Automated segmentation of the cerebrospinal fluid and the lateral ventricles from human brain MR images

Presents a computer-aided diagnosis (CAD) system which be able to segment the whole brain, brain portions, the cerebrospinal fluid (CSF) and the lateral ventricles from human brain MRI, and then to give the volumes and the 3D volume renderings. In the system, the whole brain and the brain portions are automatically segmented by using previously-developed software (Y. Hata et al., 2000). In this paper, we introduce a new concept of a representative line to segment the CSF and the lateral ventricles. The representative line is automatically detected by inferring its direction by evaluating the location in the head, its position in the CSF and the shape of the line. The inference is performed with a fuzzy inferencing technique. The developed system was applied to MRI volumes of 20 normal subjects, 20 Alzheimer disease patients and 20 hydrocephalus patients. The segmentation error ratio was 1.98% in comparison with the volumes of manually delineated regions. This allowed us to experimentally characterize the modes of variation that are indicative of disease processes