Functional Connectivity-Based Parcellation
SpeakerSang-Yun Oh, Ph.D.
The human brain is understood to be composed of distinct regions with specialized functions. With increasing interest in functional networks of the brain, reliable ways to identify brain regions motivated by functional connectivities are needed. In this work, a scalable statistical method for estimating partial correlations is used to determine voxel-wise functional connectivity networks. Previously limited by the extreme dimensionality of the fMRI data, a recent development in undirected graphical model selection can speed up the network estimation significantly. Subsequently, the recovered network structure is used to identify functionally specialized regions. The proposed parcellation results will be analyzed and compared to other brain maps to illustrate the comparative advantages and drawbacks of the proposed approach.