Abstract
The cerebral disorders often accompany the deformation of the brain shape. Many physicians use magnetic resonance (MR) images to diagnose cerebral disorders. Extraction of cerebral surface from MR images is a basic work to evaluate human brain MR images. Many methods for cerebral surface extraction have been proposed. However, it is very difficult especially for neonatal MR images to extract cerebral surface because of brain size and folded sulci. Therefore, The extraction method for neonatal MR images is proposed a little. Especially, The particle based method can expect to extract the complicated sulci with high accuracy. However, It is difficult to estimate the change probability of particle classes which are cerebrospinal fluid, gray matter, and white matter. This paper proposes a new method to estimate the change probability of particle classes using fuzzy inference technique. The feature values of fuzzy inference are ratio of particle in a voxel, particle density, and gray matter thickness. The proposed method was applied to neonatal MR images. The experimental results were evaluated by using ground truth data given by physicians.
