Computer simulation studies on the effects of input functions on the estimation of physiological parameters with positron emission tomography

Tracer kinetic modeling with Positron Emission Tomography (PET) requires measurements of the time-activity curves in both plasma (PTAC) and tissue (ITAC) to estimate physiological parameters. However, the estimation usually ignores the measurement noise in PTAC while PTAC is assumed to be in the form of the arterial plasma time-activity curve (APTAC) and the traditional sampling schedule used consists of around 20 PTAC measurement points selected during the course of the measurement to construct a smooth curve using piecewise linear approximation. The accuracy and reliability of the physiological parameters estimated by ignoring such errors are not well understood. Feng et a1. [9] recently proposed a PTAC model which fits the complex tracer kinetic behavior in blood vessels and can be used as a driving function for PET kinetic modeling. In this thesis, we apply this PTAC model, in conjunction with the [18F] 2-fluoro-2-deoxy-D-glucose (FDG) experimental data and a 5-parameter FDG model, to systematically study the effects of FDG tracer plasma concentration measurement noise on the estimation of the FDG model’s impulse response function parameters, the rate constants and the physiological parameter, i.e., Local Cerebral Metabolic Rates of Glucose (LCMRGlc). Also, the Optimal Blood Sampling Schedules (OBSS) for PTAC, based on the PTAC model, are investigated by computer simulations and applied to explore the effects of PTAC sampling schedule on the estimation of LCMRGlc. And finally, a more accurate PTAC form, the capillary plasma time-activity carve (CPTAC) is derived and used to study the effects of using APTAC to replace CPTAC on the estimation of physiological parameter. The results of the computer simulations show that all the error sources, e.g., ignoring PTAC measurement noise, using the traditional PTAC sampling schedule and using APTAC to replace CPTAC, increase the unreliability of LCMRGlc estimates