A Framework for Cross-Domain Generalization in Coronary Artery Calcium Scoring Across Gated and Non-Gated Computed Tomography

TL;DR

This paper presents an automated, cross-domain framework for coronary artery calcium scoring that works on both gated and non-gated CT scans, enabling broader cardiovascular risk assessment without extra scans.

Contribution

The authors introduce CARD-ViT, a self-supervised Vision Transformer trained solely on gated CT data, enabling accurate CAC scoring across different scan types without additional non-gated training data.

Findings

Achieves 0.707 accuracy and 0.528 Cohen's kappa on non-gated data, matching models trained on non-gated scans.

Attains 0.910 accuracy and Cohen's kappa above 0.87 on gated test sets across datasets.

Demonstrates robust cross-domain CAC scoring, supporting scalable cardiovascular screening.

Abstract

Coronary artery calcium (CAC) scoring is a key predictor of cardiovascular risk, but it relies on ECG-gated CT scans, restricting its use to specialized cardiac imaging settings. We introduce an automated framework for CAC detection and lesion-specific Agatston scoring that operates across both gated and non-gated CT scans. At its core is CARD-ViT, a self-supervised Vision Transformer trained exclusively on gated CT data using DINO. Without any non-gated training data, our framework achieves 0.707 accuracy and a Cohen's kappa of 0.528 on the Stanford non-gated dataset, matching models trained directly on non-gated scans. On gated test sets, the framework achieves 0.910 accuracy with Cohen's kappa scores of 0.871 and 0.874 across independent datasets, demonstrating robust risk stratification. These results demonstrate the feasibility of cross-domain CAC scoring from gated to non-gated domains, supporting scalable cardiovascular screening in routine chest imaging without additional scans or annotations.