Count rate dependence of local signal to noise ratio in positron emission tomography

The count rate dependence of the signal-to-noise ratio (SNR) in positron emission tomography (PET) data is frequently characterized by a global quantity such as the noise equivalent count rate (NECR). It has not been clear, however, to what extent global optimization of data variance (e.g., through varying injected dose or uptake period) also optimizes SNR in local regions of interest. In this paper we present a new methodology for predicting the dependence of local image mean and variance on count rate that can be applied when an actual measurement at only a single count rate is available, as is frequently the case in human clinical PET scanning. We define local count rates and SNR based on the method of Huesman. We demonstrate that local trues and randoms count rates can be accurately modeled from global phantom count rate curves. Using these results we compare local and global SNR count rate responses in phantom and human examples. We find only small differences between the local and global response curves, suggesting that global optimization of SNR or NECR is likely generally adequate for optimizing clinical scanning protocols.