Artificial Intelligence: A Future Tool in the Pathologist’s Toolbox

Literature on AI algorithms developed in the context of IBD histology is still limited. Moreover, to date implementation into daily clinical practice has not been reported. Here I offer a brief literature overview and discuss existing challenges.

In one of the first papers on AI in IBD histology, an algorithm was presented to quantify eosinophilic granulocytes [1]. Such an algorithm may contribute to research into the role of eosinophils in IBD. Several groups have now published algorithms that provide histological disease activity scores [2–4]. These algorithms, when completely validated, may find direct use in clinical practice as disease activity scores, such as the Nancy Index, are part of standard pathology reporting in some institutes. Another algorithm (paper under review at the time of writing) appears able to predict, based on histology alone, whether the endoscopic appearance matches Ulcerative Colitis or Crohn’s Disease [5].

The algorithms discussed above perform a single task with limited outcome options, whereas in clinical practice pathologists often perform multiple tasks with more possible outcomes. One example is the broad differential diagnosis that we encounter at first presentation. There are currently no published algorithms that address such a clinical situation. Another example is that of dysplasia detection in IBD. Although there is an algorithm for the different conventional types of dysplasia [6], the recently described non-conventional types of dysplasia in the context of IBD [7] would not be diagnosed by this algorithm.

The role of AI in IBD histology is currently limited to the research setting. As many pathology institutes have not completed the transition to digital pathology, this may not change for some time. However, the future of AI is both exciting and promising. Due to its multidisciplinary and collaborative nature, the ECCO Community is the ideal forum to discuss as well as to guide the future of AI in IBD histology.

References

1. Vande Casteele N, Leighton JA, Pasha SF, et al. Utilizing deep learning to analyze whole slide images of colonic biopsies for associations between eosinophil density and clinicopathologic features in active ulcerative colitis. Inflamm Bowel Dis 2022;28(4):539–46.
2. Peyrin-Biroulet L, Adsul S, Dehmeshki J, et al. An artificial intelligence-driven scoring system to measure histological disease activity in ulcerative colitis. J Crohns Colitis 2022;16(Suppl 1):i105.
3. Najdawi F, Sucipto K, Mistry P, et al. Artificial intelligence enables quantitative assessment of ulcerative colitis histology. Mod Pathol 2023:100124. Online ahead of print.
4. Gui X, Bazarova A, Del Amor R, et al. PICaSSO Histologic Remission Index (PHRI) in ulcerative colitis: development of a novel simplified histological score for monitoring mucosal healing and predicting clinical outcomes and its applicability in an artificial intelligence system. Gut 2022;71(5):889–98.
5. Mokhtari R, Hamidinekoo A, Sutton D, et al. Interpretable histopathology-based prediction of disease relevant features in inflammatory bowel disease biopsies using weakly-supervised deep learning. Accepted to Medical Imaging with Deep Learning 2023.
6. Byeon S-J, Park J, Cho YA, et al. Automated histological classification for digital pathology images of colonoscopy specimen via deep learning. Sci Rep 2022;12:12804.
7. Choi W-T, Yozu M, Miller GC, et al. Nonconventional dysplasia in patients with inflammatory bowel disease and colorectal carcinoma: a multicenter clinicopathologic study. Mod Pathol 2020;33:933–43.