A Comparison Of Functional Methods For Determining The Knee Flexion Axis

Abstract

The two most common approaches to finding the functional knee axis include Woltring's helical axis (HA) approach and Halvorsen's eigenvector (EV) approach. Both the HA approach and the EV approach use relative motion between the shank and thigh to establish the functional knee flexion/extension (F/E) axis. While clinical gait analyses confirm that the predominant axis of knee motion is F/E, they also show motion in the frontal (A/A) and transverse (IER) planes. The amount of A/A and IER motion varies between subjects, and the extent to which either of the algorithms tolerates this motion is unclear. PURPOSE To examine each of the functional knee axis approaches with respect to their ability to tolerate motion out of the F/E plane, and to determine their usefulness in a clinical environment. METHODS The HA and EV algorithms were tested in a clinical setting. To verify the precision and accuracy of each method, a tri-axial mechanical knee joint with a known F/E axis was fabricated and attached to a normal subject's lower extremity. The subject then performed several gait trials. The knee axis was calculated from the motion data of the artificial knee and compared to the known F/E axis. RESULTS Static Trials: Knee motion showed virtually no A/A or internal/external rotation (IER), and both algorithms estimated the knee's flexion/extension axis with <2.5° of error. Walking Trials: The minimum range of F/E required to produce estimates within a specific tolerance of the actual axis (5° for this analysis) was dependent on the amount of A/A and, for the helical approach, the amount of IER. When IER was less than 2° and F/E was approximately 60°, both approaches produced equivalent results for a given amount of A/A. When IER exceeded 2°, the eigenvector approach was more tolerant of A/A. A/A and IER had the most profound effect on estimating the knee joint axis when they appeared periodic as opposed to random in nature. Both methods stabilized at approximately three gait cycles. CONCLUSION When A/A exceeds 10°, neither method should be expected to provide estimates within 5° of the actual value. The helical approach should not be expected to produce reasonable estimates when IER exceeds 10°. It is recommended that use of either method presented here be limited to analysis of static data, unless constraints associated with A/A and IER motion can be verified in the gait data.