Iterative Scopes DDW Presentation Shows Accuracy of Automated Endoscopic Disease Scoring Model, With Potential to Accelerate IBD Clinical Trials

Machine LearningHealthcare ManagementNews

By AIT News Desk On May 25, 2022

Innovative collaboration with Eli Lilly demonstrates high-quality performance of machine learning model for identifying endoscopic disease severity in ulcerative colitis

Iterative Scopes, a pioneer in precision medicine technologies for gastroenterology, announced that data presented at Digestive Disease Week (DDW) 2022 show that the company’s proprietary machine learning model accurately and consistently predicts scoring of endoscopic disease severity in ulcerative colitis (UC), a common type of inflammatory bowel disease (IBD).

Latest Aithority Insights: Komodo Health Platform Accelerates Real-World Evidence For Life Sciences

“Iterative Scopes’ machine learning model is a positive step in addressing these challenges and is expected to positively impact both clinical trials for new therapies and directly improve the care of patients.”

Data was presented in a poster session (No. Mo1639) entitled “Development of a novel ulcerative colitis (UC) endoscopic activity prediction model using machine learning (ML)” on May 23 from 12:30-1:30 PM PDT. In their research, investigators found that analyses involving the Iterative Scopes machine learning model had better agreement rates with a baseline reference standard than multiple human interpreters typically do with each other. The results indicate that this approach to automated endoscopic assessment has potential to improve the quality and efficiency of IBD clinical trials.

The data is drawn from an innovative partnership between Iterative Scopes and Eli Lilly and Company aimed at applying machine learning models to improve the clinical trial process for developing medications focused on IBD and, specifically in this research, on ulcerative colitis. When designing UC clinical trials for new drugs, endoscopic disease activity is routinely evaluated by human readers using established scoring systems, such as the endoscopic Mayo Score (eMS). Human interpretation of endoscopic disease scoring can be highly subjective and inconsistent,¹ however, which may negatively impact the timelines and success of clinical trials. Machine learning models that automatically score endoscopic disease severity in UC have the potential to address some of these shortcomings.

The data presented at DDW showed that the Iterative Scopes machine learning model agreed with a baseline of human interpretation in 89% of cases when differentiating between active and inactive endoscopic disease severity. The model also identified endoscopic healing with 95% agreement and was in accordance with the baseline human interpretation in 85% of cases of severe disease.¹

“Despite recognition of the importance of mucosal healing in the management of ulcerative colitis, there have been a variety of challenges to the scoring systems that are used in clinical trials and in clinical practice,” said David T. Rubin, MD, Chief of the Section of Gastroenterology, Hepatology, and Nutrition at the University of Chicago and Chair of the Iterative Scopes Advisory Board. He presented “Development of a novel ulcerative colitis endoscopic activity prediction model using machine learning” as a poster of distinction at DDW on May 23. “Iterative Scopes’ machine learning model is a positive step in addressing these challenges and is expected to positively impact both clinical trials for new therapies and directly improve the care of patients.”

In their collaboration, the investigators assessed the performance of Iterative Scopes automated endoscopic disease severity scoring in UC when compared to a baseline “ground truth,” which was human interpretation. In data science, the ground truth refers to the score against which the model output is evaluated.

Browse The Complete News About Aithority : Clearlake Capital-Backed Cornerstone Completes Acquisition of EdCast

Data from the same research was featured in a separate oral presentation (No.278) at DDW entitled “Can a single central reader provide a reliable ground truth for training a machine learning model that predicts endoscopic disease activity in ulcerative colitis?” held on May 21 from 5:15-5:30 pm PDT. In the presentation, Klaus Gottlieb, MD, JD, Eli Lilly and Company, described the clinical methods that informed the research. He explained the novel approach used to develop and evaluate the machine learning algorithm, including augmentation of training of the machine learning algorithm and the parameters of the ground truth against which the model performance was evaluated.

“This data shows significant distinction between active and inactive disease, with the potential to possibly assess key levels of endoscopic activity in the future,” said Jean-Frederic Colombel, MD, Professor of Medicine, Mount Sinai Medical Center, NY, and Director of the Susan & Leonard Feinstein IBD Clinical Center and The Leona M. & Harry B. Helmsley IBD Research Center, Icahn School of Medicine at Mount Sinai, who co-authored this research. “These promising results support the potential of Iterative Scopes machine learning models to enhance accuracy and consistency of endoscopic scores and could potentially help to speed up clinical trial completion in UC.”

IBD is a category of gastrointestinal diseases characterized by chronic inflammation in the digestive tract and includes UC and Crohn’s disease (CD). Endoscopic disease activity in IBD patients is emerging as a key therapeutic metric that clinicians use to assess improvement or decline in IBD patients’ condition.

Iterative Scopes was founded in 2017 as a spin out of the Massachusetts Institute of Technology (MIT) by Jonathan Ng, MBBS, a physician-entrepreneur, who developed the company’s foundational concepts while he was in school at MIT and Harvard. In December 2021, the company and its investors closed a $150 million Series B financing, which attracted a roster of A-list venture capitalists, big pharmaceutical companies’ venture arms, and individual leaders in healthcare.

[To share your insights with us, please write to sghosh@martechseries.com]