Diagnostic Performance of In‐Procedure Angiography‐Derived Quantitative Flow Reserve Compared to Pressure‐Derived Fractional Flow Reserve: The FAVOR II Europe‐Japan Study

Background Quantitative flow ratio (QFR) is a novel modality for physiological lesion assessment based on 3‐dimensional vessel reconstructions and contrast flow velocity estimates. We evaluated the value of online QFR during routine invasive coronary angiography for procedural feasibility, diagnostic performance, and agreement with pressure‐wire–derived fractional flow reserve (FFR) as a gold standard in an international multicenter study. Methods and Results FAVOR II E‐J (Functional Assessment by Various Flow Reconstructions II Europe‐Japan) was a prospective, observational, investigator‐initiated study. Patients with stable angina pectoris were enrolled in 11 international centers. FFR and online QFR computation were performed in all eligible lesions. An independent core lab performed 2‐dimensional quantitative coronary angiography (2D‐QCA) analysis of all lesions assessed with QFR and FFR. The primary comparison was sensitivity and specificity of QFR compared with 2D‐QCA using FFR as a reference standard. A total of 329 patients were enrolled. Paired assessment of FFR, QFR, and 2D‐QCA was available for 317 lesions. Mean FFR, QFR, and percent diameter stenosis were 0.83±0.09, 0.82±10, and 45±10%, respectively. FFR was ≤0.80 in 104 (33%) lesions. Sensitivity and specificity by QFR was significantly higher than by 2D‐QCA (sensitivity, 86.5% (78.4–92.4) versus 44.2% (34.5–54.3); P<0.001; specificity, 86.9% (81.6–91.1) versus 76.5% (70.3–82.0); P=0.002). Area under the receiver curve was significantly higher for QFR compared with 2D‐QCA (area under the receiver curve, 0.92 [0.89–0.96] versus 0.64 [0.57–0.70]; P<0.001). Median time to QFR was significantly lower than median time to FFR (time to QFR, 5.0 minutes [interquartile range, –6.1] versus time to FFR, 7.0 minutes [interquartile range, 5.0–10.0]; P<0.001). Conclusions Online computation of QFR in the catheterization laboratory is clinically feasible and is superior to angiographic assessment for evaluation of intermediary coronary artery stenosis using FFR as a reference standard. Clinical Trial Registration URL: https://www.clinicaltrials.gov. Unique identifier: NCT02959814.

P hysiological assessment is the clinical standard to guide percutaneous coronary interventions of intermediate coronary stenosis. Following the FAME (Fractional Flow Reserve Versus Angiography for Multivessel Evaluation [fractional flow reserve versus angiography for guiding percutaneous coronary intervention]) trials, the adoption of fractional flow reserve (FFR) has improved with a 16-fold increase in FFRguided percutaneous coronary intervention in the United States from 2008 to 2012. 1 Globally, the use of physiological lesion assessment remains low, with large areas performing less than 15% of eligible procedures with physiology guidance. 2,3 To further expand the use of physiological-guided percutaneous coronary intervention, coronary computed tomography angiography-and invasive coronary angiography-based computation methods were developed for less-invasive FFR approximation. [4][5][6][7][8][9][10] Quantitative flow ratio (QFR) is a method for fast computation of FFR based on 3-dimensional quantitative coronary angiography (3D-QCA) and estimation of contrast flow velocity during invasive coronary angiography. The optimal approach was validated in the FAVOR (Functional Assessment by Various Flow Reconstructions) multicenter study, proving that QFR can be computed without pharmacology-induced hyperemia. 11 In FAVOR, QFR was computed post hoc in a core-lab setting. The FAVOR II China study, conducted in parallel to FAVOR II Europe-Japan (E-J), showed a high diagnostic accuracy of in-procedure QFR. 12 In FAVOR II E-J, we aimed to validate the in-procedure feasibility and compare the diagnostic performance of QFR computation with 2-dimensional quantitative coronary angiography (2D-QCA) in a multicenter setting, using FFR as a reference standard.

