Author + information
- Received April 26, 2016
- Revision received September 6, 2016
- Accepted September 8, 2016
- Published online May 1, 2017.
- Danai Kitkungvan, MDa,
- Nils P. Johnson, MD, MSa,
- Amanda E. Roby, PET, CNMT, RT(N)b,
- Monika B. Patel, MDa,
- Richard Kirkeeide, PhDc and
- K. Lance Gould, MDd,∗ ()
- aDivision of Cardiology, Department of Medicine and Weatherhead PET Center for Preventing Atherosclerosis, McGovern Medical School and Memorial Hermann Hospital, Houston, Texas
- bPET Imaging, Department of Medicine and Weatherhead PET Center for Preventing Atherosclerosis, McGovern Medical School and Memorial Hermann Hospital, Houston, Texas
- cDepartment of Medicine and Weatherhead PET Center for Preventing Atherosclerosis, McGovern Medical School, Houston, Texas
- dWeatherhead PET Center for Preventing and Reversing Atherosclerosis, McGovern Medical School, Houston, Texas
- ↵∗Address for correspondence:
Dr. K. Lance Gould, Weatherhead PET Center for Preventing and Reversing Atherosclerosis, McGovern Medical School, University of Texas Health Science Center at Houston, 6431 Fannin Street, Room MSB 4.256, Houston, Texas 77030.
Objectives Positron emission tomography (PET) quantifies stress myocardial perfusion (in cc/min/g) and coronary flow reserve to guide noninvasively the management of coronary artery disease. This study determined their test-retest precision within minutes and daily biological variability essential for bounding clinical decision-making or risk stratification based on low flow ischemic thresholds or follow-up changes.
Background Randomized trials of fractional flow reserve–guided percutaneous coronary interventions established an objective, quantitative, outcomes-driven standard of physiological stenosis severity. However, pressure-derived fractional flow reserve requires invasive coronary angiogram and was originally validated by comparison to noninvasive PET.
Methods The time course and test-retest precision of serial quantitative rest-rest and stress-stress global myocardial perfusion by PET within minutes and days apart in the same patient were compared in 120 volunteers undergoing serial 708 quantitative PET perfusion scans using rubidium 82 (Rb-82) and dipyridamole stress with a 2-dimensional PET-computed tomography scanner (GE DST 16) and University of Texas HeartSee software with our validated perfusion model.
Results Test-retest methodological precision (coefficient of variance) for serial quantitative global myocardial perfusion minutes apart is ±10% (mean ΔSD at rest ±0.09, at stress ±0.23 cc/min/g) and for days apart is ±21% (mean ΔSD at rest ±0.2, at stress ±0.46 cc/min/g) reflecting added biological variability. Global myocardial perfusion at 8 min after 4-min dipyridamole infusion is 10% higher than at standard 4 min after dipyridamole.
Conclusions Test-retest methodological precision of global PET myocardial perfusion by serial rest or stress PET minutes apart is ±10%. Day-to-different-day biological plus methodological variability is ±21%, thereby establishing boundaries of variability on physiological severity to guide or follow coronary artery disease management. Maximum stress increases perfusion and coronary flow reserve, thereby reducing potentially falsely low values mimicking ischemia.
Positron emission tomography (PET) quantifies stress myocardial perfusion in units of cc/min/g and coronary flow reserve (CFR) (1–3) as noninvasive guides to management of coronary artery disease (CAD) paralleling indirect relative flow reserve of invasive pressure-derived fractional flow reserve (4,5) originally validated by comparison to PET (6). Therefore, their test-retest precision and daily biological variability are essential for clinical decision-making based on thresholds of low myocardial perfusion causing ischemia.
Despite the long history of dipyridamole stress originated by Gould et al. (7,8), we critically examined our technology and biological variability of stress perfusion and CFR to determine whether methodology variability needed technical improvement or day-to-day biological variability were limitations to guiding or following management of CAD as recently reported for regadenoson (9). Therefore, this study aimed to establish the essential link between comprehensive quantification of perfusion measurements and their variability that sets bounds for clinical decision-making or risk stratification.
From November 2014 to August 2015, subjects 40 years or older were recruited at Weatherhead PET Center for Preventing and Reversing Atherosclerosis of University of Texas Medical School. Written informed consent was obtained from volunteers (commonly with risk factors), patients referred for clinical PET who did not have insurance coverage, or clinic patients who desired PET follow-up.
