Global Trends in Temporomandibular Joint Research (2000–2025): A Transformer-Based Topic-Modelling Map

Abstract

Background: This study aims to systematically map the thematic evolution of temporomandibular joint (TMJ)

research from 2000 to 2025, identifying shifts in clinical and methodological priorities using transformer-based topic

modelling.

Materials and Method: A total of 18,107 TMJ-related publications were retrieved from the Scopus database. Titles

and abstracts were embedded using a biomedical transformer model. Dimensionality reduction was performed via

UMAP, followed by clustering with HDBSCAN. Topics were labelled using class-based TF–IDF. Annual topic

proportions were analyzed using ordinary least squares regression with false discovery rate correction.

Results: Twenty-seven distinct topics were identified across six domains: biomechanics, clinical symptoms, imaging,

pathology, surgery, and therapies. Eleven topics showed significant upward trends, notably arthrocentesis, ankylosis,

and mouth-opening interventions. Six topics declined, including disc-focused MRI metrics and elastography. Ten

topics remained stable. The findings reveal a thematic shift from biomechanical and imaging metrics toward

minimally invasive interventions and digital health applications.

Conclusions: Transformer-based topic modelling offers a panoramic view of TMJ research trends, highlighting a

growing focus on patient-centred, conservative, and technology-assisted care. These insights provide a framework for

future research prioritization and underscore the need for interdisciplinary approaches.

Research Article (HRU Int J Dent Oral Res 2025; 5(3):159-173)

Keywords: Artificial intelligence, natural language processing, temporomandibular disorders, temporomandibular

joints.