Computer-aided Detection During Screening Colonoscopy (Experts)

Real-time Computer-aided Polyp Detection During Screening Colonoscopy Performed by Expert Endoscopists

Evaluation of the colonic mucosa with a high definition colonoscope (EPKi7010 video processor). The endoscopy images will be seen on a 27inch, flat-panel, high-definition LCD monitor (Radiance™ ultraSC-WU27-G1520 model) only by one expert endoscopist, randomly assigned. The number, location, and polyps' features (Paris classification) will be recorded by the operator. If a polyp is detected, the endoscopist will remove the polyp endoscopically with a cold snare. The same patient will be submitted to a second, the same session, computed aided real-time colonoscopy using the DISCOVERY, AI-assisted polyp detector. Colonoscopy will be performed by a same-level-of-expertise operator in comparison to the initial procedure. Any polyp or lesion detected with the AI system will be recorded and endoscopically removed and considered as a missed lesion from standard colonoscopy.

Screening colonoscopy has decreased the incidence of colorectal carcinoma in the previous decades. However, there are reports of missed polyps and interval CRC following screening colonoscopy. Several factors may affect the ADR, PDR, and missed lesions rates, such as bowel preparation, percentage of mucosal surface evaluation, and the training levels of operators. Artificial intelligence using deep-learning algorithms has been implemented in gastrointestinal endoscopy, mainly for the detection and diagnosis of GI tract lesions such as colonic polyps and adenomas. The implementation of automated polyp detection software during screening colonoscopy may prevent the missing of polyp and adenoma during screening colonoscopy. Therefore, improving the ADR and PDR during colonoscopies. All of this, with the aim of decrease the incidence of interval colorectal carcinoma (CRC), and CRC-related morbidity and mortality. The Discovery Artificial Intelligence assisted polyp detector (Pentax Medical, Hoya Group) was recently launched for clinical practice. This AI software was trained with 120,000 files from approximately 300 clinical cases. The visual aided detection (bounding box locating a polyp on the monitor) will alert the endoscopist if a polyp/adenoma was missed during the standard, screening procedure. To the best of our knowledge, this may be the first study evaluating the Discovery AI-assisted polyp detector on clinical practice in the western hemisphere. The investigators aim to evaluate the real-world effectiveness of AI-assisted colonoscopy in clinical practice. The investigators will also evaluate the role of endoscopists' levels of training in the ADR, PDR, and missed lesion rate.

A non-blinded, non-randomized prospective diagnostic trial. Two interventions: Standard colonoscopy: 1 expert AI-assisted colonoscopy: another expert

Evaluation of the colonic mucosa with a high definition colonoscope (EPKi7010 video processor). The endoscopy images will be seen on a 27inch, flat panel, high-definition LCD monitor (Radiance™ ultraSC-WU27-G1520 model) only by one expert endoscopist, randomly assigned. The number, location and polyps' features (Paris classification) will be recorded by the operator. If a polyp is detected, the endoscopist will remove the polyp endoscopically with a cold snare and forceps biopsy.

The same patient will be submitted to a second, same session, computed aided real-time colonoscopy using the DISCOVERY, AI assisted polyp detector. Colonoscopy will be performed by a same-level-of-expertise operator in comparison to the initial procedure. Any polyp or lesion detected with the AI system will be recorded and endoscopically removed and considered as a missed lesion from standard colonoscopy.

Number of examinations with at least one adenoma detected during colonoscopy while using the AI-based model

Total number of missed polyps/ (total number of missed polyps + total number of polyps on initial examination)

Total number of missed adenomas/ (total number of missed adenomas + total number of adenomas on initial examination)

Inclusion Criteria: Provided informed written consent Age greater than 45 years of age Adequate Bowel preparation Exclusion Criteria: History of inflammatory bowel disease, familial polyposis syndrome History of colorectal carcinoma, colorectal surgery History of uncontrolled coagulopathy History of previously failed attempt colonoscopy

Wang P, Berzin TM, Glissen Brown JR, Bharadwaj S, Becq A, Xiao X, Liu P, Li L, Song Y, Zhang D, Li Y, Xu G, Tu M, Liu X. Real-time automatic detection system increases colonoscopic polyp and adenoma detection rates: a prospective randomised controlled study. Gut. 2019 Oct;68(10):1813-1819. doi: 10.1136/gutjnl-2018-317500. Epub 2019 Feb 27.

Vinsard DG, Mori Y, Misawa M, Kudo SE, Rastogi A, Bagci U, Rex DK, Wallace MB. Quality assurance of computer-aided detection and diagnosis in colonoscopy. Gastrointest Endosc. 2019 Jul;90(1):55-63. doi: 10.1016/j.gie.2019.03.019. Epub 2019 Mar 26.