Primary Comparison
The primary comparison was sensitivity and specificity of QFR compared with 2D-QCA to detect hemodynamically significant coronary lesions with FFR as a gold standard. For FFR and QFR, significant obstructions were defined as FFR and QFR ≤0.80 whereas >50% diameter stenosis (% DS) was used for 2D-QCA. Sample-size calculation and a full list of secondary comparisons are provided in Data S1 and Table S1.

Patient Population
Patients with stable angina pectoris or patients scheduled for secondary evaluation of stenosis after acute myocardial infarction were eligible for enrollment when the angiographic inclusion criteria were met; indication for FFR measurement (at least 1 lesion with % DS 30-90 in a vessel with reference size >2.0 mm). Exclusion criteria were: acute myocardial infarction within 72 hours; severe asthma or severe chronic obstructive pulmonary disease; allergy to contrast media or adenosine; or atrial fibrillation. All inclusion and exclusion criteria are listed in Table S2.

Ethics
The study was approved by the Central Denmark Region Committees on Biomedical Research Ethics. Approval by local or national medical ethics committees was obtained by the local or national coordinating investigators as required for the individual sites. The Danish Data Protection Agency approved the study. All enrolled patients provided written informed consent. J.W. and N.R.H. had full access to all data in the study. All authors are responsible for integrity of the analysis. The data will not be made available to other researchers for purposes of reproducing the results or replication the procedure because of competitive reasons.

Clinical Perspective
What Is New?
• Quantitative flow ratio (QFR) estimates fractional flow reserve based on computation of 2 standard angiographic projections. • Online QFR performed during invasive angiography is feasible and can be computed within the time of conventional fractional flow reserve measurement. • QFR has superior sensitivity and specificity for detection of functional significant lesions in comparison with 2-dimensional quantitative coronary angiography using fractional flow reserve as reference.
What Are the Clinical Implications?
• QFR may broaden the access to physiological lesion assessment in diagnostic catheterization laboratories and centers with low adoption of pressure-wire-based diagnostic strategies. • Randomized trials are required to confirm that QFR provides noninferior clinical outcome compared to assessment of intermediate coronary stenosis by pressure wire.

Invasive coronary angiography
Nitroglycerine (100-200 lg IC) was administrated after acquiring the first angiographic projection. If FFR was indicated in 1 or more vessels, 2 study projections were obtained for each lesion of interest at a minimum of 12.5 frames per second. Selection of projections aimed for minimal vessel foreshortening and minimal vessel overlap by: (1) brisk, continuous and fast contrast injections and (2) no zooming and movement of the table and visualization of the entire vessel to the intended location of the pressure transducer. A table of recommended projection angles was provided for all study sites (Table S3). Images were transferred to a workstation for computation of QFR following site-specific blinding protocols (Data S1). The remaining diagnostic invasive coronary angiography and further interventions were performed per normal clinical practice.

QFR computation
QFR was computed with the CE-marked software; QAngio XA-3D/QFR solution (Medis medical imaging system bv., Leiden, The Netherlands). An end-diastolic frame was selected for each study projection and was used for the 3-dimensional reconstruction of the segmented vessel. The reference vessel was constructed by fitting to healthy segments preferably proximal and distal to the lesion of interest. The following quality checks of the reference vessel reconstruction were performed: vessel tapering; good correspondence between the 2 images used for reconstruction; the reference should not follow aneurysmatic sections; and realistic proximal sizing per sex and race. The contrast frame count was performed in an angiographic run with contrast movement clearly visualized and preferably with frames from the same cardiac cycle. 13 The detailed standard operating procedure for QFR computation is presented in Data S2. All analyses were repeated in a core-lab setting (CardHemo, Med-X Research Institute, Shanghai Jiao Tong University, Shanghai, China). Frame count based contrast-QFR was used for all analysis.

Continuous Feedback
During the enrollment period, all sites received day-to-day feedback from the QFR and FFR core labs on image acquisition quality, pressure wave-form quality, and adherence to the standard operating procedure for QFR analysis.

