Annals of Nuclear Medicine

, Volume 25, Issue 10, pp 768–776

Novel algorithm for quantitative assessment of left ventricular dyssynchrony with ECG-gated myocardial perfusion SPECT: useful technique for management of cardiac resynchronization therapy

Authors

    • Department of Radiology and Nuclear MedicineNational Cerebral and Cardiovascular Center
  • Akira Imoto
    • Department of Radiology and Nuclear MedicineNational Cerebral and Cardiovascular Center
  • Yoshihiro Nishimura
    • Department of Radiology and Nuclear MedicineNational Cerebral and Cardiovascular Center
  • Hideaki Kanzaki
    • Department of Cardiovascular MedicineNational Cerebral and Cardiovascular Center
  • Takashi Noda
    • Department of Cardiovascular MedicineNational Cerebral and Cardiovascular Center
  • Shiro Kamakura
    • Department of Cardiovascular MedicineNational Cerebral and Cardiovascular Center
  • Yoshio Ishida
    • Department of Radiology and Nuclear MedicineNational Cerebral and Cardiovascular Center
    • Department of Nuclear MedicineKansai Rosai Hospital
Original article

DOI: 10.1007/s12149-011-0525-8

Cite this article as:
Kiso, K., Imoto, A., Nishimura, Y. et al. Ann Nucl Med (2011) 25: 768. doi:10.1007/s12149-011-0525-8

Abstract

Objectives

Cardiac resynchronization therapy (CRT) is the established treatment for patients with chronic and severe heart failure, and it has been reported that the presence of left ventricular (LV) dyssynchrony is one of the most important factors which predict positive response of this therapy. In the present study, we developed new software algorithm for quantitative assessment of LV dyssynchrony from ECG-gated myocardial perfusion SPECT (GMPS), and evaluated its utility for the management of CRT.

Methods

Thirty-three patients with chronic severe heart failure were studied. GMPS was performed with 16 frame per-cardiac-cycle before and 6 months after CRT and LV end-diastolic volume, end-systolic volume (LVESV), ejection fraction (LVEF) were calculated by QGS software. We generated the time–activity curve per-cardiac-cycle in 4 myocardial segments by Fourier transform curve-fitting of the 16 serial count values, and measured the time from R-wave to the maximum-count point [time to end-systole (TES)] in each. For the evaluation of the degree of LV dyssynchrony, we used the maximum difference in TES (ΔTES) among the 4 segments which corrected for R–R time as dyssynchrony index (DI). Moreover, DI at baseline evaluated by GMPS was compared with the echocardiographic index of LV dyssynchrony; maximum difference of time to peak velocity (ΔTPV) evaluated by tissue Doppler imaging (TDI).

Results

DI before CRT showed a significant correlation with the LV function, such as LVEF, LVESV (DI vs. LVEF; r = 0.57, P < 0.0001. DI vs. LVESV; r = 0.64, P < 0.0001). The study subjects were divided into 2 groups, responder group (R-Gp) with LVEF increase >10% or LVESV decrease >10% and non-responder group (NR-Gp). DI before CRT was significantly larger in R-Gp than in NR-Gp (25.9 ± 22.2 vs. 10.8 ± 8.9, P = 0.01). In R-Gp, DI showed a significant decrease after CRT (25.9 ± 22.2 → 13.6 ± 10.9, P < 0.05). DI at baseline measured by GMPS correlated significantly with ΔTPV at baseline measured by TDI (r = 0.38, P < 0.05).

Conclusions

This new algorithm for the estimation of LV dyssynchrony might be comparable to TDI, and contributes to the prediction and the evaluation for the response of CRT.

