An efficient computational reconstruction algorithm for x-ray phase and dark-field imaging using structured illumination with conventional large spot sources and polycapillary optics

Abstract

Conventional x-ray imaging is a useful diagnostic tool that enables the visualization of internal structures through the absorption and transmission of high energy photons. X-ray phase-contrast and dark field imaging offers new contrast modes for low-density materials and materials that contain sub-resolution micro-texture. While these techniques are not new, they have yet to achieve widespread implementation due to their stringent requirements on beam coherence. Currently, these methods require synchrotrons, low power microfocus sources, or expensive gratings that require precision stepping and alignment. The computational process is also restrictive; because the collected data are intensity images, the phase and dark field are not directly observable. Of the reconstruction algorithms that currently exist, many have high computational costs, which creates an additional workflow bottleneck. In this work, we address both of these challenges. We present a rapid computational algorithm based on a Taylor expansion in small refraction angles, Phase and Intensity X-ray Linearization with Dark Field (PIXLD). Using a simple steel mesh as a reference pattern, this algorithm is capable of producing attenuation, differential phase-contrast, and dark field from images taken at just four reference pattern positions. Additionally, low coherence, large spot sources can be combined with polycapillary optics, reducing the effective spot size and improving beam coherence by creating a smaller secondary source. We demonstrate the compatibility of our algorithm with polycapillary optic data collected on a rotating anode with a 240µm spot size.