BriefCase-Quantification

BriefCase-Quantification is an AI-powered software designed to analyze non-cardiac-gated, non-contrast CT scans for the detection and categorization of coronary artery calcification in adult patients. By employing a deep learning algorithm, it provides a four-tier risk assessment for coronary artery disease, intended to assist physicians as an adjunctive tool rather than a standalone diagnostic.

BriefCase-Quantification is a software medical device designed to assist physicians in analyzing non-cardiac-gated non-contrast CT (NCCT) images of the chest that include the heart, specifically for adult patients aged 30 and older. Its primary purpose is to identify and categorize Coronary Artery Calcification (CAC), a risk factor for coronary artery disease. The device employs an artificial intelligence, deep-learning algorithm, hosted on a cloud-based Linux server, to process these CT scans. It receives filtered images, evaluates calcified plaques in the coronary arteries using a default calcium threshold of 130 Hounsfield Units, and is compatible with slice thicknesses up to 5.0 mm. The software integrates into existing workflows by sending its analysis to PACS or a desktop application for review by hospital networks and appropriately trained medical specialists. The software generates a four-category output for CAC: "very low" (Agatston score 0), "low" (1-100), "medium" (101-400), and "high" (≥400), aligning with established visual assessment categorizations and SCCT/STR guidelines. It provides a preliminary summary report including the CAC category, the number of slices with CAC, and preview axial images of the detected calcium for informational purposes. A pivotal multicenter study involving 433 cases demonstrated an overall agreement of 87.1% between the algorithm's CAC categorization and the ground truth established by radiologists. Category-specific agreements were: very low (95.1%), low (81.3%), medium (81.5%), and high (89.3%). Importantly, the device's output is not intended for stand-alone clinical decision-making and requires physicians to review full images per standard of care.

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 across different CT scanner manufacturers and imaging protocols beyond slice thickness (e.g., radiation dose, reconstruction kernels)?; What is the clinical impact of the 12.9% disagreement rate with ground truth, particularly for cases misclassified into significantly different risk categories (e.g., 'very low' to 'medium/high' or vice versa)?; How does the 'informational purposes only' preview of detected calcium influence a radiologist's interpretation time and diagnostic confidence compared to standard review without such previews?; What specific features or characteristics of calcified plaques (e.g., density, morphology, distribution) does the deep learning model primarily use for its categorization, and how robust is it to atypical plaque presentations?; Beyond the four categories, does the software provide any sub-category quantitative information or confidence scores that could aid in borderline cases or longitudinal tracking of CAC progression?