Abstract
453
Objectives: Myocardial blood flow (MBF) imaging with positron emission tomography (PET) continues to be an important non-invasive functional imaging technique for reliable and accurate characterization of coronary artery disease (CAD), but relies on relies on analysis algorithms that perform consistently and accurately with high repeatability. Previous studies by Efseaff et al tested scan-time, reconstruction, source for the blood input function, rate pressure product (RPP) normalization of flow values, and blood spillover correction (BSoC-On/Off); and studies by Ocneanu et al tested the source for the blood input function, BSoC-On/Off, right blood correction, arterial blood transit delay, and distribution volume constraint (Global/Regional). Both previous studies showed improvement in MBF repeatability using the left atrium (LA) instead of the LV cavity for the blood input function (using small (N < 40) populations). The purpose of this study was to determine the effect of the following parameters on Rb-82 rest repeatability: source for the blood input function (LV vs LA), BSoC-On/Off, bias from deadtime correction (BC-on/off), and source for LV alignment (stress images (paired) vs individual (not paired)) in a larger (N = 140) clinical population.
Methods: The study population consisted of 140 cardiac patients referred to our clinic for Rb-82 MPI. The patients were 31 % female and 69 % male with an average age of 64 [23, 90] years and an average BMI of 31 [27, 40]. All patients were instructed to fast for 6 h, and to abstain from caffeine for 24 h prior to PET imaging. For rest and repeat rest, patients were administered 82Rb Chloride saline at a dose rate of 10 and 5 MBq/kg respectively. After acquisition, all patients underwent a stress acquisition following an injection of 82Rb Chloride saline at a dose rate of 10 MBq/kg. All images were acquired on the GE Discovery 690 (D690) PET/CT scanner (GE Healthcare, Waukesha, WI). Reconstructed dynamic images were analyzed using FlowQuant to estimate LV tracer uptake and MBF and apply necessary corrections. Flow was calculated using the 1-tissue compartment model with the AHA 17-segment model using global partial-volume correction. The Kolmogorov-Smirnov test indicated that the data sets were nonparametric; therefore, significant differences in interquartile range and median were determined by Levene’s and Wilcoxon rank sum tests. Spearman's rank-order correlation was used to check for correlations in the Bland-Altman plots. Friedman's Two-Way ANOVA test was used to check for significant changes at the 95 % confidence level. In all cases, Bonferroni corrections were applied.
Results: The best repeatability achieved (coefficient or repeatability (CR) 0.25) was with the LA as the blood input function using paired LV alignment with all corrections turned off. Turning blood spill over correction on always increased variability. Lowest variability (CR = 0.39) was with the LV as the blood input function using unpaired LV alignment and RPP normalization of flow values with all corrections turned on. There is < 1 % difference in the absolute blood flow values using the LA instead of the LV for the blood input function.
Conclusions: To decrease the variability of MPI between scans, blood spill over and bias corrections should be turned off, the LA should be used as the blood input function. Using the LA instead of the LV for the blood input function should not affect absolute quantification.