Classification of inertial sensor-based gait patterns of orthopaedic conditions using machine learning: A pilot study

Area of Science:

• Biomechanics
• Gerontology
• Machine Learning

Background:

• Elderly individuals frequently present with multiple comorbidities impacting mobility.
• Objective tools are needed to identify primary causes of functional limitations for improved clinical decision-making.

Purpose of the Study:

• To investigate the efficacy of machine learning in identifying degenerative diseases based on gait patterns.
• To differentiate between healthy gait, osteoarthritis, and spinal stenosis using gait analysis.

Main Methods:

• Utilized a clinical database containing gait data (sagittal joint angles, spatiotemporal parameters) from seven inertial sensors.
• Included anthropometric data from patients with unilateral knee/hip osteoarthritis, lumbar/cervical spinal stenosis, and healthy controls.
• Employed machine learning models via MATLAB Classification Learner, selecting Support Vector Machine (SVM) as the optimal classifier.

Main Results:

• Achieved 82.3% accuracy in discriminating between healthy and pathological gait.
• Attained 51.4% accuracy in discriminating between different degenerative diseases, suggesting gait pattern similarities.
• Demonstrated that pathological gait is distinguishable from healthy gait using classical machine learning.

Conclusions:

• Distinct differences exist between pathological and healthy gait, enabling classification via machine learning.
• Current gait characteristics and anthropometric data are insufficient for reliably discriminating between specific degenerative diseases.
• Further research is necessary to explore gait pattern similarities and improve disease-specific classification.