Novel Method of Classification in Knee Osteoarthritis: Machine Learning Application Versus Logistic Regression Model

Area of Science:

• Biomedical Engineering
• Orthopedics
• Data Science

Background:

• Knee osteoarthritis (KOA) diagnosis and severity classification are crucial for patient management.
• Conventional statistical methods have limitations in accurately classifying KOA and its progression.
• Machine learning (ML) offers promising new avenues for analyzing complex medical data.

Purpose of the Study:

• To introduce novel ML-based classification methods for KOA.
• To compare the performance of ML classifiers against traditional statistical techniques.
• To identify key gait parameters for KOA classification using ML.

Main Methods:

• Recruited 84 KOA patients and 97 healthy controls.
• Classified KOA patients into three Kellgren-Lawrence (K-L) grades.
• Employed Support Vector Machine (SVM) for ML classification and logistic regression for comparison.

Main Results:

• ML classification achieved higher accuracy than logistic regression for distinguishing KOA patients from controls.
• Feature selection in ML enhanced KOA severity classification accuracy.
• Knee flexion and extension during the swing phase were identified as critical gait features.

Conclusions:

• ML presents a powerful complementary approach to logistic regression for KOA classification.
• ML-based methods can potentially improve clinical diagnosis of KOA.
• This approach may facilitate gait correction strategies for individuals with KOA.