Intrinsic Connectivity Networks in Fmri Are 1 Continuously
Total Page:16
File Type:pdf, Size:1020Kb
bioRxiv preprint doi: https://doi.org/10.1101/2021.06.23.449485; this version posted June 24, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 1 Moving Beyond the ‘CAP’ of the Iceberg: Intrinsic Connectivity Networks in fMRI are 2 Continuously Engaging and Overlapping 3 A. Iraji1, *, A. Faghiri1, Z. Fu1, P. Kochunov2, B. M. Adhikari2, A. Belger3, J.M. Ford4,5, S. McEwen6, 4 D.H. Mathalon4,5, G.D. Pearlson7, S.G. Potkin8, A. Preda8, J.A. Turner9, T.G.M. van Erp10, C. Chang11, 5 and V.D. Calhoun1,* 6 1Tri-Institutional Center for Translational Research in Neuroimaging and Data Science (TReNDS), 7 Georgia State University, Georgia Institute of Technology, and Emory University, Atlanta, GA, USA 8 2Maryland Psychiatric Research Center, Department of Psychiatry, School of Medicine, University of 9 Maryland, Baltimore, MD, USA 10 3Department of Psychiatry, University of North Carolina, Chapel Hill, NC, USA 11 4Department of Psychiatry, University of California San Francisco, San Francisco, CA, USA 12 5San Francisco VA Medical Center, San Francisco, CA, USA 13 6Department of Psychiatry and Biobehavioral Sciences, University of California Los Angeles, Los 14 Angeles, CA, USA 15 7Departments of Psychiatry and Neuroscience, Yale University, School of Medicine, New Haven, CT, 16 USA 17 8Department of Psychiatry and Human Behavior, University of California Irvine, Irvine, CA, USA 18 9Department of Psychology, Georgia State University, Atlanta, GA, USA 19 10Clinical Translational Neuroscience Laboratory, Department of Psychiatry and Human Behavior, 20 University of California Irvine, Irvine, CA, USA 21 11Department of Electrical Engineering and Computer Science, Vanderbilt University, Nashville, TN, 22 USA 23 1 bioRxiv preprint doi: https://doi.org/10.1101/2021.06.23.449485; this version posted June 24, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 1 * Correspondence address to: 2 [email protected] (Armin Iraji) 3 [email protected] (Vince Calhoun) 4 Abstract 5 Resting-state functional magnetic resonance imaging is currently the mainstay of functional 6 neuroimaging and has allowed researchers to identify intrinsic connectivity networks (aka functional 7 networks) at different spatial scales. However, little is known about the temporal profiles of these 8 networks and whether it is best to model them as continuous phenomena in both space and time or, 9 rather, as a set of temporally discrete events. Both categories have been supported by series of studies 10 with promising findings. However, a critical question is whether focusing only on time points presumed 11 to contain isolated neural events and disregarding the rest of the data is missing important information, 12 or worse, producing misleading conclusions. In this work, we argue that brain networks identified within 13 the spontaneous blood oxygenation level-dependent (BOLD) signal are not limited to temporally sparse 14 burst moments and that these event present time points (EPTs) contain valuable but incomplete 15 information about the underlying functional patterns. 16 We show that brain networks are continuously present in BOLD signal, including during their event 17 absent time points (EATs), i.e., time points that exhibit minimum activity and are the least likely to 18 contain an event. Moreover, our findings suggest that EPTs do not contain all the available information 19 about their corresponding networks. We observe distinct default mode connectivity patterns obtained 20 from all time points (AllTPs), EPTs, and EATs. We show evidence of robust relationships with 21 schizophrenia symptoms that are both common and unique to each of the sets of time points (AllTPs, 22 EPTs, EATs), likely related to transient patterns of connectivity. Together, these findings indicate the 2 bioRxiv preprint doi: https://doi.org/10.1101/2021.06.23.449485; this version posted June 24, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 1 importance of leveraging the full temporal data in functional studies, including those using event- 2 detection approaches. 