Improving TMJ Diagnosis: A Deep Learning Approach for Detecting Mandibular Condyle Bone Changes.

Area of Science:

• Dentistry and Oral Health
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Degenerative bone changes in the mandibular condyle are challenging to diagnose using traditional panoramic radiographs.
• Early detection of these changes is crucial for timely intervention and preventing temporomandibular joint (TMJ) disease progression.

Purpose of the Study:

• To evaluate the efficacy of deep learning (DL) models for detecting degenerative bone changes in the mandibular condyle.
• To develop an automated method for identifying conditions like flattening, osteophyte, and erosion on panoramic radiographs.

Main Methods:

• Utilized a dataset of 3875 condylar images from panoramic radiographs.
• Employed various deep learning architectures (DenseNets, ResNets, VGGNets, GoogleNets) with transfer learning.
• Trained and tested models using 70:30 and 80:20 data splits, considering different classification approaches for bone changes.

Main Results:

• GoogleNet architecture achieved the highest accuracy of 95.23% with an 80:20 split for the normal-flattening-deformation dataset.
• Convolutional Neural Network (CNN)-based methods demonstrated high success rates in identifying mandibular condyle bone abnormalities.
• The study confirmed the effectiveness of DL in classifying condylar bone changes.

Conclusions:

• Deep learning, especially CNNs, shows significant potential for accurate and efficient detection of TMJ-related condylar bone changes from panoramic radiographs.
• This AI-driven approach can assist clinicians in identifying patients needing further intervention for TMJ disorders.
• Future research should explore cross-sectional imaging and combined training of condyles to enhance diagnostic accuracy and disease management.