Computer Aided Detection of Polyps in the Colon

The purpose of this study is to examine the role of an automatic polyp detection software (henceforth referred to as the research software) as a support system during colonoscopy; a procedure during which a physician uses a colonoscope or scope, to look inside a patient's rectum and colon. The scope is a flexible tube with a camera-to see the lining of the colon. The research software is used to aid in the detection of polyps (abnormal tissue growths in the wall of the colon and adenomas (pre-cancerous growths) during colonoscopy. The research software used in this study was programmed by a company in Shanghai, which develops artificial intelligence software for computer aided diagnostics. The research software was developed using a large repository (database or databases) of polyp images where expert colonoscopists outlined polyps and suspicious lesions. The software was subsequently developed and validated using several databases of images and video to operate in near real-time or within minutes of photographing the tissue. It is intended to point out polyps and suspicious lesions on a separate screen that stands behind the primary monitor during colonoscopy. It is not expected to change the colonoscopy procedure in any way, and the physician will make the final determination on whether or not to biopsy or remove any lesion in the colon wall. The research software will not record any video data during the colonoscopy procedure. In the future, this software may help gastroenterologists detect precancerous areas and decrease the incidence of colon cancer in the United States.

Length of Study - The duration of the study is expected to be 8-12 months. Enrollment of study patients will cease when approximately 250 patients have been enrolled. Study Design- Design will be a multi-center, prospective, unblinded randomized control trial. Patients referred for either screening or surveillance colonoscopy will be included. Equipment: Aside from standard of care scope used, a second computer monitor that will stand behind the standard monitor used during colonoscopy. Additionally , a computer system unit with an operating system. Standard Clinical Procedure Typically, intravenous sedation using a combination of benzodiazepine and narcotic medications (with or without propofol under the supervision of a trained anesthesiologist) are used for colonoscopy. Continuous pulse oximetry and blood pressure monitoring is used throughout the procedure. Supplemental oxygen is used as needed. Patients are usually placed in the left lateral decubitus position and the colonoscope is introduced into the rectum. The colonoscope is advanced under direct visualization until the cecum and appendiceal orifice is reached. The colonoscope is usually retroflexed within the rectum. The colonoscopist carefully inspects each segment of colon during advancement and then again on withdrawal of the colonoscope. Any suspicious lesions encountered during insertion or withdrawal are inspected by the colonoscopist and a final determination is made by the clinician on whether or not to remove a given lesion. Any lesion that is deemed suspicious or polypoid is removed by en-bloc polypectomy, piecemeal polypectomy, or may be referred for endoscopic mucosal resection (EMR) at a later date. After the procedure, patients recover in the post-procedural recovery room. After the procedure, results are discussed with the patient. The ability of colonoscopy to detect lesions is discussed with the patient as well as the fact that a small percentage of polyps and other lesions may be missed during the test. Study Procedure Patients will receive a colonoscopy with a gastroenterologist. During the standard clinical procedural protocol and for the study period, colonoscopists will have the benefit of a second monitor that will project the polyp detection algorithm in real-time over the video output of the colonoscopy. The algorithm will detect suspicious, polyp-like lesions within the lumen of the colon, and during the procedure a research assistant will view the second monitor at all times and record a time stamp for any potential polyps on an intra-procedural data collection sheet. Data Collection Variables collected and measured will include colonoscopist(s) performing the procedure, number of adenomas noted per procedure, adenoma detection rate for a given colonoscopist, number of polyps detected per procedure, polyp detection rate (the proportion of colonoscopic examinations performed that detect one or more polyps), cecal intubation rate, time needed to reach the cecum, time needed to withdraw colonoscope both when polyps are identified (and thus need to be removed) and on normal colonoscopy, level of sedation, and complications: Acute if within 48 hours of procedure & delayed if within 3-30 days after procedure. Data Analysis - Normally distributed continuous variables will be summarized using means and standard deviations while non-normally distributed continuous variables will be summarized using medians and ranges.

In this study, the colonoscopist will carefully inspect segments of colon during advancement and then again on withdrawal of the colonoscope. Those who qualify will be randomized into two arms, as detailed in the bullets below: scope Insertion will be the same for both arms, without the aid of the research software. Below are two groups that qualifying subjects will be randomized into: Group of patients in Arm-1- recruited patients will receive Standard Colonoscopy followed by AI-Assisted Combined Colonoscopy Group of patients in Arm-2- recruited patients will received AI-Assisted Combined Colonoscopy followed by Standard Colonoscopy

Normal scope insertion and withdrawal first, followed by a second withdrawal with the research software running on a separate screen to catch any additional polyps missed during the first withdrawal.

Normal scope insertion but first withdrawal with the research software running on a separate screen, followed by a second withdrawal without the research software running.

The research software is deep learning algorithm used to aid in the detection of polyps (abnormal tissue growths in the wall of the colon and adenomas (pre-cancerous growths) during colonoscopy. In its current form, the automatic polyp detection system is installed on a computer system unit that utilizes an an operating system.

Adenoma Miss Rate (AMR), to determine if the combination technique identifies more adenomas compared to the standard technique. AMR will be calculated as the number of adenomas detected on the second pass or portion in either group divided by the total number of adenomas detected during both passes

To determine the accuracy of the polyp detection software by determining if the combination technique identifies more polyps compared to the standard technique: Per-patient true positive, false positive and false negative will be recorded. True positives will be defined as lesions that are detected for >2 seconds by the research software and are deemed to be consistent in appearance with a polyp by the endoscopist. False positives will be defined as lesions that are detected for > 2 seconds by the research software but are ultimately deemed by the endoscopist to have a gross appearance not consistent with polyp. False negatives will be defined as lesions that are not detected, or detected for <2 seconds by the research software, but are deemed by the endoscopist to be consistent with polyp

To determine if the combination of an automated polyp detection software and standard colonoscopy will have a higher detection rate of adenomas

Advanced adenoma miss rate will be calculated as the number of advanced adenomas [adenoma that is ≥ 10 mm in size] detected on the second pass or portion in either group divided by the total number of advanced adenomas detected during both passes.

