Low-exposure, high-quality multimodal speckle X-ray imaging via an intrinsic gradient-flow approach

We present a new approach for retrieving dark-field, phase shift, and attenuation images from speckle-based X-ray imaging data. Speckle-based X-ray imaging (SBXI) exploits sample-induced alterations to a reference near-field speckle pattern produced by a randomly structured mask. Attenuation images allow materials of different densities to be visualised. Phase-shift images are useful because they reveal how materials in a sample refract the X-ray beam, providing contrast between similar low-density structures that are difficult to reconstruct in attenuation images. Dark-field images convey information about structures that are smaller than the spatial resolution and thus invisible in both attenuation and phase-shift images. In previous works, we presented the Multimodal Intrinsic Speckle-Tracking (MIST) algorithm, which recovers the three complementary imaging modes from SBXI data by solving the associated Fokker–Planck equation. In this work, we present a variation of MIST, called ``gradient-flow MIST", which (1) reduces the amount of SBXI data required for image retrieval, (2) maintains the full generality of the X-ray Fokker–Planck equation, and (3) recovers dark-field images with higher quality than the previously proposed variants for weakly attenuating (i.e., low density) samples. We demonstrate the new gradient-flow MIST approach on experimental SBXI data of a knotted bundle of carbon fibres acquired at the Australian synchrotron. This approach is anticipated to be useful in phase-contrast and dark-field applications that require simplicity in experimentation and low sample X-ray exposure.

💡 Research Summary

Speckle‑based X‑ray imaging (SBXI) leverages the modulation of a near‑field speckle pattern, generated by a random mask, to extract three complementary contrast mechanisms: attenuation, phase shift, and dark‑field (small‑angle scattering). While earlier approaches such as XST, XSS, and UMPA performed local correlation analyses, the optical‑flow paradigm introduced a global transport‑of‑intensity‑like equation, initially limited to pure‑phase objects. The Multimodal Intrinsic Speckle‑Tracking (MIST) framework later incorporated the full X‑ray Fokker‑Planck equation, thereby accounting for dark‑field effects, but required at least four speckle image pairs and suffered from artefacts when the dark‑field varied rapidly.

The present work introduces Gradient‑Flow MIST (GF‑MIST), a variant that retains the full generality of the Fokker‑Planck model while halving the data requirement to just two speckle image pairs. The authors start from the finite‑difference form of the Fokker‑Planck equation (Eq. 1) and rewrite it as a continuity equation (Eq. 2) with a conserved “energy” flux J⊥. By applying a 2‑D Helmholtz decomposition and discarding the rotational component (justified for slowly varying attenuation), they arrive at a Poisson equation for a scalar potential V (Eq. 5). Solving this Poisson problem yields the gradient ∇⊥V, which encodes both phase‑gradient and dark‑field information.

An initial transmission estimate I_ob^(1) is obtained from a TIE‑based first‑order approximation (Eq. 6). Using the measured speckle reference I_R and sample images I_S, the authors compute the quantities α₁, α₂ (refraction angles in x and y) and α₃ (a thickness‑dependent dark‑field term) by solving the linear system formed by Eqs. 7‑8. Fourier‑domain derivative (f_d) and inverse‑Laplacian (f_L) filters provide noise‑robust differentiation and integration, respectively. Because only two image pairs are available, the system is under‑determined; however, the authors close the loop by iteratively refining the transmission image. They substitute the current α‑values into a closed‑form solution of the Fokker‑Planck equation (Eq. 15‑18), which yields an updated transmission I_ob^(2) that now includes dark‑field‑induced blur. The dark‑field map follows directly from D = −α₃·I_ob^(2) (Eq. 19), and the phase shift is retrieved via ϕ = γ ln I_ob^(2) (Eq. 20), where γ = δ/β is the known ratio of refractive‑index decrement to absorption index.

Experimental validation was performed at the Australian Synchrotron’s Imaging and Medical Beamline (IMBL). A knotted carbon‑fiber bundle was imaged with 25 keV X‑rays, a mask‑sample distance of 1 m, and a sample‑detector distance of 3 m. Twenty speckle image pairs were recorded by laterally shifting a stack of sand‑paper masks; GF‑MIST, however, uses only two of these pairs. Reconstructed images include the effective diffusion coefficient D (dark‑field), the corrected refraction angles α₁ and α₂, the refined transmission I_ob^(2), and the phase shift ϕ. Fourier Ring Correlation (FRC) analysis shows spatial resolutions of roughly 4.8 pixels for dark‑field and 5.4 pixels for phase, comparable to or better than previous MIST variants that required twice as many measurements. When an exact contact transmission image is available, the authors demonstrate a closed‑form solution (Eqs. 21‑22) that bypasses the iterative refinement, further reducing computational load.

In summary, GF‑MIST delivers three major advances: (1) a data‑efficient protocol requiring only two speckle image pairs, which shortens acquisition time and reduces radiation dose; (2) preservation of the full Fokker‑Planck physics, ensuring simultaneous retrieval of attenuation, phase, and dark‑field without simplifying assumptions; and (3) superior dark‑field image quality for weakly attenuating (low‑density) samples, addressing a known weakness of earlier MIST implementations. These attributes make GF‑MIST a compelling candidate for low‑exposure multimodal X‑ray imaging in medical diagnostics, materials science, and non‑destructive testing, where simplicity of setup and high‑contrast, high‑resolution information are paramount.