Diagnostic Performance of Artificial Intelligence-Based Angiography-Derived Non-Hyperemic Pressure Ratio Using Pressure Wire as Reference

Abstract

Background: The angiography-derived non-hyperemic pressure ratio (angioNHPR) is a novel index of NHPR based on artificial intelligence (AI) that does not require pressure wires. We investigated the diagnostic accuracy of angioNHPR for detecting hemodynamically relevant coronary artery disease.

Methods and Results: In this retrospective single-center study, angioNHPR was assessed using the invasive NHPR as the reference standard. An angioNHPR ≤0.89 was defined as indicative of physiologically significant stenosis. Two angiographic projections ≥30° difference in angulation were selected. The lumen and centerline were automatically segmented by the prototype software, allowing for the calculation of the angioNHPR. We assessed 222 vessels from 178 patients. The accuracy of angioNHPR was 76.6% (95% confidence interval [CI] 70.4–82.0), with sensitivity 66.2% (95% CI 54.0–77.0), specificity 81.5% (95% CI 74.3–87.3), positive predictive value 62.7% (95% CI 53.6–70.9), and negative predictive value 83.7% (95% CI 78.6–87.7). The angioNHPR showed good correlation with invasive NHPR (r=0.72; 95% CI 0.64–0.77; P<0.001), and the agreement between angioNHPR and invasive NHPR was −0.01 (limits of agreement: −0.13, 0.11). The area under the curve (AUC) of angioNHPR was 0.81 (95% CI 0.75–0.86), which was significantly higher than that of 2-dimensional quantitative coronary angiography (AUC 0.69; 95% CI 0.62–0.75; P=0.007).

Conclusions: AI-based angioNHPR demonstrates good diagnostic performance using invasive NHPR as the reference standard.

Physiological assessment using a pressure guidewire is currently recommended in clinical practice guidelines to guide decision making for coronary revascularization in patients with coronary artery disease (CAD).^1,2 To perform this assessment, fractional flow reserve (FFR) requires maximum hyperemia, whereas the instantaneous wave-free ratio (iFR), a non-hyperemic pressure ratio (NHPR) index, does not. The iFR has been shown to be non-inferior to FFR regarding major adverse cardiovascular events for coronary revascularization in 2 large randomized clinical trials.^3,4 In addition, new NHPRs, including the resting full-cycle ratio (RFR), diastolic pressure ratio (dPR), and diastolic hyperemia-free ratio (DFR), have demonstrated strong linear correlations with iFR.^5–7 However, the dissemination of NHPRs remains low, partly due to the risks associated with maneuvering pressure wires.

To mitigate these risks, a novel technology has been developed and prototyped as an artificial intelligence (AI)-based angiography-derived NHPR (angioNHPR; Siemens Healthcare GmbH, Forchheim, Germany). The technology uses an upfront-trained machine learning model based on reduced-order computational modeling of blood flow.⁸ It offers the advantage of eliminating the need for pressure wires. However, clinical evidence supporting its use has yet to be established. The aim of the present study was to assess the diagnostic performance of angioNHPR in detecting hemodynamically relevant CAD using invasive NHPR as the reference standard.

Methods

Patients

This single-center retrospective study included patients with CAD who had 30–90% angiographic stenoses based on visual assessment and NHPR measurements in major epicardial vessels.¹ Patients were excluded if they had undergone coronary artery bypass grafting, myocardial infarction, left main trunk disease, ostial lesions, a vessel with stents, severe aortic valve stenosis, and a vessel with collaterals. The consent form for the catheterization examination indicated that the data would be used for publication, and all patients agreed to this.

Study procedures were performed following the Declaration of Helsinki and were approved by the Ethics Committee at Gifu Heart Center.

