In this study, we compared the diagnostic performance between target-µFR and vessel-µFR using the FFR ≤ 0.80 as reference standard. The primary study findings are summarized as follows: 1) the µFR is a novel physiological assessment methods which estimates the pressure drop due to coronary stenosis according to semiautomatic delineation of target vessels and FFR simulation from single-angle. 2) both target-µFR and vessel-µFR demonstrated high correlations and great agreements with FFR; 3) the ability of target-µFR defining hemodynamic significant of coronary stenosis was similar to vessel-µFR; 4) diagnostic performance of target-µFR was slightly better than that of vessel-µFR;5) the selection of the measurement location has less influence on the accuracy of µFR. Based on these findings, it could improve the calculation algorithm for enhancing the time-efficient and showed the potential in the coronary imagine and virtual physiological evaluation of CAD. The ability of µFR highlights that by integration of the imaging information in order to enable a comprehensive assessment of the CAD.
DEFER, FAME, FAME II and FAME III establish FFR as the "gold standard" of coronary physiology for assessing coronary artery stenosis, treatment plan formulation and evaluation of treatment effect[20–23]. However, as an invasive method, the application of FFR requires expensive equipment and has potentially procedure-related complications, such as non-fatal myocardial infarction, cerebrovascular accident, and has been limited in clinical practice because of the invasive of the procedure, requirement of pressure wire, the administration of hyperemic agents and so on[3–5, 24]. To solve these limitations and reduce the complications, the QFR had been developed which is a virtual FFR technique derived from coronary angiography. And QFR derived from Murray Law as a novel angiographic-based method could enable fast computation of FFR, which provides an avenue for determining the most appropriate therapy for the intermediate lesions.
A large number of clinical studies have confirmed the accuracy of QFR in assessing coronary artery function. In FAVOR Pilot Study, the fix-flow QFR (fQFR), contrast-flow QFR (cQFR) and adenosine-flow QFR (aQFR) was compared with the gold standard FFR for evaluate the capability of those QFR in predicting coronary stenosis. The results confirm that fQFR, cQFR and aQFR had shown the great agreement and diagnostic performance (accuracy 80%, 86%, and 87%) for predicting ischemia myocardial[9]. Then, in the FAVOR II China, the sensitivity and specificity in identifying hemodynamically significant stenosis were evidently higher for QFR than for QCA (94.6% vs. 62.5%; 91.7% vs. 58.1%). The FAVOR II China also revealed that vessel-level QFR had a high diagnostic accuracy of 93.3%[10]. In a large study of FAVOR II E/J, the good diagnostic performance of QFR assessed the degree of coronary stenosis (accuracy 86.3%, specificity 86.9%, sensitivity 86.5%, AUC 0.92) and evaluated the calculation time of QFR and FFR. Furthermore, the time to complete QFR (5min) was significantly shorter than the time to complete FFR (7min) [25]. The FAVOR II China and FAVOR II E/J had proved that the diagnostic accuracy of QFR at both the patients and vessels level was better than QCA in the assessment of the relevance of functional stenoses.In a meta-analysis of 16 high quality researches comparing FFR and QFR, QFR has demonstrated good positive and significant negative predictive values in confirming the relevance of the ischemia myocardial on the basis of the FFR cut-off ≤ 0.80[18].Subsequently, Wienemann etal further verified the diagnostic performance of cQFR (sensitivity, specificity, positive predictive value, negative predictive value, and diagnostic accuracy were 82%, 95%, 97%, 92% and 91%, respectively). The area under the curve shown for cQFR is greater than that for fQFR and RFR (0.938 vs. 0.891,0.869, P < 0.01). The good diagnostic performance of cQFR was maintained in different clinical subpopulations (including gender, aortic stenosis and atrial fibrillation, etc.) and different anatomical subpopulations (including focal and non-focal lesions, etc.)[26]. In a head-to-head study, QFR showed good agreement with QFR compared with SPECT and PET. Meanwhile, the accuracy of QFR was 88%, 82% for SPECT and 78% for PET[27]. What’s more, the QFR derived from CAG is the only functional angiography system for which the value in clinical practice has been assessed in a randomised clinical trial. The patients, who enrolled in FAVOR III China, with at least one 50–90% coronary stenosis underwent QFR-guide strategies (if QFR ≤ 0.80) or angiographic-based strategies. After 1-year of follow up, patients randomised to the QFR-guide strategies demonstrated better outcomes driven by fewer myocardial infarctions and ischemia-driven revascularisations[28]. Based on those studies, QFR has demonstrated good diagnostic accuracy in detecting myocardial ischemia. Meanwhile, QFR could more quickly and conveniently calculate virtual FFR after CAG without any intervention operations, further providing clinical support for revascularization strategies.
