Artificial intelligence in colonoscopy promises better adenoma detection

A real-time artificial intelligence (AI)-assisted colonoscopy device has the potential to improve the detection of polyps, including difficult-to-detect adenomas, even for experienced endoscopists, as shown in a study from Singapore.

Over a 3-month period, the use of the GI Genius Intelligent Endoscopy Module device picked up 487 “hits” (ie, sustained detection of a mucosal abnormality) in the 298 colonoscopies performed. Complete polyp removal or polypectomy was performed for 51.3 percent of the “hits,” and 68.4 percent of the polypectomies performed were adenomatous lesions. [Surg Endosc 2022;doi:10.1007/s00464-022-09470-w]

Furthermore, 14 out of the 250 polyps excised (5.6 percent) turned out to be sessile serrated adenomas on histology.

“Our single institution experience revealed that incorporation of the real-time computer-aided detection of polyps resulted in a median 8.5-percent improvement in individual endoscopist’s adenoma detection rate (ADR) from their baseline polypectomy rate,” according to Dr Frederick Koh and colleagues from Sengkang General Hospital.

Because the baseline ADR for each endoscopists was unavailable, Koh and his team had to determine the effect of real-time computer-aided detection by comparing the changes between the ADR during the study against each endoscopist’s baseline polypectomy rate, where the baseline ADR was deemed to be equal or less than the baseline polypectomy rate.

As such, the investigators believe that the true ADR improvement after the introduction of AI technology is greater than 8.5 percent.

Taking the drudgery out of colonoscopy assessment

In total, a total of 24 (82.8 percent) endoscopists participated in the study, 18 (62.1 percent) of them performing at least five AI-aided colonoscopies. Of these 18, 13 (72.2 percent) were general surgeons.

“These results highlighted that pairing real-time AI technology with colonoscopy does help experienced endoscopists identify more adenomatous lesions and improves the detection of notoriously difficult lesions to identify with the naked eye, improving the quality of colonoscopy,” Koh said.

Numerous endoscopic society guidelines cite periodic colonoscopy as the modality of choice to assess the colon for polyps. However, the identification and characterization of colonic polyps are difficult skillsets to acquire, as Koh pointed out. He said: “Mastering these skillsets require a combination of studying endoscopic atlases and hands-on experience, the latter of which inevitably requires time.” [Unit Eur Gastroenterol J 2021;9:681-687; Endoscopy 2019;51:1155-1179; Surg Endosc 2018;32:1377-1381; Clin Endosc 2016;49:6-7]

In the present study, where 18 endoscopists had at least 10 procedures done with the AI program, the median ADR was 28.5 percent. This high sensitivity modality, according to Koh, can potentially help to minimize the variation of quality, which is highly user dependent.

“The reduction of variability of quality would ultimately benefit the patients coming for colonoscopy and gives them good confidence regardless of the endoscopists. This is especially important for endoscopists with lower procedural volume (ie, “part-time” endoscopists) to help reduce their variability and maintain high standards of colonoscopic evaluation for their patients,” Koh explained. [Endosc Int Open 2021;9:E513-E521; Best Pract Res Clin Gastroenterol 2021;doi:10.1016/j.bpg.2020.101713; World J Gastroenterol 2020;26:7436-7443]

“From the training angle, consultant endoscopists may then have more reassurance to allow trainee endoscopists perform colonoscopy on their behalf under direct or oversight supervision. From the trainees’ point-of-view, being able to perform more endoscopies would also potentially allow a quicker ascend for these trainees to attain competence,” he added. [Int J Colorectal Dis 2021;36:2237-2245; Surg Endosc 2018;32:1377-1381]

Taken together, the data indicate that endoscopy surveillance is an area ripe for AI’s assistance. Pairing an AI and a skilled endoscopists can provide better quality of care for patients and ensure that they comply with the recommended guidelines. Additionally, it eases the demands and safeguards the interests on the healthcare provider, according to Koh.

“Moving forward, future studies using real-time AI-guided computer-aid detection of polyps should focus on evaluating its utility in education for endoscopy. Evaluating the learning curve of novice endoscopist in the characterization and identification with and without AI in comparison with experience endoscopists may introduce a different dimension on the utilization of AI and a paradigm shift to how endoscopy can be taught,” he said.