EUVIMEDEUVIMED entdecken
Crossref

Exploring the Potential of Using Generative AI in Assisting Military Doctors Providing Ophthalmic Care: Based on Korean-Language Exam Performance

Minsoo Park

The Journal of the korean Military Medicine · 2025

DOI: 10.64416/jkmm.56.2.37 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The Journal of the korean Military Medicine, 2025. Verfasst von Minsoo Park.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

This study evaluated the performance of generative AI models in answering ophthalmology questions presented in Korean and assessed their utility as an assistive tool for non-ophthalmologist military doctors.

Method

Ophthalmology questions from the Korean National Medical Licensing Examination (written in Korean) were tested on three AI models (ChatGPT-4o, Gemini, Clova X). To ensure reproducibility, each model was tested in 10 independent trials. Their average accuracy rates were compared with those of 22 military doctors who took the same test in June 2025. Questions were categorized into image-based and text-only formats. In addition, explanations provided by the AI models were analyzed to identify the causes of incorrect answers.

Results

ChatGPT-4o and Gemini both achieved 100.0% accuracy, significantly outperforming all other groups across both question types (p < 0.05). Clova X scored significantly higher than the military doctors on text-only questions (87.5 ± 17.7% vs. 72.3 ± 17.1%, p = 0.031), but showed no significant difference on image-based questions (71.7 ± 11.3% vs. 80.3 ± 13.3%, p = 0.120) or overall average accuracy (78.0 ± 9.2% vs. 77.3 ± 10.3%, p = 0.191). Among Clova X’s incorrect responses, 17 resulted from image misinterpretation, 4 from incorrect information, and 1 from multiple answers.

Conclusion

Among the generative AI models, ChatGPT-4o and Gemini showed significantly higher accuracy on Korean-language ophthalmology questions than non-ophthalmologist military doctors. However, as this study was limited to comparing test results, these models should be regarded as supportive tools rather than standalone diagnostic systems in clinical ophthalmic settings.

Zitieren

APA

Park, M. (2025). Exploring the Potential of Using Generative AI in Assisting Military Doctors Providing Ophthalmic Care: Based on Korean-Language Exam Performance. The Journal of the korean Military Medicine. https://doi.org/10.64416/jkmm.56.2.37

Vancouver

Park M. Exploring the Potential of Using Generative AI in Assisting Military Doctors Providing Ophthalmic Care: Based on Korean-Language Exam Performance. The Journal of the korean Military Medicine. 2025. doi:10.64416/jkmm.56.2.37

BibTeX
@article{Park2025,
  author = {Park, Minsoo},
  title = {Exploring the Potential of Using Generative AI in Assisting Military Doctors Providing Ophthalmic Care: Based on Korean-Language Exam Performance},
  journal = {The Journal of the korean Military Medicine},
  year = {2025},
  doi = {10.64416/jkmm.56.2.37}
}
Kostenlos starten →