Identifying Robust Risk Factors for Knee Osteoarthritis Progression: An Evolutionary Machine Learning Approach

Area of Science:

• Orthopedics and Sports Medicine
• Biostatistics and Bioinformatics
• Computational Biology

Background:

• Knee osteoarthritis (KOA) is a major cause of disability, characterized by heterogeneous progression and numerous interacting risk factors.
• Existing models predict KOA onset but few identify factors driving its progression.
• High-dimensional, heterogeneous data and class imbalance in KOA progression studies present significant analytical challenges.

Purpose of the Study:

• To identify critical risk factors associated with knee osteoarthritis progression.
• To address challenges of high dimensionality, data heterogeneity, and class imbalance in KOA progression analysis.
• To develop a robust feature selection methodology for KOA risk factor identification.

Main Methods:

• Utilized a robust feature selection (FS) methodology combining evolutionary algorithms and machine learning (ML) models.
• Applied the methodology to data from the Osteoarthritis Initiative (OAI) database.
• Validated the selected 35 risk factors using comparative analysis and SHapley Additive exPlanations (SHAP).

Main Results:

• Selected a parsimonious subset of 35 risk factors for KOA progression.
• Achieved a mean accuracy of 71.25% in generalizing across the dataset.
• Demonstrated the effectiveness of the proposed FS methodology against established techniques.

Conclusions:

• The developed FS methodology successfully identifies key risk factors for KOA progression.
• Findings support the development of novel risk stratification strategies and identification of KOA phenotypes.
• Enables personalized interventions for patients with knee osteoarthritis.