Non-stationary spatial filtering and accelerated curve fitting for parametric imaging with dynamic PET

Abstract

A non-stationary spatial low pass filter was developed and implemented in combination with an accelerated non-linear curve fitting routine to create low noise-high contrast images of physiological parameters with dynamic positron emission tomography. The method was applied to 18F-2-fluoro-2-deoxyglucose (FDG) studies, and images of local blood volume, kinetic rate constants, precursor pool volume and glucose metabolism were generated. Noise reduction and contrast preservation was demonstrated in a simulated pie phantom and a study of a patient with a recent brain infarct. Considerably improvement in quantitative accuracy of pixel parameter values was observed in the phantom study in comparison with unprocessed or conventionally smoothed images.