Development of AI-based tools for assessing temporomandibular joint disorders using MRI images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Temporomandibular joint disorders (TMDs) present diagnostic challenges due to complex MRI interpretation and inter-observer variability.
• Developing automated assessment tools for TMDs using magnetic resonance imaging (MRI) is crucial for improving clinical practice.

Purpose of the Study:

• To develop and evaluate artificial intelligence (AI) based deep learning models for automated assessment of TMDs using MRI.
• To classify disc displacement, detect osteoarthritic changes, and identify joint effusions in TMJ MRI examinations.

Main Methods:

• A total of 2,847 TMJ MRI examinations were utilized to train and test deep learning models, specifically convolutional neural networks.
• The models were evaluated for diagnostic accuracy, area under the ROC curve (AUC), and agreement with expert radiologists.

Main Results:

• The convolutional neural network achieved high diagnostic accuracies: 94.2% for disc displacement, 91.8% for osteoarthritic changes, and 93.5% for joint effusion.
• AUC values exceeded 0.92, and strong agreement (κ = 0.87-0.91) was observed with expert radiologists.
• The AI system reduced interpretation time by 68%.

Conclusions:

• AI-based tools demonstrate significant potential for improving diagnostic accuracy, consistency, and efficiency in clinical TMJ evaluation.
• Automated assessment of TMDs using deep learning on MRI can aid clinicians in diagnosis and treatment planning.