A 2-D Automatic Segmentation Scheme for Brainstem and Cerebellum Regions in Brain MR Imaging

This paper describes a 2-D automatic segmentation scheme for brainstem and cerebellum regions in brain MR images by using Ada-boosted and Active Contour Model technologies. The proposed scheme includes two processing stages: the detection process for the candidate regions and the segmentation process for the final contours. In the detection process, Ada-boosted technology has been used for three times to find the large region containing the brainstem and cerebellum first and to find the individual cerebellum and brainstem candidate regions later. The partial edge information of candidate regions are used as boundary limits in the next stage. In the segmentation process, a modified Scale-based Fuzzy Connectedness algorithm (SFC), Chain code, and Active Contour Model (ACM) are used for the final boundaries. This automatic scheme provides consistent segmentation results to avoid manual errors such as the seed locations in Fuzzy Connectedness methods.