Statistical Analysis
Baseline characteristics and procedural characteristics were presented as count and percentages, continuous variables as mean and SD, if normally distributed, or otherwise reported as medians and interquartile range. Feasibility was calculated as the fraction of successful QFR computations of lesions with successful FFR measurements. The primary comparison was calculated as superiority for sensitivity and specificity of QFR (in-procedure value) in comparison with 2D-QCA (Table S1). Sensitivity and specificity for 2D-QCA and QFR were compared using McNemar's test. Negative predictive value, positive predicate value, positive likelihood ratio, and negative likelihood ratio for 2D-QCA and QFR were compared using generalized score statistics. Time to FFR and QFR were compared using Wilcoxon's rank test. The diagnostic performance of QFR compared with 2D-QCA was assessed by 2-tailed paired comparison of receiver operating characteristics curves (DeLong's method). Pearson's correlation was used to quantify the correlation between QFR and FFR. Agreement between QFR and FFR was assessed by Bland-Altman plots. Observations in patients with more than 1 study vessel were presumed independent. This assumption was evaluated by repeated analysis on a per-patient level. If multiple measurements were performed, the lowest FFR and corresponding QFR and % DS (2D-QCA) values were compared with per-patient analysis. Reproducibility was assessed as interobserver variation by Bland-Altman and scatter analysis of in-procedure QFR and core-lab QFR. The diagnostic performance of core-lab QFR compared with in-procedure QFR was assessed by 2-tailed paired comparison of receiver operating characteristics curves (DeLong's method) using FFR as a reference. Subgroup analysis for QFR accuracy was performed per FFR strata, per vessel, and for single versus tandem lesions. The diagnostic performance of 3-dimensional quantitative coronary angiography-derived % DS and area stenosis was compared with QFR with FFR as a reference standard using receiver operating characteristics curves (DeLong's method). Analysis was performed using STATA (version 13; StataCorp LP, College Stadion, TX) and R software (R Foundation for Statistical Computing, Vienna, Austria).

Results
Three hundred twenty-nine patients were included from February 22, 2017 to October 17, 2017 (Table S4). Inprocedure QFR was computed in 345 (96%) vessels with successful FFR measurements. After exclusion based on predefined FFR core-lab criteria, 272 patients and 317 vessels were included in the final analysis (patient flow chart in Figure 1 and vessel-level flow chart in Figure S1). Mean FFR was 0.83AE0.09 ( Figure S2), and mean % DS (2D-QCA) was 45AE10%. An FFR ≤0.80 was found in 104 (33%) vessels. Baseline and procedural characteristics are listed in Tables  P=0.002; Figure 2 Figure S3). Additional results of diagnostic comparisons are listed in Table 3.

Hybrid Model Limits
QFR limits to yield specificity and sensitivity >95% with FFR as a reference were 0.77 (QFR-treat) and 0.87 (QFR-defer).
Applying the 95% limits to this sample, use of pressure wires and adenosine could theoretically have been avoided in 64% of all measurements yielding 95% accuracy with FFR as a reference standard ( Figure S8). Applying a 100% limit (QFR-treat 0.64 and QFR-defer 0.93) to this sample, pressure wires and adenosine theoretically were not required in 21% of measurements yielding 100% accuracy with FFR as a reference standard. This analysis assumes that FFR is 100% accurate. The trade-off for pressure-wire-free procedures depending on aimed accuracy with FFR as a reference is illustrated in Figure S9.

Subgroup Analysis
The    Figure S10). Diabetes mellitus showed statistical significant association with increased QFR-FFR discrepancy (Table S6).

