Statistical Reconstruction For Anisotropic X-ray Dark-Field Tomography
David Frank, Cederik Höfs, Tobias Lasser · 2025
Read Statistical Reconstruction For Anisotropic X-ray Dark-Field Tomography by David Frank, Cederik Höfs, Tobias Lasser (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-03160.txt.
Introduction
David Frank, Cederik Höfs, and Tobias Lasser work on image reconstruction for a new kind of X-ray scan in this 2025 paper. The technique, called AXDT, can reveal fiber structures far smaller than ordinary CT resolves, which matters for both medicine and materials testing. Turning its raw signals into images takes heavy computation, and older methods used rough shortcuts that ignore how noise actually behaves in the scanner.
The authors implement careful statistical methods that model the noise properly, and add a new formulation that keeps the accurate noise model while running much faster. In their tests, the statistical approach beats the older one, above all on noisy data. The fast version matches the slow statistical method's quality at a fraction of the cost. They also analyze how the optimization behaves for each model, opening the door to stronger algorithms later.
Catalog record
| CREATOR | David Frank, Cederik Höfs, Tobias Lasser |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-03160
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Statistical Reconstruction For Anisotropic X-ray Dark-Field Tomography David Frank, Cederik Höfs, Tobias Lasser Abstract Anisotropic X-ray Dark-Field Tomography (AXDT) is a novel imaging technology that enables the extraction of fiber structures on the micrometer scale, far smaller than standard X-ray Computed Tomography (CT) setups. Directional and structural information is relevant in medical diagnostics and material testing. Compared to existing solutions, AXDT could prove a viable alternative. Reconstruction methods in AXDT have so far been driven by practicality. Improved methods could make AXDT more accessible. We contribute numerically stable implementations and validation of advanced statistical reconstruction methods that incorporate the statistical noise behavior of the imaging system. We further provide a new statistical reconstruction formulation that retains the advanced noise assumptions of the imaging setup while being efficient and easy to optimize. Finally, we provide a detailed analysis of the optimization behavior for all models regarding AXDT. Our experiments show that statistical reconstruction outperforms the previously used model, and particularly the noise performance is superior. While the previously proposed statistical method is effective, it is computationally expensive, and our newly proposed formulation proves highly efficient with identical performance. Our theoretical analysis opens the possibility to new and more advanced reconstruction algorithms, which in turn enable future research in AXDT. A Preprint Keywords Anisotropic X-ray Dark-Field Tomography \cdot Statistical Reconstruction \cdot Optimization 1 Introduction X-ray computed tomography (CT) revolutionized medical diagnostics. However, unlike other imaging modalities, such as magnetic resonance imaging, where Diffusion Tensor Imaging (DTI) [ 1 ] can extract directional structures, no such information can be extracted from the sample using attenuation-based X-ray CT. Methods such as DTI have provided invaluable insights into the inner workings of the human brain. A review of applications related to DTI is provided in [ 2 ] . However, X-ray CT is typically superior to MRI-based techniques in imaging speed, resolution and availability. Classical attenuation-based X-ray CT cannot image directional information and lacks the resolution to obtain fine-grained structures such as white brain matter. Nevertheless, using complementary effects of the X-ray wave such as refraction and (ultra) small-angle scattering, one can extract structures smaller than the resolution of the detector and hence microstructure information of the sample. Studies suggest that the small-angle scattering, commonly referred to as dark-field signal, provides valuable Information in the diagnostic and staging of lung diseases [ 3 , 4 , 5 , 6 , 7 , 8 ] . The dark-field signal can be extracted using grating-based interferometry. Initially developed for high brilliance X-ray sources [ 9 ] , it was extended to standard X-ray sources [ 10 , 11 ] . This makes grating-based tomography an excellent candidate for many applications. For example, [ 12 ] integrated a grating-interferometry setup into a standard clinical CT gantry, enabling X-ray Dark-field CT on the human scale. This grating-based X-ray setup only captures