Coclustering Based Parcellation of Human Brain Cortex Using Diffusion Tensor MRI

Coclustering for Cross-Subject Fiber Tract Analysis Through Diffusion Tensor Imaging

“…Moreover, BCA is able to automatically identify outliers as well as the number of cortical clusters with high efficiency; 3) The application of BCA to human DTI datasets enables automated, reproducible, and cross-subject assessment of the connectivity patterns of major fiber tracts in human brains. This paper extends our previous method [9] with a more comprehensive study and improvement: first, while our previous work only proposed the coclustering algorithm for individual brain analysis, this paper extends it to cross-subject analysis of brain fiber tracts, resulting in a totally new section, Section V; second, the original coclustering algorithm requires the user to manually specify input parameters for the analysis of each individual brain; in this paper, we have proposed a strategy to automatically decide these parameters, thus automating the parameter-setting step. The transfer operator is also improved to enhance its performance and effectiveness by introducing additional transfer conditions; third, details about image acquisition, data preprocessing procedures, cross-subject experiments, and statistical evaluation for control groups, patient groups, and their intergroups are reported in order to further evaluate our coclustering algorithm and its application to the clinical research.…”