Automated Neutrophil Quantification and Histological Score Estimation in Ulcerative Colitis

Area of Science:

• Gastroenterology
• Computational Pathology
• Artificial Intelligence

Background:

• Histological remission, defined by absent neutrophil infiltration, is a key goal in ulcerative colitis (UC) management.
• Accurate quantification and localization of neutrophils are crucial for assessing UC remission.
• Current histological assessment can be subjective and time-consuming.

Purpose of the Study:

• To develop and validate an artificial intelligence (AI) system for quantifying and localizing neutrophils in UC biopsy specimens.
• To assess the AI system's performance against expert pathologists.
• To evaluate the AI system's ability to predict clinical relapse in UC patients.

Main Methods:

• An AI system using semantic segmentation and object detection was developed for hematoxylin and eosin-stained whole slide images.
• The system identified neutrophils in the epithelium and lamina propria, predicting Nancy Histological Index and PICaSSO Histologic Remission Index components.
• Performance was evaluated against pathologists, and prediction of clinical relapse (partial Mayo score ≥3) was assessed.

Main Results:

• The AI model achieved high precision (0.77), recall (0.81), and F-score (0.79) in neutrophil identification.
• AI-derived histological scores correlated positively with pathologist diagnoses (Spearman's ρ = 0.68-0.80).
• Higher AI scores (PICaSSO, Nancy Index) were associated with increased UC relapse risk (HR 3.2-5.0).

Conclusions:

• The AI system accurately quantifies and localizes neutrophils in UC biopsies.
• This AI tool enhances histological assessment for UC remission.
• The system provides valuable prognostic information for stratifying clinical relapse risk in UC.