Restorative Neurology and Neuroscience 22 (2004) 245–260 245 IOS Press Non-invasive mapping of brain functions and brain recovery: Applying lessons from cognitive neuroscience to neurorehabilitation P.M. Matthewsa,∗, H. Johansen-Berga and H. Reddyb aCentre for Functional Magnetic Resonance Imaging of the Brain, Department of Clinical Neurology, University of Oxford, UK bMontreal Neurological Institute, McGill University, Canada Abstract. Modern cognitive neuroscience provides a powerful framework in which biological models of recovery and neurore- habilitation can be constructed and tested. The widespread availability, relatively low cost and informativeness of functional magnetic resonance imaging (fMRI) has made it the most popular of the techniques available to help with this task. Here, on the basis of functional imaging studies of stroke, diffuse microvascular disease and multiple sclerosis, we argue that processes of motor control and learning in the healthy brain share common mechanisms with those for adaptive functional reorganisation during spontaneous recovery after brain injury or with neurorehabilitation. Relatively stringent criteria can be met to confirm that adaptive functional reorganisation limits disability even in the adult brain: functional brain changes are related to disease burden, can be found in patients with demonstrable pathology but no clinical deficits and can be defined (in motor cortex) even in the absence of volitional recruitment. Initial studies of neurorehabilitation responses using fMRI and transcranial magnetic stimulation demonstrate that adaptive reorganisation can be manipulated directly with both pharmacological and behavioural interventions. The combination of strategies based on a strong biological rational with monitoring their effects using highly informative functional brain imaging methods heralds a new era of scientifically-founded neurorehabilitation. Keywords: FMRI, neurorehabilitation, learning, transcranial magnetic stimulation, motor control 1. Introduction: cognitive neuroscience and the entire central nervous system [CNS]) as a mecha- neurorehabilitation nism for adapting behaviour to changing internal state or external environment. Functional changes in the The advent of functional imaging techniques, and brain after injury or with rehabilitation thus are simply particularly functional magnetic resonance imaging specific examples of much more general changes in the (fMRI) [42], now allows the direct analysis of neurore- functional organisation of the central nervous system. Here we will focus entirely on attempts to understand habilitation in terms of functional systems in the brain. changes in the brain. Similar phenomena also must The development of modern cognitive neuroscience has occur in the brainstem and spinal cord. established a framework in which biological models We believe that it is useful to consider brain changes of recovery and neurorehabilitation can be constructed with spontaneous recovery after brain injury (or the in- and tested. At basis, these treat the brain (and, indeed, duced recovery of neurorehabilitation) in the context of a growing appreciation for the way in which the healthy ∗ brain changes with learning. Learning provides a gen- Corresponding author: Professor P.M. Matthews, John Rad- eral model for understanding the way in which brain cliffe Hospital, Headley Way, Headington, Oxford, UK OX3 9DU. Tel.: +44 1865 222729; Fax: +44 1865 222717; E-mail: structure and function can change with shifting goals [email protected]. or strategies or injury to a functional system. A broad 0922-6028/04/$17.00 2004 – IOS Press and the authors. All rights reserved 246 P.M. Matthews et al. / fMRI and neurorehabilitation grounding for a theory of learning also has been estab- 3. Adaptation involves recruitment of new systems lished across several levels of analysis (from the ge- that can activate the same final output pathway netic to that of functional systems) (see e.g., [86,114]). (in the case of movement, e.g., the same motor Observations made at a systems level with functional units) used prior to the injury. The changes re- imaging techniques thus can be linked to phenomena sponsible for this are molecular and cellular, but on a molecular or cellular level to develop hypotheses there is evidence (see below) that these are in- regarding potential new pharmacological (e.g., parox- fluenced by behaviours. Recruitment of unin- etine) [61,81] or electrophysiological (e.g., repetitive jured, parallel pathways for controlling the nor- TMS) therapies, for example. mal set of antagonist muscles in pointing with However, investigations already have begun to make a limb whose motor control has been impaired clear that there may be limitations to the reliability