Potential Impact of SPECT Resolution on Quantification of Left Ventricular Volumes and Ejection Fraction: A Phantom Study

Purpose Accuracy of quantitative assessment of the left ventricular (LV) function derived from ECG-gated SPECT may be affected by image resolution due to the partial volume effect (PVE). This study was to investigate the potential impact of SPECT resolution on quantification of the LV volumes and ejection fraction (EF). Methods Two hybrid SPECT/CT systems of different image resolutions were adopted in this study. One was a conventional dual-head SPECT scanner with low resolution (LR), and the other was a dedicated cardiac SPECT camera with high resolution (HR). Six custom-made cardiac phantoms each with a different LV volume were scanned using both scanners, and images were reconstructed with and without CT-based attenuation correction. A total of 11-bin gated SPECT images and 30 LVEFs were generated from the six phantom volumes to mimic a sequence of cardiac gated SPECT. Four commercial quantification software: Cedars-Sinai-QGS (QGS), Emory-Cardiac Toolbox (ECTB), Invia-Corridor4DM (4DM), and Yale Wackers- Liu CQ (WLCQ), were used to quantify the phantom LV volumes and EFs for comparisons. Results Both SPECT-quantified phantom volumes and EFs are correlated excellently with the true phantom values using Yale-WLCQ software, in which SPECT image resolution and CT-based attenuation show little effect on the correlations. However, the correlations of true phantom values vs. SPECT-quantified LV phantom volumes and EFs calculated from HR-SPECT with attenuation correction are either worse using ECTB, QGS, WLCQ or similar using 4DM as compared to those calculated from LR-SPECT. Quantification errors of the LV phantom EFs quantified from HR-SPECT are significantly greater than those from LR-SPECT, regardless which quantitative software was used (- 21 +/- 11% vs. - 9 +/- 8% with QGS, - 10 +/- 16% vs. - 1 +/- 11% with ECTB, - 16 +/- 10% vs. - 8 +/- 11% with 4DM, - 13 +/- 7% vs. - 1 +/- 5% with WLCQ; all p < 0.02). Conclusion Quantitative assessments of the LVEF from HR-SPECT mainly derived from the PVE results in a significant underestimation of true LVEFs as compared to that from LR-SPECT. Accordingly, current software tools for quantification of the LVEF may need to be remedied for HR-SPECT.