Angiography

A biplane angiography system (Artis zee biplane; Siemens Healthcare GmbH) was used for fluoroscopic imaging. The imaging parameters were 78 kV and 15 frames per s (fps). After the administration of intracoronary nitrates (300 μg), contrast injections were performed manually in 5-mL pulses (370 mg iodine/mL [Iopamiron; Bayer HealthCare Co. Ltd., Osaka, Japan] or 350 mg iodine/mL [Omnipaque; Daiichi Sankyo Co. Ltd., Tokyo, Japan]) following the guidelines using a 5- or 6-Fr system.⁹

The coronary angiography was visually classified as focal, tandem, diffuse, bifurcation, severe calcification, and severe vessel tortuosity.¹⁰ Lesion severity was assessed using quantitative coronary angiography (QCA; Heart II®; Ratoc System Engineering, Tokyo, Japan).¹¹

Invasive NHPR Measurement

NHPR was measured at least 20 mm distal to the stenosis by pressure wires (Philips for iFR; Abbott Vascular for RFR; Opsens Inc. for dPR; Boston Scientific for DFR).

AngioNHPR Application

The AngioNHPR prototype software (not for diagnostic use) is based on AI using a supervised machine-learning algorithm that was trained upfront at the manufacturer’s site (Figure 1). For model training, a database of 27,000 coronary representations was created, and ground-truth NHPR was calculated using a reduced-order computational fluid dynamics model.⁸ The prototype was trained using 90% of the database, with the remaining 10% used for verification. The computational fluid dynamics model computes time-varying pressures and flow rates in the epicardial coronary arteries, incorporating a lumped parameter model of the heart and specialized models for the coronary microvasculature to simulate the effects of the myocardial contraction on the coronary circulation. The computed time-varying pressures are then post-processed to determine ground-truth NHPR as the ratio between mean diastolic distal and mean diastolic aortic pressure, measured during the diastolic wave-free period.⁸ In the clinical setting, the offline-trained application of the prototype can estimate angioNHPR based on 2 angiographic scenes.

Figure 1.

Schematic flow diagram of the machine learning-based angiography-derived non-hyperemic pressure ratio (angioNHPR). (Upper) A database of 27,000 synthetic coronary anatomy models was created. The ground-truth non-hyperemic pressure ratio (NHPR) was calculated using a reduced-order computational fluid dynamics (CFD) model. The artificial intelligence (AI)-based angioNHPR was trained upfront using the data. (Lower) In the clinical setting, the 2 angiographic projections with ≥30° difference in angulation are selected and points are manually placed along the target vessel. The lumen and centerline are automatically segmented by the prototype software. Values of angioNHPR are obtained for all locations along the vessel in a few seconds.

AngioNHPR Analysis

In an offline analysis, the investigators, blinded to each other’s results and to patient history, used non-commercial prototype software (Siemens Healthcare GmbH) installed on a stand-alone workstation (Monaco; Siemens Healthcare GmbH) to determine the angioNHPR. The selection of images (i.e., individual frames) for the analysis was performed semi-automatically. Well-contrasted diastolic frames were automatically selected and proposed to the user.

The raters first select 2 angiographic projections with a ≥30° difference in angulation and manually place points along the target vessel (Figure 1). The angioNHPR analysis can incorporate a side branch with a diameter of ≥2 mm in the results.

An AI-based heart phase detection system ensures that both angiographic projections are acquired in the same heart phase.¹² The prototype software automatically segments the lumen and centerline, but manual corrections can be made if necessary. Finally, angioNHPR values are obtained for all locations along the vessel within a few seconds (Figure 2).

Figure 2.

Case example. (Left) Invasive non-hyperemic pressure ratio (NHPR) and (Right) angiography-derived NHPR at the position of the pressure wire.

Statistical Analysis

Categorical variables are expressed as frequencies (counts) and proportions (percentages), whereas continuous variables are expressed as the mean±SD. The cut-off value for 2-dimensional (2D) QCA was defined as 50%.¹³ For lesion-based analysis, the accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated as proportions with 95% confidence intervals (95% CIs), using an invasive NHPR value ≤0.89 as the reference standard for 2D-QCA >50% and angioNHPR ≤0.89.^3,4 The diagnostic performances of 2D-QCA and angioNHPR were assessed using the area under the curve (AUC) of receiver operating characteristic curves and their 95% CIs to predict invasive NHPR values ≤0.89. Per-vessel AUC comparisons were conducted using the DeLong method. Correlations were evaluated using Pearson’s correlation coefficient (95% CI). The agreement between invasive NHPR and angioNHPR was assessed using Bland-Altman analysis. Intrarater reliability was determined using intraclass correlation coefficients. Statistical significance was set at 2-tailed P<0.05. Statistical analyses were performed using MedCalc version 22.021 (MedCalc Software Bvba, Ostend, Belgium).

