Quantification of left ventricular volumes and ejection fraction from gated 99mTc-MIBI SPECT:: MRI validation and comparison of the Emory Cardiac Tool Box with QGS and 4D-MSPECT

The goal of this study was to validate the accuracy of the Emory Cardiac Tool Box (ECTB) in assessing left ventricular end-diastolic or end-systolic volume (EDV, ESV) and ejection fraction (LVEF) from gated (99m)Tc-methoxyisobutylisonitrile ((99m)Tc-MIBI) SPECT using cardiac MRI (cMRI) as a reference. Furthermore, software-specific characteristics of ECTB were analyzed in comparison with 4D-MSPECT and Quantitative Gated SPECT (QGS) results (all relative to cMRI). Methods: Seventy patients with suspected or known coronary artery disease were examined using gated 99mTc-MIBI SPECT (8 gates/cardiac cycle) 60 min after tracer injection at rest. EDV, ESV, and LVEF were calculated from gated 99mTc-MIBI SPECT using ECTB, 4D-MSPECT, and QGS. Directly before or after gated SPECT, cMRI (20 gates/cardiac cycle) was performed as a reference. EDV, ESV, and LVEF were calculated using Simpson's rule. Results: Correlation between results of gated 99mTc-MIBI SPECT and cMRI was high for EDV (R = 0.90 [ECTB], R = 0.88 [4D-MSPECT], R = 0.92 [QGS]), ESV (R = 0.94 [ECTB], R = 0.96 [4D-MSPECT], R = 0.96 [QGS]), and LVEF (R = 0.85 [ECTB], P = 0.87 [4D-MSPECT], R = 0.89 [QGS]). EDV (ECTB) did not differ significantly from cMRI, whereas 4D-MSPECT and QGS underestimated EDV significantly compared with cMRI (mean +/- SD: 131 +/- 43 mL [ECTB], 127 +/- 42 mL [4D-MSPECT], 120 +/- 38 mL [QGS], 137 +/- 36 mL [cMRI]). For ESV, only ECTB yielded values that were significantly lower than cMRI. For LVEF, ECTB and 4D-MSPECT values did not differ significantly from cMRI, whereas QGS values were significantly lower than cMRI (mean SD: 62.7% +/- 13.7% [ECTB], 59.0% +/- 12.7% [4D-M-SPECT], 53.2% +/- 11.5% [QGS], 60.6% +/- 13.9% [cMRI]). Conclusion: EDV, ESV, and LVEF as determined by ECTB, 4D-MSPECT, and QGS from gated 99mTc-MIBI SPECT agree over a wide range of clinically relevant values with cMR1. Nevertheless, any algorithm-inherent over- or underestimation of volumes and LVEF should be accounted for and an interchangeable use of different software packages should be avoided.