Discussion
The FAVOR II E-J study and the FAVOR II China study were the first multicenter studies investigating the feasibility and value of in-procedure QFR calculated in the catheterization laboratory. The main findings of FAVOR II E-J were: (1) The study confirmed the primary hypothesis with superior specificity and sensitivity of QFR compared with standard anatomical assessment by 2D-QCA with FFR as a reference standard, and (2) QFR was feasible in a multicenter setting and was faster than FFR when analyzed during coronary angiography. Diagnostic performance of QFR in FAVOR II E-J was noteworthy and comparable with the findings in the recent and almost similar FAVOR II China study. 12 SDs for mean difference FFR-QFR were identical (0.06). The higher accuracy (92.7%) in FAVOR II China may be explained by the smaller number of lesions with FFR values close to the FFR 0.80 cutpoint. Results in both studies showed improved performance of QFR compared with early validation studies on offline computation of QFR. 11,15 The improved precision may be facilitated by the online analysis setup with instant feedback between operator and analyst. The standard operating procedure (Data S2), use of recommended angulations for angiographic projections (Table S3), and day-to-day feedback on enrolled cases may further have contributed to the improved results of QFR.
Most existing FFR computation methods for invasive coronary angiography predominantly rely on computational fluid dynamics. 5,6,16 Inherited limitations of these methods may exist related to generating theoretical boundary conditions to create a "one-size-fits-all", and to long computation time for blood flow simulations. Morris et al recently presented a rapid computational fluid dynamics modality for calculation of virtual FFR with a high diagnostic precision (100% for FFR ≤0.80) and short mean time to virtual FFR (189 seconds). 17 This study was performed using rotational angiography in a limited population of 20 patients. To our knowledge, the FAVOR II studies using QFR present the first data supporting that real-time computation of FFR is feasible,   fast, and accurate in patients with stable angina pectoris and applicable stenosis. FFR is the established standard for invasive identification of flow limiting intermediate coronary lesions when no other objective evidence of lesion specific ischemia is present. 18 The clinical adaption of FFR is increasing, but remains low. 1,2,19 The underlying reasons may include the high cost of pressure wires, tortuous vessels, and the need for pharmacological hyperemia induction. Multiple studies presented approaches to avoid hyperemia for physiological lesion assessment, such as instantaneous wave-free ratio and resting distal pressure/aortic pressure measurements. The resting indices perform similar with an overall diagnostic agreement between 80% and 90% when compared with FFR depending on distribution of lesions included in the studies. [20][21][22][23][24] Still, instantaneous wave-free ratio-based strategies versus an FFR strategy resulted in comparable clinical outcomes at 1 year in 2 large, randomized clinical trials. 23,24 We found a diagnostic accuracy for QFR (87%) comparable to the early instantaneous wave-free ratio/FFR studies. Hence, the presented results support future comparison of FFR and QFR in clinical outcome trials.
Repeated core-lab QFR analysis confirmed the agreement between QFR and FFR (identical SD of 0.06). However, direct comparison of in-procedure QFR and core-lab QFR revealed a small bias. The discrepancy indicates that the standard operating procedure for QFR computation might not have been sufficiently standardized for some lesion presentations or training was insufficient before study start. Core-lab QFR showed less variation in disagreement at lower FFR values ( Figure S5), indicating that contouring tight lesions could pose a specific challenge. Computation of QFR requires user interaction at steps, such as frame selection, lumen contouring, and contrast flow evaluation, and may hence be sensitive to small differences in the approach at various steps. A more elaborate standard operating procedure, more observer training, and automatizations are likely to reduce variation.
We showed that QFR is superior to standard quantitative coronary angiography in evaluating coronary artery stenosis. QFR may extend the access to physiology-based guidance when access to pressure wires is limited by financial restrictions or inexpedient reimbursement systems. By enrolling patients where FFR is normally indicated, we included a distribution of lesions with a mean FFR approaching the clinical 0.80 cutpoint (mean FFR, 0.83AE0.09). The vast majority of binary mismatches (treat/no-treat) between QFR and FFR were cases close to the binary diagnostic cutoff, in whom the benefit of treatment approaches the percutaneous coronary intervention-related event rate. 25 Although the study was not powered to do so, the sample allowed for the predefined assessment of a QFR-FFR hybrid approach, which may reflect the true clinical application of QFR in centers with full adoption of physiology-based diagnostics awaiting results of randomized outcome trials. Applying the 95% QFR-hybrid limits (QFR-treat 0.77 and QFR-defer 0.86) to  this population could potentially save pressure wires and adenosine in 64% of all lesions ( Figure S8) and still ensure a diagnostic quality at the level of full FFR evaluation until clinical noninferiority of a QFR-based diagnostic strategy has been established.

