Cardiac CT Function Software Application

The Cardiac CT Function Software Application utilizes AI/Machine Learning algorithms for semi-automatic segmentation of the left and right ventricles from cardiac CT angiography images to aid clinicians in assessing cardiac function. Its methodology involves processing DICOM images to calculate key volumetric and functional parameters, such as ejection fraction and myocardial mass, with performance validated against expert reader consensus.

The Cardiac CT Function Software Application, developed by Circle Cardiovascular Imaging, is a software tool designed to assist qualified medical professionals in the analysis of cardiac function using cardiovascular CT angiography images. It provides qualitative and quantitative evaluation capabilities by enabling visualization, assessment, and calculation of key cardiac metrics from DICOM standard images. The software features multi-planar reconstruction (MPR) views and 3D rendering, allowing users to manually draw regions of interest or utilize semi-automatic segmentation for the left ventricle (LV) and right ventricle (RV). It is intended as an aid to the standard of care, processing images acquired from various CT scanners without direct interface to the acquisition equipment. A core component of the CT Function Module is its Artificial Intelligence/Machine Learning (AI/ML) algorithm, which detects and segments heart structures (LV cavity, LV myocardium, and RV cavity). These AI-generated segmentations are converted into editable surfaces, emphasizing that users should verify and correct any inaccuracies, with options for full manual segmentation also available. The software calculates critical cardiac functional parameters including end-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), ejection fraction (EF), cardiac output (CO), cardiac index (CI), and LV myocardial mass. Performance validation, conducted against a reference standard from expert readers, demonstrated robust segmentation accuracy for the LV cavity (Dice > 86%, EF bias 1.3%), RV cavity (Dice > 85%, EF bias -5.5%), and LV myocardium (Dice > 82%), with comparable performance across diverse imaging conditions and patient demographics.

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 performance of the AI/ML segmentation for the RV, particularly the EF bias of -5.5%, compare to established clinical acceptance limits or the predicate device's RV performance?; Given the reliance on AI/ML, what is the variability in segmentation accuracy and subsequent metric calculation when encountering images with significant artifacts or rare cardiac pathologies not extensively represented in the training data?; The document mentions validation data was 90% from US sources; how was the algorithm's performance validated on diverse global populations and imaging protocols outside the US?; While LV myocardial mass is calculated, are there plans to extend AI/ML-based segmentation and mass calculation to the RV myocardium, and what are the anticipated challenges?; The predicate device includes regional wall motion analysis and left atrium segmentation; what was the rationale for not including these features, and are they planned for future iterations?