SKOUT Software

The SKOUT system is an artificial intelligence-powered software designed to enhance colorectal polyp detection by providing real-time visual cues to gastroenterologists during white light colonoscopy procedures. Its methodology leverages a trained algorithm to analyze endoscopic video and highlight suspicious lesions, with clinical and non-clinical data supporting its effectiveness in increasing adenoma detection rates while maintaining a safe operational profile.

The SKOUTTMsystem is a software-based computer-aided detection (CADe) device designed to assist gastroenterologists in identifying potential colorectal polyps in real-time during colonoscopy examinations for adult patients undergoing screening or surveillance. It analyzes high-definition endoscopic video, utilizing a trained artificial intelligence algorithm to superimpose a two-dimensional blue rectangular outline around suspected polyps on the monitor. This visual aid deactivates if a surgical tool enters the frame or if the polyp is no longer detected, and a device status indicator displays its operational state (active or error). The system is intended for use with white light colonoscopy only and serves as an adjunct to the clinician's judgment, not as a replacement for full patient evaluation or primary diagnostic interpretation. Performance evaluations for the SKOUTTMsystem included comprehensive software verification and validation, and adherence to electrical safety and electromagnetic compatibility standards. Non-clinical testing on 79 HD videos with 94 polyps showed the algorithm detected 92 polyps, achieving an object-level True Positive Rate (TPR) of 97.87% and an average of 22.55 false positives per 15-minute interval. Marker persistence analysis indicated that markers with higher persistence were more likely to be true positives, and marker overlap analysis showed complete engulfment of polyps by the bounding box (IOGT 1.0). A multicenter, prospective, randomized controlled clinical trial involving 1,359 patients demonstrated that the SKOUTTMsystem arm had a statistically significant higher Adenomas per Colonoscopy (APC) rate (0.73 vs. 0.53 for control). The Positive Percent Agreement (PPA) was 70.9% for the SKOUTTM arm compared to 75.7% for control, meeting non-inferiority criteria. No adverse events were reported during the clinical study.

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 SKOUT system's algorithm handle variability in polyp morphology (e.g., sessile, pedunculated, flat) and how does this affect its true positive and false positive rates across different polyp types not detailed in the subgroup analysis?; What specific training data characteristics (e.g., diversity of polyp types, range of image quality, patient demographic variety) were used for the AI algorithm, and how might these influence its generalizability beyond the study population?; Given the non-clinical false positive rate of 22.55 objects per 15-minute interval, what is the observed impact on procedural time and endoscopist workflow when managing these false positives in real-world clinical practice?; The clinical trial reported an increase in Adenomas per Colonoscopy (APC) but a slightly lower Positive Percent Agreement (PPA) in the SKOUT arm, though non-inferior. What are the detailed characteristics of polyps contributing to these differing outcomes for APC and PPA?; Can the marker persistence data, indicating that longer persistence correlates with higher true positive likelihood, be translated into a user-interface feature, such as a confidence score, to help clinicians differentiate true positives from transient false positives more effectively?