Challenges and Limitations of Multimodal Large Language Models in Interpreting Pediatric Panoramic Radiographs
International Journal of Paediatric Dentistry · 2026-01 · 36 (1) · 74-80
DOI: 10.1111/ipd.70029 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Paediatric Dentistry, online erschienen am 16.09.2025. Verfasst von Yuichi Mine und 9 weiteren Autor:innen. Beteiligte Einrichtung: Project Research Center for Integrating Digital Dentistry, Hiroshima University, Hiroshima, Japan. Hauptthemen laut MEDLINE: Radiography, Panoramic, Language, Image Processing, Computer-Assisted. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: Hiroshima University.
Abstract (Originalsprache)
Background
Multimodal large language models (LLMs) have potential for medical image analysis, yet their reliability for pediatric panoramic radiographs remains uncertain.
Aim
This study evaluated two multimodal LLMs (OpenAI o1, Claude 3.5 Sonnet) for detecting and counting teeth (including tooth germs) on pediatric panoramic radiographs.
Design
Eighty‐seven pediatric panoramic radiographs from an open‐source data set were analyzed. Two pediatric dentists annotated the presence or absence of each potential tooth position. Each image was processed five times by the LLMs using identical prompts, and the results were compared with the expert annotations. Standard performance metrics and Fleiss' kappa were calculated.
Results
Detailed examination revealed that subtle developmental stages and minor tooth loss were consistently misidentified. Claude 3.5 Sonnet had higher sensitivity but significantly lower specificity (29.8% ± 21.5%), resulting in many false positives. OpenAI o1 demonstrated superior specificity compared to Claude 3.5 Sonnet, but still failed to correctly detect subtle defects in certain mixed dentition cases. Both models showed large variability in repeated runs.
Conclusion
Both LLMs failed to achieve clinically acceptable performance and cannot reliably identify nuanced discrepancies critical for pediatric dentistry. Further refinements and consistency improvements are essential before routine clinical use.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Project Research Center for Integrating Digital Dentistry, Hiroshima University, Hiroshima, Japan.
- Department of Medical Systems Engineering, Graduate School of Biomedical and Health Sciences, Hiroshima University, Hiroshima, Japan.
- Department of Pediatric Dentistry, Graduate School of Biomedical and Health Sciences, Hiroshima University, Hiroshima, Japan.
- School of Dentistry, College of Oral Medicine, Taipei Medical University, Taipei, Taiwan.
- Department of Oral and Maxillofacial Radiology, Graduate School of Biomedical and Health Sciences, Hiroshima University, Hiroshima, Japan.
ORCID-Profile: Yuichi Mine, Yuko Iwamoto
Förderung
- Hiroshima University
Zitieren
APA
Mine, Y., Iwamoto, Y., Okazaki, S., Nishimura, T., Tabata, E., Takeda, S., Peng, T. Y., Nomura, R., Kakimoto, N., & Murayama, T. (2026). Challenges and Limitations of Multimodal Large Language Models in Interpreting Pediatric Panoramic Radiographs. International Journal of Paediatric Dentistry, 36(1), 74-80. https://doi.org/10.1111/ipd.70029
Vancouver
Mine Y, Iwamoto Y, Okazaki S, Nishimura T, Tabata E, Takeda S, et al. Challenges and Limitations of Multimodal Large Language Models in Interpreting Pediatric Panoramic Radiographs. International Journal of Paediatric Dentistry. 2026;36(1):74-80. doi:10.1111/ipd.70029
BibTeX
@article{Mine2026,
author = {Mine, Yuichi and Iwamoto, Yuko and Okazaki, Shota and Nishimura, Taku and Tabata, Eimi and Takeda, Saori and Peng, Tzu-Yu and Nomura, Ryota and Kakimoto, Naoya and Murayama, Takeshi},
title = {Challenges and Limitations of Multimodal Large Language Models in Interpreting Pediatric Panoramic Radiographs},
journal = {International Journal of Paediatric Dentistry},
year = {2026},
volume = {36},
number = {1},
pages = {74--80},
doi = {10.1111/ipd.70029}
}
EUVIMED