Coronary artery calcium scoring using virtual vs. true non-contrast images from ultra-high-resolution and...

To validate the reliability of Coronary Artery Calcium Scoring (CACS) by using virtual non-contrast (VNC) images reconstructed from contrast-enhanced photon-counting coronary CT angiography (CCTA) versus true non-contrast (TNC) images, and to assess the impact on plaque burden classification.

239 patients who underwent contrast-enhanced photon-counting CCTA with TNC imaging were included. Agatston score was calculated on both (VNC and TNC) datasets. Scores ≤ 0.9 were rounded to 0. Patients were divided into plaque groups according to the CAD-RADS 2.0 classification system. Borderline cases (P0 vs. P1) were reassessed by two radiologists with blinded adjudication.

Median Agatston scores showed no significant difference (TNC: 14.2 [range: 0.0-2951.2] vs. VNC: 8.0 [range: 0.0-3069.3]; p = 0.54). Correlation was excellent (intraclass correlation coefficient, 0.97 [95% CI: 0.97, 0.98]. Robust linear regression revealed a significant relationship (intercept = 3.27 [95% CI: 1.28, 5.87]; slope = 0.97 [95% CI: 0.97, 0.98]; p < 0.001). Despite the excellent correlation, a significant difference in plaque classification was observed (p = 0.005). Classification was consistent in 160/239 (67%; κ = 0.88) patients. After a consensus reading, the agreement improved to 205/239 patients (86%; κ = 0.95).

Agatston scores derived from TNC and VNC images show no significant differences and excellent consistency. However, the agreement for plaque classification is limited, preventing VNC images from directly substituting for TNC images for this purpose. While a subjective reading significantly improved agreement, further research is required before clinical implementation.