Comparison of three methods of partial volume correction in dynamic PET cardiac imaging: a phantom and a pig study

We compared 3 methods of correcting the underestimation in measurements of myocardial radiotracer uptake due to partial volume (PV) effects in cardiac PET. Measurements were made on a static heart phantom and on 12 pigs undergoing gated cardiac MRI and 2 dynamic H₂¹⁵O PET. The 3 methods to compute PV correction factors (PVCF) were: 1. A gated MRI myocardial mask temporally and spatially blurred to the PET resolution; 2. an extra-vascular image (EV) created by subtracting the arterial phase of the H₂¹⁵O bolus from a transmission image; 3. kinetic modeling (fit) of the water data. The phantom study confirmed that 7.0 mm Gaussian blurring simulated the scanner resolution, and that the water subtraction correctly determined PVCF from EV (vs theory and vs blurred MRI). PVCF for 60 regions were computed from the 12 pigs using each method. There was no significant systematic difference between PVCF obtained from the 3 methods. When we used only one water acquisition, all the slopes (intercepts) except EV vs MRI were not significantly lower than unity (greater than zero) but there is large variability in the data. When we used two water acquisitions summed together, we saw a slight improvement in the correlations between the methods although the variability did not go down as much as expected. Several factors may contribute to the remaining variability, and to any potential systematic differences which may be present but which we were unable to detect. In general we found a good agreement between the three methods. We believe our data show that accurate partial volume corrections can be made in the myocardium.