The Effect of Time-of-Flight and Point Spread Function Modeling on Quantitative Cardiac PET of Large Patients: Phantom Studies

Abstract

The purpose of this paper was to evaluate the effects of time-of-flight (TOF) and point spread function (PSF) modeling on quantitation accuracy in cardiac positron emission tomography (PET) of large patients. Medium and large anthropomorphic cardiac torso phantoms with three anatomical configurations were used in simulating medium and large patients with: 1) arms inside the field of view (FOV); 2) breast for female anatomy; and 3) arms outside the FOV. The effects of TOF and PSF modeling were assessed under two experimental conditions, with and without mismatch in CT attenuation-correction (CTAC) maps. The PET data were reconstructed using analytical and iterative algorithms which included TOF and PSF in combination with six incremental post-reconstruction smoothing filter widths. Polar maps were created by sampling the left ventricle cardiac insert and used for further analyses. The quantitation accuracy of global and regional activity estimates were quantified with contrast recovery coefficient and coefficient of variation. Relative activity bias in the polar maps was compared between the mismatched CTAC reconstructions and their reference matched CTAC reconstructions. The results show that the global and regional quantitation accuracy generally improves with both TOF and PSF modeling; however, with PSF modeling there is an increased variability. For both phantoms, the best quantitation accuracy with least variability was achieved for reconstruction with TOF, PSF modeling, and small post-reconstruction smoothing filter. The relative degree of improvement in quantitation accuracy and uniformity was higher with TOF for the large phantom than the medium phantom. Mismatch artifacts in CTAC maps affected the quantitation accuracy; TOF had a modest effect in reducing the errors, while PSF modeling had little effect.