Hebat Kamal, MD: AI Models for Detecting Crohn's Disease Fistulas

News
Video

New data show a promising machine learning tool could help specialists detect fistula risks and provide more tailored, timely care in children.

A machine learning model designed to recognize enteroenteric fistulas could help provide timelier and more accurate diagnoses in patients with Crohn’s disease, according to new research.

In data presented in an abstract at the North American Society for Pediatric Gastroenterology, Hepatology and Nutrition (NASPGHAN) 2023 Annual Meeting in San Diego this week, a team of Florida-based investigators reported that an artificial intelligence (AI) tool was trained to help radiologists and gastroenterologists more accurately and expeditiously predict future complications due to enteroenteric fistulas.

“The finding of an enteroenteric fistula necessitates aggressive therapy to avoid disease complications including abscess formation,” investigators wrote. “With artificial intelligence already being utilized in computing bowel wall thickness and detecting intestinal narrowing in patients with CD, we believe a similar approach using our model can serve as a beneficial tool to detect internal fistulas in CD patients and improve patient outcomes.”

In an interview with HCPLive during NASPGHAN 2023, study author Hebat Kamal, MD, a third-year gastroenterology fellow with the department of pediatrics in the University of Florida College of Medicine, discussed the impact her team’s findings may have on gastric disease imaging and, ultimately, targeted treatment strategies.

“We're trying to implement a model, or an artificially intelligent or neural network, that can detect some specific findings for IBD patients on imaging— whether in MRI imaging, which is the focus for our project, or hopefully, implemented in more imaging like CAT scans and things like that,” Kamal said.

Enteroenteric fistulas are a very serious complication that can be identified in imaging versus scoping.

“Especially, it's very important for non-abdominal experts who can miss these fistulas to be picked up and helped by artificial intelligence,” Kamal said. “We're trying to measure with specific coefficients and metrics used to see how well an image segmentation which we're trying to do matches the model that we have created.”

While the coefficients of the model are “very favorable” based on these findings, Kamal said more work in refining and implementing transformations to the metrics will be needed to optimize the model’s utility in real-world patient imaging.

“In our practice, and in our IBD kids, I think this will be very important for clinicians,” Kamal said, “because we're basically telling them, 'If you have this fistula and you miss it, you can pretty much under treat patients, while if you pick it early on, with early intervention strategies,and escalating therapy when necessery, you can avoid severe disease complications and aggressive disease-like progression’.”

Reference

Kamal H, Bidani S, Desaraju S, Patel V, et al. Using Artificial Intelligence for Identifying Enteroenteric Fistulas on Cross-Sectional Imaging in Patients with Crohn’s Disease. Paper presented at: North American Society for Pediatric Gastroenterology, Hepatology and Nutrition 2023 Annual Meeting; October 4 - 7; San Diego, CA. Accessed October 7, 2023.

Related Videos
Robert Wood, MD | Credit: LinkedIn
Marla Dubinsky, MD | Credit: LinkedIn
Michelle Kittleson, MD, PhD | Credit: X.com
Marla Dubinsky, MD | Credit: LinkedIn
Ramin Tadayoni, MD, PhD | Image Credit: ASRS
© 2024 MJH Life Sciences

All rights reserved.