Exclusion criteria included contraindication to dipyridamole, pregnancy, active breastfeeding, clinical instability, and inability to undergo 2 PET protocols (Figure 1) within 2 days to 3 weeks apart in which early-late sequences within minutes were randomized before the first PET scan.
Cardiac PET acquisition and analysis
Subjects were instructed to fast for 4 h and to abstain from caffeine, theophylline, and cigarettes for at least 24 h. Cardiac PET used the Discovery ST 16-slice PET-computed tomography scanner (GE Healthcare, Waukesha, Wisconsin) in 2-dimensional mode as previously reported (2,3,7–12).
Emission images were obtained over 4 min at starting intravenous injection of 30 to 50 mCi of generator-produced rubidium 82 (Rb-82) (Bracco Diagnostics, Princeton, New Jersey). The first 2-min emission acquisition comprised arterial input images. The last 2-min emission acquisition comprised myocardial uptake images. Pharmacological stress used dipyridamole infusion (0.56 mg/kg) over 4 min (0.142 mg/kg/min).
An experienced PET cardiologist administered dipyridamole and monitored every patient throughout imaging, followed by 75 mg of intravenous aminophylline. Angina was treated with intravenous aminophylline, metoprolol, or sublingual nitroglycerin. Continuous heart rate, blood pressure, and 12-lead electrocardiographic (ECG) monitoring during stress identified significant >1-mm ST-segment depression.
Quantitative PET analysis
Computed tomography scans for attenuation correction were acquired before rest and after the last stress emission imaging at reduced radiation dose as previously reported (2,3,7–12). Coregistration was optimized for every image by shifting PET to fit attenuation data and reconstructed as previously reported (10).
As previously reported (2,3,7–12), for each radial segment of every short-axis slice, absolute myocardial perfusion (in cc/min/g) was quantified for each of 1,344 pixels in the left ventricular (LV) image with a 5-pixel smoothing noise reduction algorithm using HeartSee software (University of Texas-Houston, Houston, Texas, FDA K14366) approximating reconstructed scanner resolution of 1.5-cm full width one-half maximum with filters. This software incorporates our validated model for Rb-82 (13), reported by others to “have higher sensitivity for detection and localization of abnormal flow” (14) than multicompartmental models using time-activity curves derived from serial short (15-s) noisy images. Optimal arterial inputs were customized for individual patients from among aortic and left atrial locations as previously reported to determine rest and stress flow (15).
CFR was computed as stress-to-rest ratio for each of 1,344 pixels, synonymous with myocardial perfusion reserve to emphasize physiological concepts. Coronary flow capacity maps display stress perfusion and CFR for each pixel as a percentage of LV as illustrated in Figure 2, which maps each of 1,344 LV pixels with its stress perfusion and CFR value at its regional location (12). The histogram distribution of 1,344 pixel severities defined by both stress flow and CFR then provides a pixel-level analysis of the entire range, size, and distribution of stress perfusion and CFR for assessing the effects of maximal or submaximal stress. Global values are the average of 1,344 pixels; quadrant values are averages of all pixel values (336) in nonoverlapping quadrants.
Rest and dipyridamole stress imaging protocols
Rest perfusion precision
Sequential rest-rest-stress PET perfusion imaging was performed at 10-min intervals on day 1 and repeated 1 to 3 weeks later on day 2. The serial 10-min rest-rest scans quantified test-retest methodological precision for measuring resting perfusion (in cc/min/g), whereas repeat rest scans on different days assessed daily biological plus methodological variability in the same patient.
Stress perfusion precision
Quantifying test-retest precision of stress perfusion (in cc/min/g) and CFR requires stable constant stress perfusion for approximately 15 min to allow serial acquisition of stress perfusion images. Because the time course of myocardial hyperemia after standard 4-min dipyridamole infusion has not been defined, the protocols shown in Figure 1 were implemented in repeat paired PET studies in the same patient at minutes and days intervals.
After dipyridamole injection starting at time 0, the first RB-82 generator activation occurred at 7 or 8 min (day 1 PET 1) and again at 12, 13, 14, 15, or 16 min on day 2 (day 1 PET 2). Radiotracer delivery, PET scanner acquisition, image processing, and quantitative perfusion software remained the same for all PET scans. Subjects returned for the second PET scan (day 2 PET 1 and PET 2) within 3 weeks with similar or randomized different time intervals.
