Abstract
1570
Purpose: To compare the myocardial blood flow (MBF) values generated by two software packages for Rb-82 dynamic PET images. Methods:Dynamic Rb-82 PET images for 17 patients were collected and motion corrected across all frames. The LV endocardial cavities were automatically segmented and reviewed for accuracy. MIMcardiac uses a template based segmentation
Methods: The dynamic scans (3:5s,20:1s,6:5s,1:10s,15:15s) were then processed using MIMcardiac and QPET to generate time activity curves and flow values for the vascular segments and the entire LV myocardium. Both QPET and MIMcardiac use a kinetic 1-tissue-compartment flow model with a Rb-82 extraction fraction defined by Lortie for flow calculations. The MBF and coronary flow reserve (CFR) values for all vascular segments were recorded. During QPET data processing, 5 patients required correction of the automatic segmentation (n = 5) and 2 patients were removed due to unsatisfactory segmentation (n = 2). Results:When comparing between MIMcardiac and QPET, stress images had a correlation of 0.84, rest images had a correlation of 0.88, and CFR values had a correlation of 0.73. Comparison of vascular regions and total LV between MIMcardiac and QPET resulted in correlations of vascular segments, LAD - 0.83, LCX - 0.83, RCA - 0.87 and Total Myocardium 0.88. MBF and Bland Altman-results for each vascular region can be found in Figure 1 and Figure 2.
Conclusions: MIMcardiac showed good correlation with QPET across exam type and vascular regions. Further investigation is needed into factors that cause discrepancies including segmentation of myocardium and arterial input, scatter fraction, and partial volume correction.