Abstract
Purpose
To measure the inter- and intraobserver variability among operators of varying expertise in conducting CT-derived fractional flow reserve (CT FFR) measurements on-site by using structural and fluid analysis and to evaluate differences in reproducibility between two different training methods for end users.
Materials and Methods
This retrospective analysis of the prospectively enrolled cohort included 22 symptomatic patients who underwent both 320–detector row coronary CT angiography and catheter-derived fractional flow reserve (FFR) within 90 days. Thirteen operators of varying expertise were assigned to one of two training arms: arm 1, on-site training by a specialist in CT FFR technology; arm 2, self-training through use of written materials. After the training, all 13 operators reviewed the CT data and measured CT FFR in 24 vessels in 22 patients. Inter- and intraoperator variability and agreements between CT FFR and catheter-derived FFR measurements were evaluated.
Results
The overall intraclass correlation coefficient (ICC) among operators was 0.71 (95% confidence interval: 0.58, 0.83) with a mean absolute difference (± standard deviation) of 0.027 ± 0.022. The operators in arm 2 showed greater interoperator differences than those in arm 1 (0.031 ± 0.024 vs 0.023 ± 0.018; P = .024). Among operators who recalculated CT FFR, the mean CT FFR value did not significantly differ between the first and second calculations (ICC, 0.66; 95% confidence interval: 0.46, 0.87), with the medical specialists producing the lowest intraoperator variability (0.053 ± 0.060). The overall correlation coefficient between CT FFR and catheter FFR was r = 0.61, with a mean absolute difference of 0.096 ± 0.089.
Conclusion
Good reproducibility of CT FFR values calculated on-site on the basis of structural and fluid analysis was observed among operators of varying expertise. Face-to-face training sessions may cause less variability.
© RSNA, 2019
Summary
Good reproducibility of CT-derived fractional flow reserve values calculated on-site on the basis of structural and fluid analysis was observed among operators of varying expertise, with potentially less variability in measurements after face-to-face training sessions.
Key Points
■ Good reproducibility of CT-derived fractional flow reserve (CT FFR) values calculated on-site on the basis of structural and fluid analysis was observed among operators of varying expertise.
■ Reproducible, on-site CT FFR technologies may help improve the clinical workflow for selecting patients suitable for revascularization procedures.
■ CT FFR and catheter-derived FFR showed a moderate correlation coefficient (r = 0.61), with a mean absolute difference of 0.096 ± 0.089.
Introduction
Invasive fractional flow reserve (FFR) measurement with coronary catheter angiography is the current standard procedure for assessing pathophysiologic cardiac ischemia (1–3). In recent years, noninvasive FFR estimations derived from coronary CT angiography (eg, FFRct; Heart Flow, Redwood City, Calif) (4–7) and other CT-derived FFR (CT FFR) technologies (8–12) have been evaluated as a tool to detect hemodynamically significant coronary lesions before catheterization. CT FFR estimation technologies may reduce the number of unnecessary interventional procedures for coronary artery disease (5,7,12–15). These technologies have high diagnostic accuracy for the detection of physiologically significant lesions and are considered to be more cost-effective than catheter angiography (5,7).
A CT FFR method based on structural and fluid analysis, developed by Canon Medical Systems (Otawara, Japan), calculates CT FFR on-site at the point of care (10,12,16,17). Although currently available for research only, this model eliminates complications associated with exporting patient data off-site (eg, infrastructure for the protection of patient data, secure transfer of the data to the United States from another country) and avoids workflow roadblocks by enabling on-site CT FFR calculations in minutes. A previous study found good diagnostic accuracy of this technology for the detection of significant lesions by CT FFR, with invasive FFR used as the reference standard (10).
