Abstract ID: 194 Validation of GPUMCD for X-ray medical imaging

Abstract

GPUMCD, a fast GPU-based Monte Carlo transport code, was initially developed and validated for photon and electron dose calculations in radiation oncology. In this work, the code was adapted to X-ray medical imaging applications, notably to simulate Cone-Beam Computed Tomography (CBCT) projections. The photon interaction processes in GPUMCD, initially validated for the therapeutic (megavoltage) energy range, were revisited with imaging and therefore lower energies in mind. Comparisons with physics models available in Geant4 were conducted for the photoelectric effect, Compton and Rayleigh scattering. For Compton scattering, four Geant4 models were used: two from the Geant4 standard physics (G4KleinNishinaModel and G4KleinNishinaCompton), one from Livermore physics (G4LivermoreComptonModel) and one from Penelope physics (G4PenelopeCompton). Three out of the four models consider atomic shells and the energy of the recoil electron. This correction is not considered in G4KleinNishinaCompton and GPUMCD. Tracking of electron was turned off in GPUMCD and Geant4. The photoelectric and Rayleigh scattering algorithms yielded consistent results in GPUMCD and Geant4. However, significant differences were identified between GPUMCD and Geant4 for Compton scattering. More specifically, differences of up to 15% in energy fluence in simulated projections were observed between simulations with and without atomic shell corrections, consistent with the known overestimation of forward scatter in the plain Klein-Nishina model. Implementing shell correction in GPUMCD would significantly increase computation times but appears necessary for imaging applications.