Kinetic parameters estimation for heterogeneous tumor model

We investigate a heterogeneous kinetic model of tracer uptake in positron emission tomography (PET) for the purposes of improved characterization of small tumors or metastases. Using a mixture model we allow each voxel to contain a mixture of kinetic processes representing different tumor and normal tissue time activity curves and the plasma concentration. We then use a novel nonlinear least squares procedure to estimate the kinetic parameters of these processes from a region of interest consisting of multiple mixture voxels with unknown fractions of these processes. By investigating the Cramer Rao lower bound (CRLB) we determine conditions under which kinetic parameters and mixture fractions can be accurately identified. We also perform Monte Carlo simulations to show that the variance of the rate parameters estimated using nonlinear least squares is close to the bound predicted by our CRLB analysis.