Dual-time-point Patlak estimation from list mode PET data

We investigate the potential of using dual-time-point PET data to perform Patlak modeling. If successful this approach could be used for whole-body dynamic PET in which we compute voxel-wise estimates of Patlak parameters using two frames of data for each bed position. Our approach directly uses list-mode arrival time for each event to compute the Patlak image. We evaluate performance of the method in comparison to fractional changes in SUV values computed from the same two frames of data. The area under ROC curves is used as a figure of merit to assess the relative performance of these two approaches in differentiating tumor from background. We calculate ROC curves using (i) variance estimates computed from the Cramer-Rao bound for a simplified 1D version of the problem and (ii) 4D Monte Carlo simulations of tumors with a range of sizes in a uniform background. Both the simulation and Cramer-Rao analysis suggest that our dual-time-point Patlak estimation method can achieve superior differentiation of tumor from background in small tumors compared to using fractional changes in SUV computed from the same dual-time-point data.