Author + information
- Received February 15, 2013
- Revision received May 20, 2013
- Accepted May 24, 2013
- Published online January 1, 2014.
- Kyung-Hee Kim, MD∗,
- Joon-Hyung Doh, MD†,
- Bon-Kwon Koo, MD∗∗ (, )
- James K. Min, MD‡,
- Andrejs Erglis, MD§,
- Han-Mo Yang, MD∗,
- Kyung-Woo Park, MD∗,
- Hae-Young Lee, MD∗,
- Hyun-Jae Kang, MD∗,
- Yong-Jin Kim, MD∗,
- Sung Yun Lee, MD† and
- Hyo-Soo Kim, MD∗
- ∗Department of Medicine, Seoul National University Hospital, Seoul, Republic of Korea
- †Department of Medicine, Inje University Ilsan Paik Hospital, Goyang, Republic of Korea
- ‡Department of Medicine, Cedars-Sinai Heart Institute, Cedars-Sinai Medical Center, Los Angeles, California
- §Department of Medicine, Pauls Stradins Clinical University Hospital, Riga, Latvia
- ↵∗Reprint requests and correspondence:
Dr. Bon-Kwon Koo, Division of Cardiology, Department of Internal Medicine, Seoul National University College of Medicine, 101 Daehak-ro, Jongno-gu, Seoul 110-744, Republic of Korea.
Objectives This study sought to determine whether computational modeling can be used to predict the functional outcome of coronary stenting by virtual stenting of ischemia-causing stenoses identified on the pre-treatment model.
Background Computed tomography (CT)-derived fractional flow reserve (FFR) is a novel noninvasive technology that can provide computed (FFRct) using standard coronary CT angiography protocols.
Methods We prospectively enrolled 44 patients (48 lesions) who had coronary CT angiography before angiography and stenting, and invasively measured FFR before and after stenting. FFRct was computed in blinded fashion using coronary CT angiography and computational fluid dynamics before and after virtual coronary stenting. Virtual stenting was performed by modification of the computational model to restore the area of the target lesion according to the proximal and distal reference areas.
Results Before intervention, invasive FFR was 0.70 ± 0.14 and noninvasive FFRct was 0.70 ± 0.15. FFR after stenting and FFRct after virtual stenting were 0.90 ± 0.05 and 0.88 ± 0.05, respectively (R = 0.55, p < 0.001). The mean difference between FFRct and FFR was 0.006 for pre-intervention (95% limit of agreement: –0.27 to 0.28) and 0.024 for post-intervention (95% limit of agreement: –0.08 to 0.13). Diagnostic accuracy of FFRct to predict ischemia (FFR ≤0.8) prior to stenting was 77% (sensitivity: 85.3%, specificity: 57.1%, positive predictive value: 83%, and negative predictive value: 62%) and after stenting was 96% (sensitivity: 100%, specificity: 96% positive predictive value: 50%, and negative predictive value: 100%).
Conclusions Virtual coronary stenting of CT-derived computational models is feasible, and this novel noninvasive technology may be useful in predicting functional outcome after coronary stenting. (Virtual Coronary Intervention and Noninvasive Fractional Flow Reserve [FFR]; NCT01478100)
Revascularization of coronary artery stenoses that induce ischemia can improve the functional status and clinical outcome of patients with coronary artery disease (1,2). Measurement of fractional flow reserve (FFR) during invasive coronary angiography is the gold standard for the diagnosis of ischemia-causing stenoses and can enhance clinical decision making and reduce healthcare costs (3,4). Coronary computed tomographic angiography (cCTA) is a commonly used noninvasive test that can provide accurate anatomical information on coronary artery disease (5,6). Previous studies, however, have shown that stenosis severity interpreted on cCTA does not match well with functional severity evaluated by invasive FFR (7).
Recently, application of computational fluid dynamics technology to cCTA images has enabled computation of coronary artery blood flow and pressure, and calculation of lesion-specific FFR without the need for an invasive procedure (8). FFR can be computed from typically acquired cCTA scans without any additional image acquisition, modification of cCTA protocols, or administration of medications. Previous studies (9,10) suggest that computation of FFR from cCTA (FFRCT) can identify patients with functionally significant coronary lesions prior to invasive cardiac catheterization. The same computational modeling technology allows for modification of the coronary flow model to eliminate an ischemia-causing stenosis, thus enabling “virtual stenting” of a coronary lesion. The resulting re-calculation of coronary blood flow and FFRCT can, in turn, serve to predict hemodynamic effect of coronary stenting in a lesion-specific manner. Such prediction of revascularization benefit (or lack thereof) may be a useful tool for patient or lesion selection and treatment planning prior to invasive procedures.
