From the AGA Journals

Mucosal exposure device boosts AI-assisted detection of adenomas

The paradigm of adenoma detection is rapidly shifting within the context of screening-related colonoscopy. If one considers the various interventions available to improve one’s adenoma detection rate (ADR), the landscape is vastly different than it was 5-10 years ago. Two established interventions with robust supporting data from randomized controlled trials (RCTs) are computer-aided detection (CADe) platforms such as GI Genius (Medtronic) and distal attachment devices such as Endocuff Vision (Olympus). This RCT by Spadaccini and colleagues tested the intuitive hypothesis that these interventions applied together boost ADR, compared with CADe alone.

Nauzer Forbes, MD, MSc, FASGE, University of Calgary University of Calgary

Dr. Nauzer Forbes

In a patient cohort that was balanced across major colonoscopy indications of primary screening, positive fecal immunochemical testing, surveillance, and diagnosis, ADR was 12% higher in patients receiving colonoscopy with Endocuff Vision and CADe, compared with CADe alone, with a corresponding significant increase in the adenoma per colonoscopy rate of 26%. Detection of advanced adenomas was not significantly different between groups. Detection of serrated lesions was also similar.

Real-world studies of CADe’s effectiveness on ADR are less impressive than efficacy data from trials. Whereas CADe platforms require a significant one-time investment, distal attachment devices represent a small fraction of single procedural costs which then incrementally add up when used over large volumes. More head-to-head studies, cost-effectiveness analyses, and real-world studies are needed to elucidate the best single and/or combination strategies for optimizing ADR. In the meantime, endoscopists should be aware of all evidence-based techniques for ADR improvement, including those that can be incorporated at little to no cost.

Nauzer Forbes, MD, MSc, FASGE, is an associate professor at the University of Calgary (Alta.), where he is the training program director for advanced/therapeutic endoscopy. He is a consultant for and has received speaker’s fees from Pentax Medical and Boston Scientific, is a consultant for AstraZeneca, and has received research funding from Pentax Medical.


 

Performing colonoscopy with a mucosal exposure device and artificial intelligence (AI) software increases detection of adenomas over AI-assisted colonoscopy alone, based on results of a randomized trial.

Using the mucosal exposure device increased adenoma detection rate by 12% without impacting safety or withdrawal time, suggesting that the two approaches have a synergistic effect, reported lead author Marco Spadaccini, MD, of Humanitas University, Pieve Emanuele, Italy, and colleagues.

“Recent advances in AI, deep learning, and computer vision led to implementation of computer-aided detection [CADe] of colorectal polyps,” the investigators wrote in Gastroenterology. “CADe-assisted colonoscopy already proved its efficacy by increasing adenoma detection in randomized parallel and crossover trials. However, such benefit is mostly related to the higher accuracy in spotting lesions already within the visual field, not affecting the amount of mucosa exposed by the endoscopist during the scope withdrawal. Increasing the mucosa exposure represents a complementary strategy to CADe in order to further improve detection of colorectal neoplasia.”

To test their hypothesis, the investigators conducted a randomized trial involving 1,316 subjects undergoing routine colonoscopy at six centers in Italy and Switzerland. Participants were randomized in a 1:1 ratio to undergo colonoscopy with CADe (GI Genius, Medtronic) or CADe plus a mucosal exposure device (Endocuff Vision, Olympus).

The combination approach yielded a 49.6% adenoma detection rate, compared with a 44.0% detection rate for CADe alone (relative risk, 1.12; 95% confidence interval, 1.00-1.26; P = .04). Adding the mucosal exposure device was also associated with a higher number of adenomas detected per colonoscopy. Withdrawal time and rate of unnecessary polypectomies did not differ between groups.

“The benefit of adding [the mucosal exposure device] to AI was expected due to the complementary nature of the interventions,” Dr. Spadaccini and colleagues wrote. “The benefit of [the mucosal exposure device] is limited to increase the quantity of mucosa exposed to the lens by flatting the folds and strengthening the angulations, and the benefit of AI is only in spotting a lesion that is already displayed within the field of view. Thus, we may speculate that the additional mucosal exposure was synergistic to the AI-assisted polyp recognition by AI.”

The benefits of a combination approach were not universal, however, as the mucosal exposure device did not improve detection of either serrated lesions or advanced adenomas. This result was anticipated, the investigators noted, since the miss rate for diminutive or proximal adenomas is higher than it is for larger or distal lesions, and previous research has suggested that AI-assisted and mucosal exposure techniques, when used alone, are most effective for detecting smaller, proximal lesions.

The study was funded by a European Society of Gastrointestinal Endoscopy Artificial Intelligence Award. The investigators disclosed additional relationships with Fujifilm, Medtronic, Olympus, and others.

Next Article:

Race and ethnicity loom large in CRC screening