[Creation of a dynamic digital phantom and its application to a kinetic analysis]

A dynamic digital brain phantom was created from the MRI to evaluate visually the relationship between the noise and the error in the parameter estimates in the PET kinetic analysis. This phantom incorporates the noise level depending on administration dose, camera efficiency and the data acquisition schedule. We simulated a serial dynamic scan with 18F-FDG or 11C-flumazenil, assuming 2-tissue 3-parameter model and 1-tissue 2-parameter model, respectively, and the sampling schedule was determined according to the clinical examination. The noise in the tissue time activity curve in FDG had a peak in the first minute and decreased thereafter, whereas the noise increased gradually in the flumazenil study after the initial peak due to radioactivity decay. We examined the relationship between the noise level and the error in the parameter estimates. Both mean absolute differences between true and estimated values and standard deviation became large, and the quality of the parametric images became poor with increasing noise level. This simulation was compared with human tissue time activity curves and parametric images, which were obtained with 100 MBq administration dose in FDG study and 430 MBq in flumazenil study. We inferred that the noise level in the human study was 10-20% in FDG, and 20-40% in flumazenil, and the error in the estimated parameter of K-complex in FDG study was about 20%, that of DV in flumazenil study was 2%.