The purpose of this study is to determine whether virtual stenting of coronary stenoses identified on CT-based computational models can predict functional status of coronary lesions after stenting using measured FFR as the reference standard.
Study design and population
At 3 centers, 44 patients who had functionally significant coronary stenoses (35 men, mean age 65 years) with available pre-intervention cCTA and pre- and post-intervention FFR were enrolled. All patients were stable adults ≥18 years with suspected or known coronary artery disease who had undergone cCTA, were identified as having a ≥50% stenosis in a major coronary artery (≥2.0 mm diameter), and who underwent clinically indicated invasive coronary angiography with FFR measurement. Coronary calcium scoring was not performed at the time of cCTA, although no patient was excluded based on the upper threshold of qualitative coronary calcification, heart rate, or body mass index. Performance and timing of the invasive coronary angiography or FFR was at the discretion of the treating physician, but study cases were limited to those in which cCTA and invasive angiography was performed within 45 days without an intervening coronary event. The study protocol was approved by the institutional review boards of each participating center and all patients gave written informed consent.
cCTA and invasive coronary procedures
Each center performed cCTA in accordance with the Society of Cardiovascular Computed Tomography guidelines on performance of cCTA using a variety of different CT scanner platforms (Lightspeed VCT, GE Healthcare, Milwaukee, Wisconsin; Somatom Sensation and Definition CT, Siemens, Forchheim, Germany; Brilliance 256 and 64, Philips, Surrey, United Kingdom; Aquilion One and 64, Toshiba, Otawara, Japan). cCTA were performed by retrospective electrocardiographic helical or prospective electrocardiography-triggered methods. Oral metoprolol was administered for any patient with a heart rate ≥65 beats/min. Immediately before image acquisition, 0.2 mg sublingual nitroglycerin was administered. During the cCTA acquisition, 80 to 140 cc of iodinated contrast was injected followed by a 50-cc saline flush. Contrast timing was performed to optimize uniform contrast enhancement of the coronary arteries. The scan parameters were as follows: 64/256/320 × 0.5/0.625/0.750 mm collimation; tube voltage 100 or 120 mV; effective 400 to 650 mA. Dose reduction strategies—including electrocardiogram-gated tube current modulation and reduced tube voltage—were employed whenever feasible.
Coronary angiography and percutaneous coronary intervention were performed by standard techniques. The revascularization strategy, including the size of balloon and coronary stent, was left to the discretion of the operators. FFR was measured using a 0.014-inch pressure-monitoring guidewire (Pressure Wire Certus, St. Jude Medical Systems, Uppsala, Sweden) with the pressure sensor at the same location before and after coronary stenting. Maximal hyperemia was induced with a continuous intravenous infusion of adenosine at the rate of 140 μg/kg/min. FFR was calculated as the ratio of the mean distal pressure to the mean aortic pressure during maximal hyperemia. An FFR ≤0.8 was considered diagnostic of lesion-specific ischemia. FFR was measured in a blinded fashion, without knowledge of FFRCT values.
FFRCT and virtual stenting
FFRCT computation was performed without knowledge of pre- or post-stent FFR values by HeartFlow, Inc. (Redwood City, California) as described previously (9). To illustrate briefly, 3-dimensional models of the coronary tree were reconstructed using custom methods applied to cCTA data, and coronary flow and pressure were simulated using computational fluid dynamics principles (8). Resting flow was estimated from the myocardial mass, and resting microcirculatory resistance was distributed according to the size of the feeding vessel. A lumped parameter model representing the resistance to flow during simulated maximal hyperemia was applied to each coronary branch of the cCTA model, and FFRCT was computed for the entire coronary tree. The point on the model that corresponded to the location of the measured FFR was then selected and results were recorded. The pre-stent computational model was then marked with the location of the stent used to treat the patient. Analysts, blinded as to the results of treatment and to all FFR measurements, then performed virtual coronary intervention by modifying the computational model in the region of the stent to enlarge the lumen of the treated coronary segment according to the proximal and distal reference areas. Computational analysis of hyperemic coronary flow and pressure for the entire heart model was then repeated to determine post-treatment FFRCT. A core lab scientist, blinded to the results of measured FFR, then selected the point on the computational model that matched the location of post-stent FFR measurement to determine the corresponding FFRCT value beyond that stenosis (Fig. 1).