scattering perpendicular to the grating orientation. To capture directional information, the sample must be measured from multiple non-standard positions. The directional information can be extracted using the novel imaging modality Anisotropic X-ray Dark-Field Tomography (AXDT) [ 13 , 14 ] . It enables the reconstruction of scattering strengths and directions in arbitrary directions for each voxel. Unlike DTI, AXDT promises to provide fast scan times and higher resolution. [ 15 ] provided evidence that AXDT can successfully expose the orientation of nerve fiber connections in the human brain. As AXDT is still a novel technology, one seeks reconstruction techniques that allow for enhanced reconstruction quality, which in turn enable the reduction of the induced X-ray dose. Both are required for adoption in material testing and biomedical use cases. So far, all research conducted on AXDT has used a simplified linearized model for reconstruction. One would expect this to systematically overestimate and, as such, introduce a bias [ 16 ] . [ 17 ] presented reconstruction formulations for AXDT that make correct noise assumptions for the grating-based imaging process. However, no numerically stable implementation has been proposed until now, and as such, no detailed analysis of these models could be conducted. We contribute such a numerically stable implementation for the statistical reconstruction formulation provided by [ 17 ] . Further, we showcase its capabilities on two relevant experimental datasets, and provide a detailed analysis of the model’s optimization behavior including important tools for optimization such as bounds on the Lipschitz constants, which is a fundamental tool to ensure convergence of algorithms such as the Fast gradient method [ 18 ] . Additionally, we provide a new problem formulation that simplifies the statistical reconstruction, while retaining the same noise assumptions. This model is more efficient and easier to optimize than the previous model, which makes statistical reconstruction more approachable. Further, unlike the model provided by [ 17 ] , we can provide bounds on the Lipschitz constant as well. In the next section, we introduce the grating-based X-ray CT setup used for AXDT. AXDT itself is explained in detail in the subsequent section. We follow this up with a detailed discussion of both the linearized problem formulation and the statistical approaches. Next, we discuss the extraction of the fiber orientation from an AXDT reconstruction. We conclude the methods section with a derivation of the computation of the Lipschitz constants for the problem formulations and a short overview of optimization techniques used throughout our experiments. In Chapter 3 , we describe our experiments in detail. We follow this up with a discussion of the results in Chapter 4 and a conclusion in Chapter 5 . 2 Methods 2.1 Related Work Grating-based Tomography Grating-based tomography, as introduced by [ 9 , 10 , 11 ] , relies on the Talbot-interferometry to generate fringe patterns on the detector. These fringe patterns can be generated with a typical X-ray setup augmented with 3 gratings, as depicted in Figure 1 . Changes in the fringe pattern enable the extraction of absorption, dark-field and phase information of the sample. Figure 1: Schematic overview of grating-based interferometry setup used in Anisotropic X-ray Dark-Field Tomography. The setup consists of an X-ray source, the three gratings (source, phase and absorption), and a sample stage mounted on an Euler cradle. This setup enables acquiring non-standard acquisition viewpoints needed in AXDT. Figure by [ 19 ] is licensed under CC BY 4.0 To extract these signals, the absorption grating is moved in small increments perpendicular to the optical axis and in-plane of the gratings for each viewpoint. This series captures variations in the interference pattern. For each pixel, this series creates a sinusoidal pattern, which enables the extraction of absorption, differential phase-contrast, and dark-field images [ 11 ] . To capture the anisotropic behavior of the dark-field signal and reconstruct directional information of the sample, non-standard trajectories are used [ 20 ] . In the case of grating-based tomography, an Euler cradle, as shown in Figure 1 can be used. The absorption is related to the mean intensity change of the sinusoidal pattern in relation to a reference scan. The dark-field relates to the visibility of the interference pattern, particularly the change in amplitude. As X-rays scatter in the sample, the interference pattern is blurred, reducing its contrast. As such, the visibility is a direct measure of the dark-field signal. Given