Keywords

Cardiac resynchronization therapy (CRT)Left ventricular dyssynchronyECG-gated myocardial perfusion SPECT

Introduction

Cardiac resynchronization therapy (CRT) has been utilized for patients with advanced heart failure, and numerical studies have reported that it improved the patients’ prognosis and quality of life [14]. Although CRT is commonly applied for the selected patients according to the ACC/AHA heart failure guidelines [5, 6], 20–30% of the patients show poor functional improvement [1, 7]. Therefore, the prediction of therapeutic improvement of LV function is essential for the decision of CRT indication. Prolonged QRS duration in ECG, which is one of the major criteria of CRT indication, may be the parameter related to LV dyssynchrony, but it is an indirect estimation. Recently, echocardiographic methods, such as M-mode, pulse Doppler imaging, and tissue Doppler imaging (TDI), are utilized for the direct and quantitative assessment of LV dyssynchrony. TDI especially provides precise assessment of the difference in timing of regional myocardial contraction by measuring regional time to peak velocity, and it is now used as one of the standard methods for quantitative analysis of LV dyssynchrony [8, 9].

In radionuclide imaging techniques, ECG-gated myocardial perfusion SPECT (GMPS), which enables to examine LV myocardial perfusion and function simultaneously, will also contribute to the quantitative assessment of LV dyssynchrony. Because GMPS data contain the information about regional wall thickening in a cardiac cycle, which is estimated by the change of regional wall counts based on the partial volume effect [1013] of SPECT system. Actually, some studies with phase analysis technique, which enables the assessment of LV dyssynchrony by change of regional wall counts, have already revealed the importance of LV dyssynchrony for the prediction of response of CRT [1416]. However, the software of phase analysis is packaged to the latest workstations only; the cost of this analysis is so expensive that the availability of this technique is extremely limited, compared with echocardiographic analysis.

In this study, we developed a new software program, which is applicable to the ordinary personal computers, to quantify the degree of LV dyssynchrony using GMPS data, and evaluated its utility for the management of CRT. We additionally compared the assessment of LV dyssynchrony by GMPS with that by TDI. This study is confirmed to the 2005 version of the Ethical Guideline for Clinical Study (Ministry of Health, Labor and Welfare of Japan).

Materials

Thirty-three patients (25 males and 8 females, 58 ± 16 years) with chronic heart failure, who were scheduled for implantation of CRT device, were studied. They consisted of 21 with dilated cardiomyopathy (DCM), 2 with dilated-phase hypertrophic cardiomyopathy (d-HCM), 2 with hypertensive heart disease (HHD), 2 with ischemic cardiomyopathy (ICM), 3 with cardiac sarcoidosis, 1 with cardiac amyloidosis, and 2 with valvular heart disease. Criteria of the patients’ selection for CRT were accordant with ACC/AHA guideline [5, 6]. The baseline characteristics are summarized in Table 1. Those study subjects had severe heart failure (NYHA functional class: II in 13 patients, III in 19 patients and IV in 1 patients), prolonged QRS duration (160 ± 30 ms), and severe LV dysfunction (LVEF: 26 ± 12%, LVEDV: 290 ± 155 ml, LVESV: 229 ± 155 ml).
Table 1

Clinical characteristics of study population (n = 33)

Clinical characteristics

Value

Age (years)

58 ± 16

Gender (male/female)

25/8

NYHA functional class

2.7 ± 0.5

QRS duration (ms)

160 ± 30

Etiology (n)

 Ischemic

2

 Idiopathica

23

 Otherb

8

LVEF (%)

25.3 ± 12.0

LVESV (ml)

229.1 ± 151.1

LVEDV (ml)

289.9 ± 155.4

Data are expressed as mean ± SD

LVEF left ventricular ejection fraction, LVESV left ventricular end-systolic volume, LVEDV left ventricular end-diastolic volume

a2 with dilated cardiomyopathy and 2 with dilated phase hypertrophic cardiomyopathy

b2 with hypertensive heart disease, 3 with cardiac sarcoidosis, 1 with cardiac amyloidosis, and 2 with valvular heart disease

