Premium
SU‐FF‐I‐108: Estimation of Ejection Fractions by Three Quantitative Gated SPECT Software Packages
Author(s) -
Tiapetch O,
Tocharoenchai C
Publication year - 2007
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.2760485
Subject(s) - ejection fraction , mathematics , nuclear medicine , expectation–maximization algorithm , algorithm , gated spect , coronary artery disease , medicine , statistics , maximum likelihood , heart failure
Purpose: To investigate the accuracy of left ventricular ejection fraction (LVEF) estimated from three quantitative gated SPECT software packages, using echocardiography as a gold standard and the reliability of LVEF from ordered subset expectation maximization (OS‐EM). Method and Materials: Seventy patients with suspected coronary artery disease (CAD) were examined with gated 99m Tc‐MIBI SPECT (8 frames/cardiac cycle) at stress. The projection datasets were reconstructed using OS‐EM with 8 angles/subset. The data were smoothed with Gaussian filter with FWHM of 8 pixel width. The iteration numbers were varied for each software: 3 iterations for quantitative gated SPECT (QGS), 5 iterations for the Emory cardiac toolbox (ECTb), and 4 iterations for 4D‐MSPECT. The LVEFs were calculated using these 3 quantitative software packages. To test reliability of LVEF from OS‐EM, the same patient data were again reconstructed using filtered backprojection (FBP) with post‐filtering of Butterworth 5 th order, 0.3 cycles/pixel cutoff frequency. A two tailed pair t‐test was used to test the statistically significant difference with p‐value < 0.05. Results: There were no statistically significant differences in LVEFs for all three software packages from echocardiography with p > 0.05. The correlation between LVEF in each pair of package was high (r > 0.9). The LVEFs from OS‐EM were not statistically significantly different from that from FBP (p > 0.05) for QGS and 4D‐MSPECT, but for ECTb, there was a statistically significant difference. The LVEF from OS‐EM correlated well with that from FBP (r >0.9). Conclusion: The LVEF estimated from each software package had different characteristics and the LVEF from OS‐EM was reliable.