iii

Abstract

X-Ray Fluorescence Computed Tomography (XFCT) is an upcoming and

promising imaging modality. During the last few years a laboratory XFCT

setup has been developed at BIOX, KTH. By combining x-ray imaging (CT)

with targeted nanoparticles (NPs) it has the potential for high-resolution

molecular imaging. The aim of this thesis project is to develop a simulation

tool for this system that meet requirements on accuracy as well as on

computational speed. While simulating XFCT is possible using available

general-purpose Monte-Carlo (MC) codes, they are in general very slow.

A new simulation software using GPU parallelization, XRF-GPU, was

developed based on MC-GPU [11], originally a MC code for x-ray projection

imaging. Several new features necessary for simulating XFCT were implemented,

including atomic relaxation, object translation and the addition of

new energy-resolving detectors. Results from XRF-GPU simulations were validated

against experiments and proved to agree to good accuracy. The new

software provides a speed up factor of >2000 compared to PENELOPE [20],

a well known general-purpose reference code. The simulated scan times are

almost a factor two faster than experimental scan times.

In addition to the successful implementation, preliminary results from simulations

indicates that the previously thought multiple Compton background

might instead be caused by defects in the focusing optics.