Automatic registration of follow-up brain MRI scans

Abstract

Registration of multiple medical images is required for analysis of both pathological and normal anatomical human body structures. In this study, we defined and automatically detected some anatomical landmarks which are considered to be unaffected from brain deformation such as symmetry axis of brain, nose tips, eye ball centers, and location of center of masses of certain slices. Registration is performed using the physical coordinates of automatically detected landmarks. Registration mapping is assumed to be an affine mapping and transformation matrix is formed using a least squares solution. Registered image is obtained using the estimated transformation matrix and trilinear interpolation. Algorithm performance is demonstrated on both MR image volumes deformed synthetically with a known transformation matrix for validation and baseline and follow-up MRI volumes with unknown transformations.