Precision and variability were determined for global, average quadrant, and individual pixel values of stress flow and CFR. R version 3.1.0 (R Foundation for Statistical Computing, Vienna, Austria) and standard summary statistical tests were used. Applicable tests were 2-tailed, and p < 0.05 was considered statistically significant. Linear regression is reported between rest perfusion and rest pressure rate product (PRP). Analysis of variance (ANOVA) compared characteristics among timing protocols. Paired or unpaired Student t test was used to evaluate continuous variables where appropriate. The Pitman-Morgan F test was used to test for differences in variability of stress perfusion between 2 groups. An ANOVA model with mixed effects (to account for repeated measurements from the same subject) compared absolute flow and CFR among various timing sequences. Because an overall ANOVA p value was significant, a Tukey all-pair comparison was applied to determine which timing conditions provided a different response. To compare the histogram distribution between groups of stress perfusion and CFR for each of 1344 pixels as percentage of LV in color-coded ranges of coronary flow capacity, we used the Kolmogorov-Smirnov (KS) test for differences in histogram distribution.
One hundred twenty subjects underwent 708 PET quantitative PET perfusion scans. Thirty-one subjects underwent rest-rest-stress protocol on day 1 and 1 to 3 weeks later on day 2. Eighty-nine subjects underwent rest-stress-stress protocol on days 1 and 2. The median between paired PETs was 16 days (mean 22 ± 15 days) with no change in medical status or medications. Baseline characteristics of subjects are listed in Table 1. Four day-2 PET sessions were not obtained because subjects withdrew, and 5 PET sessions were excluded because of technical issues of scanner operations or settings, intravenous infusion of Rb-82 or dipyridamole, or venous abnormalities that invalidated arterial input.
Because definitive regional quantitative myocardial perfusion by PET for guiding management of CAD is not widely recognized, Figure 3 illustrates relative stress myocardial perfusion images in complex CAD in which angiograms did not provide guidance for revascularization compared to medical management.
The patient shown in Figure 3 is a 65-year-old man with risk factors and a right coronary artery (RCA) stent inserted in 2009 who was referred for PET because of abnormal stress test and angiogram showing in-stent RCA occlusion with distal collaterals, moderate stenosis of the first and second diagonal branches, and moderate stenosis of the first and second obtuse marginal branches.
Detailed review of images for this patient provides insight into diffuse and focal physiological severity unfamiliar to most readers and its precision. Resting relative perfusion images showed a basal inferior transmural scar composing 10% of the LV (not shown). Relative stress perfusion images show a severe defect involving 22% of inferior LV in the RCA distribution, indicating large border zones of viable myocardium with reduced CFR composing an additional 14% of LV (Figure 3A). ECG-gated perfusion images showed ejection fraction of 63% at rest and 67% during dipyridamole stress, with stress-induced inferior hypokinesis without angina or ECG changes. In view of multiple stenosis by angiogram, relative images do not identify or quantify the extent of “balanced stenosis,” low flow ischemia, or diffuse disease essential for management decisions.
Quantitative stress perfusion (in cc/min/g) was moderately reduced diffusely and severely reduced in inferior, inferolateral, and inferoseptal quadrants of LV according to the color bar scale for stress perfusion (flow) (Figure 3B). All quantitative color bars are coded red for 125 healthy young volunteers younger than 40 years without risk factors; orange for healthy subjects with risk factors but no known CAD; yellow for patients with known CAD with or without revascularization; blue for patients with stress perfusion defects, angina, and/or >1-mm ST-segment depression on ECG during dipyridamole stress; and green for a stress defect and either angina or ST-segment depression but not both, as previously reported (2,3,7–12). CFR outside the stress defect is mildly to moderately reduced diffusely but above low flow ischemic levels reflecting diffuse, nonischemic coronary atherosclerosis in those regions (Figure 3C). In the stress defect, CFR is severely reduced inferiorly to below resting perfusion in a small segment, indicating myocardial steal (dark blue) associated with collaterals beyond the RCA chronic total occlusion.
These 2 primary flow metrics, stress perfusion (in cc/min/g) and CFR, completely define physiological severity but are complex to interpret independently. Accordingly, as previously reported (2,3,7–12), they are integrated into a coronary flow capacity map (Figure 3D) with color-coded pixels for the same ranges of patient groups according to the 2-dimensional plot in Figure 2. Coronary flow capacity is severely reduced inferiorly in 26% of the LV, of which 10% is transmural scar and 14% viable border zones with reduced flow capacity. The remaining LV outside border zones of the inferior stress defect (green) has mild diffusely reduced coronary flow capacity (yellow) due to diffuse coronary atherosclerosis but is adequate above ischemic low flow. The referring cardiologist and consulting PET cardiology faculty concluded that medical treatment was a valid option without bypass surgery because of: 1) reduced but adequate coronary flow capacity without ischemia in left anterior descending coronary artery and left circumflex distributions; 2) collaterals to approximately 16% of viable inferior border zones adequate for the patient’s daily activities without angina; 3) absence of angina or heart failure; and 4) normal rest and stress ejection fraction.
