Evaluation of s-EMG Sensor Locations for Upper-limb Compensatory Movement Detection

Musculoskeletal rehabilitation is crucial for restoring upper limb function after elbow trauma or stroke. In unsupervised rehabilitation, patients may develop compensatory movements that hinder recovery. While wearable devices for home rehabilitation are a promising supplement to clinical therapy, they may overlook movement quality. Detecting compensatory motions in wearable systems is challenging, as placing sensors on all involved muscles reduces wearability, increases computational and power demands, and complicates sensor management. The objective of this study was to identify optimal locations for surface electromyography sensors for detecting compensatory movements. Data were collected from 40 healthy individuals performing various uniplanar and multiplanar tasks under conditions simulating both healthy and impaired movements. Sensor combinations that showed significant differences between healthy and compensatory patterns were identified through statistical analysis and feature importance techniques. The classification performance of these sensor combinations was then evaluated. Results indicate that 11 sensors placed on the upper trapezius, deltoids, biceps, triceps, latissimus dorsi, erector spinae, rectus abdominis, and external oblique muscles were key for accurate detection (accuracy = 81.43%, F1 score = 0.8549). Additionally, the number of sensors can be reduced to seven without compromising accuracy and F1 score, though performance for some tasks may drop. These findings can improve the design of wearable devices to detect and reduce compensatory movements in patients recovering from upper limb injuries.