Robust registration of coordinate data sets for noninvasive measurement of bone kinematics

Skeletal joint movement is studied by tracking corresponding points in a sequence of MRI or CT images. This paper describes a non-invasive registration method for tracking large movement of the forearm bones in a three-dimensional coordinate system. A triangulated surface is fitted to the series of MRI cross-sections to represent the three-dimensional volumetric data. The proposed refined iterative closest point (RICP) algorithm is a combination of shape measurement parameters with direct search in the transformation space of a series of local minima by minimizing registration errors. Real data acquired from a series of MRI images of the forearm are used to test the algorithm for estimating movement without attaching markers on the bones