and Dumoulin [33]. The radius of the stimulation area covered 6.25 deg (eccentricity) of visual angle from the fixation point.

5.2.1.2 Resting state During the resting state scans, the stimulus was replaced with a black screen and participants closed their eyes. The lights in the scanning room were oﬀ and blackout blinds removed light from outside the room. The room was in complete darkness. Thus, visual stimulation was minimized. The participants were instructed to think of nothing in particular without falling asleep.

5.2.1.3 fMRI acquisition Functional T2*-weighted 2D echo planar images were acquired on a 7 Tesla scanner (Philips, Best, Netherlands) using a 32 channel head coil at a voxel resolution of 1.98 × 1.98 × 2.00, with a ﬁeld of view of 190 × 190 × 50 mm.

75 . P BOLD

◦ TR was 1500 ms, TE was 25 ms, and flip angle was set to 80 . The volume orientation was approximately perpendicular to the calcarine sulcus. In total, eight 240 volumes functional scans were acquired, comprising 5 resting state scans (RS) interleaved with 3 VFM scans (first was an RS scan). Five dummy volumes were scanned before data acquisition began and a further eight volumes were discarded from the beginning of each scan to ensure the signal had reached a steady state. High resolution T1-weighted structural images were acquired at a resolution of 0.49×0.49×0.80 mm (1 mm isotropic resolution for the second 252 × 252 × 190 dataset), with a field of view of◦ mm. TR was 7 ms, TE was 2.84 ms, and flip angle was 8 . We compensated for intensity gradients across the image using an MP2RAGE sequence, dividing the T1 by a co-acquired proton density scan of the◦ same resolution, with a TR of 5.8 ms, TE was 2.84 ms, and flip angle was 1 . Physiological recordings were not collected.

5.2.1.4 Preprocessing First, the T1-weighted structural volumes were resampled to 1 mm isotropic voxel resolution. Gray and white matter were automatically labeled using Free- surfer and labels were manually edited in ITKGray to minimize segmentation errors [80]. The cortical surface was reconstructed at the white/gray matter boundary and rendered as a smoothed 3D mesh [87]. Motion correction within and between scans was applied for the VFM and the RS scans [61]. Sub- sequently, data were aligned to the anatomical scans and interpolated to the anatomical segmentation space [61]. Instrumental drift was removed by detrending with a discrete cosine transform (DCT) filter with cutoff frequency of 0.01 Hz. The detrended signals were used for the estimation of the empirical spatiotemporal covariances. In order to reduce the influence of high frequency variation during pRF modeling, the detrended signals were filtered with a low- pass 4th order Butterworth filter with cutoff frequency of 0.1 Hz.

5.2.2 Selection of regions of interest Since the focus of our study was modeling the propagation of BOLD activity within and between early visual field maps, we did not consider all recorded locations in the scanning volume. Instead, we applied an ROI selection and a data-reduction step. First, we identified the visual field maps of the visual cortical areas V1, V2 and V3 (section 2.2.1). Second, we averaged the BOLD signals over the foveal and peripheral quarter-fields of these maps. This resulted in a network of 24 nodes/ROIs per participant (section 2.2.2).

5.2.2.1 Population receptive field modeling The visual field maps of V1, V2, and V3 were obtained using the population receptive field (pRF) method [16] applied to our VFM data. This method provides models that summarize the visual field position to which each recording site responds as a circular Gaussian in visual space. The Gaussian pRF model for each recording site was characterized by three parameters: x and y (position), and size (sigma). These parameters were determined by taking a

76 5.2. Methods large set of candidate pRF parameters, with each set defining a different Gaus- sian. By quantifying the overlap between each candidate pRF Gaussian and the stimulus aperture at each time point, we generate predictions of the neuronal response time course each candidate pRF would produce. This predicted neuronal response time course is convolved with a two-gamma model of the hemodynamic response [8] to give a set of candidate predicted fMRI response time courses for each candidate set of pRF parameters. The best fitting predicted fMRI time course and its associated pRF parameters are then taken to summarize the visual field selectivity of each recording site [16]. Recording sites were excluded from subsequent analyses if their best fitting pRF models explained less than 30% of response variance, or had visual field eccentricities beyond 6 deg.

5.2.2.2 Grouping of data into foveal and peripheral quarter-fields We grouped the resting state (RS) and the visual field mapping (VFM) time series over the foveal and peripheral quarter-fields of V1, V2 and V3 using the eccentricity and polar angle pRF preferences of each recording site. The foveal ROIs grouped recording sites with pRF positions below 2.2 deg eccentricity, while peripheral ROIs grouped recording sites with pRF positions above 2.2 deg eccentricity. Quarter-fields were divided at the vertical and horizontal visual field meridians. The grouping process resulted in a matrix of 24 nodes/ROIs, 8 for each complete visual field map (V1, V2 and V3). Signals for each of the 24 ROIs were obtained by averaging the minimally preprocessed BOLD time series (after detrending) within the ROIs.

5.2.3 Effective connectivity model for BOLD propagation In this section we examine the propagation of BOLD activity across the foveal and peripheral quarter-fields of V1, V2 and V3 (each of the 24 ROIs previously defined) using a recently proposed method [22]. This approach uses a noise- diffusion network model of effective connectivity (EC) and intrinsic variability to account for local BOLD variability and signal propagation lags between all possible pairs of ROIs. Importantly, the model captures the empirical data covariance and its spatio-temporal structure (the time-shifted covariances), effectively accounting for the propagation of BOLD activity. This has the advantage of relying on minimal assumptions: 1) the time constant of the generative model has to match the autocovariance time constant derived empirically from the data, 2) the regional variation in the hemodynamic response shape across early visual cortex should be minimal [28, 49] and 3) for each behavioral condition, a dominant pattern of neuronal interactions should influence and reflect in the average propagation structure of the BOLD signals. To examine our hypothesis (BOLD activity propagation reflects the consequences of structured neuronal activity), we model the EC under two conditions: 1) resting state and 2) the presentation of visual field mapping stimulus, and compare the two. We iteratively tune the model parameters (directed connectivity with notation C and intrinsic variability with notation Σ and) to reproduce the empirical spatiotemporal covariance and then use the C and Σ associated with the best fitting

77 . P BOLD

model as an estimate of the EC and the local cortical excitability of the actual data. We conclude comparing the resulting diﬀerences in EC and cortical excitability between RS and VFM.

5.2.3.1 Empirical spatiotemporal covariances To identify the spatiotemporal covariance structure of the data (the BOLD signals from each of the 24 ROIs. see Methods section 2.2) we estimated the covariance with and without time-shifts. For each participant and condition, the BOLD signals were ﬁrst demeaned and then, following Gilson et al. [22], the empirical covariance was calculated for zero-lag:

0 1 t t Qˆ = (s − s¯i)(s − s¯j ) , ij T − 2 ∑ i j (5.1) 1≤t≤T −1

and a lag of 1 TR:

1 1 t t+1 Qˆ = (s − s¯i)(s − s¯j ) . ij T − 2 ∑ i j (5.2) 1≤t≤T −1

t Here si is the observed BOLD time series for ROI i and s¯i is the mean BOLD level over the session. Afterwards, for each participant and session, we estimated the empirical time constant associated with the exponential decay of the autocovariance (averaged over all regions): N τ = log(Qˆ0 ) − log(Qˆ1 ) (5.3) ∑1≤i≤N ii ii where N correspond to the number of ROis. The time constant was used to calibrate the noise diﬀusion network model.

5.2.3.2 Noise-diffusion network model of EC and parameter estimation We choose a dynamic network model that captures the spatiotemporal dynamics of the data. Here we summarize the essential ingredients of the model and its optimization (for further details see Gilson et al. [22]). The models consists of 24 interconnected nodes (as defined in Methods section 2.2) that experience fluctuating activity and excite each other [22]. The local variability is described for each node by a variance corresponding to a diagonal term of the matrix Σ. The implicated fluctuations are shaped by the network EC (denoted by themat- C rix in the following equations) to0 generate the model FC, which is quantified by the zero-lag covariance matrix Q (FC0) and the time-lag covariance matrix Q1 Qˆ0 Qˆ1 (FC1) (the counterparts0 of the empirical1 and ). Subsequently, the model covariance matrices Q and Q that better reproduce the empirical spa- ˆ0 ˆ1 tiotemporal covariances Q and Q are approximated by iteratively adjusting the directional weights ( C ) and node excitabilities (Σ) of the model using Lya- punov optimization (LO) to reduce the model error E. The parameters C and

78 5.2. Methods

Σ associated with the best fitting model correspond to maximum-likelihood es- Cij timates [22]. Importantly, because1 asymmetry in the generates asymmetry in the time-shifted covariances Q , the model captures the average propagation structure between ROIs. We choose LO because it has several advantages to other methods [22]: 1) pairwise unconditional Granger causality does not take the whole network into account; 2) multivariate autoregressive models (MVAR) models that take the whole network into account may suffer from the down-sampling due to the time resolution (TR = 1.5 s); 3) physical interpretability might be hindered by over-parameterized dynamic causal models (DCM) [24, 29, 78]. These1 advantages allowed us to estimate Cij (and the corresponding asymmetry in Q ) and Σ as accurately as possible. Our approach was also justified because the decay time constant τ in Eq. (5.3) was consistently measured across participants, suggesting a diffusion process in the empirical data; the goal of our model inversion was then to exmine whether propagation was present in the data. Now we detail the equations relating these parameters, observables and measures. Formally, the network model is a multivariate Ornstein-Uhlenbeck process where the activity xi of node i decays exponentially with the time constant τ estimated from the data in Eq. (5.3). The evolution of each xi depends on the activity of other populations and the local variability: xt dxt =(− i + C xt )dt + dB , i τ ∑ ij j i (5.4) j=1̸ where dBi is —both spatially and temporally— independent Gaussian noise Σii Σ Bi with variance0 (the matrix is diagonal); formally a Wiener process. The model Q can be calculated for known C and Σ by solving the Lyapunov equation using the Bartels-Stewart algorithm [5]:

0 0 † JQ + Q J +Σ=0 (5.5) 1 and Q is then given by

1 0 † Q = Q expm(J ) , (5.6) where expm denotes the matrix exponential, the superscript † indicates the matrix transpose, and δij is the Kronecker delta. In those equations, the Jacobian J of the dynamic system is deﬁned as δ J = − ij + C . ij τ ij (5.7)

0 1 Eqs. (5.5) and (5.6) enable the quick calculation of Q and Q , without simulating the network activity. The Lyapunov optimization (LO) starts with zero connectivity (C = 0) and uniform local variances (Σii = 1). Each iteration of LO aims to reduce the model error deﬁned as

79 . P BOLD

||∆Q0||2 ||∆Q1||2 E(C, Σ) = + , ||Qˆ0||2 ||Qˆ1||2 (5.8)

0 0 0 1 1 1 with the diﬀerence matrices ∆Q = Q − Q and ∆Q = Q − Q ; the || · || b b vertical bars0 1 indicate the L1 matrix norm. To do so, we calculate the model Q and Q for the current values of the parameters C and Σ by solving Eqs. (5.5) and (5.6). Similar to a gradient descent, the Jacobian update is given by: † 0 −1 0 1 † ∆J =(Q ) [∆Q +∆Q expm(J )] , (5.9) which gives the connectivity update:

∆Cij = ηC ∆Jij . (5.10) where ηC is the optimization rate of C (here ηC = 0.0001). To take properly the network effects in the Σ update, we adjust the Σ update from the heuristic update in [22] as was done in [23]: 0 0 † ∆Σ = −ηΣ(J∆Q +∆Q J ) . (5.11) where ηΣ is the optimization rate of Σ (here ηΣ = 1). We impose non- negativity both for C and Σ. In addition, off-diagonal elements of Σ are kept equal to 0 at all times. The optimization steps are repeated until reaching a minimum for the model error E, giving the best fit and the model estimates. The C and Σ associated with best predicting model are then taken as a proxy for the EC between the 24 ROIs and their local cortical excitability (Σ). To facilitate comparison between RS and VFM, the resulting Σ values were further grouped across foveal and peripheral ROIs in each hemisphere (giving a total of 6 new ROIs) and differences between RS- and VFM-derived values evaluated for significance using single sided permutations. A similar approach was applied to the resulting Cij values (see next section). Further details about the mathematical formalism of the model and the optimization procedures can be found in Gilson et al. [22, 23].