3 Keywords 4 functional connectivity (FC), event present time points (EPTs), event absent time points (EATs), 5 activation spatial map (ASM), independent component analysis (ICA), intrinsic connectivity networks 6 (ICNs), co-activation patterns (CAPs) 7 Introduction 8 Brain function results from the interaction among local and distributed brain areas. Thus, the temporal 9 dependency among brain regions, commonly known as functional connectivity, is a widely-used tool for 10 studying brain function (Friston, 2011). The advent of resting-state functional magnetic resonance 11 imaging (rsfMRI) revealed the presence of strong temporal dependencies between functionally related 12 brain regions (e.g., bilateral motor cortices) even without external stimulation (Biswal et al., 1995). This 13 led to a rapid growth in rsfMRI research and identification of spatial patterns of functionally connected 14 regions termed intrinsic connectivity networks (ICNs), or functional networks, at different spatial scales 15 from large-scale distributed networks to fine-grained, spatially local ones (Damoiseaux et al., 2006; 16 Allen et al., 2011; Iraji et al., 2019b). Interestingly, spatial patterns obtained by applying independent 17 component analysis (ICA) to spatial maps of thousands of different activation conditions derived from 18 the BrainMap meta-analytic tool (https://www.brainmap.org/) closely resemble large-scale networks 19 (Smith et al., 2009), as do data-driven analyses of activation maps from task data (Calhoun and Allen, 20 2013), further supporting the functional relevance of functional networks. These approaches are 21 implicitly built upon the notion of continuous information processing and interaction among brain 22 regions. 3 bioRxiv preprint doi: https://doi.org/10.1101/2021.06.23.449485; this version posted June 24, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 1 In parallel to this work, another family of approaches, that we collectively called event detection 2 approaches, was developed originally based upon a hypothesis that functional patterns and brain 3 networks emerge from discrete, neural events (Chialvo, 2010; Tagliazucchi et al., 2012). Instead of 4 using all available data, event detection approaches first select a set of extreme time points (e.g., top 5 10%) as event present time points (EPTs) and use only these EPTs to obtain corresponding brain 6 functional patterns (Tagliazucchi et al., 2012; Liu et al., 2013). These approaches assume that these time 7 points alone are sufficient to estimate the temporal dependency and to derive the functional connectivity 8 patterns observed in functional connectivity studies (Tagliazucchi et al., 2011; Tagliazucchi et al., 2016; 9 Cifre et al., 2020). As such, event-based studies shift away from typical functional connectivity and 10 mainly focus on capturing functional patterns as (co-)activation patterns of sparse EPTs using first-order 11 statistics and signal amplitude rather than statistical dependence between them (Tagliazucchi et al., 12 2011; Tagliazucchi et al., 2012; Petridou et al., 2013). Event detection approaches have shown direct 13 correspondence between such spatial co-activation patterns and large-scale brain networks obtained 14 from functional connectivity analysis (Tagliazucchi et al., 2012; Liu et al., 2013). They also 15 demonstrated that different EPTs of a given node represent different co-activation patterns reflecting 16 dynamic information of rsfMRI data (Liu and Duyn, 2013; Karahanoğlu and Van De Ville, 2015). 17 However, these findings were limited to large-scale distributed brain networks with no established 18 observations of fine-grained networks (ICNs) obtained from high-order ICA. This, in turn, may indicate 19 that event detection approaches are limited to dominant large covarying functional patterns. 20 After years of using first-order statistics in event-based studies, which followed early works utilizing the 21 temporal dependency between extreme time points to derive the functional connectivity (Tagliazucchi et 22 al., 2011; Tagliazucchi et al., 2012), recent works revisited the idea of using extreme time points to 23 study (dynamic) functional connectivity in the context of second and higher-order statistics and 4 bioRxiv preprint doi: https://doi.org/10.1101/2021.06.23.449485; this version posted June 24, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license. 1 demonstrated the potential to provide additional insights into cognition and behavior (Tagliazucchi et 2 al.,