Comparing Traditional Versus AI-Assisted TMJ Disorder Management Approaches: A Systematic Review and Meta-Analysis

This systematic review and meta-analysis compared traditional diagnostic approaches with artificial intelligence (AI)-based techniques for temporomandibular joint disorders (TMDs) and evaluated their diagnostic accuracy.

Literature searches across PubMed-MEDLINE, Scopus, Embase, Cochrane, and Google Scholar identified studies published between 1st January 2010 to 20th April 2025. Eligible studies used AI-based algorithms, such as deep learning (DL), machine learning (ML), or computer-aided diagnostic tools, for TMD diagnosis or management, reporting performance metrics including sensitivity, specificity, and accuracy. Traditional approaches, including clinical examinations, radiographic assessments, and standardized diagnostic criteria (DC/TMD), serve as comparators. Risk of bias was assessed using Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2). Meta-analysis was conducted in the form of pooled sensitivity and specificity.

Fourteen studies were included, comprising AI models trained on clinical and imaging data including cone beam computed tomography (CBCT), magnetic resonance imaging (MRI), and orthopantomogram (OPG). AI methods showed moderate-to-high diagnostic accuracy, with sensitivity ranging from 0.66 to 0.88 and specificity from 0.72 to 0.86. A meta-analysis of five studies showed pooled sensitivity and specificity estimates within these ranges. Among the included studies, AI models integrated radiomic as well as semantic features to achieve sensitivity from 0.82 to 0.93, and specificity from 0.76 to 0.90; however, evidence showed low certainty because bias risk was high (7/9 studies), sample sizes were small (mean n = 42), and external validation was absent in 8 of 9 studies.

AI-assisted techniques offer significant potential to complement traditional TMD diagnostic methods by enhancing the diagnostic precision. However, owing to methodological limitations, further high-quality prospective studies with standardized reporting are needed to validate the use of AI in TMD management.