Validation of a computational chain from PET Monte Carlo simulations to reconstructed images

Abstract

The study aimed to create a pipeline from Monte Carlo simulated projections of a Gate PET system to reconstructed images. The PET system was modelled after the GE Discovery MI (DMI) PET/CT, and the simulated projections were reconstructed with the stand-alone reconstruction software CASToR. Attenuation correction, normalisation calibration, random estimation, and scatter estimation for the simulations were computed with in-house programs. The pipeline was compared in both projection and image space with data acquired on a clinical DMI and reconstructed with GE's off-line PET reconstruction software (PET Toolbox) and CASToR. The simulated and measured data were compared for the number of prompt coincidences, scatter fraction, contrast recovery coefficient (CRC), signal-to-noise ratio (SNR), background variability, residual lung error, and image profiles. A slight discrepancy was noted in the projection space, but good agreements were generally achieved in image space between simulated and measured data. The CRC was found to be 81 % for Gate – CASToR, 84 % for GE – CASToR, and 84 % for GE - PET Toolbox for the largest sphere of the NEMA image quality (IQ) phantom, and the SNR was found to be 98 for Gate – CASToR, 91 for GE – CASToR, and 93 for GE – PET Toolbox. Profiles drawn over the spheres for the NEMA IQ phantom and the Data Spectrum (DS) phantom show a good match between measurement and simulation. The results indicate feasibility to utilise the pipeline as a tool for off-line simulation-based studies. A complete pipeline introduces possibilities to study the impact of single parameters in the whole chain from simulation to reconstructed images.