Evaluation of Left Ventricular Volumes and Ejection Fraction from Gated Myocardial Perfusion SPECT Processed with “Myovation Evolution”: Comparison of Three Automated Software Packages using Cardiac Magnetic Resonance as Reference

Abstract

Backgound: The development of resolution recovery (RR) algorithms has made it possible to preserve good quality of cardiac images in spite of reduced number of counts during study acquisition. Objective: Our purpose was to evaluate the performance of three different software packages in the quantification of left ventricular (LV) end-diastolic volume (EDV), end-systolic volume (ESV) and ejection fraction (EF) from gated perfusion SPECT, applying a resolution recovery (RR) algorithm (GE Myovation Evolution), respect to cardiac MRI (cMRI) as gold standard. Methods: We retrospectively enrolled 21 patients, with suspected or known coronary heart disease. Images at rest were reconstructed by filtered back projection (FBP) and by an iterative protocol with the RR algorithm. EDV, ESV and LVEF were automatically computed employing Quantitative gated SPECT (QGS), Myometrix (MX) and Corridor 4DM (4DM). Any difference in EDV, ESV and LVEF calculation between cMRI and the three packages (with FBP and iterative reconstruction with RR) was tested using Wilcoxon or paired t-test, with assumption of normality assessed using ShapiroWilk test. Agreement between imaging reconstruction algorithms and between gated-SPECT software packages and cMRI was studied with Pearson’s (r) or Spearman’s (R) correlation coefficients and Lin’s concordance correlation coefficient (LCC). Results: Intra-software evaluation always revealed very strong correlation coefficients (R, r ≥ 0.8) and excellent LCC coefficients (LCC > 0.95), except for LCC coefficient between MX-FBP and MX-RR in EDV evaluation, nevertheless considered very good (LCC = 0.94). EDV and ESV had significantly lower value when calculated with RR algorithm respect to FBP reconstruction in QGS and MX. LVEF estimation did not show significant differences for QGSFBP, QGS-RR, MX and 4DM-RR with respect to cMRI. Results: Intra-software evaluation always revealed very strong correlation coefficients (R, r ≥ 0.8) and excellent LCC coefficients (LCC > 0.95), except for LCC coefficient between MX-FBP and MX-RR in EDV evaluation, nevertheless considered very good (LCC = 0.94). EDV and ESV had significantly lower value when calculated with RR algorithm respect to FBP reconstruction in QGS and MX. LVEF estimation did not show significant differences for QGSFBP, QGS-RR, MX and 4DM-RR with respect to cMRI. Conclusion: All reconstruction methods sistematically understimate EDV and ESV, with higher underestimation applying only the RR. No significant differences were observed between 4DM -RR and 4DM-FBP, for each parameter, when 4DM package was used.