Methods

ECG-gated myocardial perfusion SPECT

GMPS with technetium-99m (99mTc) sestamibi was performed at rest before (2.2 ± 1.8 months) and 6 months (6.6 ± 1.3 months) after CRT. For all patients, SPECT image acquisition was performed in a supine position 1 h after intravenous injection of 600 MBq of 99mTc-sestamibi. To remove tracer retention in liver and gallbladder, the patients took light meal or milk before image acquisition. Thirty projection images were obtained over 180° in 6° increments with 50 beats per view, using a dual-headed SPECT system (VERTEX; ADAC laboratories, Milpitas, CA) equipped with a low-energy general-purpose collimator. An ECG R-wave detector provided a gate to acquire sixteen frames per cardiac cycle. The image resolution in the transaxial plane was 16 mm full-width at half-maximum. Data was stored in 64 × 64 matrix. The energy discrimination was centered on 141 keV with a 20% window. To generate transaxial tomograms from the gated projection data and reconstruct oblique angle tomograms, a ramp filter and a Butterworth filter (order 8, cutoff 0.27 cycle/pixel) were used. Acquired image data were transferred to PEGASYS workstation (ADAC Laboratories), and we measured LV volumes [LV end-systolic volume (LVESV), LV end-diastolic volume (LVESV)] and LV ejection fraction (LVEF) using QGS software developed by Germano [17, 18], and also measured the size of perfusion defects: %defect size (%DS), which were calculated from the area of perfusion abnormality lower than −2SD compared with normal files, using QPS software.

Assessment of LV dyssynchrony based on GMPS Data (Fig. 1)

From acquired GMPS data which were transferred to PEGASYS workstation, we generated a polar map displaying myocardial pixel counts distribution as %peak activity for each ECG-gated frame (16 frames per cardiac cycle). Then, we measured “the mean value of %peak activity of 4 myocardial segments (anterior, lateral, inferior, and septum)” and “the maximum counts” of the polar map in each gated frame with QPS software (Fig. 1a). Those data were transferred to normal personal computer (Endeavor MT8800, EPSON Co. operating system: Windows XP professional), then we calculated “the segmental mean value of absolute counts” of each gated frame by the following formula: “segmental mean value of %peak activity” × “the maximum counts”. Consequently, the segmental time–activity curves were generated by plotting the calibrated absolute counts of 16 gated frames, and were applied with the second-harmonic Fourier transform curve-fitting. In each transformed time–activity curve, maximum count point was determined as the end-systolic (ES) point and then the interval from R-wave to the ES point was measured as time-to ES (TES) (Fig. 1b). Finally, the degree of LV dyssynchrony among the 4 myocardial segments was estimated using the following formula as a dyssynchrony index (DI): (Max. TES−Min. TES)/R–R interval time × 100. An example of calculation of DI is shown in Fig. 1c.
https://static-content.springer.com/image/art%3A10.1007%2Fs12149-011-0525-8/MediaObjects/12149_2011_525_Fig1_HTML.gif
Fig. 1

Measurement of LV regional time-wall thickening curves from ECG-gated myocardial perfusion SPECT (GMPS) and calculation of LV dyssynchrony index (DI). a Polar maps displaying regional radioactivity in 16 gated frames per a cardiac cycle. b Segmental time–activity curves with the Fourier transform curve-fitting [black absolute count curve, red Fourier fitting curve, blue arrow time to end-systole (TES)]. c Calculation of DI (Ant. anterior segment, Lat. lateral segment, Inf. inferior segment, Sept. septal segment)

Tissue Doppler imaging

Two-dimensional (2D) TDI was performed in the left decubitus position with the echocardiography system (Vingmed Vivid Seven, GE-Vingmed) at almost the same timing and those images were stored for off-line analysis (Echopac, GE-Vingmed). For the analysis of myocardial tissue velocity curves, the sample volume was placed in the LV basal GMPS before the implantation of CRT. The images were acquired from the apical four-, three-, two-chamber views, portions of anterior, inferior, septal, and lateral wall, using apical long axis view, the 2- and 4-chamber view images (total 12 sample points). The maximum difference between time to peak systolic velocities (ΔTPV) among the four portions was used as an index of LV dyssynchrony.

Statistical analysis

Statistical analysis was performed with StatView statistical package (SAS Institute, Cary, NC). All data ware expressed as the mean ± standard deviation. Patients’ data were compared using the paired or unpaired Student’s t test when appropriate. For the evaluation of correlations, Pearson’s correlation analysis was performed. For all tests, a probability value of 0.05 or less was considered significant.

