Distinguishing Pain and No Pain in Musicians Through Machine Learning Analysis of Musculoskeletal Data

Musculoskeletal disorders are common among professional musicians and often linked to altered movement patterns. This study examined whether a combined Principal Component Analysis-Linear Discriminant Analysis (PCA-LDA) framework can identify interpretable motion features associated with shoulder-neck pain during violin performance. Kinematic data from 26 violinists (11 with pain, 15 pain-free) were recorded with a 10-camera motion capture system during playing a standardized scale. PCA was applied to 41 upper-body waveforms to extract dominant movement components, which served as inputs for LDA to classify pain status. The best model achieved 0.73 ± 0.12 cross-validated accuracy, indicating considerable group discrimination. The most influential components involved spinal lateral flexion, right shoulder rotation, and wrist abduction, suggesting multilevel patterns linked to pain. The findings show the feasibility of PCA-LDA for extracting interpretable biomechanical features from complex performance data and provide a basis for larger, multimodal studies towards objective assessment of pain-related movement adaptations in musicians.