5.2.3.3 Determination of common underlying structure in EC and its relation to topographic and anatomical connectivity Due to the nature of the model parameter estimation, the estimated EC weights (gain) may incidentally vary across participants (e.g. due to inter-participant variation, data quality, etc). However, relative diﬀerences in EC gain may still be similar across participants, suggesting a common underlying structure. To identify if there was a common structure underlying the EC of all participants, raw EC estimates were ﬁrst normalized by correcting for relative changes in EC gain across participants. Importantly, this process renders individual EC estimates comparable (since weight magnitudes are normalized) and enables the computation of an unbiased grand average. Gain correction was possible because, regardless of individual variation in EC gain across participants, the

80 5.3. Results comparison of raw EC gain values between each of the 6 possible participant pairs revealed correlated slopes (The Pearson correlation coefficient between individuals EC over the 6 possible combinations was, in RS: 0.28, 0.48, 0.43, p < 10−7 0.23, 0.23, 0.37 with −54in all cases, and, for VFM: 0.67, 0.73, 0.71, 0.58, 0.61, 0.68 with p < 10 in all cases), which we interpret as similar relative pairwise differences between EC links (and therefore similar structure). To normalizes EC weights across participants, we re-scaled the individual EC values by realigning the slopes (estimated as the Pearson correlation coefficient) of each pairwise EC difference to 1(based on an arbitrary reference participant). Subsequently, the normalized EC matrices were classified as intra- or inter-hemispheric and normalized EC links that appeared consistently across participants were detected using permutations corrected for multiple comparisons (with a threshold of p < 0.05). This allowed us to detect the common structure underlying the EC of all participants representing it as a binary matrix of significant connections. We then went on to examine differences in EC between RS and VFM. To this end, the normalized EC values that matched the matrix of significant connections were averaged across participants resulting in a grand average EC matrix. The averaged EC values were grouped across all quarter fields to give smaller EC matrices consisting of 6 ROIs (foveal and peripheral parts of V1, V2 and V3). Changes in EC between conditions were then quantified by computing the difference between the resulting VFM- and RS-derived grand average EC matrices. Finally, differences between RS and VFM were evaluated for significance across participants using permutations corrected for multiple comparisons.

5.3 Results

5.3.1 Propagation of BOLD activity across early visual cortex measured with a noise-diffusion network model of EC Figure 5.1 illustrates the propagation of an apparent wave of BOLD activity from the anterior calcarine sulcus (periphery of V1) to the occipital pole (foveal confluence of V1, V2 and V3) during rest (RS) (see Video 1S in supplementary materials). To estimate the propagation of BOLD activity across early visual cortex we first obtained visual field maps of V1, V2 and V3 using the population receptive field (pRF) method (Figure 5.2A). We then further subdivided these maps into foveal (below 2.2 deg of eccentricity) and peripheral (above 2.2 deg of eccentricity) quarter-fields (Methods section 2.2). This provided us with a functional map of cortex based on similarities in both retinotopy and hierarchy. Second, we characterized BOLD activity propagation patterns during RS and VFM through V1, V2 and V3 using a data-driven modeling approach (Methods section 2.3). Here we used the temporal autocovariance constant derived empirically from the data to calibrate a topologically agnostic (uncon- strained by anatomical connections) noise diffusion network model of EC and cortical excitability. Figure 5.2B presents the results from the analysis of the temporal autocovariance. Different time constants (τ, in seconds) for RS and VFM (mean (std) = 3.64 (1.46)) for RS and 6.66 (0.98) for VFM, paired t-test:

81 . P BOLD

Figure 5.1: Apparent propagation of BOLD activity during RS depicted in the flattened cortical surface reconstruction of the occipital pole of one participant’s cerebral hemisphere. Early visual field maps V1, V2 and V3 in one hemisphere are outlined in black (d- and v- denote dorsal and ventral). In the absence of visual stimulation (e.g. eyes closed, total darkness), spontaneous fluctuations in BOLD activity (indicated by the hot and cold colors) during RS can exhibit extensive spatiotemporal structure. This structure includes transient spatiotemporal fluctuation patterns that resemble stimulus-evoked BOLD waves as well as congruent and transient co-activations that occur across the visual field maps. Video 1S in supplementary materials shows a movie illustrating this propagation of BOLD activity.

−5 p < 10 ) demonstrate different propagation regimes. We then modeled the spatiotemporal covariance structure of the data by optimizing the noise- diffusion network parameters, namely the effective connectivity (EC) and the nodes excitabilities (Σ), to reproduce the empirical spatiotemporal covariances FC0 and FC1. Figure 5.2C illustrates one iteration step in the Lyapunov optimization (LO) procedure used to solve the model. The goodness of fit between

modeled and empirical spatiotemporal2 covariances was computed using the linear regression coeﬃcient R between the modeled and the empirical FC0 and FC1 (see Table 1).

5.3.2 Common underlying structures in EC We then asked if there was a common structure underlying the resulting ECs. Regardless of individual variations in EC values obtained from RS and VFM data (Figure 5.3A,E), a common underlying structure was revealed both for RS and VFM. Figures 5.3B and 5.3F compare relative pairwise differences between EC links in different participants pairs. The Pearson correlation coefficient between individuals EC over the 6 possible combinations was, in RS:

82 5.3. Results

Figure 5.2: Modeling the propagation of BOLD activity across visual field maps V1, V2 and V3. (A) Visual field maps in striate (V1) and extrastriate cortex (V2 and V3) were mapped using the population receptive field (pRF) modeling method [16]. Based on the visual field position selectivity estimates that the pRF method provides, these maps were further subdivided into foveal (below 2.2 deg of eccentricity) and peripheral (above 2.2 deg of eccentricity) quarter-fields in both hemispheres, giving a total of 24 ROIs (Methods section 5.2.2). We used a noise diffusion network model of effective connectivity (EC) to estimate the propagation of BOLD activity and the cortical excitability across these foveal and peripheral quarter-fields of V1, V2 and V3. (B) Logarithm of the autocovariance of the BOLD activity, averaged across participants and ROIs, as a function of the time shift (x-axis). The red lines link the mean of each box plot. The time constant τ was calculated for each participant and condition (Meth- ods section 2.3.1). The mean (standard deviation) of τ over all participants is indicated above the box plots. The empirical time constants were used to calibrate the noise-diffusion network model. (C) Schematic diagram illustrating one step in the Lyapunov optimization (LO) pro- Σ cedure. By iteratively adjusting the connectivity C0 and1 node excitability of the noise-diffusion network, the model spatiotemporal covariance (Q , Q ) approximates the empirical spatiotem- ˆ0 ˆ1 poral covariance (Q , Q ). For each participant and condition, the C and Σ corresponding to the best predicting model were taken as estimates of the underlying effective connectivity (EC) and the local cortical excitability (model optimization results for all participants are given in supplementary Figure 1S).

83 . P BOLD

Figure 5.3: Common underlying structure of EC across visual cortical areas V1, V2 and V3. (A) Diagonal and off-diagonal quadrants in each matrix represent within- and between-hemisphere EC across visual cortical areas (grouped by the colors), respectively. Inside each colored box, quarter-fields are grouped in the following order (from left to right): upper fovea, upper periphery, lower fovea and lower periphery. For each diagonal and off-diagonal quadrant, the upper triangle represents feedback connections and the lower triangle feedforward connections (rows correspond to inputs and columns correspond to outputs). White pixels represents stronger EC weights and darker pixels weaker EC weights, as indicated by the colorbars. (B) Scatter-plot comparing raw EC values between different participants. Different markers depict each of the 6 possible pairs. (C) Scatter-plot comparing the normalized EC values of each participant to the grand average EC. Different markers depict each participant. Despite individual variation in overall gain, the normalized EC for all participants was highly correlated to the grand average. (D) Common structure in EC. Black dots indicate connections that appeared consistently across participants. Significance was evaluated using permutations corrected for multiple comparisons (p< 0.05). (E-H) Same as A-D but for VFM. Differences between individual ECs were less pronounced for VFM data.

84 5.3. Results

Table 5.1: Goodness of ﬁt between modeled and empirical spatiotemporal covariances. For each participant and condition,2 we evaluated−50 the goodness of ﬁt by computing the linear regression (R , p < 10 for all cases) between the modeled and the empirical spatiotemporal covariances (FC0 and FC1).

RS VFM Subject FC0 FC1 FC0 FC1 1 0.652 0.629 0.694 0.655 2 0.473 0.329 0.702 0.689 3 0.884 0.829 0.705 0.665 4 0.847 0.819 0.754 0.731 Mean (std): 0.71 (0.19) 0.65 (0.23) 0.71 (0.027) 0.68 (0.033)

p < 10−7 0.28, 0.48, 0.43, 0.23, 0.23, 0.37 with −54in all cases, and, for VFM: 0.67, 0.73, 0.71, 0.58, 0.61, 0.68 with p < 10 in all cases. These results indicate similar relative pairwise differences between EC links (and therefore similar structure). Figures 5.3C and 5.3G compare the normalized ECs to the grand average EC. Despite individual variation in overall EC gain, the normalized ECs for all participants were highly2 correlated to the grand average EC. The linear correlation coefficient R between the normalized individual ECs and the grand average EC was: 0.60,−50 0.45, 0.46 and 0.47 for RS and 0.81, 0.67, 0.76 and 0.74 for VFM (p < 10 in all cases). These results indicate that differences between individual ECs are less pronounced in VFM. Figures 5.3D and 3H depict the EC links that appeared consistently across participants as directed binary matrices. We used these matrices as estimates of the common structure underlying the EC in RS and VFM. The common structures in EC closely matched the homotopy and hierarchy of the underlying anatomical connections (The Pearson correlation coefficient between the EC-structures and binary matrices for within-hemisphere and between-hemisphere connections p < 10−40 with ones indicating− homotopy9 was: 0.52 for VFM with and 0.24 for RS with p < 10 ). However, the homotopic organization of these was better captured in VFM (Figure 5.3H).

Σ 5.3.3 Differences in EC and between RS and VFM We then went on to examine the topology of the resulting common structures in EC and their differences between conditions. Figure 5.4 illustrates the common structures in EC (significant intra- and inter-hemispheric connections) across the foveal and peripheral quarter-fields of the visual field maps for RS and VFM. Both in RS and VFM, intra- and inter hemispheric connections linked regions with similar visual field selectivity, however, less inter-hemispheric connections were detected. Interestingly, in VFM, inter-hemispheric connections linked foveal regions only, whereas, in RS inter-hemispheric connections linked peripheral regions only. Both in RS and VFM, feedback connections outnumbered

85 . P BOLD

Figure 5.4: EC structure for RS and VFM illustrated in visual field space. The common structures in EC from Figure 5.3 (D,H) are displayed here as networks overlaying 3D representations of the visual field position selectivity of the 24 ROIs defined in section 2.2 (vertically stacked grids) by separating: 1) intra- and inter-hemispheric links, and 2) feedforward and feedback links. Colors depict the strengths of the grand average ECs (see Methods section 2.3.3 for details). In order to avoid overlaps between V1-V3 and V1-V2/V2-V3 links, we shifted the eccentricity associated to the visual position selectivity grid of V2 slightly to the periphery. The EC networks thus depicted suggest different interaction regimes in RS and VFM (see Figure 5.5 for a detailed assessment of these differences). The vertically stacked grids are oriented to match the left/right orientation of the visual field.

