Modified magnetic resonance image based parcellation method for cerebral cortex using successive fuzzy clustering and boundary detection.

Development of the accurate and reproducible parcellation of the human brain can be used to resolve the complex structure-functional relationships in the brain. We propose a modified parcellation method that provides the reliable and reproducible regions of interest using successive fuzzy c-means (sFCM) and boundary-detection algorithm. This method displays simultaneously both original brain image for identifying the sulcal landmarks and its tissue-classified image for referring to patterns of sulci. The whole cerebral region is extracted by the semiautomated region growing method and then classified to gray matter, white matter, and cerebrospinal fluid by sFCM. Referred to the other previous researches, the volume ratio of gray matter to white matter was shown to find that the efficiency of classification was improved (conventional FCM: 0.80 +/- 0.12 vs. sFCM: 1.57 +/- 0.18). Inter-rater reliability, estimated by the regression analysis, demonstrated that the proposed method was more reliable and reproducible than conventional methods [ANALYZE: correlation coefficient (CC)=0.341, Sig.=0.335 vs. proposed method: CC=0.816, Sig.=0.004]. The volume ratio of the whole cerebrum to the parceled object can be used to investigate structural abnormalities for the pathological detection of the various mental diseases such as schizophrenia, obsessive-compulsive disorder.