Clustering Analysis for fMRI Dataset based on ISODATA Algorithm

In the paper, the modified fuzzy c-means (MFc) is firstly used to treat the ill-balanced fMRI dataset to improve the efficiency, remove the redundance and reduce the population of analyzed voxels. Then the iteration self-organization data analysis techniques algorithm (ISODATA) method, as the development of data-driving methods, is utilized to find out the activated region in the brain. Therefore a multi-step strategy, including MFc and ISODATA, has been proposed to analyze a hybrid dataset and a real experimental fMRI dataset. On the whole, clustering analysis is calculated by multi-step strategy for local activity of fMRI dataset under auditory stimulation. Results show the multi-step strategy has its special characteristics in flexibility and efficiency compared with other data-driving dynamic method and SPM