Artificial Intelligence and Multimodal Imaging of the Temporomandibular Joint: From Morphology to Clinical Decision Support

The field of orofacial medicine increasingly recognizes the temporomandibular joint (TMJ) as a complex anatomical and functional unit whose disorders can profoundly affect oral health and quality of life. Despite the availability of advanced imaging modalities including cone beam computed tomography (CBCT), magnetic resonance imaging (MRI), and ultrasound, current diagnostic practices for temporomandibular disorders (TMDs) remain subjective and heavily dependent on clinician experience. This variability limits consistency across institutions and complicates the interpretation of findings such as disc displacement and degenerative joint disease. Over the past few years, artificial intelligence — encompassing radiomics, machine learning, and deep learning — has emerged as a powerful set of tools to extract quantitative features from multimodal imaging (ultrasonography, CBCT, and MRI) and enable objective, reproducible analysis of the TMJ. Notably, radiomic feature analysis, convolutional neural networks, vision transformers, and self-supervised models have shown remarkable performance in automated segmentation, morphological characterization, and texture-based classification of osseous and soft tissue pathologies. Nonetheless, critical gaps persist regarding the standardization of algorithms, external validation, interpretability, and integration into clinical workflows.

This Research Topic aims to advance the development and translation of artificial intelligence systems — including radiomics, machine learning, and deep learning — for multimodal TMJ assessment, covering the full pathway from morphological analysis to clinical decision support. The objective is to stimulate contributions that establish rigorous methodologies, evaluate diagnostic reproducibility, and link automated image analysis to patient outcomes. Specific goals include exploring whether neural networks can outperform traditional observer-based grading systems, determining how multimodal models can integrate imaging and clinical data for prognosis, and assessing how explainability and regulatory transparency can enhance clinician trust in AI driven tools. Importantly, this collection welcomes both positive and null results as well as comparative and benchmarking studies that contribute to a shared evidentiary base for future applications in clinical settings.

To gather further insights into the potential and limitations of artificial intelligence in TMJ diagnostics, we invite research spanning the entire translational spectrum from algorithm development to implementation in patient care. We particularly encourage interdisciplinary work bridging oral radiology, computer science, biomedical engineering, and clinical medicine. Potential topics include, but are not limited to:

• Automated segmentation of TMJ structures (condyle, disc, fossa, and associated muscles) on CBCT, MRI, ultrasound, or panoramic images

• Detection and classification of disc displacement, joint effusion, osteoarthritis, condylar resorption, and ankylosis

• Three dimensional morphological and statistical shape analysis for condylar and fossa modelling

• Multimodal integration of imaging with biomechanical, clinical, and patient reported outcomes

• Prediction of treatment response across conservative, orthodontic, and surgical management of TMD

• Explainable, uncertainty aware, and federated learning approaches supporting clinical deployment

• Synthetic data generation, domain adaptation, and cross center validation for robust generalization

• Ethical, regulatory, and reproducibility challenges in AI driven TMJ imaging

We accept Original Research, Systematic Reviews, Methods, Brief Research Reports, Clinical Trials, Case Reports with quantitative AI components, as well as Opinion and Perspective articles. Replication studies, benchmarking initiatives, and reports of negative findings are explicitly encouraged.