Performance of OSEM and depth-dependent resolution recovery algorithms for the evaluation of global left ventricular function in 201Tl gated myocardial perfusion SPECT.

Abstract

UNLABELLED: It is unknown whether the use of ordered-subsets expectation maximization (OSEM) and depth-dependent resolution recovery (RR) will increase the accuracy of (201)Tl electrocardiogram-gated SPECT (GSPECT) for the measurement of global left ventricular (LV) function. METHODS: Fifty-six patients having both rest (201)Tl GSPECT and planar equilibrium radionuclide angiography (planar(RNA)) on the same day were studied. Twenty-nine patients also had LV conventional contrast angiography (Rx). LV ejection fraction (LVEF), end-diastolic volume (EDV), and end-systolic volume (ESV) were calculated with the quantitative gated SPECT software (QGS) using 4 different processing methods: filtered backprojection (FBP), OSEM, RR + FBP, and RR + OSEM. LVEF calculated with planar(RNA) and LV EDV and ESV calculated with Rx were considered gold standards. LVEF and volumes provided with the GSPECT methods were compared with the gold standard methods. RESULTS: LVEF calculated with GSPECT methods (FBP, OSEM, RR + FBP, and RR + OSEM) were similar (not statistically significant) and correlated well with planar(RNA). On Bland-Altman analysis, the mean +/- SD of absolute difference in LVEF with GSPECT FBP, OSEM, RR + FBP, and RR + OSEM methods versus planar(RNA) were similar, with relatively large limits of agreement. LV volumes calculated with the 4 GSPECT methods were significantly lower but correlated well with Rx LV volumes. LV volumes calculated with FBP and OSEM were lower than those calculated with RR + FBP and RR + OSEM (P < 0.01). On Bland-Altman analysis, the mean +/- SD of absolute difference in LV volumes with FBP, OSEM, RR + FBP, and RR + OSEM versus Rx was, respectively, 56 +/- 45 mL (P < 0.01 vs. the other 3 methods), 57 +/- 45 mL (P < 0.01 vs. the other 3 methods), 43 +/- 48 mL, and 46 +/- 47 mL, with correspondingly large limits of agreement. The variance of random error did not differ between FBP, OSEM, RR + FBP, and RR + OSEM for either LVEF or volumes. CONCLUSION: OSEM and FBP presented similar accuracy for LVEF and volume measured with the QGS software. Their combination with depth-dependent RR provided similar LVEF but more accurate LV volumes.