Journal Article

Performance of GPT‐5 in the Interpretation of IBD Histopathology Reports

Ligresti, Dario, Laschi, Lucrezia, Orlando, Ambrogio, Sferrazza, Sandro, Sordo, Rachele Del, Mocci, Giammarco, Locci, Giorgia, Tacelli, Matteo, Daperno, Marco, Villanacci, Vincenzo, Vitello, Alessandro, Maida, Marcello, Calabrese, Giulio, Furnari, Manuele, Fattorini, Caterina, Macaluso, Fabio Salvatore, Rispo, Antonio, Marasco, Giovanni, Decarli, Nicola L., Facciorusso, Antonio

2026

Overview

Background Histopathological interpretation is crucial for diagnosing inflammatory bowel disease (IBD), distinguishing between Crohn's Disease (CD), Ulcerative Colitis (UC), IBD‐Unclassified (IBD‐U), and Non‐IBD colitis (NIBDC). However, interobserver variability and limited expertise can reduce diagnostic accuracy. Large Language Models (LLMs) such as GPT‐5 may offer clinical support in interpreting histology reports. Methods We analyzed 100 real‐life histological reports from ileo‐colonoscopies, equally representing CD, UC, IBD‐U, and NIBDC, collected across five Italian healthcare centers, including both IBD‐specialized and non‐specialized hospitals. A reference standard was established by an expert pathologist. Independent classifications were generated by GPT‐5, five gastrointestinal pathologists, five IBD‐expert gastroenterologists (GIs), and five non‐expert GIs. Diagnostic performance (accuracy, recall, precision, F1‐score), agreement with the reference standard (Cohen's κ), and inter‐rater reliability (Fleiss' κ) were assessed. Results GPT‐5 achieved the highest agreement with the reference standard with the highest accuracy (76.0%), compared to pathologists (68.6%), IBD‐experts (69.2%), and non‐experts (63.2%). Agreement with the reference standard was substantial for GPT‐5 (κ = 0.671) and moderate for human groups (κ = 0.508–0.588). GPT‐5 showed perfect recall for CD and UC, high recall for NIBDC (96.0%), but poor performance for IBD‐U (recall 8.0%, F1‐score 14.3%). Fleiss' κ indicated moderate agreement among pathologists and IBD‐experts, and fair agreement among non‐experts. Conclusion GPT‐5 demonstrated reliable performance in interpreting IBD histological reports, exhibiting high accuracy and strong agreement with the reference standard. While unreliable for IBD‐U, GPT‐5 may serve as a supportive tool in histopathological interpretation of IBD, particularly in centers with limited access to expert pathologists or IBD‐specialists. Key Summary Summarise the established knowledge on this subject ◦ Histopathological assessment is central to diagnosing inflammatory bowel disease (IBD), but interpretation is challenging and interobserver variability is common. What are the significant and/or new findings of this study ◦ This study shows that GPT‐5 achieved the highest agreement with the reference standard with higher overall accuracy in classifying IBD histological reports. ◦ In detail, GPT‐5 showed excellent performance in classifying Crohn's disease, ulcerative colitis, and non‐IBD colitis, but poor performance for IBD‐Unclassified (IBD‐U). ◦ GPT‐5 may support clinicians by reducing diagnostic variability and providing timely assistance, especially in settings lacking IBD expertise, while reinforcing the need for multidisciplinary review in IBD‐U.

• Publisher: John Wiley & Sons, Inc
• Subject: Accuracy / Adult / Biopsy / Colitis, Ulcerative - diagnosis / Colitis, Ulcerative - pathology / Colonoscopy / crohn disease / Crohn Disease - diagnosis / Crohn Disease - pathology / Diagnosis, Differential / Female / GPT‐5 / Humans / Inflammatory bowel disease / inflammatory bowel diseases / Inflammatory Bowel Diseases - diagnosis / Inflammatory Bowel Diseases - pathology / Male / natural language processing / Observer Variation / pathology / Reproducibility of Results / ulcerative colitis