Ventricle Registration for Inter-subject White Matter Lesion Analysis

A method is presented to non-rigidly register lateral ventricles to enable the automated analysis of peri-ventricular white matter lesions. A binary average image of the lateral ventricle system is used as a reference image for registration. To prevent false deformations of the lesions we non-rigidly register CSF segmentations to the average lateral ventricle image. The subvoxel accuracy achieved, allows accurate mapping of the peri-ventricular white matter lesions to the reference space of the average lateral ventricles. Application to patient data shows the feasibility of the presented approach.