Categorical data are expressed as the number and percentage, whereas continuous variables are presented as the mean ± SD. Mean values were compared using paired and unpaired Student t tests as appropriate. Spearman rank correlation was used to assess the relationship between FFR and FFRct. The Bland-Altman analysis and intraclass correlation coefficient were used to assess the degree of agreement between FFR and FFRCT. Statistical analyses were performed using SPSS for Windows (version 17, SPSS Inc., Chicago, Illinois).
Clinical, angiographic, and procedural characteristics
The median time between cCTA and invasive coronary angiography was 12 days (range: 2 to 40). Table 1 provides the clinical, angiographic, and procedural characteristics of the 44 patients. Mean left ventricular ejection fraction was 63.1 ± 7.4%, and 5 patients (10%) had a history of previous myocardial infarctions. The target vessels in patients with previous myocardial infarction were non-infarct-related vessels. Forty-eight coronary lesions were stented; mean stent diameter was 3.1 ± 0.4 mm and mean stent length was 26.0 ± 10.1 mm. All treated patients had TIMI (Thrombolysis In Myocardial Infarction) flow grade 3 after stenting. After stenting, angiographic percentage of diameter stenosis improved to 10.1 ± 8.5% from a pre-stent diameter stenosis of 64.5 ± 14.1%.
Comparison of FFRCT and invasively measured FFR
Measured FFR prior to intervention was 0.70 ± 0.14 and increased to 0.90 ± 0.05 after stenting. Computed FFRCT prior to intervention was 0.70 ± 0.15 and increased to 0.88 ± 0.05 after virtual coronary stenting. There was a positive correlation between FFR and FFRCT before (R = 0.60, p < 0.001) and after intervention (R = 0.55, p < 0.001) (Fig. 2). The mean difference between FFRCT and FFR was 0.006 for pre-intervention (95% limit of agreement: –0.27 to 0.28) and 0.024 for post-intervention (95% limit of agreement: –0.08 to 0.13) (Fig. 3). Intraclass correlation coefficient was 0.71 (p < 0.001).
FFRCT for the prediction of residual ischemia
Residual ischemia after stenting was found in 2 patients by FFR and predicted in 4 patients by FFRCT. In the 2 patients with false positives, FFR was 0.88 and FFRCT was 0.79 in 1 patient, and FFR was 0.84 and FFRCT was 0.80 in the other patient. Diagnostic accuracy of FFRCT to predict ischemia (FFR ≤0.8) prior to stenting was 77% (sensitivity: 85.3%, specificity: 57.1%, positive predictive value: 83%, and negative predictive value: 62%) and after stenting was 96% (sensitivity: 100%, specificity: 96%, positive predictive value: 50%, and negative predictive value: 100%) (Fig. 4).
Figure 5 illustrates the potential of this novel technology. In one case (Fig. 5A), the operator needed to decide whether to implant a stent from the distal left main segment to cover fully the ostium of the left anterior descending (LAD) coronary artery to relieve the ischemia. A stent was implanted without covering the LAD ostium, and invasively measured FFR after stenting was 0.74. Consistent with this finding, FFRCT after virtual stenting without covering the LAD ostial lesion was 0.76, which was improved to 0.81 after virtual treatment of the whole LAD lesion including the ostium. In another case (Fig. 5B), understanding the functional significance of the left circumflex ostial lesion is critical in planning the treatment strategy. Although the lesion was significant by both angiography and cCTA, FFRCT after virtual stenting of the distal left circumflex lesion was 0.83. During the invasive procedure, it was found that the ischemia was relieved by stenting the distal lesion with a post-stent FFR of 0.88.
This study demonstrates that computational modeling of the coronary tree using cCTA images and virtual stenting is feasible and can help to predict the functional outcomes of stenting prior to invasive coronary angiography. This novel technique provides a noninvasive method for planning optimal treatment strategies in a patient- and lesion-specific manner before invasive procedures. These results—though needing validation in a larger cohort—represent initial findings that demonstrate the potential of this technology to not only diagnose lesion-specific ischemia but also to predict the therapeutic benefit of coronary revascularization.
