A new landmark selection method for non-rigid registration of medical brain images

Abstract

Based on shape context and wavelet transform, a new approach to select automatically the landmarks for non-rigid registration of medical brain images is proposed in this paper. First, the two magnetic resonance (MR) images to be registered are segmented, their skull boundaries and cerebrospinal fluid boundaries are extracted, then shape context is applied to get landmarks on those boundaries and establish the corresponding relationship of landmarks between two images. Given the shape complexity of white matter, the shape context method is then improved with uniformly sampled points replaced by fewer modulus extreme points of wavelet transform on the white matter boundaries. This kind of landmarks can better reflect the shapes but decrease the landmarks number and then reduce the time for establishing their corresponding relationship. Finally, the selected landmarks are applied to make B-spline interpolation-based registration to validate the proposed method. Experimental results demonstrate the extracted landmarks can well reflect the geometry characters, have appropriate number and reasonable distribution in images; using the landmarks, the B-spline interpolation-based registration achieves ideal accuracy and faster speed.