Comparison of ejection fraction calculation between CT and SPECT at high heart rate: A dynamic cardiac phantom study

Abstract

The purpose of this study is to compare the ejection fraction (EF) calculation of CT and SPECT at high heart rate. A dynamic cardiac phantom with programmable end-systolic volume (ESV), end-diastolic volume (EDV), and heart rate was used to compare CT, which has high spatial resolution (< 1 mm) and modest temporal resolution of 175 msec, and SPECT, which has high temporal resolution of 16 bins per cardiac cycle but poor spatial resolution (> 1 cm) in EF, ESV, and EDV at the heart rates ≤ 100 bpm for EF = 30 (disease state) and EF = 60 (healthy state). EF calculations for SPECT were accurate in 2% for 40 to 100 bpm for both EF = 30 and EF = 60, and were not heart rate dependent although both ESV and EDV could be underestimated by 18–20%. EF calculations for CT were accurate in 2.2% for 40 and 60 bpm. Inaccuracy in EF calculations, ESV and EDV estimates increased when the heart rate or EF increased. SPECT was accurate for EF calculation for the heart rates ≤ 100 bpm and CT was accurate for the heart rates of ≤ 60 bpm. CT was less accurate for the high heart rates of 80 and 100 bpm, or high EF = 60.

INTRODUCTION

Myocardial perfusion imaging (MPI) using single-photon emission computed tomography (SPECT) has been shown to have good prognostic value for diagnosing coronary artery disease.¹ Electrocardiogram-gated MPI can further reduce the number of so-called borderline cases when interpreted simultaneously with the stress and rest perfusion images.² In addition, post-stress left ventricular ejection fraction (EF), end-diastolic volume (EDV), and end-systolic volume (ESV) have incremental prognostic value over perfusion information in predicting cardiac death.³ There is a growing interest in hybrid SPECT/CT imaging in which SPECT is used for functional perfusion imaging of the myocardium and CT for anatomical coronary artery imaging and calcium scores. Both modalities of SPECT and CT can also provide EF calculation.

Comparisons of EF calculation between modalities were typically performed on the same patient population without the ground truth of EF, ESV and EDV and underlying heart rate.^4–6 Correlation rather than accuracy between modalities was normally the outcome of EF comparison. In a meta-analysis of EF calculation accuracy,

Asferg et al concluded that CT of 64-slices or higher performed on the patient of heart rate < 70 bpm can provide accurate EF calculation compared to cardiac MRI and transthoracic echocardiography.⁴ They also found that CT represents a valid technique for the combined evaluation of EF and coronary artery disease.⁴ Demir et al compared gated SPECT, echocardiography (ECHO), and cardiac magnetic resonance imaging (MRI) on a group of 21 patients and demonstrated a good correlation among the three modalities.⁵ Pelletier-Galarneau et al compared two software implementations of EF calculation on 60 patients with IQ-SPECT, planar radionuclide angiography and cardiac MRI and used the best correlation as a guide for the performance of EF calculation software.⁶ It was also demonstrated in a study using a dynamic cardiac phantom⁷ that EDV and ESV can be significantly underestimated in SPECT MPI even though EF can be estimated accurately. This study showed that unlike conventional cardiac SPECT with low-energy high-resolution (LEHR) collimators, the IQ-SPECT, a cardiac imaging software and hardware option of variable-focus collimators (Siemens Healthineers, Erlangen, Germany), could significantly underestimate EDV and ESV. Dynamic cardiac phantom has also been used in the evaluation of detection of transmural and sub-endocardial defects to demonstrate the improvements of detection by attenuation correction and cardiac gating,⁸ and by correction of the respiratory motion.⁹

Although the accuracy of EF calculation has been established for both CT and SPECT,^4,7 and CT was even used as a gold standard to verify the EF calculation for SPECT for the heart rates of ≤ 70,⁷ it was not clear if CT and SPECT can both be used for EF calculation at the heart rate of > 70 bpm because most cardiac CT scans require the heart rate to be stable and ≤ 70 bpm as indicated in the ACR-NASCI-SPR practice parameter for the performance and interpretation of cardiac CT.¹⁰

In this study, we used a dynamic cardiac phantom with programmable ESV, EDV, and heart rate to compare cardiac CT with cardiac SPECT on EF calculation, ESV and EDV estimates over the heart rates of ≤ 100 bpm. Cardiac CT has high spatial resolution (< 1 mm) and modest temporal resolution of 175 msec and is designed for imaging patients with heart rate ≤ 70 bpm, while cardiac SPECT has high temporal resolution of 16 bins per cardiac cycle or up to 37.5 msec for 100 bpm but poor spatial resolution (> 1 cm).

