臨床研究等提出・公開システム

Artificial intelligence in Colonoscopy for Cancer Prevention (ACCEPT)

Artificial intelligence in Colonoscopy for Cancer Prevention

Yuichi Mori

Universitetet i Oslo

Gaustad Sykehus Bygg 20, Sognsvannsveien 21, 0372 Oslo

ibusiginjp@gmail.com

Hiroki Nakamura

Showa University Northern-Yokohama Hospital

35-1,Tsuduki,Yokohama,Kanagawa

+81-459497000

n-hiroki@td6.so-net.ne.jp

Recruiting

Nov. 27, 2025

Interventional

randomized controlled trial

open(masking not used)

no treatment control/standard of care control

single assignment

health services research

Over 20 year old who undergo colonoscopy at Showa University Northen Yokohama Hospital

Exclusion criteria for the screening programmes are applied in the present project (e.g.,
individuals with previous history of colorectal cancer). No additional exclusion criteria are used
for the project.

Both

Colorectal Neoplasms

Intervention group
Our aim is to generically assess AI detection tools during colonoscopy. Thus, individuals allocated to the intervention group will receive colonoscopy with the aid of comparable AI toolsfrom different manufacturers. The AI tools used will be ENDO-BRAIN EYE(Olympus Corp. Tokyo), WISE VISION(NEC Corp. Tokyo), CAD-EYE (Fujifilm Corp. Tokyo) according to availability. Each AI tool has cleared regulatory approval and similar functions of identifying any suspicious area for a polyp and show a bounding box on the monitor during colonoscopy. All tools work with deep learning technologies, with small difference in terms of algorithms and performance. Each tool may get updated during the intervention period (e.g., increment of images used for machine learning, addition of ancillary functions, etc).
Control group
Participants randomized to standard colonoscopy will undergo colonoscopy without use of AI.

Colorectal Neoplasms

D015179

-Short-term primary endpoints
Adenoma detection rate (ADR)
Advanced adenoma detection rate (A-ADR)

-Long-term primary endpoints
Incidence of colorectal cancers after 10-years follow-up
Mortality from colorectal cancers after 10-years follow-up

Proportion of endoscopists with ADRs 25% or higher
Serrated polyp detection rate
Mean number of adenomas per colonoscopy
Learning effect of using AI during colonoscopy
Benefits from the use of AI according to the expertise
Comparison of different AI systems
Colonoscopy performance (e.g., cecal intubation rate, insertion and withdrawal time).
Rate of adverse events during and within 30 days after colonoscopy Incidence of colorectal cancers in 15-year follow-up
Mortality from colorectal cancers in 15-year follow-up
All-cause mortality in 10-year and 15-year follow-up
Mortality due to causes other than colorectal cancer in 10-year and 15-year follow-up
Cost-effectiveness of using AI in a cancer screening programme for colorectal cancer.
Investigation of diagnostic ability of AI for colorectal lesions
Retrospective analysis of recorded videos for individuals who have cancer or relevant lesions
detected after colonoscopy screening

+81-3-3784-8129

Mar. 03, 2021

Yes

Written consent is obtained from the patient. The consent document includes a description of the AI that will be the intervention and whether it is invasive or not. Patients can decide whether they will participate or not participate. Data to be used include information related to colonoscopy, long-term prognostic information on colorectal cancer incidence and mortality, and personal information necessary to use the national cancer registry. The acquired data are stored in a secure computer in the hospital. If it is necessary to provide data outside the hospital for analysis, all personal information will be removed so that individuals cannot be identified, and then the data will be provided to an affiliated research institution.

Norway/Spain/Italy/Poland/Denmark