[Development of semi-automated segmentation of the brain and CSF Region on MR images]

MR imaging (MRI) is an important method for the diagnosis of abnormalities of the brain. In the present report, a semi-automated method is presented to segment the brain and CSF region on brain MR images. MR images were obtained from 6 subjects by SE sequence and 8 subjects by GRE sequence. The semi-automated method consisted of the following three steps: (1) segmentation of the intracranial region using the region-growing technique, (2) segmentation of the brain region using histogram analysis and mathematical morphology, and (3) segmentation of the CSF region using the top-hat transformation technique. The average ratio of a correctly recognized region (CRR) between the semi-automated method and manual method was 87.9%, 85.1% for the intracranial region (IRR), and 94.8% and 86.8% for the brain region in the SE and GRE sequences.