Results

In all, 245 vessels in 191 consecutive patients with 30–90% angiographic stenoses who underwent invasive NHPR were enrolled between November 2018 and February 2020. We excluded 17 vessels based on the exclusion criteria. In the offline analysis, we excluded 2 vessels due to insufficient contrast enhancement, 1 lesion that overlapped with another vessel, and 3 severely tortuous vessels that impeded contour detection. Finally, we analyzed 222 vessels from 178 patients (Figure 3).

Figure 3.

Flowchart showing the selection of patients for this study. CAD, coronary artery disease; NHPR, non-hyperemic pressure ratio.

The baseline characteristics of the participants are presented in Table 1. The frequency of pressure wire use was 46 (20.7%) for iFR, 68 (30.6%) for RFR, 86 (38.7%) for dPR, and 22 (9.9%) for DFR. The mean NHPR and mean angioNHPR were 0.91±0.08 and 0.90±0.08 (P=0.30), respectively. The accuracy of angioNHPR was 76.6% (95% CI 70.4–82.0), with sensitivity 66.2% (95% CI 54.0–77.0), specificity 81.5% (95% CI 74.3–87.3), PPV 62.7% (95% CI 53.6–70.9), and NPV 83.7% (95% CI 78.6–87.7; see Table 2). There was a good correlation between angioNHPR and NHPR (r=0.72; 95% CI 0.64–0.77; P<0.001), and the agreement between angioNHPR and NHPR was −0.01 (limits of agreement [LoA] −0.13, 0.11; Figure 4). The AUC of angioNHPR was significantly higher than that of 2D-QCA (0.81 [95% CI 0.75–0.86] vs. 0.69 [95% CI 0.62–0.75], respectively; P=0.007; Figure 4).

Table 1.

Baseline Patient and Lesion Characteristics

Patient characteristics (n=178)
Age (years)	72±10
Male sex	125 (70.2)
BMI (kg/m2)	23.5±3.3
Hypertension	124 (69.7)
Dyslipidemia	96 (53.9)
Diabetes	62 (34.8)
Current smoker	34 (19.1)
Previous PCI	36 (20.2)
Previous MI	20 (11.2)
Multivessel disease	80 (44.9)
LVEF (%)	62.6±11.7
eGFR (mL/min/1.73 m2)	59.6±21.3
Stable angina pectoris	106 (59.6)
Unstable angina pectoris	11 (6.2)
Silent ischemia	61 (34.3)
Lesion characteristics (n=222)
Target lesion
LAD	129 (58.1)
LCX	45 (20.3)
RCA	48 (21.6)
Focal lesion	88 (39.6)
Tandem lesion	58 (26.1)
Diffuse lesion	54 (24.3)
Bifurcation with side branch ≥2.0 mm	80 (36.0)
Severe calcification lesion	13 (5.9)
Severe vessel tortuosity	18 (8.1)
Quantitative coronary angiography
Diameter stenosis (%)	49.1±11.5
Diameter stenosis
30–50%	130 (58.6)
51–70%	83 (37.4)
71–90%	9 (4.1)
Reference diameter (mm)	2.9±1.9
Minimum luminal diameter (mm)	1.4±0.5
Length of lesions (mm	20.3±10.2

Continuous variables are presented as the mean±SD and categorical variables are presented as numbers and percentages. BMI, body mass index; eGFR, estimated glomerular filtration rate; LAD, left anterior descending artery; LCX, left circumflex artery; LVEF, left ventricular ejection fraction; MI, myocardial infarction; PCI, percutaneous coronary intervention; RCA, right coronary artery.

