Machine learning-based classification of pathological shoulder motion using phase-specific kinematic features.

Area of Science:

• Biomechanics
• Machine Learning
• Orthopedics

Background:

• Shoulder pathology diagnosis can be challenging.
• Objective assessment of shoulder movement is crucial for diagnosis and rehabilitation.

Purpose of the Study:

• To investigate the use of an upper-extremity exoskeleton for machine learning-based discrimination of orthopedic shoulder pathology.
• To identify discriminative temporal features within shoulder movements.

Main Methods:

• An upper-extremity exoskeleton was used to collect movement data.
• Twelve patients with shoulder impairments and thirty healthy controls performed eight standardized tasks.
• Logistic regression with 5-fold cross-validation was employed for classification, alongside temporal effect size and phase ablation analyses.

Main Results:

• Classification performance ranged from 0.70 to 1.00 AUROC.
• Six tasks achieved an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.90 or higher.
• Specific movement phases (mid-cycle, early, late) were identified as most informative for different movement types (flexion, abduction, rotation).

Conclusions:

• Machine learning models utilizing exoskeleton data can effectively discriminate shoulder pathology.
• Identifying key temporal movement phases enhances model interpretability and clinical relevance.
• Phase-aware machine learning approaches show promise for objective orthopedic assessment.