Bipartite Organization of Cerebellar-Cortical Networks Estimated By Intrinsic Functional Connectivity

In order to identify brain networks with homogenous cerebellar-cortical functional connectivity, a novel orthogonal nonnegative matrix tri-factorization (ONMTF) based parcellation method was proposed. Specifically, the cerebellum and cerebrum were jointly parcellated into disjoint networks with homogenous cerebellar-cortical functional connectivity based on resting-state functional MRI (rs-fMRI) data. Firstly, similarity measures among voxels of the rs-fMRI data are calculated based on their cerebellar-cortical functional connectivity measures, yielding a matrix of similarity measures with rows and columns characterizing voxels of the cerebellum and cerebrum respectively. Then, the ONMTF algorithm is applied to the similarity matrix to extract the mutually connected cerebellar-cortical functional networks. The proposed method was validated through estimation of a bipartite organization of cerebellar-cortical networks, and the validation experiment results have demonstrated that the proposed method could identify cerebellar subsystems functionally connected with the default-mode and task-positive networks. Moreover, the proposed method had better performance than state-of-the-art brain parcellation methods in terms of functional homogeneity.