The reliability of a pose-estimation-based WeChat mini-program in monitoring the motor function of patients after brachial plexus injury

Patients with brachial plexus injuries often require long-term follow-up, and the frequent hospital visits can be a burden. Remote monitoring of these patients could be facilitated using Human Pose Estimation algorithms. We evaluated the performance of seven recently released algorithms and developed a WeChat mini-program using the best performing model. The validity of this program was then assessed in brachial plexus injured patients by comparing its results with those of manual goniometry. Our results demonstrated acceptable accuracy of this model, with mean differences of 4° and 7° when compared with manual goniometry in shoulder abduction and elbow flexion, respectively. Remote monitoring using a WeChat mini-program is a reliable and convenient tool for the follow-up of brachial plexus injury patients.<b>Level of evidence:</b> II.