EUVIMEDEUVIMED entdecken
Crossref

Evaluation of ChatGPT-4 Performance in Forensic Pathology: Comprehensive Methodological Research of Myocardial Infarction-Related Questions

Aytül BUĞRA, Hüseyin Çağrı ŞAHİN, Kadriye Ebru AKAR, Beyza KESKİN ÖZTÜRK

Turkiye Klinikleri Journal of Forensic Medicine and Forensic Sciences · 2026 · 23 (2) · 116-124

DOI: 10.5336/forensic.2025-116135 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Turkiye Klinikleri Journal of Forensic Medicine and Forensic Sciences, 2026. Verfasst von Aytül BUĞRA und 3 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

Artificial intelligence (AI) applications in forensic pathology are rapidly expanding, yet their accuracy and reliability remain insufficiently evaluated. Large language models like ChatGPT-4 show promise in medical applications but require rigorous validation in forensic contexts. To evaluate the accuracy and reliability of ChatGPT-4 responses to myocardial infarction (MI) questions from a forensic pathology perspective. Material and Methods: Thirty MI-related questions were categorized into general medical knowledge (Group A) (n=15) and forensic pathology specialty knowledge (Group B) (n=15). ChatGPT-4 responses were independently evaluated by 3 forensic pathologists using a 4-point Likert scale (1=Definitely incorrect to 4=Definitely correct). Differences between 2 independent groups were evaluated using the Mann-Whitney U test. Data were presented as median (minimum-maximum), and a p value 0.05). Topic-specific analysis revealed a 100% failure rate in highly technical areas such as histopathological dating and immunohistochemical markers, whereas basic definitions and clinical prevention strategies achieved perfect accuracy (0% failure).

Conclusion

ChatGPT-4 demonstrates reliable performance in general MI knowledge but shows significant limitations in forensic pathology specialty areas, particularly histopathological dating and advanced diagnostic techniques. A risk-stratified approach is recommended for clinical implementation.

Zitieren

APA

BUĞRA, A., ŞAHİN, H. Ç., AKAR, K. E., & ÖZTÜRK, B. K. (2026). Evaluation of ChatGPT-4 Performance in Forensic Pathology: Comprehensive Methodological Research of Myocardial Infarction-Related Questions. Turkiye Klinikleri Journal of Forensic Medicine and Forensic Sciences, 23(2), 116-124. https://doi.org/10.5336/forensic.2025-116135

Vancouver

BUĞRA A, ŞAHİN HÇ, AKAR KE, ÖZTÜRK BK. Evaluation of ChatGPT-4 Performance in Forensic Pathology: Comprehensive Methodological Research of Myocardial Infarction-Related Questions. Turkiye Klinikleri Journal of Forensic Medicine and Forensic Sciences. 2026;23(2):116-124. doi:10.5336/forensic.2025-116135

BibTeX
@article{BURA2026,
  author = {BUĞRA, Aytül and ŞAHİN, Hüseyin Çağrı and AKAR, Kadriye Ebru and ÖZTÜRK, Beyza KESKİN},
  title = {Evaluation of ChatGPT-4 Performance in Forensic Pathology: Comprehensive Methodological Research of Myocardial Infarction-Related Questions},
  journal = {Turkiye Klinikleri Journal of Forensic Medicine and Forensic Sciences},
  year = {2026},
  volume = {23},
  number = {2},
  pages = {116--124},
  doi = {10.5336/forensic.2025-116135}
}
Kostenlos starten →