Estimation of kinetic parameters in dynamic FDG PET imaging based on shortened protocols: a virtual clinical study

This study investigated the estimation of kinetic parameters and production of related parametric K_i images in FDG PET imaging using the proposed shortened protocol (three 3-min/bed routine static images) by means of the simulated annealing (SA) algorithm. Six realistic heterogeneous tumors and various levels of [¹⁸F] FDG uptake were simulated by the XCAT phantom. An irreversible two-tissue compartment model (2TCM) using population-based input function was employed. By keeping two routine clinical scans fixed (60-min and 90-min post injection), the effect of the early scan time on optimizing the estimation of the pharmacokinetic parameters was investigated. The SA optimization algorithm was applied to estimate micro- and macro-parameters (K₁, k_2, k₃, K_i). The minimum bias for most parameters was observed at a scan time of 20-min, which was < 10%. A highly significant correlation (> 0.9) as well as limited bias (< 10%) were observed between kinetic parameters generated from two methods [two-tissue compartment full dynamic scan (2TCM-full) and two-tissue compartment by SA algorithm (2TCM-SA)]. The analysis showed a strong correlation (> 0.8) between (2TCM-SA) K_i and SUV images. In addition, the tumor-to-background ratio (TBR) metric in the parametric (2TCM-SA) K_i images was significantly higher than SUV, although the SUV images provide better Contrast-to-noise ratio relative to parametric (2TCM-SA) K_i images. The proposed shortened protocol by the SA algorithm can estimate the kinetic parameters in FDG PET scan with high accuracy and robustness. It was also concluded that the parametric K_i images obtained from the 2TCM-SA as a complementary image of the SUV possess more quantification information than SUV images and can be used by the nuclear medicine specialist. This method has the potential to be an alternative to a full dynamic PET scan.