Diagnosis of Tempromandibular Disorders Using Local Binary Patterns.

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Temporomandibular joint disorder (TMD) can cause structural bone changes.
• Current diagnostic tools may benefit from advanced image analysis.
• Local Binary Pattern (LBP) and histogram-oriented gradients offer novel feature extraction methods.

Purpose of the Study:

• To develop and evaluate an automated diagnostic tool for TMD using image processing techniques.
• To assess the efficacy of Local Binary Pattern (LBP) and histogram-oriented gradients in TMD detection.
• To compare the performance of various machine learning classifiers for TMD diagnosis.

Main Methods:

• Utilized Cone Beam Computed Tomography (CBCT) images from 66 TMD patients and 66 controls.
• Extracted image features using Local Binary Pattern (LBP) and histogram-oriented gradients.
• Applied Singular Value Decomposition (SVD) for dimensionality reduction and evaluated K-nearest neighbor (K-NN), Support Vector Machine, Naïve Bayesian, and Random Forest classifiers.

Main Results:

• The K-nearest neighbor (K-NN) classifier demonstrated superior performance.
• K-NN achieved an accuracy of 0.9242, sensitivity of 0.9470, and specificity of 0.9015.
• Other evaluated classifiers showed lower accuracy, sensitivity, and specificity compared to K-NN.

Conclusions:

• An automated approach for TMD detection using image processing, specifically LBP and feature extraction, was successfully proposed.
• K-NN proved to be the most effective classifier, accurately distinguishing TMD patients from healthy individuals with 92.42% accuracy.
• The developed method holds potential for assisting in the early diagnosis of TMD.