NCT07023471

Artificial Intelligence-assisted Colonoscopy, Tandem Study

Effect of an Artificial Intelligence-assisted Colonoscopy on Adenoma Miss Rate of Trainee-performed Colonoscopy: A Four-group Randomized Controlled Tandem Colonoscopy Study

• Status: Enrolling By Invitation
• Phase: N/A
• Study type: Interventional
• Enrollment: 364 (estimated)
• Sponsor: Mahidol University · Academic / Other
• Sex: All
• Age: 40 Years – 85 Years
• Healthy volunteers: Not accepted

Summary

The goal of this clinical trial is to evaluate effect of artifial intelligent (AI) system, Endoscopy as AI-powered Device (ENAD) on adenoma miss rate from colonoscopy underwent by trainee endoscopist. It will also evaluate effect of AI on adenoma and polyp detection rate from colonoscopy underwent by trainee endoscopist. The main questions it aims to answer are: • Does AI-system lower adenoma miss rate in colonoscopy underwent by trainee endoscopist? Researchers will do the tandem colonoscopy and devided the participant in 4 groups as follows: A. First pass: trainee; Second pass: expert B. First pass: trainee + AI; Second pass: expert C. First pass: trainee; Second pass: expert + AI D. First pass: trainee+AI; Second pass: expert+AI Participants will take bowel preparation in split dose regimen and nothing per oral for 4 hours. They will underwent colonoscopy as above, with sedation by anesthesiologist. Details on qualities of colonoscopy, polyps detection and pathology results will be recorded.

Detailed description

Colon cancer accounts for one of the most common cancer worldwide and also cancer-related death. Colonoscopy is accepted to be an effective tool in colon cancer screening since the polypectomy of small adenoma can prevent colon cancer. Missed adenoma is one of the causes of interval cancer between routine colonoscopy screening. Nonvisualization is the cause of missed adenoma during colonoscopy. Artificial intelligence (AI)-assisted colonoscopy was superior then routine colonoscopy from parallel study and tandem study. Previous studies often used one same endoscopist in doing tandem colonoscopy which may still have bias. Only one previous study designed to use trainee endoscopist in the first pass and expert endoscopist in the second pass, some subgroups used AI-assisted. The result revealed the lower of adenoma miss rate (AMR) in AI-assisted colonoscopy in the first pass. This study designed to evaluate AMR of AI-assisted colonoscopy in trainee endoscopist compared to expert endoscopist, the trainee will do colonoscopy in the first pass (with or without AI) and the expert will do colonoscopy in the second pass (with or without AI). The present study aimed to evaluate effect of AI-assisted colonoscopy in trainee endoscopist.

Conditions

• Artificial Intelligence (AI)
• Colonoscopy Education
• Adenoma Colon Polyp

Interventions

Type	Name	Description
DEVICE	Group A (Trainee --> expert)	The expert endoscopist insert to cecum in both passes. First pass: Trainee withdraw colonoscopy (white-light mode) without AI. The polyps detected can be removed as suitable. The pathology for polyp will be sent. Second pass: Expert withdraw colonoscopy (white-light mode) without AI. The polyps detected (which is missed from the first pass) can be removed as suitable. The pathology for polyp will be sent.
DEVICE	Group B (Trainee +AI --> expert)	The expert endoscopist insert to cecum in both passes. First pass: Trainee withdraw colonoscopy with AI. The polyps detected can be removed as suitable. The pathology for polyp will be sent. Second pass: Expert withdraw colonoscopy (white-light mode) without AI. The polyps detected (which is missed from the first pass) can be removed as suitable. The pathology for polyp will be sent. Artificial intelligent (AI) assisted colonoscopy; ENdoscopy as AI-powered Device (ENAD) ENAD system (ENdoscopy as AI-powered Device, AINEX Corporation, Seoul, South Korea) is the system using CADe system (Computer-aided detection) which developed from 66,397 images and 8,756 polyps via deep learning-based object detection algorithm (YOLOv4) . It was validated by 15,753 images of polyp from 80 colonoscopy videos and 90,144 images of non-polyps from 50 colonoscopy videos. This system decreases false positive rate from 3.2% to 0.6% and increases sensitivity from 86.4% to 87.1%.
DEVICE	Group C (Trainee --> expert + AI)	The expert endoscopist insert to cecum in both passes. First pass: Trainee withdraw colonoscopy (white-light mode) without AI. The polyps detected can be removed as suitable. The pathology for polyp will be sent. Second pass: Expert withdraw colonoscopy with AI. The polyps detected (which is missed from the first pass) can be removed as suitable. The pathology for polyp will be sent. Artificial intelligent (AI) assisted colonoscopy; ENdoscopy as AI-powered Device (ENAD) ENAD system (ENdoscopy as AI-powered Device, AINEX Corporation, Seoul, South Korea) is the system using CADe system (Computer-aided detection) which developed from 66,397 images and 8,756 polyps via deep learning-based object detection algorithm (YOLOv4) . It was validated by 15,753 images of polyp from 80 colonoscopy videos and 90,144 images of non-polyps from 50 colonoscopy videos. This system decreases false positive rate from 3.2% to 0.6% and increases sensitivity from 86.4% to 87.1%.
DEVICE	Group D (Trainee + AI --> expert + AI)	The expert endoscopist insert to cecum in both passes. First pass: Trainee withdraw colonoscopy with AI. The polyps detected can be removed as suitable. The pathology for polyp will be sent. Second pass: Expert withdraw colonoscopy with AI. The polyps detected (which is missed from the first pass) can be removed as suitable. The pathology for polyp will be sent. Artificial intelligent (AI) assisted colonoscopy; ENdoscopy as AI-powered Device (ENAD) ENAD system (ENdoscopy as AI-powered Device, AINEX Corporation, Seoul, South Korea) is the system using CADe system (Computer-aided detection) which developed from 66,397 images and 8,756 polyps via deep learning-based object detection algorithm (YOLOv4) . It was validated by 15,753 images of polyp from 80 colonoscopy videos and 90,144 images of non-polyps from 50 colonoscopy videos. This system decreases false positive rate from 3.2% to 0.6% and increases sensitivity from 86.4% to 87.1%.

Timeline

• Start date: 2025-05-13
• Primary completion: 2025-05-13
• Completion: 2026-12-31
• First posted: 2025-06-17
• Last updated: 2025-06-24

Locations

1 site across 1 country: Thailand