3-DIMENSIONAL ANATOMICAL MODEL-BASED SEGMENTATION OF MR BRAIN IMAGES THROUGH PRINCIPAL AXES REGISTRATION

Model-based segmentation and analysis of brain images depends on anato mical knowledge which may be derived from conventional atlases. Classi cal anatomical atlases are based on the rigid spatial distribution pro vided by a single cadaver. Their use to segment internal anatomical br ain structures in a high-resolution MR brain image does not provide an y knowledge about the subject variability, and therefore they are not very efficient in analysis, We present a method to develop three-dimen sional computerized composite models of brain structures to build a co mputerized anatomical atlas, The composite models are developed using the real MR brain images of human subjects which are registered throug h the Principal Axes Transformation, The composite models provide prob abilistic spatial distributions, which represent the variability of br ain structures and can be easily updated for additional subjects, We d emonstrate the use of such a composite model of ventricular structure to help segmentation of the ventricles and Cerebrospinal Fluid (CSF) o f MR brain images, In this paper, a composite model of ventricles usin g a set of 22 human subjects is developed and used in a model-based se gmentation of ventricles, sulci, and white matter lesions. To illustra te the clinical usefulness, automatic volumetric measurements on ventr icular size and cortical atrophy for an additional eight alcoholics an d 10 normal subjects were made. The volumetric quantitative results in dicated regional brain atrophy in chronic alcoholics.