86 5.3. Results

Figure 5.5: Differences in EC and Σ between RS and VFM suggest a re-configuration of feedforward and feedback interactions. (A) Average EC matrices for RS (left panel) and VFM (middle panel) obtained by grouping significant EC links into foveal and peripheral regions V1, V2 and V3 (F and P stand for fovea and periphery. Columns represents output and rows inputs. See Methods section 2.3.3 for details). The right panel illustrates the differences in EC between conditions (VFM-derived EC minus RS-derived EC). Cells corresponding to regions that showed significant changes in EC have annotated the negative logarithm of their − − p-value ( log10(p), note that log10(0.05) = 1.3). (B) Cortical excitability as estimated by Σ. For each condition, individual Σ estimates were grouped into foveal (F) and peripheral (P) regions (as in A) and represented as black dots overlaid on box-plots (the central mark is the median and the edges the 25th and 75th percentiles). A comparison between RS and VFM reveal a slight decrease in this value, particularly for V1 (p =0.033 for V1 fovea, p =0.118 for V1 periphery, p = 0.33 for V2 fovea, p = 0.46 for V2 periphery, p = 0.08 for V3 fovea, and p = 0.271 for V3 periphery. Single sided permutation test). We believe that the increased values of Σ for V1 found in RS may relate to alpha activity (see Discussion section 4.2). In VFM, Σ is slightly increased in the periphery of V1 and V2 (although not significantly). Feedback connections outweighed feedforward connections both in resting state (RS) and visual field mapping (VFM). feedforward connections. Subsequently, to interpret changes in the EC and cortical excitability between RS and VFM, we grouped the corresponding EC and Σ values across the four quadrants in each visual field map to give foveal and peripheral regions of each visual field map (the 6 ROIs defined in section 2.3.3). We then evaluated differences in EC between the two conditions across participants using permutations corrected for multiple comparisons (with a significance threshold of p< 0.05). Figure 5.5A illustrates the resulting ECs and the differences between condi-

87 . P BOLD

tions (VFM-derived EC minus RS-derived EC). These results show that strong interactions within V1 and V3 in RS are absent in VFM. Conversely, feedforward interactions between V1 and V2 were present only in VFM, whereas feedforward interactions between V2 and V3 were present both in RS and VFM, although foveal interactions were increased for VFM. Further, feedback interactions between V2 and V1 and V3 and V2 were present both in RS and VFM, although feedback interactions between V3 fovea and V2 fovea were greatly increased in VFM. Notably, homotopic feedback interactions between V3 and V1 were only detected in VFM. Figure 5.5B illustrates the diﬀerences in cortical excitability between RS and VFM, as estimated by Σ. Changes were most pronounced in V1, with higher values of Σ for RS. Cortical excitability was signiﬁcantly higher for foveal regions of V1 only (p = 0.033. Single sided permutation test).

5.4 Discussion We assessed the propagation of BOLD activity through early visual cortical areas V1, V2 and V3 during RS and VFM using a data-driven modeling approach based on a noise-diffusion network model. Informed by the empirical spatiotemporal covariance structure of BOLD cofluctuations within and between visual cortical areas, this model estimates a topologically agnostic (un- constrained by anatomical connections) effective connectivity (EC). Our model decomposes the spatiotemporal structure of BOLD fluctuations into an EC parameter and a local cortical excitability parameter Σ. Importantly, the combination of the estimated parameters explain the temporal lags between BOLD signals from all pairs of ROIs, effectively accounting for observed propagation in the data. This discussion comprises four sections. In the first section we examine the neuroanatomical substrate and the possible mechanisms implicated by the different EC interactions estimated for RS and VFM. Our focus here is to emphasize the role of recurrent feedback connectivity and non-stimulus driven inputs, as well as examine the functional implications of our findings from a theoretical perspective –touching upon the notion of predictive coding [46, 69]. In the second section, we discuss the possible mechanisms that underlie the changes in cortical excitability (Σ) observed between RS and VFM and relate those to changes in EC. In the third section, we relate the BOLD autocovariance decay constant to different behavioral states. In the last section, we discuss the methodological and theoretical limitations of our study and raise questions for future research.

5.4.1 Recurrent connectivity and its role in visual processing We demonstrate that the propagation of BOLD activity across the topography and hierarchy of (e.g. within and between) visual field maps V1, V2 and V3 reveals different directed interaction regimes for RS and VFM (Figure 5.3). We relate these differences in EC to a task-dependent reconfiguration of feedforward and feedback interactions (Figure 5.4). Across visual field maps, feedfor-

88 5.4. Discussion ward EC interactions between the foveal representations of V1 and V2 were found in VFM but not in RS. However, later in the hierarchy, feedforward interactions between V2 and V3 were observed both in RS and VFM, though foveal interactions were increased in VFM (Figure 5.5A). We attribute these increased feedforward interactions during VFM, both between V1 and V2, and between V2 and V3, to stimulus-induced changes in neuronal interactions while participants are fixating on the screen, likely reflecting increased bottom-up processing of the stimulus across the visual hierarchy. Furthermore, homotopic feedback interactions between V2 and V1, and V3 and V2, were observed both in RS and VFM, although foveal interactions between V3 and V2 were greatly increased in VFM, again pointing to stimulus-induced changes. Remarkably, homotopic feedback interactions between V3 and V1 were only observed for VFM (Figure 5.5A), evidencing the role of extra-striate feedback in visual cortical processing. At the level of individual visual field maps, we found directed EC interactions from the peripheral to the foveal representations of V1 and V3 in RS but not in VFM. In principle, these periphery-to-fovea interactions may reflect a bias in the spontaneous propagation of correlated neuronal activity along gradients of anatomically connected pathways or ’connectopies’, such as the eccentricity map in the calcarine sulcus [26]. Also, other mechanisms such as intrinsic fluctuations in the ratio between excitation and inhibition and the modulation of lateral inhibitory coupling are known to give rise to spontaneous wave propagation patterns [17, 37]. Input changes, such as the restructuring of corticothalamic network activity, might adjust the balance between excitation and inhibition in cortical neuronal populations in a state-dependent way [79]. One interesting possibility is that these periphery-to-fovea interactions reflect large- scale cortical waves traveling in the frontal-to-occipital direction. Pre-stimulus waves in the alpha range [65] as well as slow <1 Hz waves propagating in an antero-posterior direction during sleep and calmness [53, 54] have been reported and may have a functional relevance [51, 39, 41]. In the absence of stimulation, the increased periphery-to-fovea EC interactions may well reflect visual cortical operations related to memory consolidation and learning [58]. On the other hand, the absence of periphery-to-fovea interactions during VFM can be explained by the stimulus set. Since the VFM stimuli consisted on a bar drifting at various orientations, the direction of neuronal response propagation may be balanced out and not reflected in EC estimates, which aim to assess stationary propagation patterns. Another line of evidence points to feedback modulation by top-down processes associated with the predictability of a given stimulus. Previous studies have suggested that predictable stimuli (e.g. drifting bars) induce less neuronal activity in early visual cortical areas than unpredict- able stimuli [9, 30]. Concomitantly, this reduced neuronal activity induced by predictable stimuli has been shown to result in suppressed BOLD responses in early visual cortex [72, 73]. These results can also be interpreted within a predictive inference framework [86, 60, 46]. In this framework, recurrent feedforward/feedback loops serve to integrate top-down contextual priors (predictions) and bottom-up visual input by implementing a convergent probabilistic inference along the visual hierarchy. From this perspective, the increased feedback to V1 and V2 observed for VFM might be interpreted as the dampening

89 . P BOLD

of feedforward visual responses according to prediction errors originated by the mismatch between incoming signals and feedback priors [46, 69, 19, 20, 71].

5.4.2 Cortical excitability: possible mechanisms Cortical excitability in RS, as quantified by Σ, was more variable than during VFM and was increased in V1, particularly in foveal representations (Figure 5.5B). These differences in cortical excitability between foveal and peripheral regions, although not consistently reaching statistical significance (p = 0.033 for V1 fovea), may still have functional implications. Interestingly, this increase in cortical excitability was accompanied by strong directed interactions in EC from the periphery to the fovea of V1 that were absent in VFM (Figure 5.5A). One possibility is that these differences are related to changes in the power of occipital alpha oscillations, known to increase during wakeful detachment from the environment (i.e RS) [88, 32]. If changes in the power of occipital alpha oscillations are partly captured by Σ, reduced cortical excitability in foveal regions of V1 could be interpreted as reflecting surround suppression and facilitation in the fovea while participants are fixating on the screen [32, 27]. Indeed, during a visual task, alpha oscillations in V1 have been associated with increased negative BOLD responses and shown to vary as a function of stimulus position and local receptive field surround [32]. The highly localized nature of these oscillations points to a role of intra-cortical axons in surround suppression [74, 75, 32, 38]. In line with these studies, changes in Σ were identified in the foveal representation of V1 but not in extrastriate areas V2 and V3, likely reflecting the localized nature of occipital alpha oscillations. Extrastriate feedback to V1 may play a role modulating the balance between inhibition and excitation, known to change between task and rest, thereby reflecting differently in the BOLD signal during RS and VFM [45, 67]. In the absence of visual input (RS), changes in cortico-cortical connectivity may leave V1 in a ’baseline’ state of potential excitation, whereas during VFM, both external visual input and top-down feedback influences may modulate the balance between inhibition and excitation resulting in suppressed BOLD responses in V1 –thus attenuating cortical excitability [3, 72, 73]. The fact that we found lower values of Σ in the fovea of V1 for VFM corroborates this view.

5.4.3 Relation of the BOLD autocovariance decay constant to behavioral condition The temporal decay constants of the autocovariance (τ), derived empirically from the RS and VFM data, determined the rate at which the fluctuations dif- fused through the noise-diffusion network model. Higher values of τ imply longer temporal memory (e.g. the system’s past dynamics have a stronger influence on its future dynamics). Importantly, the determination of consistent different decay constants for RS and VFM demonstrated that propagation was present in both cases, albeit at different temporal scales. Estimates of τ obtained from VFM data were greater than those derived from RS data (Figure 5.2B), indicating longer temporal memory during VFM. This is in contrast to previous

90 5.4. Discussion studies showing that temporal memory decreases during task compared to RS [34, 35]. By estimating task-induced decreases in the power-law exponent of BOLD fluctuations across widespread brain regions, these studies suggest that the temporal memory is longest during RS. They relate the larger power-law exponent found in RS to higher time-lagged autocovariances and interpret this as longer temporal memory. We find that the temporal memory (estimated by τ in our study) of visual cortical BOLD fluctuations is greater during task (here VFM) than RS. One possible reason for this differing results is that He and colleagues examined widespread whole-brain interactions, whereas here we examined BOLD signal dynamics at a more local scale (the cortical surface of individual visual field maps) and with higher resolution (7T). Furthermore, our results are specific to early visual cortex and therefore may not generalize to the whole brain. Another possible explanation is that these studies were based on eyes-open RS (fixation on a white cross-hair in the center of black screen) whereas we used eyes-closed RS. These different measurement scales and task protocols may well explain the observed differences. In our study, the longer temporal memory found in VFM likely reflects stimulus induced interactions. By giving rise to slow frequency fluctuations in the BOLD signal that are spatially correlated with the stimulus position, these interactions may lead to higher temporal redundancy and therefore longer memory depth (higher τ). On the other hand, the absence of stimulus induced interactions during RS may lead to a decrease in spatiotemporally correlated slow fluctuations, leaving intrinsic fluctuations and fast transitions to dominate the temporal autocovariance structure (thus reducing τ during RS). Finally, the aforementioned studies [34, 35], computed the power-law exponent of the fMRI time series by using the low frequency range (< 0.1 Hz) of the power spectrum, whereas we computed τ from minimally preprocessed BOLD time series to which only detrending and demeaning was applied. By avoiding such low-pass filtering, we allowed faster fluctuations to influence our estimates of τ. All these lines of evidence suggest that, in early visual cortex, the temporal scale of BOLD activity propagation differs between RS and VFM. Compared to the slow and spatially widespread (long) propagation patterns evoked by the VFM stimulus, intrinsic fluctuations during RS tend to unfold locally in space and time. These shorter and more localized propagation events may dominate the spatiotemporal covariance structure and explain the increased EC within-area interactions observed during RS.

