A Bayesian framework for the extraction of input function for 18F-FDG metabolism study for both healthy and infarcted rats' hearts

Abstract

The quantitative analysis of tracers in positron emission tomography (PET) studies requires the measurement uptake and retention of tracer in tissue over time. This analysis applied to the heart allows to diagnose its state. It could provide a means to identify areas of myocardial viability and to assess myocardial ischemia. However, the input function (IF), quite commonly used in quantitative analysis, can be corrupted by undesirable effects such as spillover. In this paper, we propose a new approach to correct the cross contamination effect on PET dynamic image sequences. It is based on the decomposition of image pixel intensity into blood and tissue components using Bayesian statistics. The method uses an a priori knowledge of the probable distribution of blood and tissue in the images. Likelihood measures are computed by a General Gaussian Distribution (GGD) model. Bayes' rule is then applied to compute weights that account for the concentrations of the radiotracer in blood and tissue and their relative contributions in each image pixel. We tested the method on a set of dynamic cardiac FDG-PET of healthy and unhealthy rats. The results show the benefit of our correction on the generation of pixel-wise images of myocardial metabolic rates for glucose (MMRG).