Dynamic elasticity model for inter-subject non-rigid registration of 3D MRI brain scans

The purpose of this paper is to present a new approach for inter-subject non-rigid registration of 3D MR brain images which could assist in automatic labeling of brain structures. A physical dynamic elasticity model (DEM) is developed which tends to represent the complex non-linear deformations as elastic waves. The transformation is governed by elastodynamics wave equation. The registration process ensues in a hierarchical fashion, thus reducing the risk of obtaining a local optimal transformation. Along with the correction of local misalignments, it also removes global shape differences without any prior initialization. The proposed scheme was compared against high ranking registration methods including: DROP, SyN, ART and DRAMMS. The results were quantitatively analyzed by computing and testing the statistical significance of the volume overlap measures and Hausdorff distance for segmented structures with DROP, SyN, ART, DRAMMS and DEM registration methods. Experimental results demonstrate that the proposed DEM registration method leads to very promising results when applied to the problem of inter-subject registration and that favorably compares against classical registration approaches. Since DEM registration method is able to reduce registration errors significantly, hence it could be used to automatically label the anatomical structures for clinical studies.