Study Limitations
We only enrolled a limited portion of patients scheduled for secondary evaluation of coronary lesions after myocardial infarction. The diagnostic precision of QFR in nonculprit lesions, as recently assessed in a proof concept study by Spitaleri et al, could thus not be confirmed. 26 We excluded lesions with Medina type 1.1.1 and 1.0.1 bifurcations attributed to specific limitations of the present QFR application; hence, the diagnostic precision of QFR in bifurcation needs further developments and investigation. Despite the inclusion of tandem lesions, we did not mandate FFRpullbacks during intravenous adenosine. Thus, a direct comparison between the FFR-pullback curves and the spatially sensitive, color-coded, continuous QFR values along the 3-dimensional/angiographic roadmap could not be performed. Because FFR was the sole gold-standard, we were not able to further characterize the lesion physiology in the presence of microvascular dysfunction. Time to QFR did not include the time for angiographic acquisition that could differ from an FFR-based strategy. It is therefore not possible to determine whether use of the provided standard projections and requirement for limited overlap and foreshortening added procedure time. To emulate an integrated QFR solution, data transfer time from angiographic equipment to the QFR workstation was not included in time to QFR. In case of selection of a different view during analysis, the additional time was included in the time to QFR. Furthermore, preparing and zeroing the pressure system was not included in time to FFR because of site-specific differences in the workflow.

Conclusion
In-procedure QFR is clinically feasible and is superior to angiographic assessment for evaluation of intermediary coronary artery stenosis when FFR is used as a reference. QFR bears the potential to expand the adoption of physiological lesion assessment. acknowledges the support from Aarhus University (PhD scholarship).

Sources of Funding
The study was funded by the Department of Cardiology, Aarhus University Hospital, Skejby and by the participating institutions. The manufacturer and distributor of the QFR software (Medis Medical Imaging bv., Leiden, NL) was not involved in design, conduct, or reporting of the study and provided no funding for the study except for making the Medis Suite solution available for free in the study period and provided training for participating sites.

Disclosures
Westra received travel support and consultant fees from Medis Medical Imaging systems bv. Shengxian Tu received research from Medis Medical imaging systems bv. and Pulse medical imaging technology. Wijns received research grants (to his former institution) from stent manufacturing companies and speaker fees and honoraria from Biotronik, Mi-Cell, and MicroPort. He is a co-founder of Argonauts, an innovation facilitator. His research is supported by Science Foundation Ireland (Dublin). Holm received institutional research grants from Abbott, Boston Scientific, and Medis medical imaging. The remaining authors have no disclosures to report. Data S1.

Sample size calculation
Estimates for the sample size calculation were based on the results from the FAVOR study, where a sensitivity of 0.74 and a specificity of 0.91 for QFR were found. The null hypothesis was H0:

Feasibility
The feasibility was assessed as the fraction of lesions with successful FFR measurements where QFR was computed.

Time to QFR and FFR
Time to QFR was defined as start of frame selection for the three-dimensional reconstruction of the vessel until QFR was computed using contrast flow evaluation. Time to FFR was defined as the introduction of the pressure wire to the guiding catheter until drift check with a drift value within the specified limits.

QFR/FFR hybrid-approach limits
For a QFR/FFR hybrid strategy we used an FFR-only strategy as gold standard. QFR limits to yield a sensitivity (QFR-treat) and specificity (QFR-defer) of 90 and 95 percent were identified and used to model a hybrid approach where wire-based FFR assesment is needed between the QFR-treat and QFR-defer limits. The proportion of potential pressure wire free lesion assessments was calculated.

Prediction of QFR-FFR discrepancy
We constructed a multilevel mixed effect model including sites as level variable. Following covariates were tested individually and included in the multivariate analysis if P-value<0.10: lesion length, % DS (2D-QCA), age, BMI, adenosine route, sex, smoking, vessel, diabetes, previous PCI, and FFR.

