Original Article
Optimization and validation of radionuclide angiography phase analysis parameters for quantification of mechanical dyssynchrony

https://doi.org/10.1007/s12350-009-9119-4Get rights and content

Abstract

Introduction

Cardiac resynchronization therapy (CRT) has the potential to improve the outcome of patients suffering from mechanical dyssynchrony and heart failure. It has been suggested that accurate quantification of baseline extent of mechanical dyssynchrony may lead to pre-selection of patients likely to respond to CRT. The standard deviation from a phase histogram (phaseSD), synchrony (S) and entropy (E) are parameters obtained from phase analysis of planar radionuclide angiography (RNA) that may provide an accurate means of assessing mechanical dyssynchrony. In this paper, the ability of phaseSD, S, and E to detect mechanical dyssynchrony was investigated and optimal values for image smoothing, histogram noise thresholding, and bin size were defined. Finally, the intra- and inter-observer reproducibility of the methodology was assessed.

Methods

PhaseSD, S, and E were calculated for 37 normal subjects (LVEF > 50%, end-diastolic volume < 120 mL, end-systolic volume < 60 mL, QRS < 120 ms, and normal wall motion) and 53 patients with mechanical dyssynchrony (LVEF < 30%, QRS > 120 ms, and typical LBBB). Receiver-operator characteristics (ROC) curves were created and the area under the curve (AUC), for each parameter, was determined using three different imaging filters (no filter and an order 5 Hann filter with cut-off of 5/50 and 10/50). The AUC was also determined using histogram threshold values varying between 0% and 50% (of the max amplitude value). Finally, AUC for E was determined for bins sizes varying between 1° and 20°. Inter- and intra-observer variability was calculated at optimal imaging values.

Results

No smoothing was found to maximize the AUC. The AUC was independent of histogram threshold value. However, a value of 20% provided optimal visualization of the phase image. The AUC was also independent of bin size. At the optimal imaging values, the sensitivity and specificity for all parameters for detection of mechanical dyssynchrony was measured to be 89-100%. Inter- and intra-observer correlation coefficients >0.99 were found for phaseSD, S and E.

Conclusions

Optimized planar RNA phase analysis parameters, phaseSD, S, and E, were able to detect mechanical dyssynchrony with low inter- and intra-observer variability. Studies assessing the ability of these parameters to predict CRT outcome are required.

Introduction

Cardiac resynchronization therapy (CRT) has emerged as a new treatment for a subset of patients with congestive heart failure (CHF) and mechanical dyssynchrony. CRT has been shown to improve survival and reduce heart failure hospitalization and symptoms.1,2 However, up to 50% of apparently suitable patients do not obtain objective benefit.3,4 There are three main theories as to why some patients do not respond to this therapy: inadequate mechanical dyssynchrony,5,6 inappropriate LV lead placement,7, 8, 9 and excessive scar burden.10,11 Techniques are being developed to quantify mechanical dyssynchrony, including tissue Doppler imaging (TDI), MRI, and phase analysis of both radionuclide angiography (RNA) myocardial perfusion imaging (MPI).

A significant amount of research has focused on echocardiographic assessment of mechanical dyssynchrony, using TDI, with promising early single-centre data.12 Despite advances in TDI, recent results have shown low sensitivity and specificity values for detection of response to CRT varying between 0.5 and 0.7, and intra- and inter-observer variabilities of between 4% and 70%.13 Early results using phase analysis of MPI shows promising results that phase analysis based measures of dyssynchrony can predict CRT response with a 74% sensitivity and specificity.14 However, further large-scale, multi-centre studies are required to validate the methodology. Finally, MRI-based measures of dyssynchrony have also shown promise at predicting CRT outcome, but the techniques can be time consuming and costly.15,16 RNA phase analysis, a simple and widely available technique, may still play a role in quantifying mechanical dyssynchrony and predicting CRT outcome. However, further investigation is needed to optimize RNA techniques.

Phase analysis of radionuclide angiography (RNA) images, in which each gated-pixel in the images is fit to the first Fourier harmonic, has been investigated for its use in measuring cardiac dyssynchrony and LV pacing.17 Several studies have shown that the standard deviation of the phase values (phaseSD) can be used to assess mechanical dyssynchrony18, 19, 20; however, other studies have found that phaseSD could not differentiate between various types of mechanical dyssynchrony.19 In response to this, two newer parameters, synchrony (S) and entropy (E), have been proposed and reported to provide good distinction between various forms of mechanical dyssynchrony in a computer simulation.19 A full characterization and optimization of these variables, using patient data and including intra- and inter-observer reproducibility, has not yet been performed.

The aim of this paper was to provide a characterization of, and to optimize the previously defined phase analysis parameters (phaseSD, S, and E) at detecting mechanical dyssynchrony using ROC analysis.

Section snippets

Patient Population

A database of normal subjects and patients with known mechanical dyssynchrony was prospectively included in this study. Normal subjects were referred to our clinic for assessment of LV function. Normal LV function was defined as LVEF > 50%, EDV < 130 mL, ESV < 60 mL, QRS duration < 120 ms, and normal wall motion, as determined by an expert observer. The mechanical dyssynchrony subjects were defined as LVEF < 30%, QRS duration > 130 ms, and dyssynchronous wall motion as determined by an expert

Patient Population

A total of 37 normal subjects were included in this study. Fifty-three cardiomyopathy patients with known cardiac mechanical dyssynchrony were also included. Of the 53 mechanical dyssynchrony subjects, 45 exhibited typical LBBB, 1 with RBBB, and 8 with non-specific intraventricular conduction defect. The patient demographics are tabulated in Table 1.

Optimization of RNA Phase Analysis

Figure 3A-C plots the ROC area versus amplitude threshold value for the three parameters (phaseSD, S, and E) and using the three filters. For all

Discussion

Compared to echo, little work has been done validating and improving RNA phase analysis for quantification of mechanical dyssynchrony. Port22 recently voiced this in an editorial: “…when there was suddenly a clinical calling for the quantitative analysis of variation in the timing of contraction in and between ventricles, the majority of the nuclear community was caught sleeping, and the echocardiographers advanced tissue Doppler imaging as a tool for selection of and subsequent assessment of

Limitations and Future Directions

One of the limitations of this work is the use of the same cohort of subjects to both optimize the parameters and then determine the sensitivity and specificity of the method at detecting mechanical dyssynchrony (Table 2). This has the potential to positively bias our sensitivity and specificity results. However, given the high values for the AUC, we do not expect this bias to be large and would expect similar results from a separate cohort. In addition, the ability to detect mechanical

Conclusion

Planar RNA phase analysis parameters (phaseSD, S, and E) were optimized for detecting mechanical dyssynchrony. phaseSD, S and E displayed excellent intra- and inter-observer variability with correlation coefficients >0.99. Studies assessing the ability of these parameters to predict CRT outcome are required.

Acknowledgments

The authors would like to thank Mary Dalipaj and Brian Marvin for their assistance in collecting the data. The authors would also like to thank Benoit Galarneau and Hermes Medical Solutions for assistance in developing the phase analysis program.

References (23)

  • PortSC

    Timing is everything

    J Nucl Cardiol

    (2008)
  • Cited by (22)

    • Editorial: Imaging synchrony

      2009, Journal of Nuclear Cardiology
    View all citing articles on Scopus
    View full text