Multimodal MRI analysis of brain subnetworks in autism using multi-view EM

Autism, Asperger´s syndrome, and Pervasive Developmental Disorder - Not Otherwise Specified (referred to collectively as autism spectrum disorder or ASD) are devastating neurodevelopmental disorders. The discovery of reliable image-derived biomarkers could potentially identify people with ASD, or infants who will subsequently develop or are already developing subtle signs of ASD. We hypothesize that the quantification of regional signal change and connectivity information from particular functional subnetworks in the brain responding to ASD-related biological motion tasks, informed by anatomical and diffusion information, will provide more sensitive and robust image-derived biomarkers for studying ASD. Thus, we focus our efforts on the development of a unique mathematical approach that uses image-derived information regarding gray matter location and white matter pathways to inform the estimation of three functionally connected subnetworks related to ASD. Our computational strategy is aimed at grouping ASD-task-related activation into these functional subnetworks using a new multi-view Expectation Maximization (EM) formulation.