EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Artificial intelligence and radiologists: a comparative analysis of magnetic resonance imaging of the temporomandibular joint

Guilherme Caldeira Peixoto Gomes, Lucas Cavalieri Pereira

Advances in Oral and Maxillofacial Surgery · 2026-12 · 24 · 100743

DOI: 10.1016/j.adoms.2026.100743 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Advances in Oral and Maxillofacial Surgery, 2026-12. Verfasst von Guilherme Caldeira Peixoto Gomes und 1 weiteren Autor:innen. Die Arbeit nennt 26 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

This study evaluated the functional agreement between magnetic resonance imaging (MRI) descriptions of the temporomandibular joint (TMJ) generated by GPT-4 with vision capabilities and reports issued by human radiologists, in an observational, retrospective, cross-sectional, analytical and comparative design. Twenty-nine examinations and 58 joints made up the final sample, treated as a single, unstratified group. Composite static images built from selected slices were submitted to the model using a standardized prompt covering ten anatomical-functional items, and each response was compared item by item with the human report and classified as agreement, disagreement, or not described. Among 432 evaluable comparisons, overall agreement was 70.83%, but the central finding was the wide variability across cases: mean 68.5%, median 80%, standard deviation 30.87%, range 0% to 100%, with 41.4% of examinations excellent and 31.0% poor. Agreement was highest for retrodiscal tissues (80.6%), disc recapture (79.2%) and osseous changes of the fossa/eminence (76.7%), and lowest for joint effusion (63.8%). No difference emerged between right and left sides (p = 0.9817), while a positive correlation linked report completeness to agreement (r = 0.4949; p = 0.0063). The tool shows potential as a support resource, but its inconsistent, case-dependent performance means specialist supervision is required at every stage.

Zitieren

APA

Gomes, G. C. P., & Pereira, L. C. (2026). Artificial intelligence and radiologists: a comparative analysis of magnetic resonance imaging of the temporomandibular joint. Advances in Oral and Maxillofacial Surgery, 24, 100743. https://doi.org/10.1016/j.adoms.2026.100743

Vancouver

Gomes GCP, Pereira LC. Artificial intelligence and radiologists: a comparative analysis of magnetic resonance imaging of the temporomandibular joint. Advances in Oral and Maxillofacial Surgery. 2026;24:100743. doi:10.1016/j.adoms.2026.100743

BibTeX
@article{Gomes2026,
  author = {Gomes, Guilherme Caldeira Peixoto and Pereira, Lucas Cavalieri},
  title = {Artificial intelligence and radiologists: a comparative analysis of magnetic resonance imaging of the temporomandibular joint},
  journal = {Advances in Oral and Maxillofacial Surgery},
  year = {2026},
  volume = {24},
  pages = {100743},
  doi = {10.1016/j.adoms.2026.100743}
}
Kostenlos starten →