Machine Learning Based Diagnostic Models for Hip and Knee Prosthetic Joint Infection: A Systematic Review.

Area of Science:

• Orthopedic Surgery
• Medical Diagnostics
• Artificial Intelligence in Medicine

Background:

• Prosthetic joint infection (PJI) complicates up to 1.7% of total hip (THA) and knee arthroplasties (TKA).
• PJI significantly increases patient morbidity, mortality (up to 21% at 5 years), and healthcare costs.
• Accurate PJI diagnosis is crucial but challenging due to lack of a gold standard and variable test performance.

Purpose of the Study:

• To systematically review the literature on the application of machine learning (ML) for diagnosing PJI following THA or TKA.
• To assess the diagnostic performance and limitations of ML models in PJI detection.

Main Methods:

• A systematic literature review was conducted to identify studies using ML for PJI diagnosis.
• Included studies utilized patient demographics, clinical data, serological markers, synovial fluid analysis, and imaging.
• Analyzed 12 studies that applied ML techniques to diagnose or predict PJI.

Main Results:

• ML models demonstrated good predictive performance, with Area Under the Curve (AUC) ranging from 0.68 to 0.993.
• The reviewed ML models utilized diverse data inputs including demographics, clinical features, and imaging.
• A significant limitation identified was the lack of external validation for most models.

Conclusions:

• Machine learning offers a promising, data-driven approach to improve the accuracy of PJI diagnosis.
• Enhanced diagnostic accuracy through ML could lead to reduced diagnostic delays and more timely, appropriate treatment.
• Further research is essential to validate ML models externally and assess their generalizability for clinical PJI diagnosis.