Osteoarthritis Diagnosis Integrating Whole Joint Radiomics and Clinical Features for Robust Learning Models Using...

Area of Science:

• Medical Imaging
• Machine Learning
• Osteoarthritis Research

Background:

• Temporomandibular Joint Osteoarthritis (TMJ OA) poses diagnostic challenges.
• Accurate diagnosis is crucial for effective patient management.
• Existing diagnostic methods may have limitations in accessibility or accuracy.

Purpose of the Study:

• To develop a robust and accurate machine learning framework for diagnosing TMJ OA.
• To investigate the utility of privileged information, including clinical, imaging, and biological markers.
• To assess the performance and applicability of the proposed diagnostic model.

Main Methods:

• Utilized a machine learning model incorporating privileged information (LUPI).
• Employed Normalized Mutual Information Feature Selection (NMIFS) for optimal feature identification.
• Integrated clinical data, quantitative imaging from cone-beam computerized tomography (CBCT), and biological markers.
• Validated the model using 5-fold stratified cross-validation with hyperparameter tuning.

Main Results:

• Clinical features were identified as primary drivers for TMJ OA diagnosis.
• Quantitative imaging features significantly improved model performance.
• The LUPI model, trained with biological data, achieved high diagnostic accuracy (AUC 0.81, specificity 0.79, precision 0.77).
• Biological data was not required during the testing phase, enhancing model applicability.

Conclusions:

• The proposed LUPI model offers a promising, accurate, and widely applicable framework for TMJ OA diagnosis.
• The model's ability to perform without biological data during testing is a significant advantage for clinical settings.
• Combining clinical and imaging data effectively diagnoses TMJ OA, supported by initial biological data insights.