Pathologist-level diagnosis of ulcerative colitis inflammatory activity level using an automated histological grading...

Area of Science:

• Pathology
• Artificial Intelligence
• Gastroenterology

Background:

• Inflammatory bowel disease (IBD), including ulcerative colitis (UC), is a growing global health concern.
• Accurate histologic grading of UC inflammatory activity is crucial for patient management.
• Computational methods for IBD diagnosis from pathology images remain underexplored.

Purpose of the Study:

• To establish an artificial intelligence-assisted diagnostic system for the histologic grading of inflammatory activity in ulcerative colitis (UC).
• To develop and validate a deep learning model for analyzing whole-slide images (WSIs) of UC pathology.

Main Methods:

• A deep learning (DL) model using ResNet50 and a multi-instance learning (MIL) approach with self-attention was developed.
• The model was trained and validated on 603 UC WSIs from Nanjing Drum Tower Hospital (internal) and 212 UC WSIs from Zhujiang Hospital (external).
• The system aggregated image patch features to represent entire WSIs for predicting inflammatory activity levels.

Main Results:

• The AI system demonstrated high accuracy in distinguishing the presence/absence of inflammatory activity, with AUCs of 0.863 (internal) and 0.947 (external).
• For grading different levels of activity, the average Macro-AUC was 0.827 (internal) and 0.908 (external), and Micro-AUC was 0.816 (internal) and 0.898 (external).
• The model achieved sensitivity and specificity values exceeding 0.80 in both internal and external test sets.

Conclusions:

• The developed AI algorithm achieved diagnostic proficiency comparable to a pathologist with 5 years of experience.
• The AI system outperformed other existing MIL algorithms in grading UC inflammatory activity.
• This AI-assisted system shows promise for improving the accuracy and efficiency of UC histologic grading.