Table 2.

Diagnostic Performance of 2D-QCA and AngioNHPR

	2D-QCA	AngioNHPR
Accuracy (%)	67.1 (60.5–73.3)	76.6 (70.4–82.0)
Sensitivity (%)	63.4 (51.1–74.5)	66.2 (54.0–77.0)
Specificity (%)	68.9 (60.8–76.2)	81.5 (74.3–87.3)
PPV (%)	48.9 (41.6–56.3)	62.7 (53.6–70.9)
NPV (%)	80.0 (74.3–84.7)	83.7 (78.6–87.7)
AUC	0.69 (0.62–0.75)	0.81 (0.75–0.86)

Values in parentheses are 95% confidence intervals. An invasive non-hyperemic pressure ratio of ≤0.89 indicates ischemia. 2D-QCA, 2-dimensional quantitative coronary angiography; angioNHPR, angiography-derived non-hyperemic pressure ratio; AUC, area under the receiver operating characteristic curve; NPV, negative predictive value; PPV, positive predictive value.

Figure 4.

(A) Correlation and (B) agreement between invasive non-hyperemic pressure ratio (NHPR) and angiography-derived NHPR (angioNHPR). (C) Areas under the receiver operating characteristic curve (AUC) for angioNHPR and 2-dimensional quantitative coronary angiography (2D-QCA). The AUC of angioNHPR for the identification of invasive NHPR was ≤0.89. CI, confidence interval.

Subgroup analysis of angioNHPR accuracy showed consistent results across all subgroups, except for the vessel category (Table 3). In each subgroup, angioNHPR demonstrated a good correlation with NHPR: r=0.63 (95% CI 0.51–0.72; P<0.001) for the left anterior descending artery (LAD), r=0.58 (95% CI 0.35–0.75; P<0.001) for the left circumflex artery (LCX), and r=0.89 (95% CI 0.82–0.94; P<0.001) for the right coronary artery (RCA). The agreement between angioNHPR and NHPR was 0.01 (LoA −0.10, 0.13) for the LAD, −0.04 (LoA −0.15, 0.08) for the LCX, and −0.03 (LoA −0.12, 0.06) for the RCA (Figure 5). The intraclass coefficient was 0.79 (95% CI 0.73–0.83).

Table 3.

Subgroup Analysis of the Accuracy of AngioNHPR

Variable	Subgroup	AngioNHPR accuracy (%)	P value
Sex	Male (n=158)	78.5	0.418
	Female (n=64)	73.4
Age	≥65 years (n=185)	75.7	0.284
	<65 years (n=37)	83.8
Diabetes	Yes (n=78)	75.6	0.718
	No (n=144)	77.8
Vessel	LAD (n=129)	70.5	0.021
	LCX (n=45)	88.9
	RCA (n=48)	83.3
Focal lesion	Yes (n=88)	78.4	0.692
	No (n=134)	76.1
Tandem lesion	Yes (n=58)	75.9	0.806
	No (n=164)	77.4
Diffuse lesion	Yes (n=54)	74.1	0.553
	No (n=168)	78.0
Bifurcation	Yes (n=80)	70.0	0.062
	No (n=142)	81.0
Severe calcification	Yes (n=13)	69.2	0.491
	No (n=209)	77.5
Severe tortuosity	Yes (n=18)	88.9	0.212
	No (n=204)	76.0
Stenosis	≥50% (n=109)	76.1	0.759
	<50% (n=113)	77.9
Multivessel disease	Yes (n=113)	80.5	0.206
	No (n=109)	73.4

An invasive non-hyperemic pressure ratio of ≤0.89 indicates ischemia. Abbreviations as in Tables 1,2.

Figure 5.

Correlation and agreement between invasive non-hyperemic pressure ratio (NHPR) and angiography-derived NHPR (angioNHPR) for the left anterior descending artery (LAD), left circumflex artery (LCX), and right coronary artery (RCA). (A) Linear relationship between invasive NHPR and angioNHPR for the LAD, LCX, and RCA. (B) Bland-Altman analysis between invasive NHPR and angioNHPR for the LAD, LCX, and RCA.