METHODS AND MATERIALS

A Dynamic Cardiac Phantom (Data Spectrum Corporation, Durham, North Carolina, USA) (Figure 1) was programmed to pump at EF = 30 or EF = 60. The EF was calculated from the measured EDV and ESV using the following equation:

$$\text{EF}=\frac{(\text{EDV}-\text{ESV})}{(\text{EDV})}\times 100\text{\%}$$

Figure 1.

A dynamic cardiac phantom programmable for end-systolic volume, end-diastolic volume and heart rate.

For EF = 30, we set EDV = 130 cc and ESV = 91 cc, and for EF = 60 we set EDV = 130 cc and ESV = 52 cc. The heart rates were 40, 60, 80, and 100 bpm for SPECT and 45, 60, 80, and 100 bpm for CT. Due to a hardware limitation in CT, the lowest heart rate was 45 bpm. The stroke volume of EF = 60 was 78 cc, two times the stroke volume of 39 cc for EF = 30. As a result, the motion was larger for EF = 60 than EF = 30 because a larger volume of the “blood” was pumped out of the heart per cardiac cycle. The selection of ESV = 52 cc, EDV = 130 cc, and EF = 60 was based on a measurement by the cardiovascular magnetic resonance imaging of an average person of 70 kg.¹¹

The myocardium walls of the cardiac phantom were simulated by two layers of silicone membrane. The volume enclosed by the inner layer of the membrane can be enlarged from the systole to the diastole and reduced from the diastole to the systole by the two piston positions of a linear motor in the cardiac cycle. The difference between ESV and EDV is the stroke volume. Both the inner and outer membranes were expanded or contracted simultaneously. A myocardial defect (simulating a myocardial infarction caused by occlusion of the proximal left anterior descending coronary artery) was simulated in the anterior wall by the relative positions of the outer and inner membranes of the cardiac phantom. The myocardium was simulated by the water solution mixed with both Tc-99m for cardiac SPECT imaging and iodinated contrast for cardiac CT imaging through an injection port to the myocardium.

The parameters for clinical SPECT (Siemens Symbia E) imaging were low-energy high-resolution (LEHR) collimator, 64 angles, 128 × 128 matrix, zoom factor of 1.23, 16 gates, and auto body contouring acquisition. The reconstruction parameters of SPECT were filtered-back-projection with the Butterworth filter of order 5 and cutoff frequency of 0.57. The parameters for CT were 0.35 sec gantry rotation, 10 phases, and 2.5 mm slice thickness, filtered-back-projection of 240 degrees or half-scan projection data, and ‘standard’ reconstruction kernel. Each scan was repeated twice in SPECT and three times in CT. Cedars quantitative gated SPECT (QGS) and MIM software were used for the SPECT and CT EF calculations, respectively. Figure 2 shows an example of ESV and EDV detected by QGS and MIM.

Figure 2.

The contours for ESV and EDV were automatically determined by QGS for SPECT (A) and manually determined with Hounsfield unit thresholding by MIM for CT (B).

RESULTS

For SPECT, the EF calculations for the heart rates of 40, 60, 80, and 100 bpm were larger in average by 1.2% ± 2.0% and 1.0% ± 0.7% for EF = 30 and EF = 60, respectively, (see Figure 3A); and the ESV (Figure 3C) and EDV (Figure 3E) estimates were smaller by 19.0% ± 4.7% and 17.5% ± 5.2%, respectively for EF = 30, and smaller by 20.0% ± 3.7% and 18.5% ± 3.3%, respectively for EF = 60. There was no heart rate dependence for the EF calculations in SPECT. The EF calculations were accurate in 2% across the heart rates, although both ESV and EDV were underestimated by about 18–20%.

Figure 3.

The EF calculations of SPECT and CT are in (A) and (B), respectively, where the EF of 30 and 60 are separated by a solid vertical line in the middle. The horizontal axis is the heart rate: 40 to 100 bpm for SPECT and 45 to 100 bpm for CT. The end-systolic volume estimates of SPECT and CT are in (C) and (D), respectively; and the end-diastolic volume estimates of SPECT and CT are in (E) and (F), respectively.

For CT, the EF calculations for EF = 30 were smaller by 1.1% ± 0.4%, 1.7% ± 0.5%, 2.8% ± 0.2%, and 4.4% ± 0.7% for 45, 60, 80, and 100 bpm, respectively; and the EF calculations for EF = 60 were smaller by 1.8% ± 0.6%, 2.2% ± 0.8%, 4.6% ± 0.3%, and 7.4% ± 1.8% for 45, 60, 80, and 100 bpm, respectively. The error in EF calculation increased from EF = 30 to EF = 60, and from low to high heart rates. The errors in EF calculation were accurate in 2.2% for the heart rates of 40 and 60 bpm for both EF = 30 and EF = 60. For EF = 30, the ESV estimates for the heart rates of 45, 60, 80 and 100 were larger by 1.1% ± 0.6%, and the EDV estimates were smaller by 2.4% ± 0.4%, respectively. For EF = 60, the ESV estimates were larger by 4.0% ± 1.6%, and the EDV estimates were smaller by 5.3% ± 0.9%. The difference between the measurement and the truth in both ESV and EDV estimates increased with heart rate in CT (see Figures 3D and F).

