BriefCase

BriefCase is an AI-powered software designed to expedite the triage of chest CT scans by identifying potential acute rib fractures, thereby improving radiology workflow efficiency. Its deep learning algorithm analyzes DICOM images to provide rapid, non-diagnostic notifications to clinicians, enabling faster prioritization of critical cases without altering the standard of care.

BriefCase is a radiological computer-aided triage and notification software designed to assist medical specialists in prioritizing chest CT scans for adults (18 years and older) by flagging suspected cases of three or more acute rib fractures (RibFx). The system employs an artificial intelligence (AI) deep-learning algorithm that runs on a cloud-based, Linux server environment. It processes DICOM images from CT scans (with or without contrast) chronologically, and upon detecting potential RibFx, it sends a notification to a desktop application. This notification includes a compressed, low-quality, grayscale preview image, which is explicitly not for diagnostic use but solely for informational purposes to aid in triage. The software operates in parallel with the standard image interpretation workflow, integrating with existing PACS and radiology workstations without altering original images or changing the standard reading queue. Performance of the BriefCase software was evaluated in a retrospective, multicenter study involving 308 cases. The AI algorithm demonstrated high accuracy, achieving an Area Under the Curve (AUC) of 0.98, with a sensitivity of 98.08% and specificity of 93.14% in identifying cases with three or more acute rib fractures. This met the pre-specified primary endpoint of AUC > 0.95. A key improvement over its predicate device is a significantly reduced mean time-to-notification, now 70.1 seconds compared to the predicate's 252 seconds, facilitating quicker attention to potentially critical cases. The system also demonstrated high Negative Predictive Value (NPV) at 99.8%. Additional operating points allow users to optimize for higher sensitivity or specificity, further enhancing its utility in workflow prioritization.

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 vary with different CT scanner manufacturers, resolutions, or slice thicknesses beyond the tested range of 0.5-5.0 mm?; What is the impact on radiologist workload and overall patient turnaround time in a real-world clinical setting when implementing BriefCase, considering the separate image communication platform integration?; Are there specific types or locations of rib fractures (e.g., non-displaced, costochondral) where the algorithm's sensitivity or specificity is notably different?; Given the PPV of 61.4%, what strategies can be employed to manage the rate of false positive notifications to minimize potential alert fatigue for clinicians?; How does the algorithm account for confounding factors such as old healed fractures, surgical changes, or other thoracic pathologies that might mimic acute rib fractures on CT?