Direct EM reconstruction of kinetic parameters from list-mode cardiac PET

Abstract

The aim of this research is to extend a direct parametric reconstruction algorithm for dynamic PET to include spillover terms for use in cardiac perfusion imaging. Perfusion imaging with cardiac PET is used clinically to produce images of myocardial blood flow, aiding the diagnosis of coronary artery disease and the monitoring of condition of coronary circulation in response to treatment. PMOLAR-1T, an EM-based direct reconstruction algorithm, was adapted for parameter estimation from cardiac PET by including in the complete data space an additional dimension to account for the 3 model terms (myocardium, left ventricle, and right ventricle). The algorithm was validated with a digital phantom simulation and compared to a traditional frame-based approach. The coefficient of variation of K1 was reduced by 20% in the direct method as compared to the frame-based method, with lower bias in all parameters. The algorithm was also tested on a human 82Rb dataset, for which direct reconstruction provided lower K1 estimates in the myocardium than the frame-based method. This discrepancy may be due to errors from model fits outside the myocardium or from difference in the scatter correction implementation.