Combining Artificial Intelligence With Balloon Mucosal Exposure Device for Polyp Detection in Screening Individuals

The Italian Colorectal Cancer (CRC) screening program invites the resident population aged 50-74 years for Fecal Immunochemical Test (FIT) every 2 years. Subjects who test positive are referred for colonoscopy. Colonoscopy and polypectomy prevent incident cases of colorectal cancer by their detection at an early and curable stage, and by identification and removal of colorectal precancerous lesions (i.e. adenomas). Taking into account the number and the histological features of polyps identified, colonoscopy also allows to stratify individuals according to the risk of developing colorectal cancer over time (low-risk vs. high-risk subjects), thus allowing to plan a balanced risk-based follow-up schedule: high-risk individuals are scheduled for a surveillance colonoscopy in 3 years, whereas low-risk individuals are scheduled for a further FIT test in five years or for a new colonoscopy in 10 years. However, colonoscopy is far from being a flawless tool. Studies that have measured the incidence of colorectal cancer diagnosed in subjects who have recently undergone colonoscopy with negative results (the so-called interval cancers) have shown that a significant portion of interval cancers is related to pre-neoplastic, or frankly neoplastic colonic lesions, missed during the endoscopic procedure.

In the last few years, it has been shown that in patients undergoing colonoscopy for mixed indications AI-based systems can effectively support the endoscopist by decreasing the adenoma miss rate and by increasing the identification of colonic polyps, thus potentially lowering the rate of interval cancer and increasing the rate of individuals scheduled for a short follow-up. A recent study demonstrated that AI-assisted colonoscopy is effective also in the framework of a FIT-based organized CRC screening program resulting in an absolute difference of 8.3% and 0.23 for adenoma detection rate (ADR) and for adenomas per colonoscopy (APC), when compared to standard colonoscopy performed without the assistance of AI. Nevertheless, tandem studies have shown that despite the use of AI systems the miss rate of colonic adenomas is still about 15-20%. In fact, AI systems only allow to highlight polyps that are framed by the endoscopist within the screen. Therefore, the use of computerized AI-based system cannot disregard an accurate retraction technique, focused on exploring in an extensive and effective way the space between the folds. Several tools, mostly consisting of distal attachments or plastic caps, aimed at facilitating the evaluation of the space between the folds favoring their smoothing during the endoscope withdrawal, have been developed to maximize mucosal exposure. Among them, the G-EYE system includes a balloon integrated at the distal tip of a standard colonoscope, which flattens haustral folds, centralizes the colonoscope's optics, and reduces bowel slippage during endoscope withdrawal. Initial studies showed that G-EYE balloon system, when compared with standard colonoscopy, increases not only adenoma detection rate but also the detection of advanced, flat, and sessile serrated adenomas/polyps. Recently, a new user friendly G-EYE balloon system (G-EYE 760R endoscope) has been marketed: it consists of a standard last-generation colonoscope with a permanently integrated balloon, placed at the bending section of the endoscope, which can be gradually insufflated on-demand by the endoscopist through a pedal control. It can be hypothesized that the combination of AI systems with mucosal exposure devices might further decrease the polyp miss rate, by maximizing the detection of adenomas. Moreover, initial studies based on such a strategy, combining AI with mucosal exposure devices, yielded encouraging results. However, the combination of CADEYE with G-EYE 760R endoscopes has not been evaluated so far.

The present prospective randomized controlled trial is aimed at evaluating whether in individuals scheduled for colonoscopy in the framework of a structured FIT-based colorectal cancer screening program, the combination of an AI system (CADEYE) with a mucosal exposure device (G-EYE 760R endoscope) increases the identification of subjects at high risk to develop colorectal cancer, when compared to colonoscopy performed with the support of AI only. In addition, it will be also evaluated whether the use of an AI system combined with a mucosal exposure device, increases:

• the adenoma detection rate (i.e. the rate of individuals with at least one histologically confirmed adenoma)
• the number of adenomas per colonoscopy (i.e. the number of adenomas detected divided by the number of all colonoscopies performed)
• the serrated lesions with neoplastic potential (sessile serrated lesions-SSA; traditional serrated adenomas - TSA) detection rate.
• the detection rate of right colon adenomas
• the detection rate on non-polypoid colonic lesions. Individuals fulfilling the inclusion criteria are randomized (1:1) in two arms (see below). Randomization is based on a computer-generated randomized blocks sequence, stratified according to age (50-61 vs. 62-74) and gender (male vs. female); size of the blocks (10 individuals) is not communicated to the investigator. Allocation is concealed and kept in a sealed envelope, which is opened just before starting the procedure.

