fMRI (functional Magnetic Resonance Imaging) has long been the workhorseof cognitive neuroscience, to uncover the functional networks underlyinghuman brain function. Recently though,
EEG (Electroencephalography) and MEG (Magnetoencephalography) have gained in popularity since theirsuperior temporal resolution make it possible to track changes infunctional networks at
behaviourally relevant time-scales. However, theanalysis methods to model and analyse these time-varying functionalnetworks are only being developed.This talk will present a method we produced to analyse time-varyingnetworks in EEG task-related data. We first determined functional networksfrom short windows of post-stimulus EEG
activity, using the framework ofsparse Multi-Variate Auto-Regressive (s-MVAR) modelling. The functionalnetworks were then grouped into a few putative functional states,usingk-means clustering. We then modelled the sequence of 'states' as a Markovprocess.The talk will describe validation of the method using simulated EEG data.Then, we will present its application to EEG data from a behaviouralexperiment with two conditions, i.e. the oddball task. Using only the parameters of the Markov process of each experimental condition, we reportabove-chance classification accuracy in discriminating between trials ofthe two conditions. In discriminating between conditions from such a parsimonious representation, these results provide powerfulproof-of-concept for this approach to modeling the data.This paves the wayfor its use to test rich hypotheses on sub-second changes in functional networks as a task is being performed.

Dr. Nitin Williams completed his B.E. in Electronics& CommunicationEngineering (ECE) from Anna University, India before doing a PhD inComputational Neuroscience from University of Reading, UK. He then completed 4-years of post-doctoral research in the acclaimed MRC Cognition& Brain Sciences Unit, Cambridge, UK. He is currently funded by theflagship Human Brain Project (HBP), on a post-doctoral project to infer the brain-wide pattern of functional interactions in the human brain. Thisis being done in University of Helsinki, Finland. He is particularlyinterested in advancing understanding of human brain function by developing methods to harness the potential of MEG/EEG. He is alsointerested in building biophysical models to emulate MEG/EEG data, thereby facilitating a mechanistic understanding of cognitive function in bothhealth and disease.