Statistical modeling of human brain connectome using fMRI
Dept. of Medical Science/박사
Recent studies on human brain connectome have shown that a cortical area is not responsible for a single cognitive function but is repeatedly involved in various cognitive functions by forming functional subnetworks through various types of connectivity with other brain areas. Each subnetwork thereby includes overlaps in brain areas, and execution of various brain functions can be understood in terms of context-dependent recruitment and release of functional subnetworks. To implement the modeling of brain function, we newly propose a novel method for independent component analysis of brain graph that can identify subnetworks sharing brain regions and model brain function in units of subnetworks. The method overcomes disadvantage of existing methods that searches for subnetworks by assigning each brain region to a single functional subnetwork, hence are not appropriate for spatially overlapping functional subnetworks. We, firstly, showed the validity of the method through a simulation study reflecting overlapping subnetworks. Independent component analysis of resting-state brain graphs of 104 subjects led to discovery of 49 overlapping functional subnetworks, some of which were similar to functional subnetworks previously identified. Using the subnetworks, we introduced modeling of task performance in fMRI data of working memory, motor, and verb generation. Unlike previous methods, our technique also enables group-level comparison that was demonstrated by measuring the usage extent of each functional subnetwork between men and women.