A Novel Multi-functional Artificial Intelligence System on Treatment Efficacy and Implementation in Colonoscopy

Brief summary (registry, in English)

To overcome non-neoplastic polyp (NNP) resections, computer-aided diagnosis (CADx) systems have been developed. In a meta-analysis, the performance of CADx systems was superior to endoscopists. The proportion of incorrect predictions could be significantly reduced with CADx assistance. It was reported that autonomous AI prediction could achieve an agreement with the standard pathology-based surveillance intervals. In a prospective study targeting diminutive rectosigmoid polyps, CADx achieved a negative predictive value compared to the histology as gold standard. Lesions were amendable for a leave-in-situ strategy, suggesting a potential to reduce the burden of unnecessary polypectomies. It was estimated that the average colonoscopy cost and annual reimbursement could be reduced under the 'diagnose-and-leave' strategy. Nevertheless, almost all existing literature reporting the efficacy of CADx were simulated studies and focused on the diagnostic accuracy (i.e. all polyps were ultimately resected for histopathology). There is a lack of real-world data with hard clinical endpoints to support its implementation. A recent RCT compared leave-in-situ and resect-all strategy with real-time CADx in both arms. Using ADR as the surrogate marker, it was shown that the leave-in-situ strategy with CADx support was non-inferior. The major limitations of this RCT were that: i) CADx were activated in both arms - the pure CADx effect could not be demonstrated; ii) 'resect-all' strategy was not a real-world practice especially for diminutive rectosigmoid polyps; iii) the interaction and decision making between AI and human was not documented and uncertain. A well-designed RCT is warranted to evaluate the pure CADx effect on reducing NNP resections (enhance treatment efficacy) while maintaining the benchmark of ADR (safe implementation). In addition, AI-measured metrics may objectively validate the procedural quality for performance tracking and auditing purposes. If the above points are proven, AI-assisted colonoscopy would become the 'mainstream' in CRC prevention.

What you commit to

• Study type: Interventional
• Planned participants: 2818
• From 30 Oct 2026 to the primary completion date, 31 Dec 2028
• No sites listed in no country
• Visits and tests: ask the trial team, the registry does not list them

Built from the registry fields of ClinicalTrials.gov. It describes the study, not your situation.

Who cannot take part

Copied verbatim from ClinicalTrials.gov and grouped by topic. Only the trial team decides who can take part.

Who the study is looking for

Age in the registry: 45 years to 85 years · Sex: All · Accepts healthy volunteers

Full criteria on ClinicalTrials.gov