Shape variation analyzer: a classifier for temporomandibular joint damaged by osteoarthritis

Area of Science:

• Medical imaging analysis
• Artificial intelligence in healthcare
• Orthopedics and dentistry

Background:

• Temporomandibular joint osteoarthritis (TMJ OA) involves complex bony changes that require accurate staging for effective management.
• Current methods for assessing TMJ OA severity can be subjective and time-consuming.
• Advanced computational approaches are needed to objectively quantify disease progression.

Purpose of the Study:

• To develop and validate a deep learning neural network, the Shape Variation Analyzer (SVA), for automated disease staging of TMJ OA.
• To assess the performance of SVA in classifying TMJ OA based on 3D morphology from CBCT scans.
• To compare the accuracy of SVA against traditional machine learning algorithms.

Main Methods:

• Development of a deep learning neural network (SVA) utilizing 259 TMJ CBCT scans for training and 34 for testing.
• Application of data augmentation and SMOTE techniques to enhance training data robustness and class balance.
• Integration of geometrical features and heat kernel signature for comprehensive 3D shape description.
• Comparative analysis with nine supervised machine learning algorithms for classification accuracy.

Main Results:

• The Shape Variation Analyzer (SVA) demonstrated superior accuracy in classifying temporomandibular joint osteoarthritis (TMJ OA) compared to nine other supervised machine learning algorithms.
• The deep learning model effectively utilized 3D morphological data from CBCT scans for disease staging.
• Data augmentation and SMOTE techniques improved the model's ability to handle complex datasets and prevent overfitting.

Conclusions:

• The Shape Variation Analyzer (SVA) represents a significant advancement in the automated staging of TMJ OA, offering a robust and accurate classification method.
• SVA leverages 3D morphology analysis through a deep learning approach, providing objective insights into TMJ OA pathology.
• This 3D Sheer extension has the potential to improve clinical decision-making and patient management for TMJ osteoarthritis.