12:00 AM - 10:30 AM on Monday, November 26, 2012
Location: Virginia Tech Research Center -Arlington, Room 4-024
Invited Speaker: Dr. Rosalyn Moran of Wellcome Trust Centre for Neuroimaging
Over the past number of years, biophysical models of the brain have allowed experimental neuroscientists to pose mechanistic hypotheses regarding the neural architecture generating their empirical data. One such framework, known as Dynamic Causal Modelling (DCM) has been applied in the domains of fMRI, M/EEG and animal LFP studies, to investigate brain connectivity. This general approach has been developed to probe even finer grained physiological detail, down to the level of specific neurotransmitter activity at the synapse.
In this talk I will describe the mathematical and biological framework of DCM which allows one to infer underlying biophysical parameters generating data. In particular I will describe Dynamic Casual Models for Cross Spectral Densities which uses a variational Bayesian scheme to invert models of neuronal function using complex spectral data features from electrophysiological recordings. By incorporating the complex domain into the generative model and inversion routine, the hope is to refine the estimation of key elements of brain network organization (like synaptic time constants, connection strengths and conduction delays) and enable multi-species (rodent to human primate) application. Perhaps more importantly, the extension to complex valued data hopes to bridge the technical divide between differing approaches to multi-channel electrophysiological time series analysis, linking model based and model free analyses. The biophysical models are based on a mean field treatment of ion-channel dynamics, providing for the interaction of ensembles’ sufficient statistics.
Assaying the chemical substrates and connectivity profiles of the active brain could lead to breakthroughs in neurological and psychiatric therapies as well as better, patient-specific, pharmacological treatment. The method, outlined above provides such a platform. I will demonstrate a proof-of-principle using human magnetoencephalography data and DCM to link subject specific behaviour to dopamine dependent synaptic activity in a working memory task.