Dynamic Bayesian networks for integrated neural computation

Understanding the clinical outcomes of brain lesions necessitates knowing how networks of cerebral structures implement cognitive or sensorimotor functions. Functional neuroimaging techniques provide useful insights on what the networks are, and when and how much they activate. However, an interpretative method, explaining how the activation of large-scale networks derives from the cerebral information processing mechanisms involved in the function, is still missing. Our goal is to provide such a tool. We suggest that integrated neural computation can be best represented with dynamic Bayesian networks. Our modeling approach is based on the anatomical connectivity of cerebral regions, the information processing within cerebral areas and the causal influences that connected regions exert on each other. We use experimental results (Fox and Raichle, 1985) concerning the modulation of the striate cortex's activation by the presentation rate of visual stimuli, to show that our explicit modeling approach allows the interpretation of neuroimaging data, through the formulation and the simulation of functional and physiological assumptions.