Results

Patients’ classification

The study patients were classified into two groups based on LV functional response to CRT, 18 responders and 15 non-responders. The responders were defined as those with improvement of LVEF >10% and/or percent reduction of LVESV >10% during the 6-month follow-up. In responders, LVEF changed from 22.2 ± 12.2 to 30.0 ± 13.3% and LVESV from 278.8 ± 164.6 to 179.5 ± 135.0 ml. In non-responders, LVEF changed from 30.1 ± 10.7 to 28.2 ± 11.2% and LVESV from 169.0 ± 123.1 to 178.1 ± 118.1 ml.

Pre-CRT estimation

In examinations just before CRT (Table 2), there were no significant differences in age, gender, NYHA functional class between responders and non-responders. %DS tended to be higher in non-responders, however, significant difference was not observed between the two groups (16.3 ± 14.5 vs. 21.1 ± 15.3, P = 0.37). In terms of LV functions, LVEF was lower and LVESV was larger in responders than in non-responders. DI which we determined from GMPS data in this study showed a significantly higher value in responders than in non-responders (25.9 ± 22.2 vs. 10.8 ± 8.9, P < 0.01) (Fig. 2). When we studied for all the subjects, DI correlated inversely with LVEF and directly with LVESV (r = −0.57, P < 0.0001 and r = 0.64, P < 0.0001, respectively, Fig. 3).
Table 2

Baseline characteristics of responders (n = 18) and non-responders (n = 15)

Baseline characteristics

Responders

Non-responders

P value

Age (years)

55 ± 18

61 ± 14

NS

Gender (male/female)

16/2

9/6

NS

NYHA

2.8 ± 0.5

2.5 ± 0.5

NS

%DS

16.3 ± 14.5

21.1 ± 15.3

NS

LVEF (%)

22.2 ± 12.2

30.1 ± 10.7

<0.05

LVESV (ml)

278.8 ± 164.6

169.0 ± 123.1

<0.05

LVEDV (ml)

342.0 ± 164.3

228.9 ± 135.0

<0.05

DI

25.9 ± 22.2

10.8 ± 8.9

=0.01

LVEF left ventricular ejection fraction, LVESV left ventricular end-systolic volume, LVEDV left ventricular end-diastolic volume, DI dyssynchrony index (derived from ECG-gated myocardial perfusion SPECT)

https://static-content.springer.com/image/art%3A10.1007%2Fs12149-011-0525-8/MediaObjects/12149_2011_525_Fig2_HTML.gif
Fig. 2

Comparison of dyssynchrony index (DI) at baseline between responders (white bar) and non-responders (gray bar). Pre-treatment DI was significantly higher in responders than in non-responders (P = 0.01)

https://static-content.springer.com/image/art%3A10.1007%2Fs12149-011-0525-8/MediaObjects/12149_2011_525_Fig3_HTML.gif
Fig. 3

Relationships between DI and LVEF (a) and between DI and LVESV (b) at baseline. There was a significant correlation between pre-treatment DI and the severity of LV dysfunction. DI dyssynchrony index, LVEF left ventricular ejection fraction, LVESV left ventricular end systolic volume

Furthermore, multiple logistic regression analysis revealed that DI was the only significant parameter for the prediction of responders to CRT (Table 3).
Table 3

Multiple logistic regression analysis

 

Odds ratio

95% CI

P

Age

0.983

0.921–1.048

NS

Gender

7.358

0.649–83.75

NS

LVEDVa

1.042

0.980–1.108

NS

LVESVa

0.952

0.884–1.026

NS

LVEFa

0.923

0.721–1.08

NS

DIa

1.120

1.004–1.251

<0.05

Multiple logistic regression analysis for discriminating the predictor of CRT responders

