Novi-Sim: A fast X-ray tomography simulation software for laboratory and synchrotron systems to generate training databases for deep learning applications

Abstract

Novi-Sim is an X-ray tomography simulation software that models the attenuation contrast and scattered radiation in laboratory imaging systems, as well as phase contrast in systems with coherent X-ray sources, such as synchrotrons. It simulates multiple X-ray sources, including but not limited to X-ray reflexion and transmission tubes, as well as the image noise through the entirety of the acquisition chain. In this study, we detail some of the models already implemented or that are being implemented to Novi-Sim, particularly, the spectra models and their sampling methods and show that they generate accurate spectra compared to a real X-ray source. In addition, we present the hybrid probabilistic-deterministic model that accounts for scattered radiation and we show the great agreement between results generated in Novi-Sim, and those generated by Gate, a Monte Carlo simulation tool. Given the precision and limited computation time, Novi-Sim can be used to build large databases for training artificial intelligence algorithms.