Individual-Level fMRI Segmentation Based on Graphs

Aiming at the high complexity of fMRI data and the great spatial dependence of existing methods, a whole-brain functional segmentation algorithm with low computational overhead and low spatial structure dependence is proposed for individual-level fMRI segmentation in 3-D space. First, the spatial information and functional connectivity of each voxel in fMRI are utilized for presegmentation to create compact and functionally consistent super voxels, then extracts features, such as average spatial coordinates and average time series at the super voxel level to reduce the computational effort of the segmentation algorithm, and performs segmentation in a cut-free manner to obtain the optimal segmentation graph by minimizing the energy function. The results of contrast experiment demonstrated that the algorithm fully exploits the connectivity information of fMRI for segmentation, relies less on the spatial structure, and achieves better functional segmentation results, which is an effective whole-brain functional segmentation method.