LVEF left ventricular ejection fraction, LVESV left ventricular end-systolic volume, LVEDV left ventricular end-diastolic volume, DI dyssynchrony index (derived from ECG-gated myocardial perfusion

aAt baseline

Post-CRT estimation

Six months after the implantation of CRT, DI significantly decreased from 25.9 ± 22.2 to 13.6 ± 10.9 in responders while it did not show a significant change in non-responders (10.8 ± 8.9 to 15.1 ± 13.9) (Fig. 4). Interestingly, the reduction of DI correlated significantly with the reduction of LVESV (r = 0.47, P = 0.0055), and tended to correlate with the improvement of LVEF (Fig. 5).
https://static-content.springer.com/image/art%3A10.1007%2Fs12149-011-0525-8/MediaObjects/12149_2011_525_Fig4_HTML.gif
Fig. 4

Changes in DI from baseline to 6 months after CRT in responders (left) and non-responders (right). DI showed a significant reduction by CRT in responders, but not in non-responders. DI dyssynchrony index, 6-Mo f/u 6 months follow-up

https://static-content.springer.com/image/art%3A10.1007%2Fs12149-011-0525-8/MediaObjects/12149_2011_525_Fig5_HTML.gif
Fig. 5

Relationships between the change in LVEF (ΔEF) and that in DI (ΔDI) (left), and between the change in LVESV (ΔLVESV) and that in DI (ΔDI) (right), from baseline to 6 months after CRT. The decrease in DI tended to correlate with the improvement of LV function in CRT. DI dyssynchrony index, LVEF left ventricular ejection fraction, LVESV left ventricular end systolic volume, 6-Mo f/u 6 months follow-up

Figure 6 shows representative data in responders. In this patient, LV function estimated by QGS software demonstrated marked improvement of LVEF and reduction in LVEDV and LVESV after CRT. In association, regional TES became homogeneous after CRT and DI decreased from 30.64 to 2.97, indicating the synchronizing effect of CRT. On the contrary, Fig. 7 shows the data of non-responder without any improvements of LV function after CRT. The baseline TES in 4 regions were almost uniform, and the change of DI was not observed.
https://static-content.springer.com/image/art%3A10.1007%2Fs12149-011-0525-8/MediaObjects/12149_2011_525_Fig6_HTML.gif
Fig. 6

Representative image data in responder group. a Changes in global LV function measured by QGS software programs from baseline to 6 months after CRT. b Changes in segmental time–activity curves and DI from baseline to 6 months after CRT. DI dyssynchrony index, EF ejection fraction, ESV end systolic volume, EDV end diastolic volume, 6-Mo f/u 6 months follow-up

https://static-content.springer.com/image/art%3A10.1007%2Fs12149-011-0525-8/MediaObjects/12149_2011_525_Fig7_HTML.gif
Fig. 7

Representative image data in non-responder group. a Changes in global LV function measured by QGS software programs from baseline to 6 months after CRT. b Changes in segmental time–activity curves and DI from baseline to 6 months after CRT. DI dyssynchrony index, EF ejection fraction, ESV end systolic volume, EDV end diastolic volume, 6-Mo f/u 6 months follow-up

Comparison to TDI parameter

When DI was compared to the echocardiographic LV dyssynchrony index, ΔTPV derived from TDI, Pearson’s correlation demonstrated a fair but significant correlation between them (r = 0.38, P < 0.05) (Fig. 8).
https://static-content.springer.com/image/art%3A10.1007%2Fs12149-011-0525-8/MediaObjects/12149_2011_525_Fig8_HTML.gif
Fig. 8

Relationships between the change in TPV by TDI (ΔTPV) and that in DI (ΔDI) from baseline to 6 months after CRT. TPV time to peak systolic velocities, TDI tissue Doppler imaging, DI dyssynchrony index

Discussion

We developed a new software program to quantify the degree of LV dyssynchrony from the GMPS data, and evaluated its utility for the prediction and evaluation of therapeutic effects of CRT in this study. The results revealed that LV dyssynchrony index analyzed with this new software program contributes to the prediction of patients with functional improvement by CRT, and to the evaluation of improvement of LV dyssynchrony by CRT. Interestingly, patients with worse LV functions showed higher DI in baseline. Moreover, the degree of the DI improvement by CRT tended to correlate with that of the LV functional improvement. The result that DI correlated with the conventionally used echocardiographic index, ΔTPV derived from TDI, supported the capability of DI. Furthermore, we found that DI is the best predictor of CRT responders by multiple logistic regression analysis in comparison with other factors such as age, gender, NYHA functional class, %DS, and LV functions (Fig. 8).

