Accuracy of Computer Vision in estimating hip and knee flexion angle during squats and anterior lunges: An experimental methodological study

Abstract

Aim: Recently there have been an increase of companies using Computer Vison (CV) in managing treatment of patients with hip and knee osteoarthritis (OA). With this in mind, the objective of this study was to evaluate the accuracy of KineTech’s CV technology. The primary aim was to investigate errors of CV-based estimation of hip- and knee flexion angles (HKF) during series of squats and anterior lunges. The estimations were compared to measurements derived from marker-based motion capture system. We assessed the influence of camera position by evaluating CV-based estimations from the four different camera views.

Method: Nine students at the Norwegian School of Sport Sciences performed series of squats, short- and long step anterior lunges with external load (80% of their one repetition maximum). Series were recorded by marker-based motion capture system and three single cameras.

Results: The accuracy of CV-based estimation was assessed by examining the mean error (ME) and standard deviation (SD) of the HKF at the bottom position of each repetition. The CV-based method showed a tendency to both over- and underestimate HKF across different camera views. Overall, 54% of the bottom positions showed a acceptable ME within the referance error range (<12°), and only 15% of the total bottom positions revealed a good precision (SD <10°). The 45° back view showed consistently lower ME values at the hip flexion angle compared to other views, indicating better accuracy. For the knee flexion angle, we observed the lowest ME at the frontal view, particularly for the left knee flexion angle, though the frontal view also demonstrated the lowest ME for the right knee flexion angle in some cases.

Conclusion: Our study indicates that the accuracy CV is not yet sufficient to estimate HKF accurately when an external load is included in the form of a loaded barbell. For CV to serve as a reliable alternative for therapists during OA treatment, the accuracy of both hip and knee flexion angle estimations needs to be improved. Future development of CV technology should focus on overcoming challenges of occlusion to enhance accuracy.