A new composite technique to estimate the input function for FDG-PET brain quantitative analysis without any blood sample

The quantitative determination of LCMRGlc (local cerebral metabolic rate of glucose) is one of the advantages for FDG-PET ([¹⁸F]2-fluoro-2-deoxy-D-glucose positron emission tomography) in diagnosis of brain functions. However, in the FDG 3-compartment model, an input function made of arterial blood samples is necessary for the computation. Due to the invasion of arterial lines and inconvenience of the sampling schedule, FDG quantitative analysis is seldom performed in practice. To solve this dilemma, some methods have been used to simplify the sampling schedule. But one or two blood samples are still needed, or the estimated LCMRGlc may be seriously biased. This paper proposes a new composite technique based on two current methods to estimate the input function without any blood sample. Eighteen patients are tested with this new approach, together with current methods for evaluation. The results show that this new technique can provide reliable estimation of the input function, and the bias of K constants is also reduced when the estimated input function is used.