Particle Method Based Sub-voxel Cerebral Contour Extraction in Neonatal MR images

摘要

Precise measurement of cerebral surface area will explores a new method for evaluating cerebral disorders because the cerebrum will be deformed by the cerebral disorders. Conventional methods for segmenting the cerebral region from MR images give coarse area by counting the number of the surface voxels of the segmented region. In contrast, this paper proposes a novel method for extracting cerebral contour with sub-voxel accuracy. The new method is based on particle method, and introduces three kinds of particles that are correspond to eerebrospinal fluid, gray matter and white matter. The proposed method was applied to neonatal MR images, and the experimental results validated that the cerebral contour was extracted with a root-mean-squared error of 0.45 mm in comparison with the ground truth contour given by a physician.