For an on-site implementation of CT FFR, one would expect that inter- and intraobserver variability, experience level of the observer, and the type of training of the observer would introduce no statistically relevant variability in the CT FFR calculations. The previously mentioned studies found good reproducibility and agreement of CT FFR values obtained by two expert postprocessing analysts (10,12). Another study showed good agreement between an experienced analyst and an inexperienced operator who had undergone 40 minutes of on-site face-to-face training, but the study included only two operators and seven CT data sets (18). Furthermore, no studies have evaluated whether differences in the training method (eg, face-to-face or materials-based) cause differences in reproducibility.
The primary aim of this study was to measure inter- and intraobserver variability of CT FFR by using a larger cohort of readers across the experience spectrum of anticipated users. The secondary aim was to evaluate differences in reproducibility between two different training methods for end users.
Materials and Methods
Study Population
This study received financial support from Canon Medical Systems. Authors who are not employees of or consultants for Canon Medical Systems had full control of all data that might present a conflict of interest for those authors who are employees of or consultants for Canon Medical Systems. The employee’s involvement was related to the study design and the logistics regarding randomization of the data (described later in the article), as well as in-person training of the users.
This study retrospectively analyzed the prospectively collected data (10). The prospective study was approved by the institutional human research ethics committee at the site the study was performed. All patients gave written informed consent. The retrospective analysis of these prospective data was also approved by the institutional human research ethics committee at both of the sites described later in the article.
The study used the anonymized data collected for two CT FFR clinical studies conducted by Juntendo University Hospital, Tokyo, Japan (site 1) and by the Jacobs School of Medicine and Biomedical Sciences at the University of Buffalo, Buffalo, NY (site 2). For these studies, the methods were identical for the CT acquisition and reconstruction parameters and catheter-based FFR acquisition. Each patient underwent coronary CT angiography and invasive FFR for at least one major epicardial vessel. Patients were referred to undergo invasive angiography on the basis of the coronary CT angiographic result. The patients enrolled for prospective data collection at each site met the established inclusion and exclusion criteria described in Appendix E1 (supplement).
The initial data set comprised 23 consecutive patient data sets from site 1 and five consecutive patients from site 2. Two of the site 1 patients and four of the site 2 patients were excluded from the study because of inadequate catheter angiographic data; specifically, the location of the FFR wire with respect to the ostium could not be determined accurately. The requirement to know the location of the FFR wire measurement was considered an essential part of the current study to establish a reference standard for comparing operator results. No patients were excluded because of suboptimal CT image or high heart rate. Therefore, the study included 22 patients (mean age ± standard deviation of 68.7 years ± 8.48; mean Agatston score, 322.0 ± 397.2).
Coronary CT Acquisition
At both sites, patients underwent coronary CT angiography with a 320–detector row CT scanner (Aquilion ONE ViSION; Canon Medical Systems). All patients received sublingual nitroglycerin, and patients with a prescanning heart rate of ≥60 beats per minute were given 20 to 40 mg of metoprolol (Lopressor; Tanabe Seiyaku, Osaka, Japan) orally, and, if the heart rate remained at ≥60 beats per minute after 1 hour, landiolol (Corebeta; Ono Pharmaceutical, Osaka, Japan) (0.125 mg/kg) was intravenously administered.
The coronary CT angiographic scanning parameters were as follows: tube voltage, 100 kVp (except for one patient scanned at 120 kVp because of body mass index > 30 kg/m2); detector collimation, up to 320 × 0.5 mm; gantry rotation time, 275 msec; and tube current automatically adapted for patient size in the x-y dimension using tube current modulation (SUREExposure) with a standard deviation setting of 19. The craniocaudal range was selected from 240 detector rows (12 cm) to 320 rows (16 cm) to include the entire coronary tree. The contrast agent iohexol (Omnipaque 350 mg of iodine per milliliter; Daiichi Sankyo, Tokyo, Japan) was injected for 12 seconds at 18 mg of iodine per kilogram of body weight per second (Dual Shot GX 7; Nemoto Kyorindo, Tokyo, Japan) followed by 30 mL of saline at the same injection rate. With use of bolus tracking, scanning was started when the CT attenuation in the ascending aorta reached 300 HU. A single-heartbeat scan was obtained with prospective electrocardiographic gating, to cover 70%–99% of the R-R interval. The CT data were reconstructed for the following analyses and interpretation with a reconstruction thickness of 0.5 mm and the standard adaptive iterative dose reduction using three-dimensional processing.