Randomization groups:

• Group A (CADEYE only): all subjects receive colonoscopy with G-EYE 760R colonoscope but the balloon remains deflated for the entire procedure; the colonoscopy is performed with the support of the CADEYE system for polyp detection in both insertion and withdrawal phase; all polyps identified are removed and sent for histopathology examination.
• Group B (CADEYE+G-EYE 760R): all individuals receive colonoscopy examination with G-EYE 760R colonoscopes; once the cecum is reached the balloon is inflated, and the endoscope is withdrawn with the inflated balloon; the colonoscopy is performed with the support of the CADEYE system for polyp detection in both insertion and withdrawal phase; all polyps identified are removed and sent for histopathology examination.

For the present study a clinical outcome measure, such as the rate of "high risk" individuals across the two study arms, is calculated and compared. According to recent European Society of Gastrointestinal Endoscopy (ESGE) guidelines subjects are labelled as "high-risk" if harboring at least 1 adenoma ≥ 10 mm or with high grade dysplasia, or ≥ 5 adenomas, or any serrated polyp ≥ 10 mm or with dysplasia.

All endoscopic procedures are performed using the ELUXEOTM 7000 endoscopy platform (including videoprocessor ELUXEO TM VP-7000 and light source ELUXEO TM BL-7000; Fujifilm Co., Tokyo, Japan), high-definition 26''monitors (Barco NV, Kortijk, Belgium) and G-EYE 760R endoscopes (Fujifilm Co., Tokyo, Japan).

Bowel preparation is a split regimen of low-volume solution, according to local protocol. Quality of bowel cleansing is recorded by using the Boston Bowel Preparation Scale. Colonoscopies are performed under conscious sedation with carbon dioxide insufflation. Intubation time, overall withdrawal time, intervention time and inspection time during withdrawal are measured using a stopwatch. An inspection time of at least 6 minutes (2 minutes in each colonic segment, namely right, transverse and left colon) is mandatory in both study arms. The mucosa is carefully inspected, and all polyps are classified according to their location, size, and morphology (according to Paris classification). The pathologist's measure, when available, is considered the reference standard; the endoscopist's measure estimation is used in the remaining cases (i.e., piecemeal resection). Polyp location is defined as "proximal" if proximal to the splenic flexure). All identified polyps are removed (biopsy for non-resectable lesions), irrespective of size, color or subjective interpretation, with the possible exception of very small (1-5 mm) hyperplastic-appearing polyps located in the rectum and-according to the judgment of the endoscopists-not clinically significant.

All participating endoscopists have credentials for participating in the organized FIT-based screening program (more than 300 colonoscopies/year; cecal intubation rate >95%, adenoma detection rate>25%). All endoscopist are familiar with CADEYE features and use this system on a regular basis. All participating endoscopists receive a formal training (a 20' lecture focused on the use of the G-EYE 760R endoscope). All of them performed at least 10 colonoscopies with the G-EYE 760R endoscope, to become acquainted with the system, before entering the study.

All resected or biopsy specimens are fixed in 10% buffered formalin solution sent to pathology in separate jars. They are processed and stained for histopathology using standard methods and evaluated by expert pathologists (one in each center), with credentials to participate in the FIT-based organized screening program; pathologists are blinded to the assigned examination mode. All lesions are classified according to Vienna classification. An advanced adenoma is defined as an adenoma ≥10 mm and/or with villous component >20%, and/or high-grade dysplasia.