AI-assisted second look raises right-colon adenoma detection in 606-patient trial

An extra inspection of the right colon, supported by an AI detection system, helped doctors find adenomas in more patients in a three-centre Japanese trial. The result addresses a persistent problem in colonoscopy: small, flat growths can disappear against folds or sit in awkward positions, making them easy to miss during a single pass.[1][2]

The randomised trial, published in August 2026, analysed 606 patients. Right-colon adenomas were found in 36.4% of those assigned to an AI-assisted second inspection and 27.9% receiving standard colonoscopy: an absolute gain of 8.5 percentage points. Small, flat lesions and findings in difficult anatomical areas contributed to the improvement.[1][2]

The trial tested a combined strategy. It establishes the benefit of adding an AI-supported second look to the standard examination, while leaving open how much of the gain came from the repeat inspection and how much from the software.

What the patients actually received

First, both groups underwent an ordinary inspection of the right colon without AI. The doctors were not told which allocation their patient would receive until after that initial examination. Only then did the routes diverge. For 305 participants, the endoscopist returned to the beginning of the colon and repeated the right-side inspection using Wise Vision, a real-time computer-aided detection system. For the other 301, the procedure continued in the usual way without a repeat examination of that region.[1][2]

The distinction matters for clinical practice. A second pass gives an endoscopist another opportunity to expose folds and inspect difficult areas; AI may then flag a lesion within the new view. Comparing a second pass with and without AI would isolate the software’s additional contribution.

The primary outcome was the proportion of people in whom at least one adenoma was found on the right side, not the raw number of computer alerts. That endpoint is clinically understandable, but it has boundaries. Detection says what an examination found. It does not directly measure the subsequent risk of colorectal cancer, its mortality, or whether the detected lesion would ever have developed into a dangerous tumour.[1][2]

The result beneath the headline

An 8.5-percentage-point gain means approximately 36 patients in every hundred had a right-colon adenoma detected under the combined strategy, compared with 28 in every hundred under standard examination. The outcome counts patients with at least one adenoma, rather than the number of alerts or lesions.

The authors reported a similar direction of improvement across endoscopists with different levels of experience. That supports investigating the strategy beyond a single group of operators, although subgroup findings do not establish an identical benefit for every clinician.[1][2]

A practical choice about examination time

An additional pass takes clinical time. For a screening service, the useful follow-up is whether the extra findings justify the procedure time, equipment cost and effect on throughput, and how the strategy compares with a second inspection without AI.[1][2]

The trial’s contribution is a concrete improvement in what clinicians found during a procedure. The next comparison can make that result more useful: establish whether the software adds enough to a careful second look to earn its place in routine screening.

Report an error or suggest a correction