SKOUT® system

The SKOUT® system is a software-based device designed to assist gastroenterologists by employing real-time analysis of endoscopic video during colonoscopies to highlight potential colorectal polyps. Its core technology is an artificial intelligence algorithm, which in this iteration has been refined through retraining with extensive new clinical data and software updates, supported by non-clinical performance testing.

The SKOUT® system is an advanced software device designed to support gastroenterologists in the detection of potential colorectal polyps during colonoscopy procedures. It operates by analyzing high-definition endoscopic video in real-time, utilizing an artificial intelligence-based algorithm to identify suspicious areas. When a potential polyp is detected, the system superimposes a blue rectangular outline on the endoscopic monitor as a visual aid, but it does not provide an audible alert. The system is intended for use in adult patients undergoing colorectal cancer screening or surveillance with white light colonoscopy, and its detection feature pauses when surgical tools are present or lighting is suboptimal to avoid interference; gastroenterologists remain responsible for the final interpretation and diagnosis.\n\nThis iteration of the SKOUT® system represents a refinement of its predicate device, primarily through the retraining of its polyp detection and tracking algorithms with a significantly expanded dataset from over 30 clinical sites, including 27 new ones contributing to 61% new data. Alongside algorithm improvements, the system received updates to its operating system and software libraries, as well as various bug fixes, all while maintaining the same fundamental hardware and AI architectural design. Non-clinical performance testing, including software verification and validation, confirmed that the system meets its design requirements and special controls for algorithm performance, with the device information asserting that clinical performance remains consistent with previous versions due to the unchanged core inference architecture.

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. What specific architectural changes, if any, were made to the polyp detection algorithm beyond retraining with new data, and how do these contribute to improved performance or robustness?; Could you elaborate on the characteristics of the new clinical data from the 27 additional sites (e.g., patient demographics, polyp morphology, prevalence) and how this diversity influenced the algorithm's refinement?; Given the refinement of the polyp tracking algorithm, what metrics are used to quantify its accuracy and stability in tracking polyps across video frames, especially in challenging scenarios like rapid endoscope movement or partial occlusion?; How does the system differentiate between true polyps and polyp-like structures (e.g., folds, debris) and what is the reported false positive rate in non-clinical testing with the refined algorithm?; What are the implications of the operating system and software library updates on the system's real-time processing capabilities, particularly regarding latency and compatibility with existing endoscopy hardware?