sEMG dataset of simulated pain-related behaviours for robotic orthopaedic shoulder rehabilitation

This dataset contains surface electromyography (sEMG) recordings from 25 healthy participants aged 21–41 years performing shoulder abduction, shoulder flexion, and scapular retraction (“rower”) with a robotic manipulator. Upon oral cues, participants simulated three pain-related behaviours observed during shoulder rehabilitation and selected in consultation with an expert physiotherapist. Signals were recorded from the anterior and rear deltoids and the upper and lower trapezius using a Trigno Centro system and four wireless, single-channel Avanti sensors (Delsys Europe, Greater Manchester, UK). To the best of the authors’ knowledge, frequently used pain-detection datasets contain signals recorded during externally induced pain. However, during real orthopaedic rehabilitation sessions, pain is never voluntarily and externally induced. Pain during movement can alter muscle activation and motor capacity, while measures of muscle activity, such as the nociceptive flexion reflex (NFR), are used in pain assessment and are normally measured using sEMG signals. This motivated the use of sEMG alone. The aim of this dataset is to provide a simple yet efficient source for designing machine learning algorithms for pain detection. Moreover, to the best of the authors’ knowledge, no currently publicly available dataset contains signals recorded during rehabilitation exercises performed using a robot.