Quantitative coronary artery and aortic valve calcification by virtual non-contrast Images from dual-energy delayed CT scans in patients with severe aortic stenosis.

Abstract

With the growing use of delayed dual-energy computed tomography (DECT) imaging for extracellular volume assessment, this study aimed to assess the accuracy of coronary artery calcium (CAC) and aortic valve calcification (AVC) quantification in virtual non-contrast (VNC) images derived from delayed DECT images. Eighty-eight patients undergoing pre-transcatheter aortic valve replacement (TAVR) cardiac CTA with DECT were retrospectively enrolled. We compared CAC score (CACS), CAC volume, AVC score (AVCS), and AVC volume between true non-contrast (TNC) and VNC images. VNC images were created by subtracting iodine from delayed CT angiography using commercial post-processing software. Concordance for the likelihood of severe aortic stenosis (AS) between TNC and VNC was assessed using guideline-based AVCS thresholds: men > 3,000 AU, women > 1,600 AU (highly likely); men > 2,000 AU, women > 1,200 AU (likely). CACS was lower in VNC than TNC (363.0 [13.7-1,154.0] AU vs. 470.1 [46.8-1,569.9] AU, p < 0.001), but the correlation was excellent (r = 0.966). No significant difference between VNC and TNC was observed in AVCS (959.5 [321.2-2,475.0] AU vs. 1,079.6 [492.2-2,329.4] AU, p = 0.521), with strong correlation (r = 0.966). Application of a correction factor (1.1-fold) improved agreement between VNC- and TNC-derived CACS (weighted κ = 0.839 to 0.869). Diagnostic concordance for severe AS between VNC and TNC was 92.1% (weighted kappa = 0.785) at the "highly likely" threshold and 97.7% (weighted kappa = 0.952) at the "likely" threshold. Delayed VNC images from DECT demonstrate strong agreement with TNC images for CACS and AVCS, suggesting that delayed VNC could potentially reduce the need for TNC acquisition in pre-TAVR assessment.