5.4.4 Limitations and interpretability of the model An important limitation in the present study was the fact that we only considered a subset of all existing connections. This might have neglected the contribution of indirect connections to estimated changes in the EC, as dependencies may arise from indirect interactions in the underlying anatomy. Similarly, the increased Σ values identified during RS in the foveal representation of V1 may reflect input from other brain areas as well. Other possible limitations derived from the fact that our approach was different from the original implementation of the noise-diffusion network model [22, 23] in two aspects: 1) Diffusion tensor imaging (DTI) can’t estimate struc-

91 . P BOLD

tural connectivity at the spatial scales involved here. Therefore, our implementation was topologically agnostic compared to these previous studies: no structural connectivity matrix (e.g. DTI-derived) was used to constrain the EC; 2) we apply the approach to the scale of individual visual field maps whereas it was originally devised for whole-brain analyses (ROI ∼ 500-1000 voxels instead of ∼ 50 here) [22, 23]. However, we think the approach is still valid since there are known strong anatomical connections between V1, V2 and V3. Also, regional variation in the hemodynamic response is less pronounced at this scale [49], which further justifies the implementation of the model. Another limitation is that we only allowed positive weights to be adjusted in the EC, which lead us to interpret the intrinsic variability of the model (Σ) to the aggregate changes in cortical excitability across both inhibitory and excitatory neuronal populations. We justify this decision based on the metabolic underpinnings of the BOLD signal: inhibitory functions, which are supported more by oxidative mechanisms than by excitatory signaling, may contribute less than excitatory functions to the measured BOLD activity [13]. Therefore, excitatory glutamatergic input to principal neurons might influence EC more than modulatory functions, which are exerted by a mostly inhibitory interneuronal network [13]. Furthermore, allowing negative correlations between one voxel and another allows connections between a voxel where the stimulus is in the centre of the pRF and a voxel where the stimulus is in the suppressive surround. Suppressive surrounds are large, so this is likely to lead to widespread spurious EC. Based on these reasons, we believe that positive weights in the EC are enough to capture the underlying neuronal interactions that shape BOLD responses. We note that our aim was not to infer the detailed causal mechanism that give rise to the propagation of BOLD activity. Rather, our aim has been to assess the utility of a specific effective connectivity framework (a topologically agnostic noise-diffusion network) to quantify BOLD activity propagation at the level of the individual cortex. However, we note that local variation in neurovascular coupling profiles may hinder our analysis [4, 64], as they would affect the EC values (but less likely their modulations across conditions). Nev- ertheless, we do not model the hemodynamic response function for a number of reasons. First, we assume the HRF to be relatively constant across V1, V2 and V3, even though the underlying vascular network may introduce certain non-uniformity [31, 28, 82, 49]. Second, our model reproduces the empirical spatiotemporal covariance analytically, and therefore does not rely on generative models of neuronal activity and neurovascular coupling to simulate the BOLD time series. While such an approach would be valuable to address questions of mechanistic causality, we think that the current temporal and spatial resolution of fMRI leaves such questions out of reach. The time- and task-dependent nature of BOLD activity propagation patterns poses the question of how closely directed interactions map onto structural connections [1]. An emerging view suggests that structural connection patterns are indeed major constraints for the dynamics of brain activity [15], which are partly captured by functional and effective connectivity. However, whether BOLD propagation is generated only through temporally ordered processes of neuronal origin unfolding through underlying neuroanatomical networks, or

92 5.4. Discussion also through additional changes in physiological, metabolic or vascular variables remains an issue of debate [54]. Indeed, the precise neuronal mechanisms that determine the spatial and temporal distribution of BOLD signal cofluctuations and propagation are not yet fully understood. If BOLD fluctuations reflect the consequences of spiking activity, aggregate sub-threshold fluctuations [50], or metabolic relationships among neurons, astrocytes and the supporting capillary network (e.g. neuro-vascular coupling) [62, 63], is an area of ongoing research. On the one hand, BOLD activity propagation patterns within areas during VFM, together with transient, retinotopically selective, co-activation between different visual areas, appears to reflect retinotopically organized switching of spiking input activity between voxels sharing similar visual field position selectivity and tuning characteristics [42, 7, 47, 85]. On the other hand, during RS, BOLD propagation patterns may reflect the footprint of slow subthreshold fluctuations in local field potentials, which can be retinotopically organized and are known to be good predictors of the BOLD signal spatiotemporal covariance structure [52, 14]. Indeed, a recent study by Matsui and colleagues [54] using neuronal calcium signals and simultaneous hemodynamic recordings brings together these lines of evidence by demonstrating that both global fluctuations, in the form of waves propagating across cortex, and transient local co-activations in calcium signals are necessary for setting the spatiotemporal covariance structure of hemodynamic signals [54]. Another line of evidence points to neuronal mechanisms of inter-areal coupling and modulation reflected in the estimated changes in EC. A recent study analysed simultaneous recordings from V1 and V4 in monkeys and showed that feedforward interactions from V1 to V4 were based on frequencies around the gamma band [83, 56] whereas feedback interactions from V4 to V1 were supported by alpha activity [11, 90]. These study highlights the important fact that different temporal processes may be used as channels over which ’information’ flows between visual cortical areas. Therefore, care should be taken when interpreting the nature of the estimated interactions in EC. Furthermore, non-neuronal mechanisms such as the wave-like propagation of hemodynamic activity originating in pial arterioles has been described [70] and related to oscillations in systemic blood pressure (the so called ’Mayer waves’) [40]. Similarly, large veins draining to the dural sinuses near the occipital pole, which are known to modulate the phase of nearby hemodynamic fluctuations with little effect on signal amplitude [55, 89], may also play a role in shaping, for instance, the periphery-to-fovea interactions observed during RS. Moreover, acting as a temporal low pass filter, neurovascular coupling mechanisms involving the activity of astrocytes [63] and the diffusion of vasodilatory signalling molecules (e.g. nitric oxide) may play an important role in setting the pace of BOLD activity propagation patterns. Together, all these lines of evidence point to the importance of considering the multiple physiological factors implicated in shaping the BOLD signal when interpreting patterns of BOLD activity propagation. The different propagation patterns of BOLD activity during RS andVFM, as assessed with the present EC analysis, demonstrate distinct cortical dynamics during visual stimulation and in its absence [43, 48]. Nevertheless, in a previous analysis of the present dataset we found that BOLD activity across cortical

93 . P BOLD

locations of V1, V2 and V3 sharing similar visual field selectivity (functionally homotopic) can co-fluctuate during RS, enabling the estimation of connective field models from RS data (see Figure 3 in Gravel et al. (2014)) [25] that resemble those obtained from VFM data. The fact that retinotopically congruent cofluctuations in BOLD activity across visual cortical areas can also occur during RS [36, 25, 12] suggests that non-retinal and ’top-down’ influences, such as feedback modulation of V1 responses, may play a role generating structured patterns of BOLD propagation [59]. A variety of behavioral processes, such as memory consolidation and learning, may recruit V1 into a processing stream, even without external visual stimulation [44, 77, 67]. Together, these studies suggest that, during RS, periods of highly organized neuronal activity in the visual cortex give rise to transitory periods of retinotopically organized BOLD activity propagation. Cofluctuations within and between early cortical visual field maps may follow, likely reflecting different states of cortical processing [21]. Finally, the current study assesses BOLD propagation in four healthy participants. Although our results are consistent across participants, further studies involving more participants are advised. Moreover, the EC models were estimated based on entire RS and VFM scans. As such, they estimate average BOLD propagation patterns and do not capture specific intervals of variation in these. To establish the neuronal mechanisms underlying the observed changes in EC, further research is still necessary.

5.5 Concluding remarks

We have shown that the propagation of BOLD activity through early visual cortex reveals different directed interaction regimes across the topography and hierarchy of visual cortical areas V1, V2 and V3 during both RS and VFM. We relate these differences in the estimated EC to a task-dependent reconfiguration of feedfoward and feedback interactions throughout the visual system, and changes in Σ to a task-dependent neuronal modulation of local cortical excitability. Our results add to a growing body of evidence suggesting that recurrent connectivity and cortico-cortical feedback plays an important role in visual processing. They are consistent with the hypothesis that directed influences (e.g. feedback to V1), as well as intrinsic connectivity (e.g. cortical excitability), inter- act differently during visual stimulation and rest as a consequence of the visual system using an efficient predictive strategy to process incoming stimuli. We conclude by answering our original question of how the propagation of BOLD activity can also reveal relevant aspects of brain activity at a more local scale. By acknowledging the existence of propagated disturbances in BOLD activity, our approach provides a simple method to infer the local excitability of visual cortical areas and the directed influences unfolding among them during distinct behavioral states.

94 5.6. Supplementary Material

5.6 Supplementary Material Please refer to the online version of this article (doi.org/10.1101/172452) for the supplementary material.

5.7 References [1] Yusuke Adachi, Takahiro Osada, Olaf Sporns, Takamitsu Watanabe, Tep- pei Matsui, Kentaro Miyamoto, and Yasushi Miyashita. Functional connectivity between anatomically unconnected areas is shaped by collective network-level effects in the macaque cortex. Cereb. Cortex, 22(7):1586– 1592, July 2012. [2] G K Aguirre, E Zarahn, and M D’Esposito. The variability of human, BOLD hemodynamic responses. Neuroimage, 8(4):360–369, 1998. [3] Alessandra Angelucci and Paul C. Bressloff. Contribution of feedforward, lateral and feedback connections to the classical receptive field center and extra-classical receptive field surround of primate V1 neurons. Progress in Brain Research, 154:93–120, 2006. . [4] Kevin M Aquino, Mark M Schira, P A Robinson, Peter M Drysdale, and Michael Breakspear. Hemodynamic traveling waves in human visual cortex. PLoS Comput. Biol., 8(3):e1002435, 22 March 2012. [5] R. H. Bartels and G. W. Stewart. Solution of the Matrix Equation AX + XB = C. Commun. ACM, 15(9):820–826, September 1972. . [6] Bharat Biswal, F Zerrin Yetkin, Victor M Haughton, and James S Hyde. Functional connectivity in the motor cortex of resting human brain using echo-planar mri. Magn. Reson. Med., 34(4):537–541, 1995. [7] Barak Blumenfeld, Dmitri Bibitchkov, and Misha Tsodyks. Neural network model of the primary visual cortex: from functional architecture to lateral connectivity and back. J. Comput. Neurosci., 20(2):219–241, April 2006. [8] Geoffrey M. Boynton, Stephen A. Engel, Gary H. Glover, and David J. Heeger. Linear Systems Analysis of Functional Magnetic Resonance Ima- ging in Human V1. Journal of Neuroscience, 16(13):4207–4221, July 1996. [9] Oliver J Braddick, Justin M D O’Brien, John Wattam-Bell, Janette Atkin- son, Tom Hartley, and Robert Turner. Brain Areas Sensitive to Coherent Visual Motion. Perception, 30(1):61–72, January 2001. . [10] D H Brainard. The psychophysics toolbox. Spat. Vis., 10(4):433–436, 1997. [11] Elizabeth A. Buffalo, Pascal Fries, Rogier Landman, Timothy J. Busch- man, and Robert Desimone. Laminar differences in gamma and alpha coherence in the ventral stream. Proceedings of the National Academy of Sciences, 108(27):11262–11267, July 2011. .