Procedure training
Participating sites were requested to have operators and dedicated staff trained on QFR computation. The staff received instructions and training from Medis medical imaging bv. Only staff with QFR certificates obtained from Medis could perform the study computation of QFR.
Besides the QFR training from Medis, all sites were required to submit at least two complete and fully anonymized training datasets for approval by the respective core-labs before study enrolment.

FAVOR II standard operating procedure for QFR computation in FAVOR II Europe-Japan
The QFR standard operating procedure (SOP) applied by all sites in the FAVOR II multicenter study by Aarhus University Hospital, Skejby, Denmark. The set of instructions do not constitute a manual, neither partly nor in full, for clinical use of QFR.

Aorto-ostial stenosis
Aorto-ostial stenosis is not analyzable by QFR at present due to the requirement of two optimal projections, the guiding catheter intrusion and back flow of contrast in aorta overlapping the ostium.

Low angiographic quality or poor contrast filling
In some cases, the application is not able to recognize the vessel contours due to excessively low angiographic quality or poor contrast filling and exclusion of the case can be necessary ( fig. 1).
With experience the operator may decide to exclude the case even before transmitting runs to the QFR work station for analysis.

Figure 1
Low angiographic quality. The QFR application has difficulties finding the vessel lumen and vessel borders.

Overlap
If correct lumen contouring is impossible due to severe overlap of the stenosed segment, the case should be excluded.

Nitroglycerin administration
When nitroglycerin is not administered neither systemic nor intracoronary, vessel spasms cannot be ruled out, and the case should be excluded ( fig. 2). Without prior nitroglycerin, both the QFR analysis and FFR measurement can be unreliable.

Step-by-step manual
The Medis Suite QAngio XA 3D/QFR solution (Medis medical imaging system bv, Leiden, The Netherlands) is used for computation of QFR in FAVOR II. The Medis Suite QAngio XA 3D/QFR solution requires installation on a Windows-based computer. QFR computation is described stepby-step below.

Coronary angiography
Two good projections at least 25 degrees apart are required for the 3D vessel reconstruction.
Angiographic procedure: • Inject I.C. nitroglycerin as early as possible  a. If several lesions are located in the same vessel, a compromise must be made to ensure that most of the lesions and the most severe lesions are seen in the same

Image transfer
The angiographic runs are transferred to the QFR-computer using an angiographic equipment specific protocol.

Angiographic run selection
Optimal projections are chosen according to the following criteria: • Minimal overlap of the target vessel

Frame selection
The best frames for analysis are selected by ensuring: • The lesion site(s) is not overlapped When the best end diastolic frame is found in both panels, the 3D reconstruction is initiated by pressing the Create single vessel analysis-button ( fig. 6, red arrow) in the top panel.

Figure 6
Create single vessel analysis (marked by red arrow) initiates the 3D reconstruction.

3D target vessel reconstruction
To link the two projections, corresponding landmarks near the lesion are identified by a pair of offset points in both projections ( fig. 7). Make sure to: • Identify a landmark that is easily identified in both projections (i.e. a bifurcation, a localized stenosis or the off-spring of a side branch) • If using a side branch:  The path line is verified visually for both projections. If it deviates from the target vessel, it is dragged into position using support points.
When the position of the vessel pathline is accepted, the pathline is "locked" (fig. 8, red box). It is visible when the mouse is shifted over it. The pathline is fixated by ticking Lock pathlines '(red box).

Lumen contouring
The yellow contour lines ( fig. 9) are adjusted to follow the lumen border. Pay special attention to: • Indication of non-existing narrowings in the proximal and distal ends • Correct contouring of the target lesion(s) • Side branches and overlap • Ensure that contours are correct in all segments -also non-target segments as it influences

QFR calculation
The lines are corrected by dragging them into position with correction points. If a correction needs to be reverted, right-click the created correction-point and it will be deleted.  NOTE: Focus on getting the lesions and proximal vessel segments to correspond. Use the lesions markers to check the correspondence at this step before proceeding.

