Deep learning-based automated diagnosis of temporomandibular joint anterior disc displacement and its clinical application

Yue Yu ¹, Shu Jun Wu ¹, Yao Min Zhu ¹

⁰Department of Oral & Maxillofacial Surgery, Shenzhen Stomatology Hospital, Affiliated to Shenzhen University, Shenzhen, Guangdong Province, China.

Frontiers in Physiology +

|

December 30, 2024

View abstract on PubMed

Summary

Deep learning models accurately detect temporomandibular joint anterior disc displacement (TMJ ADD) using MRI scans. This automated system enhances diagnostic precision for clinical practice.

Area Of Science

Medical Imaging
Artificial Intelligence
Orthodontics

Background

Temporomandibular joint anterior disc displacement (TMJ ADD) is a common condition affecting jaw function.
Accurate diagnosis of TMJ ADD is crucial for effective treatment planning.

Purpose Of The Study

To develop a deep learning-based method for interpreting MRI scans of TMJ ADD.
To create an automated diagnostic system for clinical application.

Main Methods

Four deep learning models (ResNet101_vd framework) were trained on 618 TMJ MRI cases.
Models performed region of interest identification, anatomical segmentation, and TMJ ADD classification.
Performance was validated internally and externally using distinct datasets.

Main Results

Deep learning models achieved high precision rates (>92%) in classifying TMJ ADD.
Segmentation-based models outperformed non-segmentation models in diagnostic metrics.
Satisfactory results were obtained despite variations in performance between hospital datasets.

Conclusions

Deep learning models demonstrate high accuracy and interpretability in detecting TMJ ADD.
The developed system can be integrated into clinical practice to improve diagnostic precision.
Automated interpretation of TMJ MRI scans holds significant potential for patient care.

Keywords:

anterior disc displacement automated diagnosis deep learning magnetic resonance imaging temporomandibular joint