A univariate linear regression with heart rate as the predictor and the response variable of EF, ESV, or EDV was performed for the estimated linear regression coefficient for the effect of heart rate, i.e., the expected change in the mean of the response variable for every one unit increase in the heart rate. The results are in Table 1. The associations for CT between EF, ESV, or EDV and heart rate were all significant, while the associations for SPECT were all non-significant (not shown). This suggested that in CT there was a decreasing trend of EF calculation or EDV estimate when the heart rate increased, and that there was an increasing trend of ESV estimate when the heart rate increased.

Table 1.

The estimated linear regression coefficient for the effect of heart rate as the predictor

SPECT/CT	Response variable	Estimated coefficient	P-value
CT	EF (EF = 30)	− 0.060	2.6e–06***
CT	EF (EF = 60)	− 0.105	3.7e–05***
CT	ESV (EF = 30)	0.038	0.0014**
CT	ESV (EF = 60)	0.071	0.0005***
CT	EDV (EF = 30)	− 0.053	3.4e–05***
CT	EDV (EF = 60)	− 0.130	9.1e–06***

The P-value and stars indicate the significance of this linear association, where

^***P-values< 0.001

^**P-values between 0.001 and 0.01

^*P-values between 0.01 and 0.05. The associations for CT were all significant, while the associations for SPECT were all non-significant

DISCUSSIONS AND CONCLUSION

In this study, we used a dynamic cardiac phantom with programmable ESV, EDV, and heart rate to compare (1) cardiac CT, which has high spatial resolution (< 1 mm) and modest temporal resolution of 175 msec, and (2) cardiac SPECT, which has high temporal resolution of 16 bins per cardiac cycle or up to 37.5 msec for 100 bpm but poor spatial resolution (> 1 cm) in EF calculation, ESV and EDV estimates of the heart rates ≤ 100 bpm for EF = 30 and EF = 60.

The EF calculations for SPECT were accurate in 2% across the heart rates of 40 to 100 bpm for both EF = 30 and EF = 60, and were not heart rate dependent although both ESV and EDV could be underestimated by 18–20%. The poor spatial resolution of SPECT did not impact the EF calculation but did impact the individual estimates of EDV and ESV. Our results were consistent with the results by Hippelainen et al⁷ who also used a dynamic phantom to demonstrate the accuracy of SPECT EF calculations to within 1% at the EF of 45 to 70 for the heart rates of 40 to 70 bpm. However, our results were not consistent with the results by Bailliez et al¹² who used the Amsterdam gated dynamic phantom at one constant heart rate to compare the EF, EDV and ESV for the GE Discovery NM 530c, Biosensors International D-SPECT and Siemens IQ-SPECT. In their findings, the EF calculations were 39, 42, and 48 for EF = 33.3 and were 66, 71, and 73 for EF = 60 for NM 530c, D-SPECT, and IQ-SPECT, respectively.

The EF calculations for CT were accurate in 2.2% for the heart rates of 40 and 60 bpm. The errors increased when the heart rate increased or EF increased from 30 to 60. The errors also increased for the estimates of ESV and EDV when the heart rate increased or EF increased from 30 to 60. CT has high spatial resolution, which helps to identify the boundaries for ESV and EDV. However, the temporal resolution of 175 msec of CT was not sufficient to freeze the heart motion in particular for the high heart rates of 80 and 100 bpm and high EF of 60, leading to a larger error at a higher heart rate or a higher EF.

The most advanced CT, which has the temporal resolution of 75 msec, and which is a dual-source dual-detector CT,^7,13,14 could improve the results of CT for high heart rate and high EF. However, this type of CT scanner is not available on hybrid SPECT/CT scanners. The 64-slice CT scanner with the temporal resolution of 175 msec in this study is capable of coronary artery CT imaging at the diastolic phase, which is the quiescent period of the cardiac cycle and is the best CT available for SPECT/CT.

The limitation of the study was that iodinated contrast was injected in the myocardium instead of the ventricle of the phantom. This caused negative contrast rather than positive contrast between the ventricle and the myocardium. This difference in contrast should not impact our study of EF because it was the volume determined by the boundary between the ventricle and the myocardium used in the EF calculation, EDV and ESV estimates.