iCAC Device

The iCAC Device is a software tool that employs a deep-learning algorithm to opportunistically estimate coronary artery calcium presence and quantity from routine non-gated, non-contrast chest CT scans. It provides physicians with adjunctive, quantitative (exact Agatston score) and four-category estimates of calcium burden in Agatston units to aid in patient assessment.

The iCAC Device is a standalone software as a medical device (SaMD) intended to assist physicians by estimating the presence and quantity of coronary artery calcium (CAC) in patients aged 30 years and older. It analyzes non-gated, non-contrast chest computed tomography (CT) images, typically acquired during routine care like lung cancer screenings, using a deep-learning-based computer vision algorithm. The software, which accepts DICOM-formatted images from clinical imaging databases, automatically processes these scans to output a visual representation of the estimated CAC segmentation. Crucially, it provides both an exact Agatston score and classifies the patient's CAC burden into one of four quantitative categories, serving as adjunctive information for the physician. The device aims to leverage the frequent use of routine chest CTs for opportunistic CAC identification and quantification, providing informational support to the reviewing physician without replacing the original scan or report. Its efficacy was assessed in a retrospective study using 762 paired CT studies (a non-gated, non-contrast chest CT and a gated coronary CT for ground truth) from nine geographically diverse sites. This pivotal testing demonstrated an overall agreement of 79.7% between the iCAC Device's calcium score group outputs and the ground truth categories. Furthermore, the positive predictive values for each of the four calcium score groupings ranged from 68.1% to 90.1%, meeting the study's predefined acceptance criteria.

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 iCAC deep-learning algorithm's performance vary across different CT scanner models and acquisition parameters commonly used for routine non-gated chest CTs?; What is the anticipated clinical impact or benefit of the iCAC device providing an exact Agatston score in addition to four detection categories, compared to predicate devices that may offer fewer categories or no exact score?; What were the specific demographic and comorbidity characteristics of the 762 patient studies in the pivotal testing dataset, and how might these affect the algorithm's generalizability to broader patient populations?; Beyond overall agreement and positive predictive value, what were the negative predictive value, sensitivity, and specificity of the iCAC Device for detecting any coronary artery calcium (e.g., Agatston score > 0) in the retrospective study?; How does the iCAC device handle common image artifacts or variations in image quality often present in non-gated, non-contrast chest CTs, and what is its reported failure rate or the need for manual review in such scenarios?