P6179Correlation of computed tomography-based fractional flow reserve with instantaneous wave free ratio to detect hemodynamically significant coronary stenoses

Abstract

Abstract Background Based on coronary computed tomography angiography (cCTA), stenoses can be detected but provides anatomical assessment solely. Fractional flow reserve based on coronary CT angiography (ML-cFFR) is gaining in importance for non-invasive hemodynamic assessment of obstructive coronary artery disease (CAD), as several large trials demonstrated significantly improvements in diagnostic accuracy to cCTA. Comparably instantaneous wave free ratio (iFR) is a novel resting index for the invasive determination of haemodynamic relevant stenoses, finds consideration in the ESC guideline on myocardial revascularization and is now of equal standing with FFR as a class IA recommendation. Purpose The aim of our study was to evaluate the on-site ML-cFFR in terms of diagnostic accuracy and clinical practicability in comparison to the iFR as the current invasive gold standard to detect hemodynamically significant coronary artery stenoses. Methods In our prospective, multi-center study, patients with CAD who had a clinically indicated cCTA and subsequent invasive coronary angiography with iFR-measurement were included. To analyse the acquired cCTA dataset we used a third-generation dual-source CT with on-site prototype ML-cFFR software that is based on a machine-learning algorithm, to determine the hemodynamic relevance of coronary stenoses. Results Between July 2017 and December 2018, in 40 of 42 cases (95%), the on-site ML-cFFR calculation was successful. Finally we enrolled 40 patients (72.5% males, mean age 66.7±11.9 years) with ML-cFFR calculation based on cCTA and iFR-measurement during ICA. The mean calculation time of the ML-cFFR values was 10.6±1.9minutes. 57 vessel specific lesions were analysed, of which 15 (26%) were determined as hemodynamically relevant stenoses by iFR (iFR≤0.89) whereas ML-cFFR classified only 14 (24.5%) as hemodynamic significant coronary stenoses (ML-cFFR≤0.80). We observed that cCTA overestimated the severity of stenoses in 27 of 40 cases, which might lead to unnecessary coronary angiographies. However, ML-cFFR detected no obstructive CAD in 26 of 40 patients (65%) and this would have resulted in a reduction of initially performed pure diagnostic coronary angiography. Estimated values sensitivity, specificity, PPV and NPV were 86.7%, 97.4%, 92.9% and 95.0%. The diagnostic accuracy of ML-cFFR in terms of iFR on a per-patient and per-lesion level was 95.0% and 96.5%. The area under the curve (AUC) on a per-lesion and per-patient basis by ML-cFFR to detect lesion specific ischemia was 0.97 and 0.96. The analysis of the correlation (Pearson's product-moment) on a per-lesion level was r=0.82 (p<0.0001) between the ML-cFFR algorithm and iFR. Conclusion(s) On-site ML-cFFR correlates excellently with the novel gold standard iFR to non-invasively detect hemodynamic significant coronary stenoses in routine clinical practice. Acknowledgement/Funding Doctor S. Baumann receives research support from Siemens and Philips Volcano. All other authors declare that they have no financial disclosure.