95 . P BOLD

[12] Omar H Butt, Noah C Benson, Ritobrato Datta, and Geoffrey K Aguirre. Hierarchical and homotopic correlations of spontaneous neural activity within the visual cortex of the sighted and blind. Front. Hum. Neurosci., 9: 25, 10 February 2015. [13] György Buzsáki, Buzsáki György, Kaila Kai, and Raichle Marcus. Inhib- ition and brain work. Neuron, 56(5):771–783, 2007. [14] Matteo Carandini, Daisuke Shimaoka, L Federico Rossi, Tatsuo K Sato, Andrea Benucci, and Thomas Knöpfel. Imaging the awake visual cortex with a genetically encoded voltage indicator. J. Neurosci., 35(1):53–63, 7 January 2015. [15] Gustavo Deco, Morten L Kringelbach, Viktor K Jirsa, and Petra Ritter. The dynamics of resting fluctuations in the brain: metastability andits dynamical cortical core. Sci. Rep., 7(1):3095, 8 June 2017. [16] Serge O Dumoulin and Brian A Wandell. Population receptive field estimates in human visual cortex. Neuroimage, 39(2):647–660, 15 January 2008. [17] G B Ermentrout and J D Cowan. Temporal oscillations in neuronal nets. Journal of mathematical biology, 7(3):265–280, 1979. . [18] K Friston. Dynamic causal models for fMRI. In Statistical Parametric Mapping, pages 541–560. 2007. [19] Karl Friston. A theory of cortical responses. Philos. Trans. R. Soc. Lond. B Biol. Sci., 360(1456):815–836, 29 April 2005. [20] Karl Friston, James Kilner, and Lee Harrison. A free energy principle for the brain. J. Physiol. Paris, 100(1-3):70–87, July 2006. [21] Charles D Gilbert and Mariano Sigman. Brain states: top-down influences in sensory processing. Neuron, 54(5):677–696, 7 June 2007. [22] Matthieu Gilson, Ruben Moreno-Bote, Adrián Ponce-Alvarez, Petra Ritter, and Gustavo Deco. Estimation of directed effective connectivity from fMRI functional connectivity hints at asymmetries of cortical connectome. PLoS Comput. Biol., 12(3):e1004762, March 2016. [23] Matthieu Gilson, Gustavo Deco, Karl J. Friston, Patric Hagmann, Dante Mantini, Viviana Betti, Gian Luca Romani, and Maurizio Corbetta. Effective connectivity inferred from fMRI transition dynamics during movie viewing points to a balanced reconfiguration of cortical interactions. NeuroImage, October 2017. . [24] Rainer Goebel, Alard Roebroeck, Dae-Shik Kim, and Elia Formisano. Investigating directed cortical interactions in time-resolved fMRI data using vector autoregressive modeling and granger causality mapping. Magn. Reson. Imaging, 21(10):1251–1261, 2003.

96 5.7. References

[25] Nicolás Gravel, Ben Harvey, Barbara Nordhjem, Koen V Haak, Serge O Dumoulin, Remco Renken, Branislava Curčić-Blake, and Frans W Cor- nelissen. Cortical connective field estimates from resting state fMRI activity. Front. Neurosci., 8:339, 31 October 2014. [26] Koen V. Haak, Andre F. Marquand, and Christian F. Beckmann. Con- nectopic mapping with resting-state fMRI. NeuroImage, June 2017. . [27] Saskia Haegens, Verónica Nácher, Rogelio Luna, Ranulfo Romo, and Ole Jensen. α-Oscillations in the monkey sensorimotor network influence discrimination performance by rhythmical inhibition of neuronal spiking. Proc. Natl. Acad. Sci. U. S. A., 108(48):19377–19382, 29 November 2011. [28] Daniel A. Handwerker, John M. Ollinger, and Mark D’Esposito. Vari- ation of BOLD hemodynamic responses across subjects and brain regions and their effects on statistical analyses. NeuroImage, 21(4):1639–1651, April 2004. . [29] L Harrison, W D Penny, and K Friston. Multivariate autoregressive modeling of fMRI time series. Neuroimage, 19(4):1477–1491, August 2003. [30] L. M. Harrison, K. E. Stephan, G. Rees, and K. J. Friston. Extra-classical receptive field effects measured in striate cortex with fMRI. NeuroImage, 34(3):1199–1208, February 2007. . [31] Robert V Harrison, Noam Harel, Jaswinder Panesar, and Richard J Mount. Blood capillary distribution correlates with hemodynamic-based functional imaging in cerebral cortex. Cereb. Cortex, 12(3):225–233, March 2002. [32] B M Harvey, M J Vansteensel, C H Ferrier, N Petridou, W Zuiderbaan, E J Aarnoutse, M G Bleichner, H C Dijkerman, M J E van Zandvoort, F S S Leijten, N F Ramsey, and S O Dumoulin. Frequency specific spatial interactions in human electrocorticography: V1 alpha oscillations reflect surround suppression. Neuroimage, 65:424–432, 15 January 2013. [33] Ben M Harvey and Serge O Dumoulin. The relationship between cortical magnification factor and population receptive field size in human visual cortex: constancies in cortical architecture. J. Neurosci., 31(38):13604– 13612, 21 September 2011. [34] Biyu J He. Scale-free properties of the functional magnetic resonance imaging signal during rest and task. J. Neurosci., 31(39):13786–13795, 28 September 2011. [35] Biyu J He, John M Zempel, Abraham Z Snyder, and Marcus E Raichle. The temporal structures and functional significance of scale-free brain activity. Neuron, 66(3):353–369, 13 May 2010. [36] Jakob Heinzle, Thorsten Kahnt, and John-Dylan Haynes. Topographic- ally specific functional connectivity between visual field maps in the human brain. Neuroimage, 56(3):1426–1436, 1 June 2011.

97 . P BOLD

[37] Stewart Heitmann and G. Bard Ermentrout. Synchrony, waves and ripple in spatially coupled Kuramoto oscillators with Mexican hat connectivity. Biological Cybernetics, 109(3):333–347, June 2015. .

[38] Rikkert Hindriks, Michel J A M van Putten, and Gustavo Deco. Intra- cortical propagation of EEG alpha oscillations. Neuroimage, 103:444–453, December 2014.

[39] Shantanu P Jadhav, Caleb Kemere, P Walter German, and Loren M Frank. Awake hippocampal sharp-wave ripples support spatial memory. Science (New York, N.Y.), 336(6087):1454–8, June 2012. .

[40] Claude Julien. The enigma of Mayer waves: Facts and models. Cardiovas- cular Research, 70(1):12–21, 2006. .

[41] Raphael Kaplan, Mohit H. Adhikari, Rikkert Hindriks, Dante Mantini, Yusuke Murayama, Nikos K. Logothetis, and Gustavo Deco. Hippo- campal Sharp-Wave Ripples Inﬂuence Selective Activation of the Default Mode Network. Current Biology, 26(5):686–691, March 2016. .

[42] Tal Kenet, Dmitri Bibitchkov, Misha Tsodyks, Amiram Grinvald, and Amos Arieli. Spontaneously emerging cortical representations of visual attributes. Nature, 425(6961):954–956, 30 October 2003.

[43] Tal Kenet, Dmitri Bibitchkov, Misha Tsodyks, Amiram Grinvald, and Amos Arieli. Spontaneously emerging cortical representations of visual attributes. Nature, 425(6961):954–956, 30 October 2003.

[44] S M Kosslyn, W L Thompson, I J Kim, and N M Alpert. Topographical representations of mental images in primary visual cortex. Nature, 378 (6556):496–498, 30 November 1995.

[45] V A Lamme, H Supèr, and H Spekreijse. Feedforward, horizontal, and feedback processing in the visual cortex. Curr. Opin. Neurobiol., 8(4):529– 535, August 1998.

[46] Tai Sing Lee and David Mumford. Hierarchical bayesian inference in the visual cortex. J. Opt. Soc. Am. A Opt. Image Sci. Vis., 20(7):1434–1448, July 2003.

[47] Christopher M Lewis, Conrado A Bosman, Thilo Womelsdorf, and Pas- cal Fries. Stimulus-induced visual cortical networks are recapitulated by spontaneous local and interareal synchronization. Proc. Natl. Acad. Sci. U. S. A., 113(5):E606–15, 2 February 2016.

[48] Christopher M Lewis, Conrado A Bosman, Thilo Womelsdorf, and Pas- cal Fries. Stimulus-induced visual cortical networks are recapitulated by spontaneous local and interareal synchronization. Proc. Natl. Acad. Sci. U. S. A., 113(5):E606–15, 2 February 2016.

98 5.7. References

[49] Fa-Hsuan Lin, Jonathan R Polimeni, Jo-Fu Lotus Lin, Kevin W-K Tsai, Ying-Hua Chu, Pu-Yeh Wu, Yi-Tien Li, Yi-Cheng Hsu, Shang-Yueh Tsai, and Wen-Jui Kuo. Relative latency and temporal variability of hemodynamic responses at the human primary visual cortex. Neuroimage, 22 January 2017.

[50] N K Logothetis, J Pauls, M Augath, T Trinath, and A Oeltermann. Neurophysiological investigation of the basis of the fMRI signal. Nature, 412(6843):150–157, 12 July 2001.

[51] N K Logothetis, O Eschenko, Y Murayama, M Augath, T Steudel, H C Evrard, M Besserve, and a Oeltermann. Hippocampal-cortical interaction during periods of subcortical silence. Nature, 491(7425):547–53, November 2012. .

[52] Nikos K Logothetis and Brian A Wandell. Interpreting the BOLD signal. Annu. Rev. Physiol., 66:735–769, 2004.

[53] Marcello Massimini, Reto Huber, Fabio Ferrarelli, Sean Hill, and Giulio Tononi. The Sleep Slow Oscillation as a Traveling Wave. Journal of Neuro- science, 24(31):6862–6870, August 2004. .

[54] Teppei Matsui, Tomonari Murakami, and Kenichi Ohki. Transient neuronal coactivations embedded in globally propagating waves underlie resting-state functional connectivity. Proc. Natl. Acad. Sci. U. S. A., 113 (23):6556–6561, 7 June 2016.

[55] Ravi S. Menon, Seiji Ogawa, and K฀฀amilUgûrbil. High-temporal- resolution studies of the human primary visual cortex at 4 T: Teasing out the oxygenation contribution in fMRI. International Journal of Imaging Systems and Technology, 6(2-3):209–215, June 1995. .

[56] Georgios Michalareas, Julien Vezoli, Stan vanÃ฀ Pelt, Jan-Mathijs Schof- felen, Henry Kennedy, and Pascal Fries. Alpha-Beta and Gamma Rhythms Subserve Feedback and Feedforward Inﬂuences among Human Visual Cortical Areas. Neuron, 89(2):384–397, January 2016. .

[57] Anish Mitra, Abraham Z Snyder, Enzo Tagliazucchi, Helmut Laufs, and Marcus E Raichle. Propagated infra-slow intrinsic brain activity reorgan- izes across wake and slow wave sleep. Elife, 4, 9 November 2015.

[58] Anish Mitra, Abraham Z. Snyder, Carl D. Hacker, Mrinal Pahwa, Enzo Tagliazucchi, Helmut Laufs, Eric C. Leuthardt, and Marcus E. Raichle. Human cortical-hippocampal dialogue in wake and slow-wave sleep. Proceedings of the National Academy of Sciences, 113(44):E6868– E6876, November 2016. .

