Assessing Aortic Valve Calcification with Contrast-Enhanced Computed Tomography via a Novel Contrast-adjusted Model

S. Uhr (Göttingen)1, S. Gersch (Göttingen)1, K. Toischer (Göttingen)1, H. Dathe (Göttingen)2, J. Gronwald (Göttingen)1, L. Wüstefeld (Göttingen)1, D. Krefting (Göttingen)2, N. Spicher (Kgs. Lyngby)3
1Universitätsmedizin Göttingen Herzzentrum, Klinik für Kardiologie und Pneumologie Göttingen, Deutschland; 2Universitätsmedizin Göttingen Institut für Medizinische Informatik Göttingen, Deutschland; 3Technical University of Denmark Department of Health Technology Kgs. Lyngby, Dänemark

Objective:

This study introduces a novel method for quantifying aortic valve calcification (AVC) using contrast-enhanced computed tomography (CECT) data. Standard quantification of AVC is typically performed using non-contrast-enhanced CT (NCCT), expressed as the Agatston score (AGS). Previous attempts to assess AVC using CECT have been hindered by methodological challenges, including potential over- or underestimation due to fluctuations in blood density within the ascending aorta and variations in contrast agent concentration.

Methods:

To overcome these limitations, this study proposes a novel, contrast-adjusted approach for AVC quantification which is evaluated retrospectively on a novel, single-center dataset of N=234 patients (56.8% females, age: 80.9 ± 6.6 years) with suspected severe aortic valve stenosis who underwent NCCT and CECT. A log-linear regression model for AVC quantification for individual patients at specific contrast agent concentrations was developed. For population-level assessment, a multivariate regression model was employed, and validated through repeated random subsampling in different 70%/30% train/test splits.

Results:

Results outperform state-of-the-art methods by reaching an adjusted R2 value of 0.95. Robust performance in relevant subgroups; e.g. in patients with low flow, low-gradient aortic stenosis are observed with residuals not differing from the rest of the cohort. Moreover, in patients at diagnostically critical range with AGS ≤ 2000, residuals were smaller in median and interquartile range compared to higher AGS patients.

Conclusions:

Our results demonstrate the potential of assessing AVC using CECT alone in a single center study and highlight the added value of contrast adjustment.