Skip to Main Content
Being able to measure deformations precisely in tissue by magnetic resonance (MR) imaging is very useful for many medical imaging applications. While a variety of different algorithms have been formulated for this purpose, few are based on tracking features in the images. Here, we propose an approach of automatically extracting feature points and matching them to measure local deformation as seen in 3D MR volumetric images. Ccorrelation scores (cs) are given to pairs of high curvature points in a 3D cubic region to ensure that they are well matched. Those with scores above a given threshold are considered as candidate points. The strength of matching of the candidate points are evaluated using an iterative energy function, and then the well matched points are used to estimate the deformation.The approach was very effective when applied to actual MR volumetric images of a person's calf.