[59] Lars Muckli and Lucy S Petro. Network interactions: non-geniculate input to V1. Curr. Opin. Neurobiol., 23(2):195–201, April 2013.

99 . P BOLD

[60] David Mumford. Pattern theory: A unifying perspective. In First European Congress of Mathematics, Progress in Mathematics, pages 187– 224. Birkhäuser Basel, 1994. [61] O Nestares and D J Heeger. Robust multiresolution alignment of MRI brain volumes. Magn. Reson. Med., 43(5):705–715, May 2000. [62] Philip O’Herron, , Pratik Y Chhatbar, Levy Manuel, Shen Zhiming, Ad- rien E Schramm, Lu Zhongyang, and Kara Prakash. Neural correlates of single-vessel haemodynamic responses in vivo. Nature, 2016. [63] J C Pang, P A Robinson, K M Aquino, and N Vasan. Effects of as- trocytic dynamics on spatiotemporal hemodynamics: Modeling and en- hanced data analysis. Neuroimage, 14 October 2016. [64] J C Pang, P A Robinson, K M Aquino, and N Vasan. Effects of as- trocytic dynamics on spatiotemporal hemodynamics: Modeling and en- hanced data analysis. Neuroimage, 147:994–1005, 15 February 2017. [65] Timothy M. Patten, Christopher J. Rennie, Peter A. Robinson, and Pulin Gong. Human Cortical Traveling Waves: Dynamical Properties and Cor- relations with Responses. PLOS ONE, 7(6):e38392, June 2012. . [66] D G Pelli. The VideoToolbox software for visual psychophysics: trans- forming numbers into movies. Spat. Vis., 10(4):437–442, 1997. [67] Lucy S Petro, Luca Vizioli, and Lars Muckli. Contributions of cortical feedback to sensory processing in primary visual cortex. Front. Psychol., 5: 1223, 6 November 2014. [68] Marcus E Raichle. A brief history of human functional brain mapping. In Brain Mapping: The Systems, pages 33–75. 2000. [69] R P Rao and D H Ballard. Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. Nat. Neurosci., 2(1):79–87, January 1999. [70] Aleksandr Rayshubskiy, Teresa J. Wojtasiewicz, Charles B. Mikell, Mat- thew B. Bouchard, Dmitriy Timerman, Brett E. Youngerman, Robert A. McGovern, Marc L. Otten, Peter Canoll, Guy M. McKhann, and Eliza- beth M. C. Hillman. Direct, intraoperative observation of ~0.1hz hemodynamic oscillations in awake human cortex: Implications for fMRI. NeuroImage, 87(C):323–331, February 2014. . [71] Yulia Revina, Lucy S. Petro, and Lars Muckli. Cortical feedback signals generalise across different spatial frequencies of feedforward inputs. NeuroImage, September 2017. . [72] Wouter Schellekens, Richard J. A. van Wezel, Natalia Petridou, Nick F. Ramsey, and Mathijs Raemaekers. Predictive coding for motion stimuli in human early visual cortex. Brain Structure and Function, 221(2):879–890, March 2016. .

100 5.7. References

[73] Andreas Schindler and Andreas Bartels. Connectivity Reveals Sources of Predictive Coding Signals in Early Visual Cortex During Processing of Visual Optic Flow. Cerebral Cortex, 27(5):2885–2893, May 2017. . [74] Lars Schwabe, Klaus Obermayer, Alessandra Angelucci, and Paul C Bressloff. The role of feedback in shaping the extra-classical receptive field of cortical neurons: a recurrent network model. J. Neurosci., 26(36): 9117–9129, 6 September 2006. [75] Lars Schwabe, Jennifer M Ichida, S Shushruth, Pradeep Mangapathy, and Alessandra Angelucci. Contrast-dependence of surround suppression in macaque v1: experimental testing of a recurrent network model. Neuroimage, 52(3):777–792, September 2010. [76] Anil K Seth, Adam B Barrett, and Lionel Barnett. Granger causality analysis in neuroscience and neuroimaging. J. Neurosci., 35(8):3293–3297, 25 February 2015. [77] Scott D Slotnick, William L Thompson, and Stephen M Kosslyn. Visual mental imagery induces retinotopically organized activation of early visual areas. Cereb. Cortex, 15(10):1570–1583, October 2005. [78] Stephen M Smith, Karla L Miller, Gholamreza Salimi-Khorshidi, Mat- thew Webster, Christian F Beckmann, Thomas E Nichols, Joseph D Ramsey, and Mark W Woolrich. Network modelling methods for FMRI. Neuroimage, 54(2):875–891, 2011. [79] M. Steriade. Corticothalamic resonance, states of vigilance and mentation. Neuroscience, 101(2):243–276, 2000. [80] P C Teo, G Sapiro, and B A Wandell. Creating connected representations of cortical gray matter for functional MRI visualization. IEEE Trans. Med. Imaging, 16(6):852–863, December 1997. [81] Yunjie Tong, Tong Yunjie, and Blaise Deb Frederick. Studying the spatial distribution of physiological effects on BOLD signals using ultrafast fMRI. Front. Hum. Neurosci., 8, 2014. [82] Yunjie Tong, Lia M Hocke, Kimberly P Lindsey, Sinem B Erdoğan, Gordana Vitaliano, Carolyn E Caine, and Blaise Deb Frederick. Sys- temic Low-Frequency oscillations in BOLD signal vary with tissue type. Front. Neurosci., 10:313, 30 June 2016. [83] Timo van Kerkoerle, Matthew W Self, Bruno Dagnino, Marie-Alice Gariel-Mathis, Jasper Poort, Chris van der Togt, and Pieter R Roelfsema. Alpha and gamma oscillations characterize feedback and feedforward processing in monkey visual cortex. Proc. Natl. Acad. Sci. U. S. A., 111(40): 14332–14341, 7 October 2014. [84] J L Vincent, G H Patel, M D Fox, A Z Snyder, J T Baker, D C Van Es- sen, J M Zempel, L H Snyder, M Corbetta, and M E Raichle. Intrinsic functional architecture in the anaesthetized monkey brain. Nature, 447 (7140):83–86, 3 May 2007.

101 . P BOLD

[85] Martin Vinck and Conrado A Bosman. More gamma more predictions: Gamma-Synchronization as a key mechanism for eﬃcient integration of classical receptive ﬁeld inputs with surround predictions. Front. Syst. Neur- osci., 10:35, 25 April 2016. [86] H Von Helmholtz. Handbuch der physiologischen optik (english trans.). 1867. [87] B A Wandell, S Chial, and B T Backus. Visualization and measurement of the cortical surface. J. Cogn. Neurosci., 12(5):739–752, September 2000. [88] S J Williamson, L Kaufman, Z L Lu, J Z Wang, and D Karron. Study of human occipital alpha rhythm: the alphon hypothesis and alpha suppression. Int. J. Psychophysiol., 26(1-3):63–76, June 1997. [89] Jonathan Winawer, Rory A Sayres, and Brian A Wandell. Mapping hV4 and ventral occipital cortex : The venous eclipse Hiroshi Horiguchi. 10: 1–22, 2010. . [90] Dajun Xing, Chun-I. Yeh, Samuel Burns, and Robert M. Shapley. Lam- inar analysis of visually evoked activity in the primary visual cortex. Pro- ceedings of the National Academy of Sciences, 109(34):13871–13876, August 2012. .

102 General discussion 6

The aim of my thesis was to investigate the extent to which there is neuroanatomical organization and functional meaning to be found in resting-state functional magnetic resonance imaging (RS-fMRI) recordings from the human visual cortex. To this end, I used advanced anatomical MRI and fMRI techniques and novel fMRI analysis methods that allow for characterization of the spatial and temporal organization of intrinsic fluctuations in BOLD activity across the striate (V1) and extrastriate (V2 and V3) visual cortex. My results speak to the link between intrinsic patterns of intra- and inter-areal functional interactions and the underlying neuroanatomical organization of the visual cortex. They relate changes in those interactions to different behavioral states. My findings provide evidence that intrinsic fluctuations in fMRI activity recorded from early cortical visual field maps are pointers to functionally relevant processes and support the use of RS-fMRI for characterizing visual cortical function and connectivity in health and disease.

6.1 Summary of ﬁndings

6.1.1 Visuotopic maps from resting state In chapter 2, I mapped cortico-cortical population receptive fields between V1, V2 and V3 using connective field (CF) modelling and showed that retinotopically organized CF maps —like the those based on visual field mapping data— can be obtained from RS data. Previous studies have found cortico-cortical connectivity patterns in RS data [12, 20], but did not reveal the detailed visuotopic maps directly from RS data that I found. However, I found that only part of the RS data supported the derivation of retinotopically organized maps. One way to interpret this finding is that the retinotopically organized co-fluctuations

103 . G

between cortical visual field maps occur in the form of spatially and temporally discrete events. Due to the non-stationary nature of RS-fMRI activity, retinotopically organized RS-derived visuotopic maps —which in my case were computed for 6-minute recordings— might reflect the density of such local events. Alternatively, physiological factors such as motion, respiration and changes in input from the autonomous nervous system might contribute differentially from recording to recording [18, 26]. Another interesting possibility is that this type of activity in early visual areas reflects non-retinal neuronal input to the visual cortex as well. However, based on the present data, I cannot determine the extent to which the variability is explained by genuine changes in neuronal activity or such other factors.

6.1.2 Similar synchronization clusters in resting state and visual field mapping In chapter 3, I examined synchronization patterns of fMRI activity across V1, V2 and V3 in RS and visual field mapping (VFM) fMRI data. I found synchronization clusters that were similar, regardless of whether they were derived from RS or VFM data. However, phase synchronization is more extensive in VFM, likely reflecting stimulus induced interactions between local responses. I also found that synchronization clusters with similar retinotopic selectivity are synchronized across the hierarchy of the visual cortical areas and hemispheres, reflecting homotopic anatomical connections. My findings suggest that synchronization clusters obtained from RS and VFM share a common neuroanatomical origin. My work corroborates previous studies that have found similar patterns of neuronal interactions using different methods [20, 2, 19]. Importantly, my findings justify and facilitate the direct comparison of RS and stimulus-evoked activity.

6.1.3 It is feasible to obtain connective field maps and synchronization clusters from 3T fMRI activity In chapter 4, I studied the feasibility of reproducing the findings of chapter 2 & 3 using 3T instead of 7T fMRI. I found that the results obtained at 3T are in fair agreement to those obtained previously with 7T data, despite the lower resolution and reduced signal-to-noise ratio of the 3T data. I categorized the limitations affecting the feasibility of reproducing the results into factors related to the acquisition and preprocessing of the data (e.g., artifacts resulting from excessive movement, poor segmentation or misaligned volumes that may hinder an accurate pRF mapping) and factors related to the nature of the measurements (lower resolution, tissue specificity and SNR) and related the differences between 3T- and 7T-derived results mostly to the second category. I suggest that the quality of the measurements at 3T could be further improved by avoiding excessive movement, adding a denoising step, and achieving a good alignment and segmentation.

104 6.2. Discussion

6.1.4 Propagation of BOLD activity reveals task-dependent changes in effective connectivity In chapter 5, I modeled cortical excitability and directed interactions across V1, V2 and V3 to explain task-dependent propagation patterns in BOLD activity in RS and VFM. I found that the differences in cortical excitability and directed interactions between RS and VFM point to a task-dependent reconfiguration of local, feedforward and feedback interactions across the visual system. Previous studies have shown task-dependent reconfiguration of directed interactions a the scale of the whole brain [15, 10]. Here, my main contribution is that I show that this type of change in interactions can also be found at the mesoscopic scale of cortical visual field maps.

