SKOUT® system

The SKOUT® system is an adjunctive software tool employing an artificial intelligence algorithm for real-time analysis of endoscopic video, including pre-processing steps like de-interlacing, during colonoscopies to aid gastroenterologists in identifying potential colorectal polyps. Its methodology involves processing the video feed to superimpose a visual marker on suspected lesions under white light conditions, with the physician retaining ultimate diagnostic responsibility.

The SKOUT® system is a software-based medical device designed to assist qualified gastroenterologists in the real-time detection of potential colorectal polyps during colonoscopy examinations for adult patients undergoing cancer screening or surveillance. It operates as a computer-aided detection (CADe) tool, analyzing high-definition endoscopic video by processing the input with features like de-interlacing. The system provides an informational visual aid on the endoscopic monitor, highlighting suspected lesions to augment the physician's observational capabilities during the procedure. Utilizing an artificial intelligence-based algorithm, SKOUT® processes the video feed and, when a potential polyp is identified, generates a two-dimensional blue rectangular outline around it. The polyp detection function automatically pauses if endoscopic tools are detected in the video feed or if lighting conditions are deemed inadequate. The system also features a graphical user interface that indicates its operational status (active, tool present, or error). It is intended for use with white light colonoscopy only, and the responsibility for confirming the presence or absence of a polyp and making any diagnostic or treatment decisions remains with the gastroenterologist. Non-clinical testing confirmed the software meets design requirements and performance benchmarks for true/false positives, pixel degradation, and video delays, with its core inference algorithms remaining consistent with previously evaluated versions.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the introduction of de-interlacing during SDI video pre-processing quantitatively affect the AI algorithm's input image quality and, consequently, its polyp detection sensitivity and specificity under various clinical conditions?; What were the specific quantitative benchmarks for true positive rate, false positive rate, pixel degradation, and video delay that the SKOUT® system met during non-clinical performance testing?; Since the core inference algorithms remain unchanged, how did the software verification and validation specifically address the integration of the new de-interlacing step to ensure it does not negatively impact the algorithm's established performance or introduce new failure modes?; What mechanisms are in place to ensure the system's reliability when encountering subtle variations in endoscopic image quality, such as minor motion blur or suboptimal insufflation, that don't trigger an outright pause in polyp detection?; What studies have evaluated the potential for "automation bias" or over-reliance on the SKOUT® system's polyp detection markers by gastroenterologists, and what training or usage guidelines are recommended to mitigate this risk?