Details of catheter angiography and catheter FFR measurements are described in the Appendix E2 (supplement).
Operators
Thirteen operators of varying levels of expertise used the CT FFR software (Table E1 [supplement]): three medical specialists, three CT technology specialists, and seven trainees (residents and fellows). Medical specialist was defined as having completed level 2 or level 3 training in cardiac CT as defined by the Certification Board of Cardiovascular Computed Tomography. The CT technology specialists had some or no cardiac CT experience. The cardiac CT experience level of the trainees ranged from some to no CT experience. The CT FFR software was processed for all eligible patient data sets at site 2. Included in the CT FFR analyses were 24 vessels (16 left anterior descending arteries, three left circumflex arteries, and five right coronary arteries) in 22 patients who had also undergone catheter FFR. Images used to train the software were not used in the inter- or intraoperator variability study. No operators had experience using the CT FFR software before the study.
All 13 operators read the full set of 22 cases in a random blinded fashion. The order of cases differed from reader to reader in a random manner. Operators were instructed not to discuss the CT FFR technology with anyone except the study coordinators.
To analyze intraoperator variability, five of the 13 operators read 11 preselected CT images a second time, also in random order. The time between the first and second readings of any given image was 7 days or longer.
Training
All 13 operators underwent basic training on a Vitrea 7.6 workstation (Vital Images, Minnetonka, Minn) to ensure they were comfortable using it. The basic training did not include training with any CT FFR application.
Operators were first categorized into three experience levels (ie, medical specialist, CT technology specialists, and trainees), after which they were randomly (through use of the “draw a straw” method) assigned to one of two training arms within their experience category (Table E1 [supplement]); the arms were equivalently composed, with operators from each level of training evenly distributed among them.
Arm 1 received hands-on training by a product applications specialist, who had been trained in the use of the CT FFR analysis software. The purpose of the training was to teach the operator the workflow (ie, how to use the software, review the contours and centerlines, and edit when appropriate). The on-site training included the established training materials along with four cases used for training. Operators were also given training materials. Arm 1 allowed the operator to ask questions in real time as they performed the test cases.
Arm 2 relied only on the established training materials provided with the CT FFR technology (no on-site training). This design was selected to be representative of a worst-case scenario in which individuals must learn to use the technology by using only the materials provided and the option to ask the vendor (eg, online help, telephone assistance).
Each operator (both in arm 1 and arm 2) was given the same training materials and the same four training cases with which to practice. The training cases increased in difficulty. An example of an easy case was one that required little or no contouring (ie, the software contoured everything correctly). An example of a challenging case was one that required vessel editing (eg, the software selected a branch instead of the main vessel and substantial calcium or irregular vessel anatomy was encountered). Reporting the results required little expertise because the result was a quantitative value at a specific location. For this study, the operators were also provided an additional training document that described the method for measuring CT FFR at a predefined location (matching the location of the wire-based FFR).
CT FFR Analysis
All data were anonymized and installed on two Vitrea 7.6 workstations equipped with the CT FFR analysis software and located at site 2. CT FFR analysis was performed by using the CT FFR software (Canon Medical Systems) after each operator had completed training. First, the operator selected a coronary CT angiographic image from the cardiac phase that showed the least motion. Vessel centerlines and lumen contours were then automatically processed by using the CT FFR software and edited manually by the operator as needed. The operator measured the CT FFR value of the “target location” and recorded it in the CT FFR results sheet (Figure). The target lesion for CT FFR measurement was predetermined by the study coordinator on the basis of the catheter FFR.