6.2 Discussion In order to establish the functional relevance of intrinsic fMRI activity across visual field maps (V1, V2 and V3), I used a data-driven modeling approach to compare the spatiotemporal characteristics of intrinsic co-fluctuations in fMRI recordings acquired while subjects were in RS to those obtained during VFM. I showed that fMRI activity recorded from these areas during RS followed the underlying retinotopic organization. Intrinsic variations in BOLD activity and inter-areal coupling occurred in a coordinated manner between cortical visual field maps. Importantly, differences in the structure of directed interactions between RS and VFM reveal a task-dependent reconfiguration of feedforward and feedback interactions across the visual system. Hence, they constitute state- dependent interactions between distinct parts of the visual cortices. Therefore, these patterns suggest that the intrinsic variations are partly a consequence of functionally relevant brain events. The variations itself need not be relevant, but can simply be an epiphenomenon related to functional events. They are like the ripples that indicate that a stone was thrown in the water. In what follows, I will relate the findings of my thesis to different perspectives.

6.2.1 Evolution, development and homeostasis Animal nervous systems are intricate and complex, yet smoothly interface the animal’s internal state to its sensorium [29]. How an organism coordinates its internal and external affairs efficiently —thereby narrowing an immense number of possibilities to behaviorally relevant ones— is one of the most challenging questions in contemporary neuroscience. Any attempt to answer this question would be impossible without situating the organism in its ecological context. Constrained by evolutionary and developmental expediency forces, the neuroanatomical circuitry of the human sensory cortices has been shaped according to the structural properties of our perceptual habitats. Hence, from this broader naturalistic perspective, finding patterns of highly organized intrinsic activity in the visual cortex —as I did— is not a mystery but something to be expected.

105 . G

For example, it has been shown that —during the development and morpho- genesis of the nervous system— spontaneous retinal waves contribute to the formation of ocular dominance columns and retinotopic maps [1]. Moreover, patterns of intrinsic neuronal activity in the visual cortex can be remarkably similar to those evoked by visual input, suggesting that sensory input modulates pre-existing patterns rather than evoking the neuronal activity all by itself. Additional insight on the role of intrinsic brain activity comes from homeo- static mechanisms [? ]. The balance between excitatory and inhibitory functions within neuronal populations maintains a stable working point in which the populations can optimally process incoming signals according to the demands posed by different behavioral states and tasks [11]. As a result, neuronal populations experience transitions between quiescent (Down) and depolarized (Up) states [25]. Sculpted by the underlying neuroanatomical backbone, intrinsic transitions between Down and Up states may result in spatiotemporally correlated population activity. Indeed, using neuronal calcium and intrinsic signal imaging in vivo, Matsui and colleagues have found that transient neuronal coactivations embedded in widespread propagating waves may underlie RS hemodynamic fluctuations [14]. Therefore, an interesting possibility is that these fluctuating dynamics reflect in the RS-fMRI signal as retinotopically organized fluctuations and patterns of inter-areal coupling. If this is the case, the retinotopically organized CF maps that I derived from RS-fMRI data might be the footprint of such dynamic processes.

6.2.2 Evoked versus intrinsic activity Compared to understanding the relevance of externally evoked activity (e.g., during stimulation), contemporary research has paid relatively little attention to understanding the meaning of changes in intrinsic activity. In a way, this is strange since the changes in energy consumption that stimulation evokes in the brain are small compared to the energy spent in sustaining intrinsic fluctuations in brain activity [3]. Moreover, the common approach of averaging across many trials effectively allows the identification of consistent responses, but neg- lects the potential informativeness of variability. Nevertheless, since evoked responses can be controlled and are relatively robust, intrinsic activity has been typically regarded as and treated as noise. This has the advantage of being able to separate meaningful ”signals” from the “noisy” baseline fluctuations. But this focus holds a risky assumption: the brain is conceived as a passive receiver of incoming signals transmitted by the world, with the visual system being just a channel. This view implicates the notion of an ulterior observer, which brings us back to disparate metaphors such as the Cartesian theater [5]. On the other hand, if the brain is conceived as being primarily concerned with maintaining its internal dynamics —and the environmental input modulating rather than driving it— the spontaneous endogenous dynamics by necessity hold meaning and are not simply noise. Seeing the brain in this light, I would like to advocate that —in addition to externally triggered responses— there is a constant stream of internal processes that we should not ignore. Drawing from Shannon’s information theory, I propose that any responses to external input are affected by the current state

106 6.3. Future research and applications of the receiver and the channel, and not just by the sender [23]. In this sense, there is no such thing as objective information because information is always subject to interpretation by a receiver. Put diﬀerently, for a signal to be truly informative, it has to make a diﬀerence to a receiver –in this case the brain. In most contemporary neuroimaging studies, brain signals are interpreted from the perspective of an external observer (‘experimenter-as-receiver’). However, the question is whether those signals are actually used by the rest of the brain (‘cortex-as-receiver’) [4]. Retinotopic maps are a prime example of this.

6.2.3 Predictive coding and enaction Traditional models of brain function and information processing can be regarded as ‘sandwich models’ in the sense that they assume three stages of information processing. Akin to the reﬂex-arc paradigm, perception, cognition and action follow each other in a sequential fashion. This view of brain function has proved useful to understand brain activity and its neuronal correlates in a variety of experimental and clinical conditions. However, over the past decades, its limitations have also become apparent and a paradigm shift has taken place. The new paradigms suggest that information processing can be explained from a more constructivist and enactive perspective that speaks to recurrent neuronal processing and a circular causality among processing stages and the world [27, 9]. From this perspective, the visual cortex is not just driven by feedforward external input but by a hierarchical inference machinery built upon internal models derived from learned associations [28, 16, 21, 13, 30]. According to the predictive coding hypothesis, the main function of this machinery is to predict sensory input [8, 21]. The diﬀerence between the incoming sensations and the internally generated predictions gives rise to prediction errors. Subsequently, these prediction errors are used to update the neuronally encoded models of the causes that generated the expected sensation [7]. The preponderance of feedback connections in the visual cortex supports this view [22, 24]. In chapter 5, I observed state-dependent changes in feedforward, feedback, and local interactions. Interestingly, I found that feedback from V3 to V1 only could be observed during stimulation. This feedback may well relate to modulatory signals generated by the putative predictive coding mechanisms [6]. If this is the case, the signals underlying these feedforward and feedback interactions are neither purely sensory, nor purely intrinsic or modulatory, but have sensory and intrinsic components that reveal us as organisms actively involved in the world, rather than as its passive observers [17, 30].

6.3 Future research and applications In the introduction, I stated the expectation that a better understanding of the relationships between neuronal activity, brain anatomy and hemodynamics would help develop RS-fMRI as a valuable tool for fundamental and applied —non-invasive— research in humans. I believe my work has contributed to realizing this goal. I have implemented new techniques to estimate visuo-

107 . G

topic connectivity from RS-fMRI data, assessed the similarities between RS and VFM fMRI data and implemented an approach to measure directed interactions across early visual cortical areas. The novel analysis techniques and analyses methods implemented inmy thesis have made it possible to establish similarities and differences between RS and stimulus-evoked fMRI activity. Future studies could assess task-dependent changes at the mesoscopic scale of cortical visual field maps in different experimental and behavioral conditions in health and disease. With appropriate analysis techniques, even more details could be extracted from the RS data, such as the temporal density of retinotopically organized fluctuations. This would allow to relate moments of organized activity in early visual cortex to activity in other brain regions and the interpretation of these as the consequences of non-retinal visual processing. Future studies could apply and extend the techniques used in my thesis to study the properties of deafferented parts of visual cortex or cortical lesion projection zones. This could reveal changes in cortical function in patients with glaucoma, macular degeneration and other eye or brain diseases. Deficient top-down processing is implicated in neuropsychiatric diseases, such as schizophrenia. My methods could be used to shed light on interactions across cortical visual field maps in various patients with visual hallucinations. This might reveal possible imbalances in their feedforward-feedback integration mechanisms. This could be combined with eye-tracking and stimuli that invite predictive responses to examine aberrant precision control in such clinical populations. Of interest to future studies of RS activity in visual cortical areas is the correction of physiological factors such as head motion, respiration and changes in the autonomic nervous system input, which might contribute differentially from recording to recording [18]. More sophisticated acquisition and analyses approaches could allow examining the quality of processing in the visual cortex with an eye on predicting the success of rehabilitation or retinal or cortical implants.

6.4 Conclusions

For this thesis, I posed as my main question: What can we learn about the neuroanatomical organization of the human visual cortex from RS-fMRI recordings? First of all, I found similar visuotopic connective field maps and synchronization clusters in resting state and visual field mapping. Moreover, I found that it is feasible to obtain such connective field maps and synchronization clusters from 3T fMRI activity. I also made suggestions for how the quality of measurements at 3T could be further improved. Finally, I found that the propagation of BOLD activity reveals task-dependent changes in the effective connectivity in the visual cortex. Here, my main contribution is that I showed that these types of change in the interactions can also be found at the mesoscopic scale of cortical visual field maps. They point to a task-dependent reconfiguration of local, feedforward and feedback interactions across the visual system.

108 6.5. References

Hence, my main answer to my question is that RS-fMRI recordings can reveal a remarkable amount of detailed properties regarding the functional neuroanatomical organization, also at the relatively fine scale of the human visual cortex. Further, my findings justify and facilitate the direct comparison of RS and stimulus-evoked activity. While I have focused on RS activity, the methods presented in this thesis can also be applied to study task-dependent changes in a variety of experimental and behavioral conditions, both in health and disease. Finally, my results establish RS-fMRI recordings as an important tool to query the neuroscientific framing of our inner mental life.

6.5 References

[1] James B. Ackman and Michael C. Crair. Role of emergent neural activity in visual map development. Curr Opin Neurobiol, 0:166–175, February 2014. .

[2] Omar H Butt, Noah C Benson, Ritobrato Datta, and Geoﬀrey K Aguirre. The ﬁne-scale functional correlation of striate cortex in sighted and blind people. J. Neurosci., 33(41):16209–16219, 9 October 2013.

[3] György Buzsáki, Buzsáki György, Kaila Kai, and Raichle Marcus. Inhib- ition and brain work. Neuron, 56(5):771–783, 2007.

[4] Lee de Wit, David Alexander, Vebjørn Ekroll, and Johan Wagemans. Is neuroimaging measuring information in the brain? Psychon. Bull. Rev., 23(5):1415–1428, October 2016.

[5] Daniel C. Dennett. Styles of Mental Representation. Proceedings of the Aristotelian Society, 83:213–226, 1982.

[6] Grace Edwards, Petra Vetter, Fiona McGruer, Lucy S. Petro, and Lars Muckli. Predictive feedback to V1 dynamically updates with sensory input. bioRxiv, page 180539, August 2017. .

[7] K Friston and S Kiebel. Predictive coding under the free-energy principle. Philos. Trans. R. Soc. Lond. B Biol. Sci., 364(1521):1211–1221, 2009.

[8] Karl Friston. A theory of cortical responses. Philos. Trans. R. Soc. Lond. B Biol. Sci., 360(1456):815–836, 29 April 2005.

[9] Karl J. Friston. Precision Psychiatry. Biological Psychiatry: Cognitive Neuroscience and Neuroimaging, 2(8):640–643, November 2017. .

[10] Matthieu Gilson, Gustavo Deco, Karl J. Friston, Patric Hagmann, Dante Mantini, Viviana Betti, Gian Luca Romani, and Maurizio Corbetta. Eﬀective connectivity inferred from fMRI transition dynamics during movie viewing points to a balanced reconﬁguration of cortical interactions. NeuroImage, October 2017. .

