Use of Artificial Intelligence for Duodenal Biopsy for Celiac Disease: Developing a Prediction Model for Histopathology Diagnosis.

BACKGROUND AND AIMS: Duodenal biopsy combined with tissue transglutaminase (tTG-IgA) antibodies is the gold standard for celiac disease diagnosis. Issues with biopsy include artifacts, poorly oriented specimens, inter-observer reliability, and pathologist shortages. Artificial intelligence may address these challenges. We aimed to determine if deep learning-based image analysis can be used on duodenal biopsies to accurately diagnose celiac disease. METHODS: The study population was individuals with a duodenal biopsy and tTG-IgA antibody from 2008-2021. Inclusion criteria were adults ≥18 years. Exclusion criteria were those following a gluten-free diet. Patients were grouped into 5 categories: normal, celiac disease, increased intraepithelial lymphocytes (IELs) only, IELs with a positive tTG-IgA antibody, and seronegative villous atrophy. Whole-slide images were obtained. Clustering-constrained attention multiple-instance learning was the deep learning algorithm used. The dataset was split into training, validation, and testing splits with 10-fold cross validation. RESULTS: A total of 2,406 whole-slide images from >2100 patients were included: 1290 normal, 414 celiac disease, 450 IELs only, 89 IELs with a positive TTG-IgA, and 163 with seronegative villous atrophy. For all categories, the average AUC was 0.872 with an accuracy of 74%. The model reliably differentiated celiac disease from normal with average AUC 0.992 and accuracy 97%. To distinguish celiac disease from seronegative villous atrophy the average AUC was 0.810 with accuracy 76%. CONCLUSIONS: This deep learning model can discriminate between celiac disease and normal histology. It can also separate patients into 5 classifications, representing a promising tool that would benefit from further optimization to improve diagnostic accuracy, reduce the manual pathology review burden, and enhance resource management.