‘Keep our hands on the wheel’: Caution, supervision urged as AI enters the IBD space
Click Here to Manage Email Alerts
DENVER — Although the use of artificial intelligence for inflammatory bowel disease can reasonably replicate expert judgment and predict outcomes, clinicians must retain the ability to intervene and “take control” when the AI is wrong.
“I don’t have to tell any of you that artificial intelligence is everywhere right now, doing some truly amazing tasks and starting to integrate itself seamlessly into our personal lives,” Ryan W. Stidham, MD, MS, associate professor of medicine and computational medicine and bioinformatics at the University of Michigan, told attendees at the Crohn’s and Colitis Congress. “We have machines that are doing simple tasks using vision technologies, that are able to see and comprehend the visual world the way we do, and then act to do simple things, like robots that vacuum for you, as well as increasingly complex tasks, such as autonomous driving.”
Source: Healio
Stidham added: “We are beginning to see AI technologies make their way into our IBD practice, and we have seen tremendous progress in these tools in our space. Just about all of us are familiar with the automated polyp detection technology that has been brought to market.”
In the gastroenterology space, Stidham noted that AI technologies have found a particular niche in replicating expert interpretation of imaging, notably in endoscopic and histologic scoring, as well as reading cross-sectional imaging and capsule endoscopy.
“There are so many implications for artificial intelligence, and it will be coming to your clinic soon so you have to get ready,” Stidham told attendees. “In the coming 5 years, we are going to see these artificial intelligence methodologies implemented to help us make decisions, not just measure disease.”
To manage a booming IBD population, not just in North America but around the world, gastroenterologists will need AI technologies to handle time-consuming tasks, such as reviewing laboratory results or imaging, while the clinician interfaces with the patient. However, even Stidham, who sees the tremendous advantage of AI technologies, is not without reservations about the risks of allowing the technology to go unchecked by the clinician.
“These networks are not magic — it’s just math,” he said. A poorly designed AI system can misdiagnose, and software that incorporates cultural bias from the data sets given to it will simply perpetuate those biases. Clinicians will need to remain cognizant of the factors that go into AI decision-making and clinical predictions.
“As AI becomes available, and it will be pushed into our clinics, we have to keep our hands on the wheel; we cannot lose sight of who is in control here,” Stidham said. “Unsupervised AI can make some catastrophic mistakes. Even when you think that you’re paying attention to the AI, if you don’t, if you take your eyes off the road for one second, something bad can happen. We want to make sure that this does not happen for our patients as we integrate these tools.”
Stidham proposed that, in the not-too-distant future, selecting AI models may be similar to evaluating nutritional labels on food products at the grocery store.
“We should all have AI model transparency facts,” he said. “Who made the model? How are you supposed to use it? What population is it used in? How is it developed? We should have something like this for any AI model that we use, and if your hospital starts using these models, I think that you should demand it provide this information.”
Stidham acknowledges that there is a bright future in AI for gastroenterology, with these technologies standardizing practice, elevating the level of care for all patients and offloading a sizeable portion of labor from the clinician. However, with that time savings, implementation of AI will also require thorough attention.
“We have to be careful how much we trust it; we should always be at least a little suspicious and verify what the AI is doing,” Stidham said. “I think that if we use AI in the right ways, we can not only preserve but improve the patient-caregiver relationship, because AI is going to give us back a lot of time so that we can do what we love doing.”
Collapse
Stidham R. Presentation Sp147: Keynote presentation: Don't fall asleep at the wheel: When to overrule AI in IBD. Presented at: Crohn’s and Colitis Congress; Jan.19-21, 2023; Denver (hybrid meeting).
Read more about
Click Here to Manage Email Alerts