Imaging dynamical brain networks using hybrid dynamical models

Dynamic reorganisation of a functional brain network may be related to shifts in brain state, such as those associated with adaptation and learning. From a clinical perspective, changes in the relative contribution of brain areas (plasticity) or their interactions (effective connectivity) underpin the early stages of a number of developmental (e.g Dyslexia) and psychiatric (e.g. depression, psychosis), but most prominently neurodegenerative (e.g. Alzheimer's or Parkinson's) conditions. Due to plasticity, changes in effective connectivity may precede any changes in brain structure or in observed behaviour. Early detection and monitoring of such changes by means of non-invasive imaging techniques like EEG (or MEG) is vital if early intervention and early rehabilitation strategies were to be successful . Most current analyses of effective connectivity (EC) assume that the architecture and connection strengths of the functional network are static in time, or differ between task conditions at known time points. Thus, current approaches cannot be used to detect dynamic changes in network organisation at arbitrary time points or in the absence of a cognitive task, such as during on-going EEG monitoring and diagnosis of the above neurological conditions. I propose to develop a novel model and a toolbox for estimating the time-dependent effective connectivity of the dynamical brain network underlying the on-going EEG, and to test its accuracy and limits in detecting changes in effective brain connectivity induced by Transcranial Magnetic Stimulation (TMS). Prior anatomical connectivity information and probabilistic atlases, derived from Diffusion Weighted Magnetic Resonance Imaging (DWMRI), will be used to inform the model about the likely architecture of the functional network and about the likely anatomical location of sources, respectively. The goal is to provide a low-cost and fast application to detect, track and predict early changes in brain causal networks and their dynamics from the on-going EEG. Given the current emphasis on reducing the social and economic impact of neurodegeneration as highlighted by the Prime Minister's dementia challenge, the specific focus here is on using TMS to induce small changes in two exemplars of distributed networks that simulate semantic and motor neurodegeneration as a demonstration of what this method can detect. The proposed model has the potential to change strategies for screening and early diagnosis of neurodegenerative conditions. It opens the possibility for developing new clinical applications that impact on the study of the aging brain and mental health, as well as the analysis of a wide variety of brain activity including normal during decision-making, resting-state and social behaviour.

This fellowship develops a new mathematical method for the analysis of brain function that will allow neuroscientists to study the dynamics of brain state transitions during spontaneous (not evoked by experimental tasks) brain activity, which are inaccessible using conventional techniques. The project transfers stochastic dynamical systems methods and models that have been successful in the study of non-biological complex systems, to the study of the living brain. Under very general assumptions, such dynamical models can respond in a way that resembles brain behaviour in a wide variety of situations, which will provide valuable insights into the brain mechanisms underlying normal as well as abnormal behaviour. The most immediate impact is expected through uptake of the developed method by researchers in clinical and cognitive Neuroscience and engineers/neuroimagers/modellers working on the development of clinical applications. The fellow will advance this uptake through direct collaboration with leading Neuroscientists (Lambon-Ralph) and Neuroimagers/modellers (Parker, Cootes, El-Deredy) at the host institution (University of Manchester) and will explore the potential for collaboration with other groups (attendance to conferences and workshops targeting different impact groups are planned).

Two main directions in which the proposed research will find its way into applications are currently on the horizon.

(1) Clinical and cognitive Neuroscience investigates underlying mechanisms and new treatments for developmental (e.g Dyslexia) and psychiatric (e.g. depression, psychosis), but most prominently neurodegenerative (e.g. Alzheimer's or Parkinson's) conditions. In most of these conditions, the interesting data are recorded during spontaneous (non-task related) brain activity. Therefore, the development of a tool for the non-invasive detection and monitoring of brain state transitions, that allow differentiating normal from abnormal brain function in the absence of an experimental task is vital if early intervention and early rehabilitation strategies were to be successful. Furthermore decoupling experimental design from the analysis method extends the range of feasible laboratory testing conditions.

(2) By using spontaneous EEG (and potentially other non-invasive Neuroimaging techniques with a time component) to infer brain network dynamics, the proposed method provides a new way to make feedback control non-invasive (that is, applying control without changing the natural behaviour). The proposed method then provides a new avenue for interaction with the neural system without disturbing its natural state. In doing so, it also opens a range of possibilities for the development of biomedical applications based on non-invasive control, such as Neuro-feedback and Brain Computer Interfaces. Furthermore, with the recent development of low-cost portable EEG devices, the proposed method can also allow for the provision of personalised therapeutic treatment.