But none of the QFR mentioned above studies is a lesion-specific QFR. In previous studies, they demonstrated that the choice of virtual FFR measurement locations is particular importance when identifying ischemic lesions or guiding treatment strategies[13–15]. Series studies demonstrated that lesion-specific FFR, such as lesion-specific CT-FFR, can reclassify positive patients defined by the vessel-derived FFR value, and that lesion-specific FFR has higher diagnostic performance than vessel-derived FFR[29–32]. The possiable causes of the above phenomennon were as follows: 1) the virtual FFR measurement at far distal segments may overeatimate coronary ischemia, 2) these differences between vessel territories in pressure gradients for segments 1–2 cm distal to the stenosis versus far distal segments relate to the larger territory of perfused myocardium, 3) the virtual FFR were only assessed in the main coronary arteries, which may have disregarded the impact of collateral stenosis on myocardial ischemia[6, 14]. Then, Kołtowski Ł et al analysised the diagnostic performance of index QFR, vessel QFR (assessemnt for entire segmented vessel) and lesion QFR (assessment for the target lesion) to identify the best measurement location for optimal accuracy of QFR. Thie researche demostrated the index QFR value which obtained at the pressure transducer position (R 0.85, accuracy 85.4%, AUC 0.94) had better diagnostic performance than vessel QFR (R 0.78, accuracy 78.5%, AUC 0.90) and lesion QFR (R 0.70, accuracy 76.1%, AUC 0.82)[33]. Of note, the µFR had been showed gteat agreement and correlation with standard three-dimensional quantitative flow ratio (R 0.996)[34]. All the above results are comparable to those shown in our analysis.
In previous studies, the µFR demonstrated powerful and superior diagnostic performance for lesion-specific ischemia compared with angiography alone regardless of coronary calcification[35]. Hence, this paper explores whether target-µFR could further improve the diagnostic ability of myocardial ischemic by comparing the diagnostic performance of target-µFR and vessel-µFR. The diagnostic performance (accuracy 86.40% ,sensitivity 86.44%, specificity 88.51%, and Youden index 0.750) and AUC (0.936) of vessel-µFR were similar to those reported in previous studies. Then, those indexes of target-µFR (accuracy 92.98%,sensitivity 92.98%, specificity 91.01%, Youden index 0.840, and AUC 0.937) were slightly better than vessel-µFR and previous research. At the same time, we found the calculation time of target-µFR (4-5min) and vessel-µFR (4-5min) was also similar to the QFR reported in previous papers. Becaose of that, the µFR could assess the degree of coronary stenosis which is a time-efficient and accurate method, the visualized anatomic geometry of the coronary artery can provide guidance for subsequent therapeutic regimens. By the way, the µFR calcualted the pressure loss by the frictional loss along the lesion entrance and stenotic segment as well as the inertial loss stemming from the sudden expansion of the flow as it emerged from the stenosis were calculated, based on the stenosis geometry and the hyperemic flow rate. What’s more, based on the U-Net architecture, Murray Law and artificial intelligence, µFR automatically outlines the lumen of the target vessels and their collaterals through artificial intelligence[12]. All in all, we suggest the accuracy of µFR have almost less influence by the selection of the remote measurement locations. At the same time, µFR behaved similarly well in sexes and has great diagnostic performance, indicating its potential as a reliable wireless tool for identifying functional ischemia[36].
However, our study had several limitations. First, it was a single-center and retrospective study with a small sample size. This might have introduced selection bias even though consecutive patients were included. The limited number of enrolled patients due to the low adoption rate of patients undergoing FFR in clinical practice also affected the statistical efficiency of the study. Secondly, not all the vessels were interrogated for the enrolled patients. The vessels with diameter stenosis < 30% or > 90% were not assessed because performing physiological assessments in such lesions was unnecessary. Thirdly this is a retrospective analysis in which one-third of the data were excluded because the QFR assessment was not applicable, the study should more likely be viewed as a hypothesis-generating study, and further prospective studies would provide more evidence. The availability of QFR can be improved by requiring careful attention to the projection angle and location of the target lesions in coronary angiography; however, the extent to which this can be improved remains to be assessed. Forth, target-µFR and vessel-µFR computation require automatic reconstruction of 3D anatomical models of coronary vessels, and further studies should be consider that analyze the impact of anatomical features on diagnostic accuracy in target vascular lesions. Fifth, there may be inter-operator differences and previous PCI operation effect in the target-µFR and vessel-µFR calculation process, so further evidences from larger studies are needed. Sixth, the selection of the measurement location depends on the location recorded during the FFR evaluation procedure. Influenced by the real world, some of the measurement locations cannot be accurately positioned ar rhe 2-3cm distal to the target vessels. Hence, further large-sample. Multicenter, prospective, randomized sthudies are needed to further confirm the feasibility of target-µFR in clinical practice.