156 Chapter 7 Sub-analysis of the CE-MARC study showed that in ischemic myocardium perfused by a significantly obstructed coronary artery (defined by quantitative coronary angiography; QCA, as ≥ 70% stenosis in any coronary artery of ≥ 2 mm diameter or ≥ 50% in the left main stem) rest and stress MBF were 1.23 ± 0.41 and 2.16 ± 0.70 ml/g/min, respectively.37 In the study by Hsu et al., ischemic myocardium supplied by a vessel with significant obstruction (defined by QCA as ≥ 70% stenosis) had mean rest MBF 0.74 ± 0.25 ml/g/min and stress MBF 0.92 ± 0.36 ml/g/min.19 Kotecha et al.24 showed, that the mean stress MBF in myocardial territories supplied by arteries with FFR ≤ 0.8 was 1.47 ± 0.48 ml/g/min, while in the study by Everaars et al.16, mean rest and stress MBF in myocardium supplied by non-culprit stenosed vessels with FFR ≤ 0.8 were 0.91 ± 0.29 ml/g/min and 1.90 ± 0.59 ml/ min/g, respectively.24, 38 Although comparable to the study by Everaars et al.16, the mean value of stress MBF within ischemic myocardium in the current analysis is slightly higher when compared to the study by Kotecha et al.24 This discrepancy may partly be related to differences in ischemic burden, severity of obstruction, low number of patients and other factors, including pharmacotherapy, which has not been taken into account in calculating the average value of stress MBF within ischemic coronary territory. Nevertheless, the proposed workflow provides QP results consistent with values reported in the literature on dual-sequence approach. Finally, the results suggest that the presented workflow is promising for achieving high performance in the diagnosis of obstructive CAD. The optimal cut-off value of stress MBF to detect significant coronary stenosis obtained in this study was comparable to the study by Kotecha et al.24 and Everaars et al.16 where FFR was used to define physiologically relevant obstruction24,38. However, Kotecha et al. showed higher AUC (0.90) and specificity (81%)24. One of the causes may be higher prevalence of CMD in the studied population, although this has not been routinely assessed. Moreover, it has to be noted, that one of the major limitations of perfusion CMR is the presence of the dark rim artifact. Since the dark rim artifacts are associated with loss of SI, they may give false-positive results in QP analysis and reduce the specificity and overall accuracy of the method. To date, the diagnostic accuracy and clinical utility of pixel-wise color maps generated by the QP CMR software has not been widely studied. Biglands et al. showed in a CEMARC sub-study, that QP CMR (ROC curve analysis of stress MBF and MPR) has high diagnostic accuracy for detecting obstructive CAD, although this was not superior to visual analysis of conventional gray-scale first-pass perfusion images.37 However, Villa et al. have shown, that although diagnostic accuracy of QP CMR is not significantly different to visual conventional assessment performed by a level 3 CMR expert, it significantly outperforms the assessment made by level 1 and 2 readers.32 Our results show that analysis of only color pixel-wise QP maps yield higher sensitivity to detect obstructive CAD than when combined with assessment of gray-scale images. On the other hand, the co-assessment of QP pixel maps and conventional first-pass perfusion images provides higher specificity, PPV, NPV and overall diagnostic accuracy. Presence of dark-rim artifacts may give a
RkJQdWJsaXNoZXIy MTk4NDMw