AI-Based Low-Dose Chest CT Assessment of Coronary Artery Calcification Score: An Analysis of Accuracy-Relating Factors

Objective: This study aimed to assess the accuracy of AI-based coronary artery calcium scoring on non-gated low-dose chest CT and to identify factors influencing calcium risk classification, using ECG-gated non-contrast cardiac CT as the reference standard. Methods: A total of 312 patients with chest pain were prospectively included and underwent ECG-gated non-contrast cardiac CT plus an additional low-dose chest CT. Images from the low-dose chest CT were reconstructed using both a lung kernel and a standard reconstruction kernel. AI software automatically generated coronary artery calcium scores (Agatston, mass, volume and total CACS) on both scan types. Patients were stratified by Agatston-based calcium risk, heart rate (≤ / > 65 bpm) and BMI (< / ≥ 25 kg/m²). Agreement and systematic bias between non-gated low-dose CACS and the reference ECG-gated CACS were evaluated with paired tests and linear regression, and confusion matrices were used to assess risk-stratification performance in each subgroup and HR–BMI combination. Results: Pulmonary-kernel LDCT yielded Agatston scores consistent with ECG-gated CT, without significant differences across heart-rate, BMI, or calcium-severity subgroups. Under the pulmonary kernel, measurement bias was weakly but significantly correlated with heart rate and BMI (r = 0.145, r = 0.131; both P < 0.05). Overall, CACS derived from non-gated low-dose chest CT showed excellent correlation with the reference standard (r > 0.94). The low-HR/low-BMI subgroup achieved the best performance for calcium risk stratification, with 80% accuracy and good agreement (κ = 0.72). Conclusion: Reconstruction kernel, heart rate, BMI, and calcium severity significantly influence the accuracy of AI-based CACS derived from non-gated low-dose chest CT. Compared with standard reconstruction, the pulmonary kernel provides more accurate calcium scoring and better coronary artery disease risk stratification, particularly in patients with low heart rate and low BMI. Integrating AI-based coronary calcium assessment into low-dose chest CT for lung cancer screening is feasible, offering dual benefits from a single scan and supporting early cardiovascular disease prevention. Trial registration: Chinese Clinical Trial Registry (2025.04.02, ChiCTR2500100073)