A Fast Monte Carlo-Based Forward Projector with Complete Physics Modeling of Y-90 Bremsstrahlung

摘要

Intra-arterial administration of Y-90-microspheres is an established technique for radioembolization of hepatic tumors. We have previously developed a new approach for rapid simulation of bremsstrahlung from Y-90 beta particles; this method is being incorporated into the forward projector of a Monte Carlo (MC)-based SPECT reconstruction program. Conditional probability densities, derived from EGSnrc simulations, were first used to obtain multi-dimensional tables of equally likely parameters for bremsstrahlung production. Rapid simulation then consisted of randomly selecting a beta energy from a table of 100 equiprobable energies, a radial distance to the location of bremsstrahlung production for the given beta energy, and an equiprobable photon energy for the given beta energy and range. The rare nuclear de-excitation and internal pair-production processes yielding 1760- and 511-keV photons were also included. Simulated photons were used to build tables of distance- and energy-dependent resolution kernels (PSFs), including all collimator and detector interactions, and were also propagated through an attenuation map for 8 orders of scatter, yielding scatter-map (S-map) images in 40 energy bins from 59 to 859 keV. Primary and S-map photons were projected into 6 detector energy windows from 59 to 563 keV using convolution-forced detection (CFD) with the precomputed PSFs. Data were simulated and acquired from a 2.5-cm-diameter sphere of Y-90 centered within a water cylinder (7.2-cm diam. × 10.3 cm long). Energy spectra and projection images in the 6 windows were recorded with this phantom positioned ～20 cm from the high-energy general-purpose collimators on a Siemens Symbia SPECT-CT scanner. Acquired and simulated spectra and images were in good agreement. The average absolute difference of counts between fully-MC-simulated and Smap+CFD-computed images was 13.6%; discrepancies can be further reduced by using wider CFD kernels for the highest energy windows.