Automated segmentation of the lateral ventricle in MR images of human brain

Segmentation of cerebral ventricle in 3D magnetic resonance images (MRI) of human brain is a crucial task for neuroimaging researches, because abnormal changes in size, shape and volume of the lateral ventricle are closely related to the progression of many neurodegenerative diseases. However, the major obstacles for achieving the goal of accurate segmentation of cerebral ventricle in brain MRI are the presence of imaging noise, magnetic field inhomogeneities, and anatomical variation among individuals. In this paper, a novel method for automated segmentation of cerebral ventricle in 3D MRI of human brain is presented. This method combined the Bayesian framework with the state-of-the-art super-pixel technique to accurately segment the lateral ventricle in brain MRI. Quantitative comparison has been made between the segmentation results of the proposed method and expert´s manual delineation. The promising results suggested this method can be a viable choice for the clinical studies involving ventricle morphometry.