Temporomandibular Joint Disc Exploration, and Its Pathologies diagnosis by using Artificial Intelligence: A Systematic Review

Abstract

Artificial intelligence is invading almost every field, including health. Scientific publications concerning the medical field and AI have been increasing steadily in recent years, covering almost all medical specialities. This literature review aims to systematically synthesize current research on the development and validation of multimodal machine and deep learning models for automating TMJ disc displacement diagnosis. It highlights how integrating automatic MRI segmentation with clinical data correlation can significantly reduce diagnostic errors. Several databases (PubMed, Web of Science, Scopus, Science Direct, Springer, and ResearchGates) were consulted for articles on the subject of « articular disc and AI », from 2003 to 2025. One thousand eight hundred seventy-six (1876) studies were identified, of which eighteen (27) were included, totaling more than 6869 patients between controls and patients with articular disc displacement and more than 44000 images. Most of the studies used MRI imaging (considered the gold standard) alone or in combination with other types of data to explore the articular disc. To achieve this goal, various artificial intelligence models were used, including Machine Learning (Random forest, XG Boost, KNN, SVM…), and Deep Learning (ANN, CNN, U-Net, Seg- Net, RestNet 34…). Artificial intelligence can help in this sense by determining the condition of the disc in order to avoid not only the serious stages of disc displacements (ADD without Reduction), but also disc perforations.