For years, artificial intelligence in gastroenterology has been synonymous with polyp detection, but some leaders say that framing misses the bigger transformation underway.
The next wave of disruption will instead be from predictive analytics, chronic disease management and AI-driven clinical decision support. These areas could fundamentally reshape how gastroenterologists manage inflammatory bowel disease, stratify oncology risk and interpret patient data at scale.
As these tools move from pilot programs to everyday practice, the skillset required of new physicians will shift as well.
Neil Parikh, MD, a gastroenterologist with Farmington, Conn.-based Connecticut GI, joined Becker’s to discuss how AI is poised to change GI training and why awareness, data fluency and clinical judgment may become just as critical as procedural expertise.
Question: If AI becomes a physician partner, what skills do you think the next generation of GIs will need that maybe weren’t emphasized five or 10 years ago?
Editor’s note: This interview was edited lightly for clarity.
Dr. Neil Parikh: I think the first thing I’ve been harping on is awareness. Gastroenterologists in training need to understand that artificial intelligence will be part of their clinical practice, and that it’s more than just polyp detection. Right now, I don’t think there’s great awareness. We use “AI” as a buzzword, but we don’t always know what it means for GI specifically. So that’s step one.
Second, we’re going to need to learn how to manage all this data. AI will generate a lot more information than we’re used to — not just patient messages, but computerized, algorithmic scores. We’ll need workflows for how those scores get incorporated. If you’re already managing, say, 1,000 IBD patients and you get 1,000 scores every week, that system isn’t going to work as-is. We’ll need new processes. Trainees will need to learn not only what these scores mean, but how to interpret their severity. A lot of that may be automated, but clinical understanding will still matter.
On the polyp detection side, you still need human judgment. The role of AI right now is to help ensure you’re not missing anything that shouldn’t be there. That’s how I explain it to my fellows: it will flag anything that wouldn’t appear in a normal colon. If what it flags isn’t actually a polyp, you still need the expertise and judgment to know not to remove it.
That judgment will become even more important as more automated tools enter the workflow.
At some point, I wonder if medical education will shift, too. We’ve historically spent the first two years of medical school memorizing large textbooks — organic chemistry, and so on — then moved into clinical training, with residency being even more clinical experience.
I think we’re shifting toward more clinical exposure because with large language models 0 — GPTs, OpenEvidence — you may not need as much pure recall as before. But you will need the ability to take that information and translate it into sound clinical decisions and execution.
Q: Is there a new generation of gastroenterologists that is unprepared for this shift?
NP: I think they’re getting more prepared now. If you asked me a year ago, we weren’t very prepared, and we were underusing resources that were readily available — which puts you at a disadvantage in real time.
If you have one trainee using large language models and one not, it’s going to be more efficient and potentially more informationally accurate for the trainee who’s using the LLM.