Discussion

The present study was a single-center retrospective study assessing the diagnostic performance of angioNHPR. The main finding of the study is that angioNHPR had good diagnostic performance in detecting hemodynamically significant CAD using invasive NHPR as the reference standard.

Advent of AI-Based AngioNHPR

Physiology-guided decision making of coronary artery stenoses plays an important role in CAD. As an alternative to invasive FFR using a pressure wire, several angiography-derived FFR software have emerged, offering less-invasive maneuvers, shorter procedural times, and cost-effectiveness; however, NHPR has not been included in this trend. The accuracy of AI-based FFR (angioFFR; Siemens Healthcare GmbH) has been reported previously.¹⁴ This software uniquely allows for the measurement of both angioNHPR and angioFFR. In the present study, we analyzed the accuracy of angioNHPR using invasive NHPR as the reference. AngioNHPR is the first AI-based angiography-derived NHPR software using an upfront-trained machine learning model based on 2 angiographic projections. The introduction of angioNHPR will provide physicians with additional information to help determine whether to proceed with revascularization.

Diagnostic Performance of AngioNHPR

AI-based angioNHPR demonstrated good accuracy (76.6%), with a correlation coefficient (r) of 0.72, a mean difference of −0.01, and an AUC of 0.81 for detecting hemodynamically relevant CAD, using invasive NHPR as the reference standard. The accuracy for LAD appeared to be lower than for non-LAD vessels. One contributing factor is the frequency of positive invasive NHPR results: 59 (45.7%) for LAD vs. 6 (13.3%) for LCX and 6 (12.5%) for RCA. The high invasive NHPR values observed in non-LAD vessels may contribute to the increased accuracy for non-LAD vessels.

Although this software is not for diagnostic use due to the non-commercial status of the angioNHPR prototype, there is still room for improvement in terms of its accuracy. To develop software with better accuracy, it is essential to investigate the strengths and weaknesses of the current version. In this respect, this study is significant because it is the first to compare the accuracy of angioNHPR with invasive NHPR, which is crucial for clinical practice. Furthermore, to enhance the software in the future, it will be necessary to incorporate clinical data from this study using AI, ultimately leading to the creation of software with higher accuracy. One reason for the decrease in accuracy is unclear imaging due to severe stenosis, calcified lesions, or ulcer lesions, which may result in larger contour detection errors in the vessel. To improve the accuracy of angiography-derived functional assessments, more precise angiographic imaging may be necessary. Another approach is to enhance the software’s accuracy. The greatest advantage of angioNHPR can be further refined using current clinical data and AI algorithms; we anticipate that the next generation of angioNHPR will become a more useful tool.

Efficacy of AngioNHPR

AI-based angioNHPR offers several advantages. It does not require the use of pressure wires and hyperemic drugs, which may make physiological assessments more accessible and widespread. AngioNHPR is measured using resting coronary physiology, which may offer additional advantages over invasive FFR and functional coronary angiography (FCA).¹⁵ In a non-hyperemic state, coronary blood flow remains constant and does not change significantly with the degree of coronary stenosis due to the autoregulation of microvascular circulation. However, during hyperemia, the presence of tandem and diffuse lesions can cause coronary blood flow to alter after passing through the stenosis, making the pressure gradient across each stenosis unpredictable. Therefore, NHPR may provide a more accurate assessment of the pressure gradient at each stenosis than FFR.^16–18 This software can generate pull-back curves of angioNHPR based on angiographic views, greatly benefiting physicians by visualizing the pressure gradient during coronary interventions. Using the features of NHPR, it is feasible to predict the hemodynamic values after treating a segment.^16–18 The angioNHPR curve-guided strategy holds potential for predicting NHPR values after stenting. Unfortunately, we cannot currently assess the accuracy of the predicted NHPR values, because invasive NHPR pull-back measurements before and after stenting were not performed. We plan to evaluate this accuracy in future trials.

