Multimodal Intrinsic Speckle-Tracking (MIST) to extract rapidly-varying diffuse X-ray scatter

TL;DR

This paper introduces an extended MIST formalism for speckle-based phase-contrast X-ray imaging that accurately reconstructs multimodal signals without assuming spatially slowly varying dark-field signals, improving image quality.

Contribution

The authors extend the MIST formalism to handle non-slowly varying diffusive-dark-field signals, enabling better reconstruction of unresolved microstructures in speckle-based X-ray imaging.

Findings

Reconstructed superior diffusive-dark-field images compared to previous methods.

Successfully applied the extended MIST to samples with distinct X-ray properties.

Enhanced potential for applications in biomedical, engineering, and geological imaging.

Abstract

Speckle-based phase-contrast X-ray imaging (SB-PCXI) can reconstruct high-resolution images of weakly-attenuating materials that would otherwise be indistinguishable in conventional attenuation-based imaging. The experimental setup of SB-PCXI requires only a sufficiently coherent source and spatially random mask, positioned between the source and detector. The technique can extract sample information at length scales smaller than the imaging system's spatial resolution; this enables multimodal signal reconstruction. ``Multimodal Intrinsic Speckle-Tracking'' (MIST) is a rapid and deterministic formalism derived from the paraxial-optics form of the Fokker-Planck equation. MIST simultaneously extracts attenuation, refraction, and small-angle scattering (diffusive-dark-field) signals from a sample and is more computationally efficient compared to alternative speckle-tracking approaches. Hitherto, variants of MIST have assumed the diffusive-dark-field signal to be spatially slowly varying. Although successful, these approaches have been unable to well-describe unresolved sample microstructure whose statistical form is not spatially slowly varying. Here, we extend the MIST formalism such that there is no such restriction, in terms of a sample's rotationally-isotropic diffusive-dark-field signal. We reconstruct multimodal signals of two samples, each with distinct X-ray attenuation and scattering properties. The reconstructed diffusive-dark-field signals have superior image quality compared to our previous approaches which assume the diffusive-dark-field to be a slowly varying function of transverse position. Our generalisation may assist increased adoption of SB-PCXI in applications such as engineering and biomedical disciplines, forestry, and palaeontology, and is anticipated to aid the development of speckle-based diffusive-dark-field tensor tomography.