A Geant4 simulation of X-ray emission for three-dimensional proton imaging of microscopic samples.

ABSTRACT

PURPOSE: Proton computed microtomography is a technique that reveals the inner content of microscopic samples. The density distribution of the material (in g·cm-3) is obtained from proton transmission tomography (STIM Scanning Transmission Ion Microscopy) and the element content from X-ray emission tomography (PIXE Particle Induced X-ray Emission). A precise quantification of chemical elements is difficult for thick samples, because of the variations of X-ray production cross-sections and of X-ray absorption. Both phenomena are at the origin of an attenuation of the measured X-ray spectra, which leads to an underestimation of the element content. Our aim is to quantify the accuracy of a specific correction method that we designed for thick samples. METHODS: In this study, we describe how the 3D variations in the mass density were taken into account in the reconstruction code, in order to quantify the correction according to the position of the proton beam and the position and aperture angle of the X-ray detector. Moreover, we assess the accuracy of the reconstructed densities using Geant4 simulations on numerical phantoms, used as references. RESULTS: The correction process was successfully applied and led, for the largest regions of interest (little affected by partial volume effects), to an accuracy ≤ 4% for phosphorus (compared to about 40% discrepancy without correction). CONCLUSION: This study demonstrates the accuracy of the correction method implemented in the tomographic reconstruction code for thick samples. It also points out some advantages offered by Geant4 simulations i) they produce projection data that are totally independent of the inversion method used for the image reconstruction; ii) one or more physical processes (X-ray absorption, proton energy loss) can be artificially turned off, in order to precisely quantify the effect of the different phenomena involved in the attenuation of X-ray spectra.