For clinical reliability, this comprehensive quantification of complex physiological severity of focal and diffuse CAD is highly reproducible, as shown by serial PETs shown in lateral and inferior views for simplicity in Figure 4. Subjects were studied over the full range, from low rest perfusion or severe stress impairment to high coronary flow capacity in healthy young volunteers with no risk factors (Figure 5). These examples also illustrate that stress perfusion at 13 min was higher than at 8 min and at 15 min was lower than at 8 min (Figure 5).
Test-retest precision in the same patient of rest perfusion
Table 2 lists test-retest precision of serial resting perfusion (in cc/min/g) over minutes and day-to-day variability in the same patient. For same-day rest 1–rest 2 perfusion minutes apart in the same patient, the SD of differences was ±0.093 cc/min/g, and the coefficient of variance (COV) was 10.7% (0.093 of 0.87). For the day-to-different-day difference in rest perfusion in the same patient, the SD of differences was 0.2 and COV was 21.1%, significantly higher by Pitman-Morgan F test (p < 0.001).
Therefore, the 21.1% variability observed on repeat rest perfusion measurements on different days is partly due to 10.7% imprecision of methodology with comparable additional contribution from biological variability. Resting perfusion was previously reported to have a modest direct linear relation to PRP (11,12). Similarly, in this study rest perfusion and rest PRP were linearly related: rest PRP = 3,565 × rest perfusion + 4,196 and R2 = 0.25, indicating that PRP accounted for 25% of variation in rest perfusion.
However, differences between day 1 and day 2 PETs in resting heart rate (1 ± 6 beats/min), systolic blood pressure (3 ± 11 mm Hg), and diastolic blood pressure (0 ± 6 mm Hg) were small. In contrast, some patients may have greater differences in pressure rate product because of anxiety, medications, caffeine, and labile blood pressure altering resting perfusion.
Time course of myocardial perfusion after 4-min dipyridamole infusion
With time 0 for starting the standard 4-min dipyridamole infusion and stress PET 1 as the reference, stress perfusion of PET 2 at 12, 13, or 14 min was significantly higher than PET 1 at 7 or 8 min; was highest at 12 and 13 min; and decreased at 14, 15, and 16 min with no significant difference between perfusion at 8 versus 15 min (Table 3). Stress perfusion at each time interval as a ratio to perfusion at minute 7 or 8 is graphed in Figure 6, which show a peak of 1.10 at minute 12 to 13 falling thereafter to below 1.0 at minute 16.
Stress perfusion changes significantly over time after the 4-min dipyridamole injection, with a global p < 0.001 by the mixed-effects ANOVA model for stress flow, accounting for repeated measurements in the same subjects. Because the global ANOVA test is significant, pairwise comparisons at different time intervals can be compared using the Tukey test for multiple comparisons. Table 4 lists the Tukey test for multiple comparisons of differences among PET 1–PET 2 stress perfusion (cc/min/g) for the row minus the column for minute intervals between stress PET 1 and stress PET 2. For example, the 12-min dipyridamole perfusion is 0.243 cc/min/g higher than the 8-min dipyridamole perfusion (p = 0.0044). The 16-min perfusion is 0.362 cc/min/g lower than the 12-min dipyridamole perfusion (p ≤ 0.02). All bold differences are significant with p ≤ 0.022; all other values (italics or roman) are not significantly different.
Test-retest precision in the same patient of stress perfusion
The subjects with 2 8- to 15-min stress pairs provide the most compelling analysis of pure methodological test-retest precision of stress perfusion. For 15 to 8 min PETs on day 1, the mean difference was 0.02 ± 0.26 cc/min/g with COV of 10.8% (0.259 of 2.393) (Table 5). For 15- to 8-min PETs on day 2, the difference was –0.03 ± 0.23 with COV of 9.6% (0.23 of 2.393).