Whereas cCTA can be helpful in identifying patients with coronary stenosis and selecting patients for invasive angiography, it is well known that the functional significance of stenosis cannot be accurately assessed by cCTA (7,11–13). In the present study, we sought to assess the accuracy of treatment planning using the computational model to simulate a virtual stent and to determine whether computation of anticipated post-stent FFRCT could predict the success of stenting prior to the invasive procedures. It is important to note that virtual stenting is performed by modification of the computational model derived from the original cCTA taken before the invasive procedure, and that post-stent FFR can be predicted from this same model without any additional noninvasive or invasive procedures. We found that FFRCT had a diagnostic accuracy of 96% in predicting or ruling out myocardial ischemia after stenting as defined by a post-stent FFR of >0.80. The mean difference between FFR after stenting and FFRCT after virtual stenting was 0.02 ± 0.05. Thus, it appears that comprehensive planning of a revascularization strategy and selection of the optimal target coronary lesion(s) for revascularization is possible using this novel technology, which can provide both anatomical and functional information for each lesion before the invasive procedure. This technique can be especially helpful for determining the revascularization strategy in patients with complex disease. For example, in patients with serial stenoses, even using invasive FFR, the true functional significance of each stenosis cannot be precisely assessed until the elimination of the other stenoses (14,15). However, it is possible to discriminate the functional significance of each stenosis using virtual stenting and computed FFRCT. Therefore, clinical application of this noninvasive “all-in-one” diagnostic and treatment planning approach may reduce clinically unnecessary interventions, procedural time, radiation dose, and costs. Importantly, we did not examine the potential diagnostic performance of FFRCT to identify lesions that may be best treated by medical therapy and/or coronary artery bypass graft. Whereas the FFRCT is hypothetically capable of determination of these findings, it remains unknown whether FFRCT can determine post-medical therapy and/or post–coronary artery bypass graft FFR with as high a diagnostic performance as reported in this study of percutaneous revascularization. Future diagnostic and outcomes studies evaluating these concepts now appear warranted.
First, the number of patients enrolled in this study was relatively small, with no systematic derivation/validation cohorts. Future studies doing so will be needed to corroborate these initial study results and, until then, these findings should be considered as “proof-of-concept.” Second, because this technology depends on the image quality of cCTA, it can be applied only to patients with diagnostic quality cCTA images obtained prior to invasive catheterization. Third, as this technology does not incorporate the microvascular injury during percutaneous coronary intervention, the accuracy of post-stent FFRCT can be lower in patients with severe microvascular injury during stent implantation. Fourth, as the FFR calculation is based on resting coronary anatomy, possible anatomical changes during actual stress cannot be reflected. However, to minimize this influence, all patients were given sublingual nitroglycerin immediately before cCTA. Finally, FFRCT is a new technology and further studies are needed to determine its clinical utility. Nonetheless, this novel technique is promising, and the present study demonstrates its potential as a noninvasive method for treatment planning prior to invasive coronary angiography and interventional procedures.
Treatment planning using noninvasive FFRCT and virtual stenting is feasible and may be helpful in determining optimal revascularization strategies before invasive procedures.
This study was supported by grants from the Innovative Research Institute for Cell Therapy (A062260); the Korea Healthcare Technology Research and Development Project, Ministry of Health and Welfare, Republic of Korea (A102065, A070001); and the Seoul National University Hospital Research Fund (03-2010-0270). The fractional flow reserve from coronary computed tomographic angiography analysis was performed by HeartFlow, Inc. Dr. Min has received research support from GE Healthcare and Philips Healthcare; has served on the medical advisory boards of GE Healthcare, Arineta, AstraZeneca, and Bristol-Myers Squibb; has served on the Speakers' Bureau of GE Healthcare; has received consulting fees from AstraZeneca, Bristol-Myers Squibb, and HeartFlow Inc.; and has equity interest in TC3 and MDDX. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. K.-H. Kim and Doh contributed equally to this paper.
- Abbreviations and Acronyms
- coronary computed tomographic angiography
- fractional flow reserve
- computed fractional flow reserve from coronary computed tomographic angiography
- left anterior descending
- Thrombolysis In Myocardial Infarction
- Received February 15, 2013.
- Revision received May 20, 2013.
- Accepted May 24, 2013.
- American College of Cardiology Foundation
- Shaw L.J.,
- Berman D.S.,
- Maron D.J.,
- et al.,
- for the COURAGE Investigators
- Fearon W.F.,
- Bornschein B.,
- Tonino P.A.,
- et al.,
- for the FAME Study Investigators
- Min J.K.,
- Shaw L.J.,
- Berman D.S.
- Meijboom W.B.,
- Van Mieghem C.A.,
- van Pelt N.,
- et al.
- Koo B.K.,
- Erglis A.,
- Doh J.H.,
- et al.
- De Bruyne B.,
- Pijls N.H.,
- Heyndrickx G.R.,
- Hodeige D.,
- Kirkeeide R.,
- Gould K.L.
- Pijls N.H.,
- De Bruyne B.,
- Bech G.J.,
- et al.