CRT has been proved to be effective for improving LV function for patients with chronic severe heart failure in several large-scale clinical trials [14], but it has been noted that up to 30% of the patients receiving CRT show poor response [1, 7]. Since CRT is an invasive and expensive therapy, the prediction of therapeutic outcome should be cautiously performed before therapy. Regarding the selection of candidates for CRT, Bax et al. [8] reported that LV dyssynchrony investigated by TDI is the most useful parameter which predicts the effectiveness of CRT, other than the conventional indication criteria such as prolonged QRS duration, etc. However, the echocardiographic analysis has limitations in terms of reproducibility and inter-observer variability in measurement. Chung et al. [20] revealed in PROSPECT trial that no single echocardiographic measure of LV dyssynchrony improve the patient selection for CRT beyond the conventional guidelines and speculated that it would be due to poor reproducibility and inter-observer variability in the echocardiographic measurement.

On the other hand, our newly developed software program using GMPS data is based on the automatically performed QGS software developed by Germano et al. [17, 18], and therefore it may provide better reproducibility and less inter-observer variability than the echocardiographic measurement. For that matter, it will promise as a mean for the objective investigation of LV dyssynchrony. Moreover, it is highly contributable in making the assessment of LV dyssynchrony in combination with that of myocardial viability. Since it has been previously reported that ischemic heart failure patients with large nonviable segments show poor response for CRT [19], myocardial viability is also recognized as the important factor which influence the response of CRT. For that matter, the GMPS method, which provides simultaneous assessment of LV dyssynchrony and myocardial viability, has a value as one-stop shop for the management of CRT. However, in contrast to the previous reports, %DS calculated from the area of perfusion abnormality lower than −2SD compared with normal files did not show a significant difference between responders and non-responders in this study. We suppose that it may be attributed to a small number of patients with large defect size in our study.

In radionuclide techniques, phase analysis with GMPS was firstly developed by Emory university group [21, 22]. Henneman et al. [15] reported the importance of LV dyssynchrony evaluated by phase analysis, which was comparable to TDI, for the prediction of CRT response. Moreover, they also investigated the predictive cut-off value with ROC analysis [16]. As I have described, phase analysis is one of the useful tools for the estimation of LV dyssynchrony, however, this software has been bundled with latest workstations only. So, thinking of the cost and availability, there is a big disadvantage in this phase analysis. On the contrary, our new software program is installable and available to the ordinary personal computers. Therefore, this software is superior to phase analysis technique in terms of cost and availability.

Yamamoto et al. has also developed “CardioGRAF” software for the assessment of LV dyssynchrony. The algorithm of this software, which is different from that of our method and phase analysis, is based on the changes of volumes of regional LV cavity during a cardiac cycle [23, 24]. Furthermore, the clinical usefulness of this software for the evaluation of CRT response has been also reported [25, 26]. However, this software is applicable to the GMPS data obtained by p-FAST (Perfusion-Function Assessment for myocardial SPECT) analysis only. On the contrary, our new algorithm can be utilized for every software, not only QGS software, which can measure the maximum counts and segmental %peak activity. Thinking of availability, our new method might have advantage over this cardioGRAF.

Several limitations of this study must be considered. Firstly, this study involved small number subjects. Therefore, the results must be validated with larger study population. Secondly, this study was analyzed with the 4-segment myocardial model. Phase analysis and TDI can estimate LV dyssynchrony in more detail with larger myocardial segments. So, we have to improve our software program to deal with more segments, such as 17- or 20- segment model. And it is necessary to investigate the suitable number of myocardial segment model for the prediction and estimation of CRT response.

Conclusion

Our newly developed software program for quantitative estimation of LV dyssynchrony, which is based on the differences of timing in LV regional wall thickening derived from GMPS, can provide important information about the decision-making and evaluation of CRT. And this new algorithm might be comparable to TDI which has been utilized as the only modality for the estimation of LV dyssynchrony.

Copyright information

© The Japanese Society of Nuclear Medicine 2011