Nonlinear effect of pile-up in the quantification of a small animal PET scanner

Abstract

Accurate and reliable quantitative analysis of PET images are necessary for pre-clinical studies. To derive quantitative information from PET images, determination of the calibration curve, that is, the relationship between the pixel values in the reconstructed image and radiotracer concentrations is required. In a typical PET acquisition, several effects, such as random counts and pile-up contributions make this relationship nonlinear. We find that for PET detectors based upon relatively large PMTs like the Hamamatsu H8500, pile up may become the dominant nonlinear effect, with random contributions playing a minor role. We confirm this by means of detailed simulations of small and large cylinders in the rPET small animal scanner as well as with real acquisitions. The simulations allow us to study the impact of pile-up as a source of nonlinearity in the calibration curve of this commercially available small animal PET scanner. We compare the results obtained from images for both real and simulated data. The results show that for the activities considered in this study the quantitative results can be affected by pile-up by more than 20%. We find that pile-up, which shifts counts to the center of the FOV and attenuation, which removes activity from the center of the FOV, may cancel each other for moderate activity values. This would cause quantification errors if attenuation corrections were attempted for acquisitions without pile-up corrections. The pile-up correction software improves the linearity of the calibration curve, extending the range of activity values for which a linear calibration curve can be reliably applied.