Semi-automated algorithm for rotational axis determination of the ulno-humeral joint.

Preoperative simulation of the elbow joint motion has been proposed for prosthesis alignment, ligament reconstruction or treatment of impingement-inducing osteophytes. However, its daily application remains seldom. This study proposes an algorithm to simulate the ulno-humeral joint motion in flexion/extension. Four observers placed reference points on 3D surface models of elbows. The algorithm generated five spheres representing specific joint surface: medial and lateral trochlea humeri (MT and LT), capitellum (CAP), medial and lateral trochlear notch (MED- and LAT NOTCH). Three rotational axes were defined: MT-LT, CAP-MT and MED-LAT NOTCH. A fourth axis, COMB, was computed using the average 3D distance between MT-MED NOTCH and LT-LAT NOTCH. Interobserver average distance between the reference points and the computed sphere as well as the average interobserver 3D angle between the axis were analysed. The dynamic articular congruence of the axes in relation to the MED-LAT NOTCH axis was assessed by calculating their respective 3D angle variation from 0° extension to 150° flexion. The number of patients needed to reach stable dynamic articular congruence was assessed. The computed spheres exhibit lower interobserver average translation compared to the reference points. The CAP-MT axis shows the lowest interobserver variation of 3D angle (4.8°). However, COMB axis has the lowest dynamic articular incongruency (3D angle variation of 7.4°, p < 0.001). Once a learning curve of six patients is reached, an average congruence of 4.8° can be achieved. An algorithm based on multiple articular references can reduce observer-induced inaccuracies in simulation of elbow joint motion.