Fuzzy medical image processing for segmenting the lateral ventricles from MR images

This paper shows an automated method for segmenting the lateral ventricles from human brain MR images. It enables automatic volumetry and construction of 3D rendering images. First, we segment the whole brain and the cerebrospinal fluid (CSF) using 3D MR image processing software developed by Y. Hata et al. (see IEEE Trans. Syst., Man, Cybern. C, vol.30, no.3, p.381-95, 2000). Our method for segmenting the lateral ventricles is based on fuzzy inference techniques and is able to represent expert knowledge and to introduce the knowledge to image processing. The employed knowledge on the lateral ventricles consists of the location, the intensity and the shape. The proposed method was applied to MR volumes of 20 normal volunteers, 20 Alzheimer disease and 20 hydrocephalus patients. The experimental results validated that this method was able to segment the lateral ventricles with high accuracy.