Deep Learning-Based Joint Effusion Classification in Adult Knee Radiographs: A Multi-Center Prospective Study

Area of Science:

• Radiology
• Artificial Intelligence
• Medical Imaging

Background:

• Knee effusion is a key indicator of joint diseases like osteoarthritis.
• Magnetic resonance imaging (MRI) is superior for effusion detection but less accessible than radiographs.
• Radiographs offer a cost-effective and accessible method for early knee effusion detection.

Purpose of the Study:

• To develop and evaluate a deep learning model for automatic knee effusion detection on radiographs.
• To compare the AI model's diagnostic performance against a baseline model and human physicians.
• To enhance the interpretability of AI-driven radiographic analysis for knee effusion.

Main Methods:

• A multi-center prospective study analyzed 1281 knee radiographs.
• A state-of-the-art deep learning classification model with novel preprocessing was employed.
• Explainable artificial intelligence (XAI) was used for result visualization.

Main Results:

• The proposed AI method achieved an AUC of 0.892, accuracy of 0.803, sensitivity of 0.820, and specificity of 0.785.
• The AI model significantly outperformed a baseline model and two non-orthopedic physicians.
• The XAI method successfully highlighted effusion areas, improving interpretability.

Conclusions:

• The AI-driven approach enables early and accurate classification of knee effusions using radiographs.
• This method has the potential to reduce healthcare costs and improve patient outcomes through prompt intervention.
• AI-powered radiographic analysis offers a promising, accessible tool for diagnosing knee joint diseases.