The day-to-different-day test-retest reproducibility of stress perfusion for day 1–day 2 in the same patient for the 15-min PETs (Table 5) had COV of 19% to 21% that is significantly greater than minute differences between PET 1–PET 2 having COV of 9.6% to 10.6%, a significant difference with Pitman-Morgan p ≤ 0.011. Therefore, daily variability on serial stress-stress perfusion measurements is due to approximately ±10% methodological imprecision with an additional comparable component of biological variability. The rest day 1–day 2 differences (SD of Δ = ±0.10) (Table 2) were smaller than the stress day 1–day 2 difference (SD of Δ = ±0.50) (Table 5) with Pitman-Morgan p < 0.001 but the COVs were similar, 21% for both rest and stress flows. COV for minute and day differences for CFR are similar to those for stress perfusion.
Table 6 lists minute-to-minute precision and day-to-day variability for regional average quadrant values of stress perfusion and CFR that are comparable to global values because both are determined from the primary 1,344 pixel flows averaged for 336 pixels in each quadrant and 1,344 pixels for the entire LV.
Figure 7 illustrates the clinical relevance of maximal versus submaximal stress. Relative stress images show little difference (Figures 7A and 7B). With submaximal stress, global stress perfusion was 1.5 cc/min/g and CFR was 1.8 compared to maximum stress with global stress perfusion of 2.2 cc/min/g and CFR of 2.7.
Submaximal stress lowers stress perfusion and CFR, with a larger percentage of LV having lower flows by pixel colored percentage of LV, which might be interpreted as abnormal for both regional and diffuse CAD (Figures 7C and 7D). Maximal stress increases flows into higher ranges of color-coded perfusion, thereby reducing apparent severity compared to submaximal stress and hence reducing potential false-positive results due to inadequate stress. Therefore, submaximum stress (Figure 7C) erroneously suggests more severe focal and diffuse disease (yellow and green) than true coronary flow capacity with maximal stress (red and orange) (Figure 7D).
In addition to comparable precision for global and regional quadrants, Figure 8 shows mean individual pixel distribution of stress perfusion by comparing histogram distribution of all 1,344 perfusion pixels in LV for all serial PET histograms acquired within minutes. There is no difference in pixel distributions in the 2 histograms by KS statistic (KS = 0.06; p = 0.30).
However, in Figure 9, the mean histogram of 1,344 pixel distribution for all subjects with submaximum stress is significantly different than for all subjects with maximum stress (KS statistic = 0.18; p < 0.0001). Submaximal stress incurs a higher percentage of LV in the middle perfusion ranges (yellow and orange) and a lower percentage of LV in the highest range of perfusion (red) compared to maximal stress with a lower percentage of LV in middle ranges and greatest percent of LV in the highest range of perfusion. Therefore, submaximum stress causing lower stress perfusion (yellow) may be misinterpreted as showing diffuse CAD, small vessel disease, or even ischemia compared to higher perfusion with maximum stress that in some patients may bracket the low flow ischemic threshold as shown in Figure 7.
Test-retest methodological precision of serial myocardial perfusion in the same patient on serial imaging minutes apart without daily biological or intersubject variability is approximately ±10%, accounting for about one-half of the approximately ±20% day-to-day variability due to added biological variability. Our results show that myocardial perfusion is maximal at 8 min after completing the 4-min dipyridamole infusion, averaging 10% higher than perfusion imaging at the standard 4 min after dipyridamole. For some patients with CAD severity near the low flow ischemic threshold, this 10% increased stress flow may substantially change the relative importance of focal and diffuse CAD for integrated physiological severity as illustrated in Figure 7.
Minute-to-minute methodological precision of quantitative myocardial stress perfusion by PET has not been previously reported. Our day-to-day variability of dipyridamole stress perfusion compares with day-to-day variability in our regadenoson study (9).
Our reported low stress flow ischemic threshold of 0.9 cc/min/g is the stress flow that maximizes the area under the receiver-operating characteristic curve (98%), predicting ischemia defined as significant regional stress defect with ECG depression >1 mm or moderate-to-severe angina requiring aminophylline reversal (11,12). This threshold of 0.9 cc/min/g reflects a high probability of ischemia during dipyridamole stress in a large group of patients, including methodology and day-to-day biological variability in the same patient and between different patients (11,12). The methodology plus biological variability of ±20% indicates that for true flow of 0.9 cc/min/g, repeated measurements would give a range of flows between 0.9 ± 20% or 0.72 to 1.08, averaging 0.9 cc/min/g.
