BriefCase-Triage

BriefCase-Triage is an AI-powered software designed to expedite the identification of suspected Pulmonary Embolism in adult CTPA scans by providing rapid notifications to clinicians, thereby enabling prioritized review within existing radiology workflows. Its methodology relies on a deep learning algorithm trained on extensive image data to analyze DICOM images, demonstrating high sensitivity and specificity in clinical evaluations for PE detection and significantly reducing time-to-notification.

BriefCase-Triage, developed by Aidoc Medical, Ltd., is a radiological computer-aided triage and notification software designed to analyze Computed Tomography Pulmonary Angiography (CTPA) images in adults (18 years and older). Its primary purpose is to assist hospital networks and medical specialists in workflow triage by flagging and communicating suspected cases of Pulmonary Embolism (PE). The system utilizes an artificial intelligence deep-learning algorithm, trained on manually labeled CT images exhibiting PE pathology, to analyze DICOM images chronologically. When a suspected PE case is detected, BriefCase-Triage provides a pop-up notification to the user, accompanied by a compressed, low-quality, grayscale preview image intended solely for informational and prioritization purposes, not for primary diagnosis. The device operates in parallel with standard image interpretation workflows, does not alter original medical images, and is explicitly not a diagnostic tool. The performance of BriefCase-Triage was evaluated in a retrospective, blinded, multicenter pivotal study involving 499 CTPA cases from six U.S. clinical sites, compared against a ground truth established by three senior board-certified radiologists. The study demonstrated high efficacy, with a sensitivity of 94.39% and specificity of 94.39%, exceeding the predefined 80% performance goal. A key improvement noted was a significant reduction in the mean time-to-notification to 26.42 seconds. The software also showed strong Negative Predictive Value (98.96%) and Positive Predictive Value (74.79%). The algorithm was trained on a larger dataset compared to its predicate, and its software architecture was updated to allow integration with various image communication platforms. The study included diverse patient demographics (mean age 62.1 years) and scans from multiple CT scanner manufacturers, with additional operating points demonstrating flexibility in optimizing for sensitivity or specificity.

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 algorithm's performance (sensitivity, specificity, time-to-notification) vary across different CT scanner manufacturers and imaging protocols not explicitly detailed in the pivotal study summary?; What specific features or image characteristics does the deep learning algorithm prioritize when identifying suspected Pulmonary Embolism, and how robust is it against common image artifacts or atypical PE presentations?; Given the separation of the algorithm analysis module from the image communication platform, what are the minimum technical specifications and validation requirements for third-party image communication platforms to ensure optimal and reliable performance of BriefCase-Triage?; The document mentions "pathologies present in negative cases"; what was the prevalence and nature of these co-existing pathologies, and did they impact the algorithm's false positive or false negative rates for PE detection?; Beyond the primary and additional operating points presented, what is the full receiver operating characteristic (ROC) curve for the algorithm, and how was the "default" operating point chosen in relation to the trade-offs between sensitivity and specificity for PE triage?