Two-dimensional rigid-body kinematics using image contour registration.

Abstract

A method for calculating two-dimensional rigid-body kinematic parameters using shape features is presented. Proposed applications include the noninvasive quantification of planar joint motion in vivo. By using digitized images (computed tomographs, radiographs, etc.) of a bone contour at two positions, the contour curvatures can be 'best-fit' to obtain a one-to-one mapping or registration of the bone images. This produces a dense field of displacement vectors from which planar rigid-body kinematic parameters can be estimated. Accuracy was studied using radiographic images of cadaveric femoral bone. The two motions of pure rotation with a fixed center of rotation and of pure translation were simulated. For pure rotation, error in rotation was independent of the rotation magnitude, with an average (n = 10) error of 0.3 +/- 0.8 degrees. The translation error averaged 0.9 +/- 0.5 mm. For pure translation, the error in rotation was -0.01 +/- 0.69 degrees and the error in translation was -0.62 +/- 0.98 mm (n = 10). This novel method has broad applications in the field of planar kinematics, especially in cases for which marker fixation is neither possible nor practical.