Reference vessel
Every part of the contoured lumen that is narrower than the reference vessel is marked yellow as plaque ( fig. 11). These yellow markings can be removed by ticking off the box Show plaque ( Figure   11, red box).

Figure 11 Reference vessel (red contours on 2D images). Show/hide yellow plaque (red box).
The proximal and distal ends are supposed to approximately match the healthy vessel parts. The reference function should obey the following: • Always tapering reference function. A straight function is allowed in short segments.
• Should not follow stenosed or aneurysmatic sections.
• Sizes should be realistic according to gender and body mass index. See 2.9.3 If the abovementioned criteria are not fulfilled, the reference contours can be edited as follows:

Selection of normal areas
Select "Normals" under the "check reference wizard" (text fig. 12, red box). Select two normal areas, using the green areas. The reference function is now calculated as a linear regression based on the two selected normals. Note the slightly adjusted reference contours after using the "Normals" function ( fig. 12 compared to fig. 11).

Figure 12
Reference function editing using the "Normals" function (red box). The green normal areas are selected to indicate two healthy vessel segments.

Fixed proximal reference
To impute a reference size for a particular segment use the fixed reference tool. Select "Fixed prox" under the "Check reference" (text fig 13, red box). A fixed proximal reference size is selected with a 0.25 mm interval from 2 to 5 mm. Place the proximal green marker where the vessel should have the indicated value. A normal distal area is chosen to adjust the slope of the linear function (see fig. 13).

Figure 13
Reference function editing using the "Fixed Prox" function (red box). The normal areas is moved to indicate a healthy distal vessel segment.

Reference diameter strategy
If the automatic generated reference function based on the 3D-recontruction follows the criteria (2.9), it is used as the first choice. If not satisfied, selection of normal areas (2.9.1) is recommended in vessels with: • Clearly identifiable heathy segments • Realistic proximal reference size of the vessel according to gender and body mass index A fixed proximal reference (2.9.2) is recommended in cases with: • Proximal LAD disease defined as a proximal LAD reference size < 2.5 mm for women and < 3.0 mm for men in Caucasians with healthy segments distally • Diffuse LAD disease with segments in mid/distal vessel parts exceeding the proximal reference size The fixed proximal diameter is set to 3.0-3.5 mm men and 2.5-3.0 mm for women depending on the size of LM and LCx) and the patient (age, MBI).
Verify the reference diameter (criteria from 2.9) by looking at the diameter graph in the lower right panel. The red line indicates the reference lumen diameter, and the two graphs the minimum and maximum lumen diameters from the two images.
NOTE: it is more important to have a correct reference function by manual adjustments than preserving a wrongly automatic generated reference function to aim for reduced variability frames, the first of the two frames is chosen, and the +1/2-box is ticked off e. Another option is to relocate the proximal and the distal vessel delimiters, to get a better correspondence between the chosen start or end frame and the contrast position PLEASE NOTE that projections where the contrast seems to appear uniformly in most of the analysed segment simultaneously are not appropriate for frame count. Remember to enter patient state (the angiographic run for frame counting is acquired during resting condition or hyperaemia).

Documentation
After finalizing analysis, it is saved by two steps for the study purpose. • After an analysis is finalized, Medis Suite QAngio XA 3D/QFR solution creates a report summarizing the analysis, including 2D images of the vessel reconstruction, the 3D reconstruction, results and more ( fig. 17).

Figure 17
Report. Access the report by selecting the Report pane (red box).

Left main coronary artery (LMCA)
• Stenosis in LMCA can only be assessed if the aorto-ostium is not involved (see 1.1) • QFR of LMCA stenosis in combination with proximal Cx stenosis is not recommended with the present version of QFR (see 1.5)

Ostial stenosis in Cx or SBs with healthy main vessel but large diameter difference
• See 1.5 if the main vessel is diseased, otherwise the following applies: o The ostium must be visible in both angiographic views to be able to segment the entire stenosis o The proximal marker should be placed in the ostium o It is important to optimize the size of the reference diameter. In most cases, reference diameter editing is required (2.8.1 + 2.8.2) to ensure tapering of the reference diameter.