Figure:
Sample case shows detection of the target location in CT-derived fractional flow reserve (CT FFR) measurement. With use of this method, distance from the ostium to the measurement location was determined. Each operator received a spreadsheet for recording the results, which provided the subject code, vessel(s) of interest, and the wire position as distance from the ostium. Shown are three-dimensional (3D) coronary tree from CT, stretched multiplanar reconstruction (MPR) image from CT, stenosis measurements on stretched MPR image, CT FFR measurement (bottom right), and catheter angiographic images (bottom middle and left) with FFR measurement. Black arrow = segment where FFR was measured. White arrow = pressure wire to measure FFR.
The location for catheter FFR measurement was identified by using an image from the catheter FFR performed in the original prospective trial. The study coordinator identified the same location (by visual comparison) on the CT image by using vessel branching or calcifications as landmarks. The CT application was then used to measure the length from the ostium to that measurement point. The measurement location (as denoted by the length from the ostium) for each vessel of interest was provided in a spreadsheet to the operators before the CT FFR analysis.
Statistical Analysis
Summary statistics for the continuous variables are presented as the mean ± standard deviation. Study outcomes were the levels of agreement of CT FFR values between operators, measured as intraclass correlation coefficients (ICCs) and mean absolute differences. We also evaluated agreements by κ statistics after CT FFR, which were classified as abnormal (≤0.8) or normal (> 0.8). For the evaluation of the agreement between CT FFR and invasive FFR at relevant points, Pearson correlation coefficients, mean absolute differences, and correct or incorrect classification rates (≤0.8 or >0.8) were calculated. These analyses were performed in the samples from overall population, as well as in the subgroups stratified by training method (arm 1 or 2) and by specialty (medical specialists, CT technology specialists, or trainees). Statistical analyses were performed by using Stata software, version 14.1 (Stata, College Station, Tex), with P < .05 considered to indicate a statistically significant difference.
Results
Demographic, clinical, and scan characteristics of the 22 patients are listed in Table 1. Significant stenosis (>50% in diameter) was observed in 18 of the 22 patients. CT FFR was successfully calculated for all 24 vessels by all operators. The mean distance from the ostium to the point where CT FFR was measured was 98.0 mm ± 21.2.
Table 1:
Baseline Patient Characteristics
Note.—Unless otherwise noted, values are expressed as number of patients, with percentages in parentheses. DLP = dose-length product, LAD = left anterior descending artery, LCx = left circumflex artery, RCA = right coronary artery.
*Data are means ± standard deviations.
Interoperator Variability
The overall ICC was good (0.71 [95% confidence interval: 0.58, 0.83]). The ICC tended to be higher among the operators in arm 1 and lower among the trainees (Table 2). The κ statistics for overall population indicated significant fair agreement (κ = 0.29) (Table 2).
Table 2:
Intraclass Correlation Coefficient of CT-derived Fractional Flow Reserve Measurements as Indicator of Interoperator Variability
Note.—ICC = intraclass correlation coefficient. * Data in parentheses are 95% confidence intervals
The overall mean absolute difference of CT FFR between operators was 0.027 ± 0.022. The operators in arm 2 showed larger interoperator differences than those in arm 1 (0.031 ± 0.024 vs 0.023 ± 0.018; P = .024) (Table 3). The trainees showed greater variability than the medical specialists and CT technology specialists, although not by a statistically significant margin; the largest mean absolute difference between trainees and other trainees was 0.036 ± 0.030 (Table 4).
Table 3:
Mean Absolute Difference between Operators Stratified by Training Method
Table 4:
Mean Absolute Difference between Operators Stratified by Specialty
Intraoperator Variability
Five operators remeasured the CT FFR values in 11 images after a delay of 1 week; the ICC was 0.66 (95% confidence interval: 0.46, 0.87) and the κ statistic was 0.29. The mean CT FFR value did not significantly differ between the first and second readings: 0.798 ± 0.148 versus 0.792 ± 0.153 (P = .83). The medical specialists produced the least variability between readings, but no group showed a statistically significant difference: The absolute differences between first and second readings were 0.053 ± 0.060, 0.116 ± 0.096, and 0.121 ± 0.092 (P = .095) for medical specialists, CT technology specialists, and trainees, respectively.
