Carpe Algorismum in gastroenterology clinics

The Irish Society of Gastroenterology Summer Meeting 2025 heard from Prof Michael Byrne, Clinical Professor of Medicine at the Division of Gastroenterology, Vancouver General Hospital, Canada, who delivered a talk titled ‘Carpe Algorismum: Practical implementation of AI [artificial intelligence] into clinical practice and clinical trials in gastroenterology’.

“‘AI’ is an all-encompassing term and it’s a loosely used term,” Prof Byrne told the conference. “Some appropriate subdivisions would be machine learning, which is what we are doing in the imaging space and the advent of deep learning, which is essentially a set of neural networks. And of course there is currently much hype, but also much promise, for generative AI. We all use it every day when we do a Google search, and how we are going to use it in medicine is exciting, but also fraught with some issues.”

Prof Byrne provided an outline of the many processes that are involved in developing an AI tool, but also cautioned that “in the GI [gastrointestinal] space, an algorithm is not the same as a product”.

“The gap between developing an algorithm with some data and having something that can be used safely and effectively in your practice is absolutely huge… on the business side of AI, I have seen that significant gap,” he said.

“Also, your solution needs to add value,” Prof Byrne continued. “No matter how cool it is, if it is not going to provide value, it’s not going to be adopted.”

To make the transition from an idea to a product, he advised that important considerations include clinical validation, regulatory barriers, having security built-in, creating an effective user interface, and market considerations, among many others. “It’s an incredibly long and expensive journey with many phases, but also a very exciting one,” he said.

Radiology has so far led the way in the use of AI in medicine, and Prof Byrne provided examples of how to handle raw data, ultrasound AI, and using AI for polyp detection, with the goal of precision surveillance. He used an example of the potential value of AI in inflammatory bowel disease and how it has the potential to detect subtle changes in colitis, for example. “It is really hard for the human eye to pick up that subtle, or sometimes not-so-subtle, change,” he said. “Also, our scoring systems to date have limitations and they are not fully capturing [some clinical]  improvement.”

He provided an overview of research on AI in endoscopy, which includes heat maps to detect disease activity in the bowel and which can capture the areas of the colon that may be unreadable, as well as detecting the presence of ulcers, erosions, or vascular pattern change, bleeding, erythema, and other clinical manifestations. “With that kind of deep-level approach, maybe we can change how we do things.”

Prof Byrne concluded: “AI in healthcare has multiple applications and if we harness the power of it well, it’s not a case of ‘if’, it’s a case of ‘how’ and how quickly and how iteratively this technology will be in our lives.”