🤖 Generated with AI assistance · Does not replace individualized medical advice.

Real-Time AI-Assisted Detection of Colonic Lesions Using the iIDEAS Intelligent-C Module: A Prospective Single-Center Validation Study

AI Transparency Notice: This post was generated by an AI assistant to summarize and contextualize published medical literature. It is intended for educational and informational purposes only and does not constitute medical advice, diagnosis, or treatment recommendations. Always consult a qualified healthcare provider for personal medical concerns.

Lead

In a prospective single-center study of 200 patients, the iIDEAS-C AI module achieved a 0% lesion-level miss rate during a second inspection, significantly improving adenoma detection rates compared to conventional endoscopy alone PMID: 42761157.

Key Facts

Metric	Value
Study Design	Prospective, nonrandomized, single-center validation PMID: 42761157
Sample Size	200 consecutive adults (18–75 years) PMID: 42761157
Primary Endpoints	Adenomas per colonoscopy (APC) and lesion-level miss rate PMID: 4276157
Key Result	0% per-polyp and per-adenoma miss rate with iIDEAS-C PMID: 42761157

Why It Matters / Context

Colorectal cancer (CRC) remains a leading cause of cancer mortality, largely driven by the failure to detect and remove precancerous adenomas during screening colonoscopy. Conventional colonoscopy is subject to human variability, with studies suggesting that up to 26% of adenomas may be missed during a standard withdrawal PMID: 42761157. These missed lesions, particularly small and proximal ones, are associated with interval cancers—cancers that develop between routine screenings.

Computer-aided detection (CADe) systems represent a technological intervention designed to mitigate this risk by providing real-time visual cues to endoscopists. While retrospective studies have shown promise, prospective validation is critical to determine if AI assistance translates into improved clinical outcomes in real-world settings. This study addresses that gap by evaluating the iIDEAS-C module in a consecutive cohort of patients undergoing screening, surveillance, or diagnostic colonoscopy. The focus on a “second look” protocol allows for a direct comparison of detection capabilities between the unaided endoscopist and the AI-assisted system, isolating the specific contribution of the AI to lesion identification.

Details

The study enrolled 200 consecutive patients between February and May 2023. The cohort had a mean age of 44.1 ± 12.7 years, with 64% male participants PMID: 42761157. Each patient underwent a two-step inspection process: an initial blinded withdrawal without CADe, followed by a second inspection with the iIDEAS-C module active. Lesions identified only during the second inspection and confirmed endoscopically were classified as “missed” by the initial unaided pass.

A total of 82 polyps were detected across 42 patients, resulting in a polyp detection rate (PDR) of 21% PMID: 42761157. During the initial unaided inspection, endoscopists identified 61 of the 82 polyps, yielding a per-polyp miss rate of 25.6% PMID: 42761157. In contrast, the iIDEAS-C module detected all 82 lesions during the second inspection, achieving a 0% per-polyp and per-adenoma miss rate PMID: 42761157.

The study also evaluated secondary outcomes, including the polyp per colonoscopy (PPC) rate. The PPC increased from 0.305 in the unaided inspection to 0.410 with AI assistance PMID: 42761157. Analysis of the missed lesions revealed that they were predominantly diminutive (small) and located in the proximal colon, areas where visual detection is often more challenging due to anatomical complexity and shorter inspection times PMID: 42761157.

Outcome Measure	Unaided Inspection	iIDEAS-C Assisted Inspection
Total Polyps Detected	61	82
Per-Polyp Miss Rate	25.6%	0%
Polyps Per Colonoscopy (PPC)	0.305	0.410
Polyp Detection Rate (PDR)	N/A (Baseline)	21% (Cohort)

Note: The PDR of 21% refers to the proportion of patients in the cohort who had at least one polyp detected during the combined process. The table above compares detection counts and rates between the two inspection phases.

Limitations

Critical Limitations:

* Single-Center Design: The study was conducted at a single institution, which may limit the generalizability of the findings to other clinical settings with different patient populations or endoscopist expertise levels PMID: 42761157.

* Nonrandomized Protocol: The “second look” design, while useful for identifying missed lesions, does not reflect standard clinical practice where AI is used in real-time during the initial withdrawal. This may overestimate the practical benefit of AI in routine screening PMID: 42761157.

* Small Sample Size: With only 200 patients and 82 polyps, the study may lack the statistical power to detect subtle differences in detection rates for specific lesion subtypes or to assess long-term clinical outcomes such as interval cancer reduction PMID: 42761157.

* Short Follow-up: The study did not track patients for long-term outcomes, such as the incidence of interval colorectal cancer, which is the ultimate clinical endpoint of interest in screening programs PMID: 42761157.

Clinical Takeaway

While the iIDEAS-C module demonstrated a 0% miss rate in this prospective validation, clinicians should await larger multicenter randomized controlled trials to confirm its impact on reducing interval cancer rates before integrating it into routine screening protocols.

Full Reference

Rughwani H, Garg R, Memon SF, Yeung CK, Sreedhar B, Lai KP. Real-Time AI-Assisted Detection of Colonic Lesions Using the iIDEAS Intelligent-C Module: A Prospective Single-Center Validation Study. JGH open : an open access journal of gastroenterology and hepatology. 2026. doi:10.1002/jgh3.70470. PMID: 42761157. ClinicalTrials.gov Identifier: NCT05784935.