Agreement with Catheter FFR
The mean value of catheter FFR for the 24 vessels was 0.85 ± 0.08. With use of a cutoff of 0.8, catheter FFR was abnormal in eight of the 24 vessels (30%).
CT FFR and catheter FFR showed an overall correlation coefficient of r = 0.61, with a mean absolute difference of 0.096 ± 0.089; this was not statistically significantly different between the training methods (arm 1 vs arm 2; P = .20) or specialties (medical specialists, CT technology specialists, and trainees, P = .34) (Table 5). We found 68.9% (215 of 312) of the measurements were classified into the correct categories (≤0.8 or >0.8).
Table 5:
Agreement between CT-derived Fractional Flow Reserve and Catheter-derived Fractional Flow Reserve
*Data are absolute difference ± standard deviation.
Discussion
The current study showed good reproducibility of CT FFR values calculated on-site on the basis of structural and fluid analysis among operators with different levels of experience. Interreader correlation in dichotomous CT FFR results (≤0.8 or >0.8) was fair, and the correlation between CT FFR and invasive FFR was modest. On-site calculation of CT FFR is expected to expedite decision making in the management of coronary artery disease. The current study suggests that the reproducibility of the values might be better after face-to-face training than after training with written materials and that trainees (residents or fellows) may need to receive more careful training than do medical or CT technology specialists.
As with other CT FFR simulation techniques (4,6,9,11), determination via segmentation of the patient’s coronary anatomy is one of the most important steps in computing reliable CT FFR values. The previous data showed that incorrect identification of the vessel’s centerline results in substantially different vessel lengths, and thus incorrect CT FFR values (18). In the current study, data obtained by observers who already had a fundamental knowledge of CT acquisition tended to be more reproducible than those obtained by residents or fellows. We propose that when the software is commercially distributed, candidate operators should receive face-to-face training sessions from an expert before the software is implemented in the institution’s clinical workflow. For vessels with severe or diffuse calcification, based on the previous study, the training session should include how to correctly identify the vessel’s centerline and how to define the lumen in calcified segments (18). Such training is especially important for people who are unfamiliar with CT postprocessing (eg, residents or fellows).
The mean absolute differences (<0.04) between operators were similar to the variability of FFRct (19). Other on-site CT FFR techniques have limited reproducibility data, especially for comparing operators with various expertise. Donnelly et al (8) reported an ICC of 0.95 for on-site CT FFR calculation between two experienced cardiologists. Although invasive FFR is the reference standard, this method is reported to also have measurement errors and be affected by biological factors (20). In another study (3), the absolute difference between two separate FFR measurements at the same location was 0.03, which is similar to the variability in the current CT FFR study.
In the current study, the mean analysis time was approximately 25 minutes, which included manual adjustment of the centerlines and contours plus the calculation time for fluid analyses. This is approximately the same as the previously reported value of 30 minutes for CT FFR calculated by experienced technologists (12) and is shorter than durations reported for inexperienced operators (18). The average analysis time of other on-site CT FFR technologists varies from 30 minutes to 2 hours (8,9,11). Although the underlying algorithms differ among different on-site CT FFR simulation techniques, the most time-consuming step is vessel segmentation (9). We expect the analysis time to decrease as operators accumulate experience with the procedures. The results of the current study show that CT technology specialists achieved similar or even higher reproducibility and correlation with invasive FFR compared with medical specialists. These data support that, in the clinical practice, experienced CT technologists can perform CT FFR analysis instead of busy medical doctors.
