Using deep learning to predict temporomandibular joint disc perforation based on magnetic resonance imaging

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Orthopedics

Background:

• Temporomandibular joint (TMJ) disc perforation is a significant condition affecting joint function.
• Accurate prediction of TMJ disc perforation is crucial for effective treatment planning.
• Current diagnostic methods for TMJ disc perforation have limitations.

Purpose of the Study:

• To develop and validate a deep learning-based algorithm for predicting TMJ disc perforation.
• To assess the algorithm's performance against conventional methods and prior research using MRI findings.
• To leverage advanced machine learning techniques for enhanced diagnostic capabilities in TMJ disorders.

Main Methods:

• Retrospective review of 299 TMJ joints from 289 patients (January 2005 - June 2018).
• Feature extraction from magnetic resonance imaging (MRI) by experienced observers.
• Development and validation of prediction models using random forest and multilayer perceptron (MLP) with the Keras framework.
• Performance evaluation using the area under the receiver operating characteristic curve (AUC).

Main Results:

• The MLP deep learning model achieved the highest performance with an AUC of 0.940.
• Random forest model showed strong performance with an AUC of 0.918.
• Both MLP and random forest models significantly outperformed previous MRI-based prediction results (AUC 0.808) and nomograms (AUC 0.889).

Conclusions:

• Deep learning algorithms demonstrate superior performance in predicting TMJ disc perforation compared to conventional methods.
• The developed MLP model offers a highly accurate and reliable tool for diagnosing TMJ disc perforation.
• This AI-driven approach has the potential to significantly improve the diagnostic accuracy and management of TMJ disorders.