Imaging of each compartment in PET dynamic study with dual tracer injection and 2-input compartment model

Aim: PET kinetic analysis with dual tracer injection is challenging to assess two different functions, such as cerebral metabolite rate of glucose and distribution volume of benzodiazepine receptor, at the same time, in the same condition. We investigated the reliability of parameter estimates for various injection protocols, and evaluated the possibility of generating the each compartment kinetic images with 2-input compartment model by means of a computer simulation. Methods: The relation between the reliability of parameter estimates and injection interval of two tracers or between the reliability of parameter estimates and administration dose ratio of two tracers was investigated in the computer simulation study, and the injection protocol for reliable estimation was evaluated. Simulated decaying tissue time activity curves were generated with true k-values and input functions by using the 2-input compartment model for various injection intervals and administration dose ratio of two tracers, and the noise depended on the collected total count was added to each time activity curve. From these activity curves, the rate constants were estimated by nonlinear least square method, and the reliability was evaluated by the mean of absolute differences (MAD) between the true and estimated value for one thousand runs. With the optimal injection protocol determined from the simulation study, dynamic digital phantom was created, parameters were estimated pixel by pixel, and each compartment image of both tracers were created. Results: The reliability of parameter estimates for various injection protocols were compared, and it was found that parameters were able to be estimated reliably when the second injection was started 15 minutes later. The reliability did not depend on the administration dose ratio of two tracers. In case of [18F]FDG and [11C]flumazenil, the MAD of parameters reflected cerebral metabolite rate of glucose or distribution volume of receptor was under 10% in the 5% noise level. The radioactivity image of each compartment represented the accumulation of two tracers. Conclusion: In the simulation study, the rate constants of two tracers were able to be estimated simultaneously by 2-input compartment model, and the possibility of kinetic analysis for dual tracer injection was shown