Mental Disorder Diagnosis using fMRI Images by Deep State-Space Model

Abstract

Machine learning-based accurate diagnosis of mental disorders is expected to support finding their biomarkers and understanding their underlying mechanism. Recent studies employed dynamic and generative models due to the time-varying nature of the brain activities. Though it is difficult to extract complex features due to the simpleness of the model. In this paper, we model fMRI data using dynamic and deep generative model. The proposed deep state-space model is flexible, dynamic and generative(interpretable). Hence it can extract complex feature, capture time-varying nature of the brain activities and identifies brain regions related to the disorders.