The time it takes to withdraw the colonoscope from the end of the colon back to the rectum. This is the time that your gastroenterologist will be looking for polyps most.

We will be monitoring the rate of adverse events related to the procedure for the duration of the study.

Inclusion Criteria: Patients age: ≥ 22 years Patients presenting for routine colonoscopy for screening and/or surveillance purposes. Willingness to undergo two withdrawals with and without the use of computer-aided software while undergoing conventional colonoscopy with sedation Ability to provide written, informed consent and understand the responsibilities of trial participation Exclusion Criteria: Minors aged < 22 years. People with diminished cognitive capacity Patients undergoing diagnostic colonoscopy (e.g. as an evaluation for active gastrointestinal bleed, referring collectively to the stomach and the small and large intestine). Patients with incomplete colonoscopies (those where endoscopists did not successfully intubate the cecum due to technical difficulties or poor bowel preparation) Patients that have standard contraindications to colonoscopy in general (e.g. documented acute diverticulitis, fulminant colitis and known or suspected perforation). Patients with inflammatory bowel disease Patients with any polypoid/ulcerated lesion > 2 cm concerning for invasive cancer on endoscopy Patients referred for endoscopic mucosal resection (EMR), which is a procedure to remove early-stage cancer and precancerous growths from the lining of the digestive tract.

There is a plan to make IPD and related data available to researchers involved in the study at other centers. Data sharing is subject to data sharing agreement signed by participating institutions with BIDMC. The polyp detection software does not save or store any study data.

All collected data is to be analyzed in support to the study's hypothesis and endpoints. This data includes other variables, which will be obtained shortly after the procedure via chart review, including intra-procedural data points such as time needed to reach the cecum and scope withdrawal time. Data will be collected and stored in an encrypted and anonymized database such as REDCap or in an excel spreadsheet with de-identified information and encryption. All collected de-identified data (data which is stripped off all personal information) will be shared with other sites via REDCap.

Siegel RL, Miller KD, Jemal A. Cancer Statistics, 2017. CA Cancer J Clin. 2017 Jan;67(1):7-30. doi: 10.3322/caac.21387. Epub 2017 Jan 5.

Winawer SJ, Fletcher RH, Miller L, Godlee F, Stolar MH, Mulrow CD, Woolf SH, Glick SN, Ganiats TG, Bond JH, Rosen L, Zapka JG, Olsen SJ, Giardiello FM, Sisk JE, Van Antwerp R, Brown-Davis C, Marciniak DA, Mayer RJ. Colorectal cancer screening: clinical guidelines and rationale. Gastroenterology. 1997 Feb;112(2):594-642. doi: 10.1053/gast.1997.v112.agast970594. No abstract available. Erratum In: Gastroenterology 1997 Mar;112(3):1060. Gastroenterology 1998 Mar;114(3):625.

Ferlitsch M, Reinhart K, Pramhas S, Wiener C, Gal O, Bannert C, Hassler M, Kozbial K, Dunkler D, Trauner M, Weiss W. Sex-specific prevalence of adenomas, advanced adenomas, and colorectal cancer in individuals undergoing screening colonoscopy. JAMA. 2011 Sep 28;306(12):1352-8. doi: 10.1001/jama.2011.1362.

Winawer SJ, Zauber AG, Ho MN, O'Brien MJ, Gottlieb LS, Sternberg SS, Waye JD, Schapiro M, Bond JH, Panish JF, et al. Prevention of colorectal cancer by colonoscopic polypectomy. The National Polyp Study Workgroup. N Engl J Med. 1993 Dec 30;329(27):1977-81. doi: 10.1056/NEJM199312303292701.

Rex DK, Schoenfeld PS, Cohen J, Pike IM, Adler DG, Fennerty MB, Lieb JG 2nd, Park WG, Rizk MK, Sawhney MS, Shaheen NJ, Wani S, Weinberg DS. Quality indicators for colonoscopy. Am J Gastroenterol. 2015 Jan;110(1):72-90. doi: 10.1038/ajg.2014.385. Epub 2014 Dec 2. No abstract available.

Corley DA, Jensen CD, Marks AR, Zhao WK, Lee JK, Doubeni CA, Zauber AG, de Boer J, Fireman BH, Schottinger JE, Quinn VP, Ghai NR, Levin TR, Quesenberry CP. Adenoma detection rate and risk of colorectal cancer and death. N Engl J Med. 2014 Apr 3;370(14):1298-306. doi: 10.1056/NEJMoa1309086.

Glissen Brown JR, Mansour NM, Wang P, Chuchuca MA, Minchenberg SB, Chandnani M, Liu L, Gross SA, Sengupta N, Berzin TM. Deep Learning Computer-aided Polyp Detection Reduces Adenoma Miss Rate: A United States Multi-center Randomized Tandem Colonoscopy Study (CADeT-CS Trial). Clin Gastroenterol Hepatol. 2022 Jul;20(7):1499-1507.e4. doi: 10.1016/j.cgh.2021.09.009. Epub 2021 Sep 14.