Correction of partial volume effect on rate constant estimation in compartment model analysis of dynamic PET study

The estimates for the rate constants in compartment model analysis of ¹⁸F-FDG PET are affected by the fraction of tissue mixture in the ROI, which depends both on the spatial resolution and on the size of the ROI. At first we have evaluated the relationship between the Recovery Coefficient (RC) and estimates of ¹⁸F-FDG rate constants in a ROI without tissue mixture by using a segmented striatum phantom with various resolution. The RC was calculated as the percentage of ROI value to the true activity in the ROI. Next, we have investigated the relationship among the estimates of ¹⁸F-FDG rate constants, spatial resolution and ROI size for the striatum and cerebral cortex, using dynamic digital brain phantom. The activity time course in each tissue was determined with a three-parameter model for ¹⁸F-FDG. Noise was generated by Poisson process according to the collected count for each frame. Sinograms with various spatial resolutions were produced by forward projection, and the images were reconstructed by filtered backprojection method with spatial resolution of 2-12 [mm FWHM]. The parameters (K1, k2, k3, CMRGlc) were estimated by Modified Marquardt method. The fractional tissue component (FTC) was determined as the percentage of the target tissue volume in the total volume of the ROI in the smoothed images. The values of estimates linearly correlated with the FTC for the ROIs with various sizes. The true value was estimated as an extrapolation to 100 [%] FTC on the linear regression line. When this method was applied to the human data for three normal subjects, the estimates also linearly correlated with the FTC and the value at 100 [%] FTC was obtained as a partial volume corrected parameter estimate