Quantifying performance and joint kinematics in functional tasks crucial for anterior cruciate ligament rehabilitation using smartphone video and pose detection.

Abstract

The assessment of performance during functional tasks and the quality of movement execution are crucial metrics in the rehabilitation of patients with anterior cruciate ligament (ACL) injuries. While measuring performance is feasible in clinical practice, quantifying joint kinematics poses greater challenges. The aim of this study was to investigate whether smartphone video, using deep neural networks for human pose detection, can enable the clinicians not only to measure performance in functional tasks but also to assess joint kinematics. Twelve healthy participants performed the forward reach of the Star Excursion Balance Test 10 times, along with 10 repetitions of forward jumps and vertical jumps, with simultaneous motion capture via a marker-based reference system and a smartphone. OpenPifPaf was utilized for markerless detection of anatomical landmarks in video recordings. The OpenPifPaf coordinates were scaled using anthropometric data of the thigh, and task performance and joint kinematics were computed for both the marker-based and markerless systems. Comparing results for marker-based and markerless systems revealed similar joint angles, with mean root mean square errors of 2.8° for the knee, 3.1° for the hip, and 3.9° for the ankle. Excellent agreement was observed for clinically pertinent parameters, i.e., the performance, the peak knee flexion, and the knee range of motion (intraclass correlation coefficient > 0.97). The results underscore the feasibility of using markerless methods based on OpenPifPaf for assessing performance and joint kinematics in functional tasks crucial for ACL patients' rehabilitation. The simplicity of this approach makes it suitable for integration into clinical practice.