AI-Assisted Endoscopy Module Shows Limited Impact on Serrated Polyp Detection
PHILADELPHIA—A recent study presented at the ACG Annual Scientific Meeting unveiled modest results from the use of an AI-assisted endoscopy module in colonoscopy procedures. Conducted by Dr Rajesh N. Keswani and his team at Northwestern Medicine, the research aimed to evaluate the effectiveness of the GI Genius (Medtronic) module in enhancing polyp detection during colonoscopies for patients who had previously tested positive through stool-based methods.
The investigation occurred at a single urban academic centre, utilising four computer-aided detection (CADe) units across 12 endoscopy rooms. In this setup, a group of 51 providers performed colonoscopies over six months, rotating between procedures with and without the CADe system, based on availability. A total of 15,719 colonoscopies were conducted during this time, with 194 patients referred following a positive result from a multitarget stool DNA test (MT-sDNA, known as Cologuard by Exact Sciences) and 82 patients following a positive fecal immunological test (OC-Auto FIT by Polymedco).
The study concentrated on determining the CADe system's influence on several key detection rates: adenoma detection rate (ADR), serrated polyp detection rate, and neoplasia detection rate (NDR). Results indicated that the CADe system only marginally improved neoplasia detection (74.6% with CADe vs. 67.9% without, with a P-value of .4) and adenoma detection (70.1% with CADe vs. 60.3% without, with a P-value of .2). It did not significantly impact serrated polyp detection rates (23.9% with CADe vs. 24.9% without).
Dr Keswani noted, "In our single center analysis, AI had a non-significant improvement in the detection of adenomas, approximately by 10%, but had no impact on serrated polyp detection." The study was prompted by existing knowledge that computer-aided polyp detection tools generally enhance polyp identification in screening and surveillance contexts. However, their effectiveness among higher-risk patients, specifically those with positive stool-based test results, remains insufficiently understood.
This research adds to the understanding of artificial intelligence's role in maintaining high-quality colonoscopy procedures. However, the study reflects the necessity for further investigation into AI's efficacy in detecting advanced neoplasia in patients who have already indicated potential issues through positive stool tests. The findings suggest a modest advancement, highlighting the need for additional research to explore potential improvements and refinements in AI-assisted colonoscopy techniques.
Source: Noah Wire Services