HealthCCSng

HealthCCSng is an AI-powered software tool that analyzes existing non-gated, non-contrast CT scans to automatically detect and quantify coronary artery calcium, thereby identifying potential risk for coronary artery disease in adult patients. Its science relies on a deep-learning algorithm to process DICOM images, segment the heart, measure calcification based on Hounsfield Units, and output an exact score and risk category for clinical review.

HealthCCSng by Nano-X AI Ltd. is a non-invasive, artificial intelligence-driven post-processing software designed to evaluate calcified plaques in coronary arteries, which are indicative of coronary artery disease risk. It operates by analyzing routine non-gated, non-contrast CT studies that encompass the entire heart, specifically for adult patients aged 30-85. The software employs an AI algorithm for the automated detection and estimation of coronary calcium, receiving DICOM studies from imaging platforms and integrating its output into the standard radiological workflow. This methodology aims to leverage the widespread use of CT scans for other purposes, such as lung cancer screening, to enable opportunistic cardiac risk assessment, functioning as a non-diagnostic aid for clinicians. Upon analysis, HealthCCSng generates an exact coronary calcium score and classifies it into one of four detection categories: Zero (0 HU), Low (1-99 HU), Medium (100-399 HU), or High (≥400 HU). It also provides preview axial images showing annotated calcifications (130 HU threshold) for informational purposes, though these results are not intended for standalone diagnostic use. Performance validation on 427 CT scans showed an overall agreement of 89.46% with ground truth categories (defined by radiologists), with specific category agreements of 86.63% (Zero), 87.65% (Low), 87.36% (Medium), and 98.85% (High). The software-derived calcium scores demonstrated a strong correlation (Pearson's r=0.959) with ground truth scores and showed good agreement in a sub-study comparing non-gated to gated CT scan analyses.

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, models, and acquisition parameters (e.g., radiation dose, reconstruction kernels) beyond the validated range of slice thickness (up to 5.1 mm)?; What is the clinical impact of the 'Success No Category - Cause: metal artifact(s) suggestive of known cardiac disease is suspected' output, and how frequently does this occur in a real-world clinical setting?; Given the 89.46% overall agreement with ground truth, what are the common characteristics of CT scans or patient conditions that lead to discordant results between HealthCCSng and radiologist assessment, particularly for misclassifications between Zero and Low categories?; How does the 'opportunistic' use of HealthCCSng on non-gated, non-contrast CTs compare in terms of long-term patient risk stratification and clinical outcomes versus dedicated cardiac CT calcium scoring with ECG-gating?; What is the training methodology for the AI algorithm, including the size and diversity of the dataset used for its development, and how does it mitigate potential biases related to patient demographics or imaging characteristics not fully represented in the validation set?