Design of temporal sampling schedules for dynamic cardiac FDG SPECT

Abstract

Dynamic cardiac SPECT is being explored for use in measuring myocardial viability using F-18 labelled deoxyglucose (FDG). However, poor photon sensitivity due to high crystal penetration and the collimator used results in very high statistical noise. The temporal sampling protocol currently used by our group for dog and human subject FDG SPECT imaging with the fitting routine often fails to converge. The goal of this work was to design a sampling schedule for noisy measurements obtained with dynamic cardiac FDG SPECT whereby the bias and variance in the parameter estimates are minimized. This was done by performing computer simulations to obtain an effective sampling protocol for four different tracer infusion lengths. The simulated input function was acquired by arterial sampling from a canine study. A three-compartment model was used to represent the kinetics of FDG in the myocardium. The constraint for minimum bias in the parameter estimates was achieved by limiting the intra-sample temporal changes in the blood and the tissue uptake, while integrating uniform 10 sec samples done to obtain minimum covariance in the parameter estimates. This method gave sampling schedules with minimum bias and variance in parameter estimates. This approach may be useful in formulating a data acquisition technique that would yield more accurate parameter estimates thus improving the quantitation of myocardial viability.