of by a stroke would be one example. However, inferences from studies of the normal brain. Additional adaptive changes involve several, distinct mecha- mechanisms may come into play with injury that are nisms [107], including the unmasking of existing not important in the healthy adult brain. For example, latent corticocortical connections [38], synaptic inflammatory mediators and other local responses may rearrangement, and axonal growth coupled with provide additional stimuli for neurogenesis or func- new synapse formation [17]. tional reorganisation after injury [12,105]. Rehabilitation may particularly support compen- satory strategies or stimulate adaptive responses. Viewed in this way, in order to understand rehabilitation 2. Fundamental issues regarding biological it is essential to define the neurobiological changes as- mechanisms of recovery and rehabilitation sociated with given behaviour, as well as the behaviour itself. Only this allows the distinction between repair, The application of fMRI and related techniques to compensation and adaptation to be defined clearly. A understanding brain recovery and neurorehabilitationis a second goal also can be achieved simultaneously: still in its early stages. The focus of this review will be development of a method for objective monitoring of on understanding motor recovery, because it has been changes related to outcome. This is an important goal in this area that most effort has been placed thus far. in its own right: neurorehabilitation studies in the past However, extensions of the strategy to other areas (e.g., have been confounded (or, at the least, had their ex- cognitive rehabilitation) already have begun [82,84]. planatory power severely limited) by a lack of sensitive The ways in which rehabilitation may reduce func- and objective markers of change [118]. tional impairments can be considered in the context of general molecular, cellular and behavioural mech- anisms for recovery after brain injury (whatever the 3. New tools for studying human cognitive cause): neuroscience and neurorehabilitation 1. Repair refers to molecular and cellular changes A variety of methods have been developed over the that can restore functions of the damaged sys- past few decades to allow patterns of brain activity asso- tem itself after an injury. For example, follow- ciated with specific behaviours to be defined. Two basic ing white matter ischaemia, oligodendroglial cell classes of such “functional brain mapping” techniques death leads to local demyelination and conduc- have evolved: i. those techniques that directly map tion block, which can give rise to functional im- electrical activity of the brain; and, ii. those that map lo- pairment. Repair can occur with generation of cal physiological or metabolic consequences of this al- new oligodendroglial cells from progenitors and tered brain electrical activity. Among the former are the remyelination, restoring normal conduction and non-invasive neural electromagnetic techniques of elec- glial trophic support for axons [35]. troencephalography (EEG) and magnetoencephalogra- 2. Compensation involves behavioural changes le- phy (MEG). Transcranial magnetic stimulation (TMS) ading to an altered strategy for completing a task. can be used to assess physiological characteristics of For example, there may be increased use of trun- the cortex, such as excitability, and to map its organ- cal movements to point accurately with a hemi- isation by assessing behavioural responses to cortical paretic limb [10,29]. stimulation at different locations [5]. These methods al- P.M. Matthews et al. / fMRI and neurorehabilitation 247 low exquisite temporal resolution of neural processing it is not the increased energy use itself that directly (typically on a 10 to 100 ms time scale), but suffer from triggers the increase in blood flow Instead, increased relatively poor spatial resolution (between one and sev- blood flow appears to be a direct consequence of neu- eral centimet res). Functional MRI (fMRI) methods are rotransmitter action [3]. The increased blood flow thus in the second category, along with positron emission reflects local neuronal signalling. Electrophysiologi- tomography (PET). They can be used to detect changes cally, increases in the BOLD signal are correlated most in regional blood perfusion, blood volume (e.g., using strongly with the local field potential (reflecting input to injected magnetic resonance contrast agents), or blood neurons and local processing) rather than the neuronal oxygenation that accompany neuronal activity. PET firing rate [59]. The volume over which blood flow demands injection of radioactive tracers and highly spe- increases associated with neuronal activity are found is cialised equipment, limiting the