CONNECTOMICS IN PSYCHIATRIC CLASSIFICATION

Washington University, Saint Louis MO

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): Traditional conceptualization of mental disorders based on phenomenology is increasingly recognized as limited, but to date, we have lacked a clear path forward toward a more valid approach. Clinical heterogeneity and the imprecise nature of nosological distinctions represent fundamentally confounding factors limiting a better understanding of etiology, prevention and treatment. Neural connectivity of the major psychiatric disorders such as schizophrenia (SZ) and bipolar disorder (BP) has been variable across studies, which inarguably reflect multiple disease processes with distinct etiologies and overlapping clinical manifestations. Connectomics is an umbrella term that refers to scientific attempts to accurately map the set of neural elements and connections comprising the brain collectively referred to as the human connectome. Our application promises to uncover latent, homogenous, connectivity phenotypes using neuroimaging tools, which are free from the limitations of traditional diagnostic boundaries, and which correlate with clinical manifestations. SZ, BP and healthy control subjects will be scanned using the state-of-the-art Connectome Skyra, an optimized MRI scanner used by the NIH Human Connectome Project at Washington University, to obtain exceptionally high-resolution brain diffusion and functional connectivity images. We aim to identify brain signatures and network patterns that relate to psychosis, affectivity and cognitive deficits across all groups using diffusion MRI and resting-state functional connectivity MRI. Our classification methods will employ computational tools that include graph theory and support vector machine based pattern classification to derive multiple segregate clusters of individuals with unique patterns of behavioral/cognitive profiles and brain connectivity. We will also use the novel unsupervised method of Non-Negative Matrix Factorization-Based Biclustering, which we developed for use on neuroimaging datasets to identify subgroups based on patterns of whole brain connectivity following voxelwise deconstruction of the entire brain's white matter tracts. Our application benefits from its multi-disciplinary collaborators and consultants, including several key investigators from the Human Connectome Project.