Although this study aimed to evaluate reproducibility between operators, not the “validity” of the technique, we included invasive FFR data in the analysis as a reference standard method. The correlation coefficient was better than the previous data from the validation study of this technique (r = 0.57) (12), but the accuracy (68.9%) was worse than reported by other groups (5,8,11). One likely reason is the higher mean calcium score in the present population (322.0 ± 397.2); severe calcification led to false-positive CT FFR findings, in keeping with the results of studies of other CT FFR algorithms (4,5,8,12), probably due to overestimation of stenosis. Because the current study demonstrated high reproducibility, correction of the CT FFR calculation at calcified segments should be considered as a next step. Another reason for the modest correlation may be the location of the CT FFR measurements. Invasive FFR was constantly measured at a distal segment, irrespective of the stenosis location. Because we measured CT FFR where the invasive FFR was measured, the mean distance from the ostium to the point of CT FFR measurement was 98.0 mm. Because of spatial resolution, measurements that are more distal are expected to be less accurate.
A strength of the current study is that it used 13 operators with diverse experience to evaluate the reproducibility of structural and fluid analysis–based CT FFR, thus simulating clinical practice after commercial distribution of the software. These 13 operators were part of a typical health care team taking care of the patients. The involvement of CT technologists, residents, cardiology fellows, and cardiologists makes our study relevant to current clinical practice. In addition, we compared two different training methods.
However, this study had some limitations. First, although the effect size, calculated on the basis of the mean absolute differences reported in Table 3, indicated sufficient sample size (21) for the current analyses, a modest number of CT data sets and vessels did not allow for subanalyses based on patient characteristics (eg, stratification based on calcium score). Second, all operators were from the same institution in the United States, and thus further study may need to include other institutions in other countries. In addition, the usefulness of the core laboratory should also be investigated. Future studies comparing different CT FFR algorithms with regard to reproducibility, accuracy, and costs would be ideal to further explore the clinical utility of CT FFR technology. Third, the “target location” for CT FFR measurement was predefined by the study coordinator. When CT FFR is used in clinical settings, the investigators need to prospectively identify the point to measure the CT FFR value. Fourth, the CT FFR technique proposed in the current study needs CT data covering 70%–99% of the R-R interval, which increases a radiation dose. However, the mean dose-length product in the current study was 255.5 mGy · cm. Finally, it would be advantageous to compare the CT FFR analyses of two different CT scans from the same patient to evaluate the effects of scanning parameters and image quality on reproducibility, which was beyond the scope of this study.
Good reproducibility of CT FFR measurements calculated on-site on the basis of structural and fluid analysis was observed among operators of varying expertise. Face-to-face training sessions, rather than self-training, may lead to more reproducible values. Reproducible, on-site CT FFR technologies may enable improvement in the clinical workflow for selecting patients suitable for revascularization procedures.
APPENDIX
Acknowledgments
Acknowledgment
We thank James Sayre for assistance with statistical analysis.
Study supported by Canon Medical Systems.