Therefore, day-to-day variability of ±20% does not negate this threshold but rather reinforces its validity within objective probability bounds, which has not been previously demonstrated for coronary blood flow. The ±10% due to methodology and the added 10% due to biological variability leave somewhat limited opportunity for further improvement, thereby defining the probability bounds for measuring low perfusion (in cc/min/g) as the ischemic threshold to guide, follow, or risk-stratify CAD management.
Although data are from a single experienced center, our results exemplify a standard for variability of routine quantitative myocardial perfusion imaging by PET to guide or follow CAD management. The total radiation dose for 2 3-PET sequences was 16 mSv, comparable to an average Tc-99m sestamibi rest stress study (16).
Methodological test-retest precision of serial quantitative myocardial perfusion by PET within minutes apart in the same patient is ±10% and days apart is ±20% due to added biological variability, thereby establishing the boundaries of variability for physiological severity to guide or follow CAD management. Therefore, this study provides an essential link between comprehensive quantification of perfusion measurements and their variability that set the bounds for clinical decision-making or risk stratification affected by measurement variability.
COMPETENCY IN MEDICAL KNOWLEDGE: Personalized medical management increasingly depends on complex quantitative measurements integrated with traditional history, examination, blood tests, functional testing, or visually interpreted images. For laboratory testing, the test-retest precision and biological variability are well defined and indeed are criteria for laboratory accreditation. However, for most cardiac imaging, particularly quantitative myocardial perfusion, this systematic, objective, definitive measurement of test-retest methodology precision and day-to-day biological variability are rarely determined because of the time, expense, and complexity involved in doing so, replaced with an emotional bias that “our methods are the best,—are adequate,—are satisfactory for our purposes.” As quantitative myocardial perfusion increasingly guides management of CAD for optimal personalized outcomes, this study sets the standard for variability of quantitative perfusion to guide invasive procedures that may help or harm our patients: ±10% test-retest methodology precision plus ±10% biological day-to-day biological variability for ±20% total variability on repeat measurements on different days for severity thresholds or after changes to guide management of CAD.
TRANSLATIONAL OUTLOOK: For guiding clinical management, quantitative myocardial perfusion imaging needs validation by its integration with management to predict personalized optimal outcomes, which is the greatest benefit with the least harm based on some objective evidence-based, threshold, or risk-to-benefit balance documented for the size and severity of quantitative abnormalities. However, measurement variability profoundly affects correlation with subsequent MACE and, hence, optimal management based on MACE predictions. This study on the “mundane” methodology of quantitative myocardial perfusion is the essential link between comprehensive integration of all myocardial perfusion measurements, their variability, and their bounds for predicting MACE or for clinical decision-making affected by measured variability. It provides the essential technical basis for variability-dependent, evidenced-based prediction of MACE, which requires different future studies.
Dr. Johnson has received internal funding from the Weatherhead PET Center for Preventing and Reversing Atherosclerosis; has received significant institutional research support from St. Jude Medical (for NCT02184117) and Volcano/Philips Corporation (for NCT02328820), makers of intracoronary pressure and flow sensors; and has an institutional licensing and consulting agreement with Boston Scientific for the smart minimum fractional flow reserve (FFR) algorithm. Dr. Gould has received internal funding from the Weatherhead PET Center for Preventing and Reversing Atherosclerosis; and is the 510(k) applicant for CFR Quant (K113754) and HeartSee (K143664), software packages for cardiac positron emission tomography image processing and analysis, including absolute flow quantification. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- analysis of variance
- coronary artery disease
- coronary flow reserve
- coefficient of variance
- left ventricle
- positron emission tomography
- pressure rate product
- right coronary artery
- Received April 26, 2016.
- Revision received September 6, 2016.
- Accepted September 8, 2016.
- 2017 American College of Cardiology Foundation
- Taqueti V.R.,
- Hachamovitch R.,
- Murthy V.L.,
- et al.
- Gould K.L.,
- Johnson N.P.,
- Kaul S.,
- et al.
- Gould K.L.,
- Johnson N.P.,
- Bateman T.M.,
- et al.
- De Bruyne B.,
- Baudhuin T.,
- Melin J.A.,
- et al.
- Gould K.L.,
- Goldstein R.A.,
- Mullani N.A.,
- et al.
- Johnson N.P.,
- Gould K.L.
- Gould K.L.,
- Pan T.,
- Loghin C.,
- et al.
- Yoshida K.,
- Mullani N.,
- Gould K.L.
- Vasquez A.F.,
- Johnson N.P.,
- Gould K.L.
- Jerome S.D.,
- Tilkemeier P.L.,
- Farrell M.B.,
- Shaw L.J.