AI segments TMJ disc on MRI using deep learning

Title

Automatic segmentation of the temporomandibular joint disc on magnetic resonance images using a deep learning technique

Authors

Michihito Nozawa; Hirokazu Ito; Yoshiko Ariji; Motoki Fukuda; Chinami Igarashi; Masako Nishiyama; Nobumi Ogi; Akitoshi Katsumata; Kaoru Kobayashi; Eiichiro Ariji

Summary

This study aimed to construct a deep learning model for the automatic segmentation of the temporomandibular joint (TMJ) disc on magnetic resonance (MR) images and to evaluate its performance using both internal and external datasets. The researchers used 1200 MR images from two different hospitals to train and test a modified U-Net architecture. The model achieved recall rates above 80% for both internal and external validation sets, suggesting potential utility for identifying disc positions clinically.

Key Words

Artificial intelligence, Deep learning, TMJ disc, MRI, Segmentation

Extracted Data

Year: 2022

Modality: Magnetic Resonance Imaging (MRI)

Dataset: 1200 MR images (600 TMJs, closed/open mouth)

Dataset Split: Training+Validation 800 (Hospital A); Test 200 internal (Hospital A); Test 200 external (Hospital B)

Network Architecture: Modified U-Net CNN

Metrics: Recall, Precision, F-measure (Dice), IoU

AP - Professional Qty: 2

AP - Supervisor Presence: Yes

AP - Experience Level: 6 and 20 years

AP - Expertise Area: Oral and Maxillofacial Radiology

**AP - Tool or System: Adobe Photoshop CS6;

ML Task: Semantic Segmentation

Project Objective: Automatic identification and segmentation of the TMJ disc on MRI

Clinical Relevance: Clarification of disc position is essential for evaluating TMJ disorders, and this tool may assist observers with minimal experience in identifying disc positions or serve as an educational training tool.