Current address: Department of Radiology, University of Cincinnati, Cincinnati, Ohio
Disclosures of Conflicts of Interest: K.K.K. disclosed no relevant relationships. E.A. Activities related to the present article: disclosed no relevant relationships. Activities not related to the present article: employed by Canon Medical Systems USA. Other relationships: disclosed no relevant relationships. K.N.S. disclosed no relevant relationships. V.I. disclosed no relevant relationships. M.F.W. Activities related to the present article: institution receives grant from Jacobs Institute. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. N.A. disclosed no relevant relationships. A.B. Activities related to the present article: institution receives grant from Jacobs Institute. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. S.B.K. disclosed no relevant relationships. S.K. Activities related to the present article: institution receives grant from Jacobs Institute. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. S.M. disclosed no relevant relationships. C.M. Activities related to the present article: author did not receive payment or services from a third party for any aspect of the submitted work; author is uncertain if his institution received payment or services from a third party for any aspect of the submitted work (defers to lead author). Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. K.P. disclosed no relevant relationships. T.R. Activities related to the present article: author received payment from the University of Buffalo for participation in review activities. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. A.C.S. disclosed no relevant relationships. M.M.S. Activities related to the present article: institution receives grant from Jacobs Institute. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. S.W. Activities related to the present article: author paid by Toshiba for participating in the data collection of this study. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. S.F. disclosed no relevant relationships. U.C.S. Activities related to the present article: author received consulting fee from Jacobs Neurological Research Institute for interpretation of CT FFT and was fully blinded and not influenced by any means regarding the outcome. Activities not related to the present article: disclosed no relevant relationships. Other relationships: disclosed no relevant relationships. F.J.R. Activities related to the present article: institution received grant from Canon Medical Systems. Activities not related to the present article: medical director of Imagia (there is no overlap between this work and any company activity). Other relationships: disclosed no relevant relationships. C.N.I. Activities related to the present article: University of Buffalo received grant from Canon Medical Systems; institution receives travel funds for graduate and undergraduates for international conferences. Activities not related to the present article: author paid consultant fee for studies on vascular device evaluation from Jacobs Institute; institution receives NIH grants, Cumming Foundation grant, Canon Medical Systems grant, author paid for lecture on 3D printing advances from Canon Medical Systems; paid for patent for x-ray system design (University of Buffalo); University of Buffalo gives travel grants to conferences. Other relationships: disclosed no relevant relationships.
Abbreviations:
- FFR
- fractional flow reserve
- ICC
- intraclass correlation coefficient
References
- 1.Pijls NH, De Bruyne B. Coronary pressure measurement and fractional flow reserve. Heart 1998;80(6):539–542. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Pijls NHJ, van Schaardenburgh P, Manoharan G, et al. Percutaneous coronary intervention of functionally nonsignificant stenosis: 5-year follow-up of the DEFER Study. J Am Coll Cardiol 2007;49(21):2105–2111. [DOI] [PubMed] [Google Scholar]
- 3.Bech GJW, De Bruyne B, Pijls NHJ, et al. Fractional flow reserve to determine the appropriateness of angioplasty in moderate coronary stenosis: a randomized trial. Circulation 2001;103(24):2928–2934. [DOI] [PubMed] [Google Scholar]
- 4.Min JK, Taylor CA, Achenbach S, et al. Noninvasive fractional flow reserve derived from coronary CT angiography: clinical data and scientific principles. JACC Cardiovasc Imaging 2015;8(10):1209–1222. [DOI] [PubMed] [Google Scholar]
- 5.Nørgaard BL, Leipsic J, Gaur S, et al. Diagnostic performance of noninvasive fractional flow reserve derived from coronary computed tomography angiography in suspected coronary artery disease: the NXT trial (Analysis of Coronary Blood Flow Using CT Angiography: Next Steps). J Am Coll Cardiol 2014;63(12):1145–1155. [DOI] [PubMed] [Google Scholar]
