Support Vector Machines for detecting recovery from knee replacement surgery using quantitative gait measures

Area of Science:

• Biomechanical analysis
• Machine learning applications in healthcare

Background:

• Knee osteoarthritis (OA) is a primary cause of disability in the elderly, often necessitating surgical intervention.
• Knee replacement surgery improves pain and physical function, but gait dysfunction can persist post-operatively.
• Altered spatio-temporal measures and gait asymmetry are common gait dysfunctions in knee OA patients.

Purpose of the Study:

• To apply Support Vector Machines (SVM) for classifying knee OA gait patterns using spatio-temporal parameters before surgery.
• To evaluate SVM's capability in assessing gait improvement 2 months after knee replacement surgery.

Main Methods:

• Utilized spatio-temporal gait parameters to train and test SVM models.
• Employed leave-one-out (LOO) cross-validation for accuracy assessment.
• Applied feature selection techniques to identify key gait parameters.

Main Results:

• SVM achieved a maximum LOO accuracy of 94.2% in distinguishing knee OA gait from healthy gait.
• Feature selection improved accuracy to 97.1% using only two symmetry index features.
• Post-surgery SVM analysis identified persistent gait alterations in 4 out of the evaluated patients.

Conclusions:

• SVM is effective in classifying knee OA gait and assessing post-surgical recovery.
• Gait symmetry monitoring is crucial for evaluating treatment outcomes and recovery progress after knee replacement.
• Identifying persistent gait alterations can guide further clinical interventions.