On Oct. 23, Iterative Health released research findings that demonstrated the effectiveness of artificial intelligence for use in the treatment and management of inflammatory bowel disease and in improving adenoma detection during colonoscopies.
In May, Iterative Health announced plans to debut its own artificial-intelligence polyp detection solution for use in colonoscopies, in partnership with software provider Provation.
While research is increasingly showing the promise of AI-assisted colonoscopies, with Medtronic's GI Genius increasing adenoma detection rates by 14.4%, there are still relatively few AI-assisted options available on the market.
Currently, Skout by Iterative Health and GI Genius by Medtronic are making the biggest waves in the industry.
Aasma Shaukat, MD, a gastroenterologist at New York City-based NYU Langone Health and a member of the Iterative Health scientific board, co-authored the study "Endoscopist and procedure-level factors associated with increased adenoma detection with the use of a computer-aided detection device," which was awarded an outstanding research award by the American College of Gastroenterology.
Dr. Shaukat spoke with Becker's about the benefits of AI-assisted colonoscopies and how new AI technologies are changing the field of gastroenterology for the better.
Question: How are AI-assisted colonoscopy devices changing patient care for the better?
Dr. Aasma Shaukat: AI devices provide a unique opportunity in the realm of endoscopy and colonoscopy to help us improve the quality of our exams. We know that endoscopy is very operator dependent. In terms of colonoscopy that means we can miss very important things, undermining the effectiveness of colonoscopy. Having tools available to us that give us confidence in our exam and also reassure us that we haven't missed anything provides for better care and better patient outcomes.
Q: Are you seeing any concerns from patients when it comes to the implementation of AI-assisted colonoscopy technology?
AS: Patients are always concerned about if endoscopists do anything less than they normally would do. I can understand, because I would be concerned about that too. We provide them reassurance that when we use AI, particularly for polyp detection with colonoscopy, the endoscopist is still doing an extremely careful and thorough examination. The AI is an addition, it's a supplementary tool that ensures there aren't areas missed by the endoscopist not paying attention. The AI directs the endoscopist by putting boxes around potential polyps to emphasize the area. AI is actually supplementing what the endoscopist is doing rather than replacing or substituting. That really assures patients. Once they understand that, I have patients who are specifically asking if I can use AI during the colonoscopy because they've heard it will help with polyp detection, which provides them confidence that they've had a thorough and high-quality exam.
Q: Do you think practices will be able to afford AI-assisted colonoscopy technology? What about patients in rural/underserved areas?
AS: The first step is to develop tools that can help our colonoscopy and then understand that they are effective in doing what we want. The second step is making them available, and then making them accessible and helping with that implementation piece. These technologies are new; the first system was approved just last year. We are in the very early stages of applying them outside of clinical trials into real-world settings. As we do that, there are a lot of lessons we are learning. At the moment, there is no additional reimbursement for AI. Obviously, the technology does add cost to endoscopy units. One of the first decision points for systems and endoscopy units is to figure out if this makes sense for their practice. Some of the reasons it may make sense are that even though it's an upfront cost, benefits include fewer patients that need to come back in short intervals and fewer missed cancers, which obviously has implications. Those are the balance or the tipping points. The other aspect is if AI helps detect more adenomas that we remove then the idea is colonoscopies that would otherwise just be a diagnostic exam will now be upcoded because they would now be colonoscopies with a polypectomy. There are some ways to capture funds back, but what would really move the needle is if we were able to get an additional modifier so that endoscopists can get reimbursed when we use AI. For that we will have to demonstrate that there's long-term clinical impacts, but I think there is movement in that direction because there are now several health companies with AI-enabled software helping in this effort.
Q: What payer battles are holding clinicians back from expanding AI-assisted colonoscopies even more?
AS: Obviously, it's an out-of-pocket cost, so for very small practices we've heard reimbursement has been going down and adding costs is challenging. However, there are models that show that endoscopy units can still, even without additional reimbursement, break even. If a unit thinks that the technology is applicable to them and helpful, it might be worth adapting. IBD codes are a long process where additional effort has to be demonstrated, and that will take some time and more data to accumulate. And we've seen that with other technologies, so I'm hoping AI can go that way too so we can truly make it mainstream and widespread.
Q: Do you think more medtech companies will start to roll out AI-assisted colonoscopy devices?
AS: It takes a long time to develop software and algorithms that are precise and also sensitive and specific. We also want them tested in clinical trials. With Skout, we did conduct a very thorough clinical trial with 1,400 patients across five centers to get to the end points of clinical benefit. Those kinds of studies take time and so far, there are three FDA-approved AI systems and there are more developed around the world, but I am not sure where they are in terms of their approval processes.
Q: What other developments are you keeping an eye on in GI right now?
AS: AI-assisted polyp detection is the first step. The additional thing that AI-enabled technology can do for us is categorize polyp histology, which would be a big benefit. There are a lot of polyps that we endoscopically recognise that do not increase risk of colon cancer, that we could really just leave in place; however, because of legal risks we still end up taking those polyps out. If AI could tell us the histology prediction with enough confidence, we could leave these nonconsequential polyps in the colon. That would obviously have an impact on our field by reducing costs. The other application is in patients with IBD where patients might have colitis. Right now, we tend to score colitis using an endoscopic scoring system and it's pretty subjective. But that score is used to decide what type of therapy or treatment patients receive. Having AI tell us what that scoring should be and give us an objective assessment will really help us clinically because we can reduce variation and more confidently assign patients to a specific treatment or even refer them for clinical trials. There are other enhancements we're looking at as clinicians for AI to solve, like sizing of polyps. Again, that tends to be subjective and difficult to do. I think the applications of AI are going to continue to improve and benefit our practices and patient outcomes.
Q: Is there anything you want to expand on?
AS: I want to place AI in the realm of colonoscopy quality. The quality of colonoscopy is extremely important, and there's multiple ways to improve the quality of colonoscopy. AI definitely seems to be an effective tool. The idea is every endoscopist in the healthcare system should think about what's available to them and think about whether AI could benefit their unit and then economic modeling would come later. If the technology is there it behooves us to think if this will help with patient care, with the understanding that every day we have to keep the quality of the colonoscopies that we do front and center of our practice.