- 6.Taylor CA, Fonte TA, Min JK. Computational fluid dynamics applied to cardiac computed tomography for noninvasive quantification of fractional flow reserve: scientific basis. J Am Coll Cardiol 2013;61(22):2233–2241. [DOI] [PubMed] [Google Scholar]
- 7.Douglas PS, Pontone G, Hlatky MA, et al. Clinical outcomes of fractional flow reserve by computed tomographic angiography-guided diagnostic strategies vs. usual care in patients with suspected coronary artery disease: the prospective longitudinal trial of FFR(CT): outcome and resource impacts study. Eur Heart J 2015;36(47):3359–3367. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Donnelly PM, Kolossváry M, Karády J, et al. Experience with an on-site coronary computed tomography-derived fractional flow reserve algorithm for the assessment of intermediate coronary stenoses. Am J Cardiol 2018;121(1):9–13. [DOI] [PubMed] [Google Scholar]
- 9.Coenen A, Lubbers MM, Kurata A, et al. Fractional flow reserve computed from noninvasive CT angiography data: diagnostic performance of an on-site clinician-operated computational fluid dynamics algorithm. Radiology 2015;274(3):674–683. [DOI] [PubMed] [Google Scholar]
- 10.Fujimoto S, Kawasaki T, Kumamaru KK, et al. Diagnostic performance of on-site computed CT-fractional flow reserve based on fluid structure interactions: comparison with invasive fractional flow reserve and instantaneous wave-free ratio. Eur Heart J Cardiovasc Imaging 2019;20(3):343–352. [DOI] [PubMed] [Google Scholar]
- 11.Kishi S, Giannopoulos AA, Tang A, et al. Fractional flow reserve estimated at coronary CT angiography in intermediate lesions: comparison of diagnostic accuracy of different methods to determine coronary flow distribution. Radiology 2018;287(1):76–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ko BS, Cameron JD, Munnur RK, et al. Noninvasive CT-derived FFR based on structural and fluid analysis: a comparison with invasive FFR for detection of functionally significant stenosis. JACC Cardiovasc Imaging 2017;10(6):663–673. [DOI] [PubMed] [Google Scholar]
- 13.Kawaji T, Shiomi H, Morishita H, et al. Feasibility and diagnostic performance of fractional flow reserve measurement derived from coronary computed tomography angiography in real clinical practice. Int J Cardiovasc Imaging 2017;33(2):271–281. [DOI] [PubMed] [Google Scholar]
- 14.Koo BK, Erglis A, Doh JH, et al. Diagnosis of ischemia-causing coronary stenoses by noninvasive fractional flow reserve computed from coronary computed tomographic angiograms. Results from the prospective multicenter DISCOVER-FLOW (Diagnosis of Ischemia-Causing Stenoses Obtained Via Noninvasive Fractional Flow Reserve) study. J Am Coll Cardiol 2011;58(19):1989–1997. [DOI] [PubMed] [Google Scholar]
- 15.Duguay TM, Tesche C, Vliegenthart R, et al. Coronary computed tomographic angiography-derived fractional flow reserve based on machine learning for risk stratification of non-culprit coronary narrowings in patients with acute coronary syndrome. Am J Cardiol 2017;120(8):1260–1266. [DOI] [PubMed] [Google Scholar]
- 16.Hirohata K, Kano A, Goryu A, et al. A novel CT-FFR method for the coronary artery based on 4D-CT image analysis and structural and fluid analysis. In: Hoeschen C, Kontos D, eds. Proceedings of SPIE: Medical Imaging 2015—Physics of Medical Imaging. Vol 9412. Bellingham, Wash: International Society for Optics and Photonics, 2015; 94122O. [Google Scholar]
- 17.Kato M, Hirohata K, Kano A, et al. Fast CT-FFR analysis method for the coronary artery based on 4D-CT image analysis and structural and fluid analysis. In: ASME 2015 International Mechanical Engineering Congress and Exposition: Volume 3: Biomedical and Biotechnology Engineering. New York, NY: American Society of Mechanical Engineers, 2015; V003T03A023. [Google Scholar]
- 18.Ri K, Kumamaru KK, Fujimoto S, et al. Noninvasive computed tomography-derived fractional flow reserve based on structural and fluid analysis: reproducibility of on-site determination by unexperienced observers. J Comput Assist Tomogr 2018;42(2):256–262. [DOI] [PubMed] [Google Scholar]
- 19.Gaur S, Bezerra HG, Lassen JF, et al. Fractional flow reserve derived from coronary CT angiography: variation of repeated analyses. J Cardiovasc Comput Tomogr 2014;8(4):307–314. [DOI] [PubMed] [Google Scholar]
- 20.Petraco R, Sen S, Nijjer S, et al. Fractional flow reserve-guided revascularization: practical implications of a diagnostic gray zone and measurement variability on clinical decisions. JACC Cardiovasc Interv 2013;6(3):222–225. [DOI] [PubMed] [Google Scholar]
- 21.Cohen J. Statistical power analysis for the behavioral sciences. 2nd ed. New York, NY: Academic Press, 1988; 30. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.