Regarding FCA, the quantitative flow ratio is one of the most validated FCA software options.¹⁵ The quantitative flow ratio is derived from coronary angiography performed under resting conditions by predicting microvascular resistance and blood flow during hyperemia. Because coronary angiography is conducted in a resting state, angioNHPR, also derived from resting conditions, may have an advantage over the quantitative flow ratio.

This Siemens software is unique and appealing because it provides both angioFFR and angioNHPR.¹⁴ In a clinical setting, acquiring both values may enhance accuracy in detecting ischemic lesions (i.e., when both are positive). We plan to investigate how to use both indices in future research.

Study Limitations

This study had several limitations. First, it was conducted at a single center, which may have introduced selection bias. Second, only 23 (9.4%) of 245 vessels met the exclusion criteria, meaning the findings may not be generalizable. Third, the processing time was not measured because a prototype was used; theoretically, AI-based angioNHPR has the potential to shorten processing time. Further studies are needed to address these limitations. Fourth, each NHPR algorithm varies depending on the vendor. A previous study reported that all NHPRs were identical to iFR, both numerically and in terms of agreement with FFR, without being restricted to wave-free periods.⁶ Although some minor errors may exist, we believe they are not significant. Fifth, we did not analyze the influence of the selected frame on the results; however, we expect the influence to be minimal as long as the arteries are well contrasted. Sixth, if the role for revascularization for ischemia lesions changes in the future, the significance of this study may also be affected. Finally, future trials are warranted to assess the clinical outcome of angioNHPR-guided revascularizations.

Conclusions

AI-based angioNHPR demonstrated good diagnostic performance for CAD using invasive NHPR as the reference standard.

Acknowledgments

The authors thank radiology technicians Syunsuke Imai, Yuki Nakashima, Takahito Mizusaki, and Takuya Kagatani. The authors also thank medical engineers Shinya Iwata, Airi Kimata, and Tetsuya Hiraiwa for quantitative analysis of the coronary angiography.

Sources of Funding

This study was supported, in part, by a research grant (ID C00236395) from Siemens Healthcare K.K. Siemens Healthcare provided a workstation equipped with the non-commercial angiography-derived non-hyperemic pressure ratio prototype (not for diagnostic use) on loan. The concepts and prototype software presented in this study are based on research. Due to regulatory reasons, its future availability cannot be guaranteed. Siemens Healthcare was not involved in the conduct of the study, nor in data collection, analysis, or interpretation.

Disclosures

The authors declare no conflicts of interest associated with this manuscript.

IRB Information

The study was approved by the local ethics committee (No. 2019014) of Gifu Heart Center and was performed in accordance with the Declaration of Helsinki.

