Optimizing contrast and spatial resolution in hard X-ray tomography of medically relevant tissues

TL;DR

This study compares filtering techniques in hard X-ray tomography, showing Gaussian filtering can enhance contrast and resolution in soft tissue imaging more effectively than Paganin's phase retrieval, especially when assumptions are invalid.

Contribution

It demonstrates that Gaussian filtering outperforms Paganin's phase retrieval in certain conditions, offering improved image quality and density measurements without relying on false assumptions.

Findings

Gaussian filter yields higher CNR at high SR

Paganin's filter provides larger CNR at low SR

Simple absorption plus Gaussian filtering improves image quality

Abstract

Hard X-ray tomography with Paganin's widespread single-distance phase retrieval filter improves contrast-to-noise ratio (CNR) while reducing spatial resolution (SR). We demonstrate that a Gaussian filter provided larger CNR at high SR with interpretable density measurements for two medically relevant soft tissue samples. Paganin's filter produced larger CNR at low SR, though \emph{a priori} assumptions were generally false and image quality gains diminish for CNR $>1$. Therefore, simple absorption measurements of low-$Z$ specimens combined with Gaussian filtering can provide improved image quality and model-independent density measurements compared to single-distance phase retrieval.