Graph-based Deep Learning for MRI-based Brain Network Analysis

Emerging research indicates that most brain diseases lead to alterations in brain networks, often preceding the onset of clinical symptoms. The advancement of neuroimaging technology enables the non-invasive acquisition of magnetic resonance imaging (MRI) data, encompassing both functional and structural MRI, facilitating the comprehensive mapping of the brain’s structural and functional networks. In this context, graph-based deep learning methods become important in uncovering changes within brain networks and enabling early disease prediction. Graph-based deep learning has emerged as a cornerstone in neuroscience research by effectively characterizing the topology of brain networks and extracting pertinent discriminative information. This chapter begins with a succinct overview of various graph-based deep learning models and their underlying mathematical principles. Subsequently, we delineate the latest advancements in brain network analysis facilitated by graph-based deep learning models, spanning diverse applications such as functional and structural brain network analysis, as well as multi-modal brain network analysis.