Data Availability

The deidentified participant data for this clinical trial will not be shared

References

• 1.   Patel MR, Calhoon JH, Dehmer GJ, Grantham JA, Maddox TM, Maron DJ, et al. ACC/AATS/AHA/ASE/ASNC/SCAI/SCCT/STS 2017 appropriate use criteria for coronary revascularization in patients with stable ischemic heart disease. J Am Coll Cardiol 2017; 69: 2212–2241.
• 2.   Neumann FJ, Sousa-Uva M, Ahlsson A, Alfonso F, Banning AP, Benedetto U, et al. 2018 ESC/EACTS guidelines on myocardial revascularization. Eur Heart J 2019; 40: 87–165.
• 3.   Davies JE, Sen S, Dehbi HM, Al-Lamee R, Petraco R, Nijjer SS, et al. Use of the instantaneous wave-free ratio or fractional flow reserve in PCI. N Engl J Med 2017; 376: 1824–1834.
• 4.   Götberg M, Christiansen EH, Gudmundsdottir IJ, Sandhall L, Danielewicz M, Jakobsen L, et al. Instantaneous wave-free ratio versus fractional flow reserve to guide PCI. N Engl J Med 2017; 376: 1813–1823.
• 5.   Svanerud J, Ahn JM, Jeremias A, Veer M, Gore A, Maehara A, et al. Validation of a novel non-hyperaemic index of coronary artery stenosis severity: The Resting Full-cycle Ratio (VALIDATE RFR) study. EuroIntervention 2018; 14: 806–814.
• 6.   Van’t Veer M, Pijls NHJ, Hennigan B, Watkins S, Ali Z, Bruyne B, et al. Comparison of different diastolic resting indexes to iFR: Are they all equal? J Am Coll Cardiol 2017; 70: 3088–3096.
• 7.   Johnson NP, Li W, Chen X, Hennigan B, Watkins S, Berry C, et al. Diastolic pressure ratio: New approach and validation vs. the instantaneous wave-free ratio. Eur Heart J 2019; 40: 2585–2594.
• 8.   NPasserini T, Itu L, Sharma P. (2017) Patient-specific modeling of the coronary circulation. In: Itu L, Sharma P, Suciu C, editors. Patient-specific Hemodynamic Computations: Application to Personalized Diagnosis of Cardiovascular Pathologies. Springer International Publishing 2017; 61–88.
• 9.   Naidu SS, Rao SV, Blankenship J, Cavendish JJ, Farah T, Moussa I, et al. Clinical expert consensus statement on best practices in the cardiac catheterization laboratory: Society for Cardiovascular Angiography and Interventions. Catheter Cardiovasc Interv 2012; 80: 456–464.
• 10.   Sianos G, Morel MA, Kappetein AP, Morice M, Colombo A, Dawkins K, et al. The SYNTAX score: An angiographic tool grading the complexity of coronary artery disease. EuroIntervention 2005; 1: 219–227.
• 11.   Pompa J, Almonacid A, Burke D. Qualitative and quantitative coronary angiography. In: Topol EJ, Teirstein PS, editors. Textbook of interventional cardiology, 6th edn. Philadelphia: Saunders, 2011; 757–775.
• 12.   Ciusdel C, Turcea A, Puiu A, Itu L, Calmac L, Weiss E, et al. Deep neural networks for ECG-free cardiac phase and end-diastolic frame detection on coronary angiographies. Comput Med Imaging Graph 2020; 84: 101749.
• 13.   Westra J, Andersen BK, Campo G, Matsuo H, Koltowski L, Eftekhari A, et al. Diagnostic performance of in-procedure angiography-derived quantitative flow reserve compared to pressure derived fractional flow reserve: The FAVOR II Europe–Japan Study. J Am Heart Assoc 2018; 7: e009603.
• 14.   Omori H, Kawase Y, Mizukami T, Tanigaki T, Hirata T, Okubo M, et al. Diagnostic accuracy of artificial intelligence-based angiography-derived fractional flow reserve using pressure wire-based fractional flow reserve as a reference. Circ J 2023; 87: 783–790.
• 15.   Asano T, Tanigaki T, Ikeda K, Ono M, Yokoi H, Kobayashi Y, et al. Consensus document on the clinical application of invasive functional coronary angiography from the Japanese Association of Cardiovascular Intervention and Therapeutics. Cardiovasc Interv Ther 2024; 39: 109–125.
• 16.   Nijjer SS, Sen S, Petraco R, Escaned J, Echavarria-Pinto M, Broyd C, et al. Pre-angioplasty instantaneous wave-free ratio pullback provides virtual intervention and predicts hemodynamic outcome for serial lesions and diffuse coronary artery disease. J Am Coll Cardiol Intv 2014; 7: 1386–1396.
• 17.   Kikuta Y, Cook CM, Sharp ASP, Salinas P, Kawase Y, Shiono Y, et al. Pre-angioplasty instantaneous wave-free ratio pullback predicts hemodynamic outcome in humans with coronary artery disease: Primary results of the international multicenter iFR GRADIENT registry. J Am Coll Cardiol Intv 2018; 11: 757–767.
• 18.   Omori H, Kawase Y, Mizukami T, Tanigaki T, Hirata T, Kikuchi J, et al. Comparisons of nonhyperemic pressure ratios: Predicting functional results of coronary revascularization using longitudinal vessel interrogation. JACC Cardiovasc Interv 2020; 13: 2688–2698.