109 . G

[11] Bilal Haider, Alvaro Duque, Andrea R. Hasenstaub, and David A. Mc- Cormick. Neocortical Network Activity In Vivo Is Generated through a Dynamic Balance of Excitation and Inhibition. J. Neurosci., 26(17):4535– 4545, April 2006. . [12] Jakob Heinzle, Thorsten Kahnt, and John-Dylan Haynes. Topographic- ally specific functional connectivity between visual field maps in the human brain. Neuroimage, 56(3):1426–1436, 1 June 2011. [13] Tai Sing Lee and David Mumford. Hierarchical bayesian inference in the visual cortex. J. Opt. Soc. Am. A Opt. Image Sci. Vis., 20(7):1434–1448, July 2003. [14] Teppei Matsui, Tomonari Murakami, and Kenichi Ohki. Transient neuronal coactivations embedded in globally propagating waves underlie resting-state functional connectivity. Proc. Natl. Acad. Sci. U. S. A., 16 May 2016. [15] Anish Mitra and Marcus E Raichle. How networks communicate: propagation patterns in spontaneous brain activity. Philos. Trans. R. Soc. Lond. B Biol. Sci., 371(1705), 5 October 2016. [16] D. Mumford. On the computational architecture of the neocortex. Biol. Cybern., 65(2):135–145, June 1991. . [17] S Murray Sherman and R W Guillery. Functional Connections of Cortical Areas: A New View from the Thalamus. MIT Press, 9 August 2013. [18] Hyeong-Dong Park, Stéphanie Correia, Antoine Ducorps, and Cather- ine Tallon-Baudry. Spontaneous fluctuations in neural responses to heart- beats predict visual detection. Nature neuroscience, 17(4):612–8, 2014. . [19] Andrew Parker, Parker Andrew, Bridge Holly, and Coullon Gaelle. Spa- tial scale of correlated signals in 7T BOLD imaging. In Proceedings of the 9th EAI International Conference on Bio-inspired Information and Commu- nications Technologies (formerly BIONETICS), 2016. [20] Mathijs Raemaekers, Raemaekers Mathijs, Schellekens Wouter, Richard J A van Wezel, Petridou Natalia, Kristo Gert, and Nick F Ramsey. Pat- terns of resting state connectivity in human primary visual cortical areas: A 7T fMRI study. Neuroimage, 84:911–921, 2014. [21] R P Rao and D H Ballard. Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. Nat. Neurosci., 2(1):79–87, January 1999. [22] R P Rao and D H Ballard. Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. Nature neuroscience, 2(1):79–87, January 1999. . [23] C. Shannon. The bandwagon. IRE Transactions on Information Theory, 2 (1):3–3, March 1956. .

110 6.5. References

[24] Fraser W Smith and Lars Muckli. Nonstimulated early visual areas carry information about surrounding context. Proc. Natl. Acad. Sci. U. S. A., 107 (46):20099–20103, 16 November 2010. [25] M. Steriade. Corticothalamic resonance, states of vigilance and mentation. Neuroscience, 101(2):243–276, 2000. [26] Catherine Tallon-Baudry, Florence Campana, Hyeong-Dong Park, and Mariana Babo-Rebelo. The neural monitoring of visceral inputs, rather than attention, accounts for ﬁrst-person perspective in conscious vision. Cortex, June 2017. . [27] Francisco J Varela. Autopoiesis and a Biology of Intentionality. 1991. [28] H Von Helmholtz. Handbuch der physiologischen optik. 1867. [29] Jakob von Uexküll. A Foray into the Worlds of Animals and Humans. [30] Alan Yuille and Daniel Kersten. Vision as bayesian inference: analysis by synthesis? Trends Cogn. Sci., 10(7):301–308, July 2006.

111

Thesis summary

The existence of intrinsic neuronal activity has been demonstrated atmany scales using a variety of methodological approaches, yet its biological relevance is only beginning to be understood. One way to study intrinsic neuronal activity in humans is by using resting-state (RS) functional magnetic resonance imaging (fMRI). In this thesis, I investigated the neuroanatomical organization and the biological relevance of RS-fMRI activity recorded from the human visual cortex. Using novel analysis methods, I examined the spatial and temporal organization of intrinsic fluctuations in fMRI activity across striate (V1) and extrastriate visual cortex (V2 and V3). First, I asked whether cortico-cortical neuronal interactions between different cortical visual field maps could be characterized using RS-fMRI. I show that it is possible to map, based on RS-fMRI recordings, the cortico-cortical neuronal interactions between V1, V2 and V3. My results corroborate the view that intrinsic neuronal activity reflects underlying neuroanatomical organization. Next, I asked whether intrinsic fMRI activity was comparable to that evoked by visual field mapping (VFM) stimulation. By detecting patterns of spatially- localized synchronized activity in fMRI activity, I found spatial patterns of synchronized fMRI activity that were similar in RS and VFM. However, for the activity obtained during stimulation, synchronization was spatially more extensive, reflecting the stimulus driven interactions between neighboring locations in the visual cortex. The resemblance of the synchronization patterns derived from RS and VFM suggest that their underlying causes share common organizational principles. Building on these first two experiments, I also examined the feasibility of reproducing the findings of the previous two experimental chapters using data acquired with a 3T rather than a high-resolution 7T scanner. Despite the lower resolution and signal-to-noise ratio of the 3T data, I find that the results obtained with it are in agreement to those obtained previously with 7T data. I also made suggestions for how the quality of measurements at 3T could be further improved. Finally, I asked whether the propagation of fMRI activity within and between V1, V2 and V3 relates to structured neuronal activity. To explore this question, I implemented a descriptive model aimed at disentangling various contributions to measured fMRI activity. Applying this approach to 7T fMRI data revealed changes in cortical excitability and directed interactions in RS and VFM.

113 T

Moreover, my results pointed to a task-dependent reconfiguration of local, feedforward and feedback interactions within and across V1, V2 and V3. Here, my main contribution is that I show that it is possible to separate the contribution of local cortical activity and directed influences at the scale of cortical visual field maps. I conclude my thesis by interpreting and discussing my findings in the light of various theoretical perspectives. I stress that –in addition to externally triggered responses– there is a constant stream of internal processes that we should take into account as well. Drawing from Shannon’s information theory, I propose that, for a signal to be truly informative, it has to make a difference to a receiver –in this case the brain. In most contemporary neuroimaging studies, brain signals are interpreted from the perspective of the experimenter as receiver. How- ever, the key question is how those signals are actually received and used by the rest of the brain. Retinotopic maps are a prime example of this. In summary, my studies show that RS-fMRI recordings can reveal a remarkable amount of detail regarding the functional neuroanatomical organization of the human visual cortex. Furthermore, my findings justify and facilitate the comparison of RS and stimulus-evoked activity. While I have focused on RS and VFM activity, the methods presented in this thesis can also be applied to study task-dependent changes in a variety of experimental and behavioral conditions, both in health and disease. From this, I conclude that the brain is not only concerned with the demands imposed by the environment but also with internally generated dynamics.

114

Dutch summary

Het bestaan vanintrinsieke neuronale activiteit op vele schalen is aangetoond met behulp van verschillende methodologische benaderingen, maar de biolo- gische relevantie ervan begrijpen we nog niet goed. Een manier om intrinsieke neuronale activiteit bij mensen te bestuderen is door gebruik te maken van rusttoestand (resting-state; RS) functionele magnetische resonantie beeldvorm- ing (fMRI). In mijn proefschrift, getiteld ”The Neuroanatomical Organization of Intrinsic Brain Activity Measured by fMRI in the Human Visual Cortex”, beschrijf ik nieuwe RS-fMRI analyses die een opmerkelijke mate van detail onthullen over de functionele neuro-anatomische organisatie van de menselijke visuele cortex. Ik heb deze analyses toegepast om de lokale en gedistribueerde hersendynamiek binnen en tussen verschillende delen van de visuele cortex te onderzoeken. Ik laat daarbij zien hoe deze dynamiek als gevolg van verschillende manieren van stimulering verandert. Ik besluit mijn proefschrift door mijn bevindingen te interpreteren en te bespreken in het licht van verschillende theoretische perspectieven. Ik benadruk dat er, naast extern opgewekte reacties, ook een constante stroom van interne processen bestaat. Daarmee moet reken- ing worden gehouden omdat ze ook bijdragen aan de neuronale activiteit in de visuele cortex. Hoewel ik me heb gericht op het visuele systeem, kan de benadering die ik heb ontwikkeld worden toegepast op elk corticaal netwerk om veranderingen in hersendynamiek te bestuderen onder verschillende exper- imentele omstandigheden. Samenvattend, de bevindingen verkregen met behulp van mijn nieuwe fMRI-analyses leveren bewijs dat de intrinsieke ﬂuctu- aties in activiteit in de vroege corticale visuele cortex wijzen op functioneel rel- evante processen. Mijn bevindingen ondersteunen het gebruik van RS-fMRI voor het karakteriseren van de corticale functie en connectiviteit bij gezonde personen en mensen met een aandoening.

117

Acknowledgements

I would like to dedicate this thesis to the Chilean matorral, shrublands that cover the area between the western slopes of the Cordillera de los Andes and the coastal ranges of the southern Pacific Ocean. Without the influence they exert on my dreams, I may have missed big part of the joy and sense of mystery that observation of natural phenomena educe. A similar feeling I dedicate to the Cochayuyo1 (Durvillaea antarctica), a wig- gly seaweed covering the coastal rocks of the southern Pacific. Their graciously elongated tubular shapes makes one think about the rare abyssal creatures of an unrecorded underwater sci-fi epic. Incessantly clashing with the ocean, they seem to have been waiting eons for us to awash our deepest ruminations about underwater life. Furthermore, I would like to express my admiration and gratitude to all the wonderful people who made this journey possible. To Frans Cornelissen for his kind assistance and the amount of tolerance he demonstrated when directing me towards the correct path. I could not have imagined a better supervisor for my project. To Remco Renken, for always keeping a careful eye on my research as well as for his optimistic perseverance and full dedic- ation to creative scientific thought. To Nomdo Jansonius, for his grounded wisdom and generous advocacy for research. To Ben Harvey, for mentoring me throughout the project as well as for his personal and professional advice. To Serge Dumoulin and Branislava Curčić-Blake, for their worthy advice and enthusiasm. To Koen Haak, for opening the venues from which my research branched. To Gustavo Deco, for helping me understand the complex dynamics emerging from the brain. To Matthieu Gilson, for patiently guiding me through complicated mathematical concepts. To Diana Koopmans, for her pa- tience and support. To Barbara Nordhjem, Funda Yildirim, Joana Carvalho, Alessandro Grillini, Hinke Halbertsma, Sandra Hanekamp, Michelle Servaas, Azzurra Invernizzi, Tharcila Chaves, Ruud Kortekaas, Mendel Kaelen, Leor Roseman, Christopher Timmermann, Alessio Fracasso, Etienne Hugues, Rik- kert Hindriks, Thomas Pfeffer, Joao Barbosa, Simón Guendelman, Natalia Bielczyk, Vinod Kumar, Juan Carlos Letelier, Jorge Mpodozis, Gonzalo Marín, Daniel Margulies, Mauricio Toro, Tomás Ossandón and José Hurtado, for their openness to share their knowledge and enthusiasm for science. To the reading

1From Quechua: cocha = water, yuyo = weed.

119 A

committee, for their time to read this thesis and for their valuable remarks. To my fairy godmothers Mariette Borg, Marianne Jaspar, Clementine Heijden and Olga Meza-Lehman, for their boundless wisdom and generosity. My heartfelt thanks are extended to all those who stood beside me and supported me throughout this journey. To Tarek, Barbara, Elía, Joaquín, Nicolás, José, Paul, Jinsu, Felipe, Igor, Gijs, Adriaan, Tino, Roma, Jacek, Philippe and Sander. There are so many more people I would like to thank, please forgive me if I forgot someone. My deepest gratitude goes to my parents —Rai and Graciela— and family members —especially Elisa—, for their wisdom, strength and loving support.

Nicolás Gaspar Gravel Araneda Santiago, Chile March 2018

120