Classification of hemodynamically significant stenoses from dynamic CT perfusion and CTA myocardial territories
Marco Giordano ( Erasmus Universiteit Rotterdam, TU Delft - ImPhys/Quantitative Imaging)
Dirk Poot (TU Delft - ImPhys/Quantitative Imaging, Erasmus Universiteit Rotterdam)
Adriaan Coenen ( Erasmus Universiteit Rotterdam)
Theo van Walsum ( Erasmus Universiteit Rotterdam)
Michela Tezza (University of Verona)
Koen Nieman ( Erasmus Universiteit Rotterdam)
Wiro Niessen (TU Delft - ImPhys/Quantitative Imaging, Erasmus Universiteit Rotterdam)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
Purpose
Myocardial blood flow (MBF) obtained by dynamic CT perfusion (CTP) has been recently introduced to assess hemodynamic significance of coronary stenosis in coronary artery disease. The diagnostic performance of dynamic CTP MBF is limited due to subjective interpretation of MBF maps and MBF variations caused by physiological, methodological, and technical issues. In this paper, we introduce a novel method to quantify the hypoperfused volume (HPV) in myocardial territories derived from CT angiography (CTA) to overcome the limitations of current dynamic CTP MBF analysis methods.
Methods
The diagnostic performance of HPV in classifying significant stenoses was evaluated on 22 patients (57 vessels) that underwent CTA, CTP and invasive fractional flow reserve (FFR). FFR was used as the standard of reference to determine stenosis significance. The diagnostic performance was compared to that of the mean MBF computed in regions manually annotated by an expert (MA-MBF). HPV was derived by thresholding the MBF in myocardial territories constructed from CTA by locating the closest artery. Diagnostic performance was evaluated using leave-one-case out cross-validation. Inter-observer reproducibility was assessed by performing annotations of coronary seeds (HPV) and manual regions (MA-MBF) with two users. In addition, the influence of different parameter settings on the diagnostic performance of HPV was assessed.
Results
Leave-one-case out cross-validation showed that HPV has an accuracy of 72% (58–83%) with sensitivity of 72% (47–90%) and specificity of 72% (58–83%). The accuracy of MA-MBF was 70% (57–82%) with a sensitivity of 50% (26–74%) and a specificity of 79% (64–91%). The Spearman correlation and the kappa statistic was (ρ = 0.94, κ = 0.86) for HPV and (ρ = 0.72, κ = 0.82) for MA-MBF. The influence of parameter settings on HPV based diagnostic performance was not significant.
Conclusions
The proposed HPV accurately classifies hemodynamically significant stenoses with a level of accuracy comparable to the mean MBF in regions annotated by an expert. HPV improves inter-observer reproducibility as compared to MA-MBF by providing a more objective criterion to associate the stenotic coronary with the supplied myocardial territory.