By Alex Houlbrook, Senior Director, Business Development; Kasia Moscicka, Head, Feasibility; Jenna Zalud, Manager, Feasibility; and Alex Zorin, Program Manager
Artificial intelligence (AI) was a major topic at Digestive Disease Week (DDW) 2025, and it was even referred to as “the new electricity.” Just as in the early days of electricity, the potential is high, and everyone has their ideas of how it could work.
So, what’s out there poised to make an impact right now?
1. Predicting Inflammatory Bowel Disease Flares
AI has the opportunity to realize what was previously just a dream for many patients, who have lived their lives at the mercy of unpredictable but deeply uncomfortable flares.
Using deep learning analysis of wearable data and patients undergoing flares, researchers found significant alterations in heart rate variability (HRV), longitudinal heart rate (HR), resting heart rate (RHR), steps, and oxygenation up to seven weeks before inflammatory and symptomatic flares.[1] The potential for patients is considerable, from the simple planning of daily life around flares to the advanced planning of treatment or prophylaxis.
Many people incorporate wearable devices into their daily routines, constantly collecting physiological and activity data. As these devices become more sophisticated and collect more information, there is the likelihood of further advances in predictions.
Source: Hirten RP, Danieletto M, Sanchez-Mayor M, et al.
2. Addressing Bias with AI
AI can help correct inherent biases in existing medical scoring systems.
For example, a new AI model for knee osteoarthritis, trained on diverse X-rays and patient pain scores, proved more equitable in guiding total knee replacement (TKR) allocations across different racial groups, unlike the traditional Kellgren-Lawrence score, which showed bias against Black patients.[2],[3]
To maximize the impartiality of AI, we need to plan carefully and collect data in a standardized way where possible.
3. Precision Medicine in IBD
Precision medicine is rapidly advancing in inflammatory bowel disease (IBD) to achieve more individualized treatment approaches. To treat with precision, work is being done to profile and measure with precision. As subpopulations are defined, researchers can develop more targeted treatments.
Research is delving into the molecular roots underlying IBD phenotypes using tissue transcriptomics, microbiome profiling, serum proteomics, and cellular profiling. Meanwhile, AI is serving to model interactions and create subgroups of IBD patients, including:[4]
- Patient-Level Trajectories
Analyzing individual symptom-response trajectories in patients enables the identification of subpopulations that share similar responses to treatment. Rapid responders (superfast or fast) tend to achieve better disease control, highlighting the utility of early response assessment.[5] This information can guide treatment optimization, therapy changes, and the use of combination therapies. - Advanced Imaging and Monitoring
Modern advanced imaging and AI-assisted disease monitoring, such as video colonoscopy for ulcerative colitis (UC), are being developed to better understand and track IBD phenotypes.[4] - Radiomics
This quantitative approach uses advanced mathematical analysis of medical images to extract textural information and define phenotypes, thereby enabling greater reliance on cross-sectional imaging and contributing to the development of automated precision medicine. Challenges currently remain with image quality and variations between equipment.
4. AI-Driven Endoscopic Scoring
Everyone working in gastroenterology is familiar with the Mayo Endoscopic Score (MES); however, using AI to analyze images and generate a cumulative disease score allowed both ustekinumab and ozanimod to demonstrate a greater healing effect than with MES alone.[6]
The Cumulative Disease Score (CDS) reflects differences in endoscopic activity that aren’t measurable using MES and is more strongly associated with clinical remission in UC. Combining endoscopic disease maximum intensity (by MES) with total burden (by CDS) could serve as an AI-enabled endpoint defining sufficient mucosal healing for appropriate symptomatic control.
Researchers at DDW presented suggestions for a new model, EASE-CD, which combines some of these techniques and removes more ambiguous terms, such as “affected area,” and hard-to-evaluate aspects, such as “narrowing,” focusing instead on a new scoring index.[7]
Source: Ma C, Khanna R, Maguire BR, et al.
5. Analyzing Huge Datasets
Whether it’s at the hospital level, for a clinical study, or for a global trend analysis, a large amount of data is available; however, before the advent of AI, it was challenging to draw conclusions quickly.
Electronic Medical Record (EMR) data is a great example; by training models on these large datasets, researchers are now close to being able to take a patient’s EMR, run it through AI, and have the physician rapidly provided some suggested treatment options, be it approved drugs, clinical studies, or lifestyle changes.
6. The “Gray Zone” and Human Oversight
For AI models to be truly useful, they must assist in “gray zone” cases where clinical diagnosis is not immediately obvious, rather than just confirming clear-cut cases. This is the next step that is to be worked on to further benefit from these models.
AI isn’t coming to GI—it’s already here. For maximum adoption, however, both physicians and patients must have a high level of trust in the rationale and decisions these models generate. We hope to see clinical guidelines and ethical standards developed in parallel with the evolution of these new tools to ensure their integrity is preserved.
References
[1] Hirten RP, Danieletto M, Sanchez-Mayor M, et al. Physiological Data Collected from Wearable Devices Identify and Predict Inflammatory Bowel Disease Flares. Gastroenterology. Published online January 1, 2025. doi:10.1053/j.gastro.2024.12.024
[2] Obermeyer Z, Powers B, Vogeli C, Mullainathan S. Dissecting racial bias in an algorithm used to manage the health of populations. Science. 2019;366(6464):447-453. doi:10.1126/science.aax2342
[3] Pierson E, Cutler DM, Leskovec J, Mullainathan S, Obermeyer Z. An algorithmic approach to reducing unexplained pain disparities in underserved populations. Nature Medicine. 2021;27(1):136-140. doi:10.1038/s41591-020-01192-7
[4] Iacucci M, Santacroce G, Zammarchi I, et al. Artificial intelligence and endo-histo-omics: new dimensions of precision endoscopy and histology in inflammatory bowel disease. The Lancet Gastroenterology & Hepatology. 2024;9(8):758-772. doi:10.1016/s2468-1253(24)00053-0
[5] Schreiber SW, Mueller-Breckenridge A, Cheng J, et al. P0694 Clinical response trajectories identify distinct subpopulations of ulcerative colitis patients and are linked to disease outcome across different therapy classes. Patient level analysis of 2378 participants from 6 randomized controlled trials in moderate-severe UC representing 8 different mechanisms of action. Journal of Crohn’s and Colitis. 2025;19(Supplement_1):i1357-i1358. doi:10.1093/ecco-jcc/jjae190.0868
[6] Stidham RW, Cai L, Cheng S, et al. Using computer vision to improve endoscopic disease quantification in therapeutic clinical trials of ulcerative colitis. Gastroenterology. 2023;166(1):155-167.e2. doi:10.1053/j.gastro.2023.09.049
[7] Ma C, Khanna R, Maguire BR, et al. P0390 Development, internal, and external validation of a novel endoscopic activity score for evaluation of Crohn’s disease (EASE-CD). Journal of Crohn’s and Colitis. 2025;19(Supplement_1):i881-i882. doi:10.1093/ecco-jcc/jjae190.0564
