arXiv subjects

Tilo Baumbach

Publications and source records attributed to Tilo Baumbach.

14 recordsLinked to original sources

arXiv2023

The Bragg demagnifier: X-ray imaging with kilometer propagation distance within a meter

We introduce a new X-ray imaging technique to facilitate propagation-based phase contrast of large, centimeter-sized samples. The diffracted X-ray wavefield behind the sample is demagnified by asymmetric Bragg crystal optics, thereby virtually increasing the propagation distance and thus enhancing the image contrast. We demonstrate the significant increase in image contrast compared to conventional phase contrast imaging at the same short physical propagation distance. Additionally, the Bragg demagnifier enables the reduction of image blur caused by the finite X-ray source size. In combination with a subsequent Bragg magnifier, the method will allow for an even higher dose efficiency, rendering this technique a potential candidate for, e.g., low-dose (bio)medical diagnostics.

physics.optics↗

arXiv2023

Shot Noise Reduction in Radiographic and Tomographic Multi-Channel Imaging with Self-Supervised Deep Learning

Noise is an important issue for radiographic and tomographic imaging techniques. It becomes particularly critical in applications where additional constraints force a strong reduction of the Signal-to-Noise Ratio (SNR) per image. These constraints may result from limitations on the maximum available flux or permissible dose and the associated restriction on exposure time. Often, a high SNR per image is traded for the ability to distribute a given total exposure capacity per pixel over multiple channels, thus obtaining additional information about the object by the same total exposure time. These can be energy channels in the case of spectroscopic imaging or time channels in the case of time-resolved imaging. In this paper, we report on a method for improving the quality of noisy multi-channel (time or energy-resolved) imaging datasets. The method relies on the recent Noise2Noise (N2N) self-supervised denoising approach that learns to predict a noise-free signal without access to noise-free data. N2N in turn requires drawing pairs of samples from a data distribution sharing identical signals while being exposed to different samples of random noise. The method is applicable if adjacent channels share enough information to provide images with similar enough information but independent noise. We demonstrate several representative case studies, namely spectroscopic (k-edge) X-ray tomography, in vivo X-ray cine-radiography, and energy-dispersive (Bragg edge) neutron tomography. In all cases, the N2N method shows dramatic improvement and outperforms conventional denoising methods. For such imaging techniques, the method can therefore significantly improve image quality, or maintain image quality with further reduced exposure time per image.

cs.CV↗

arXiv2023

Optimizing the Procedure of CT Segmentation Labeling

In Computed Tomography, machine learning is often used for automated data processing. However, increasing model complexity is accompanied by increasingly large volume datasets, which in turn increases the cost of model training. Unlike most work that mitigates this by advancing model architectures and training algorithms, we consider the annotation procedure and its effect on the model performance. We assume three main virtues of a good dataset collected for a model training to be label quality, diversity, and completeness. We compare the effects of those virtues on the model performance using open medical CT datasets and conclude, that quality is more important than diversity early during labeling; the diversity, in turn, is more important than completeness. Based on this conclusion and additional experiments, we propose a labeling procedure for the segmentation of tomographic images to minimize efforts spent on labeling while maximizing the model performance.

cs.CV↗

arXiv2023

A Knowledge Distillation framework for Multi-Organ Segmentation of Medaka Fish in Tomographic Image

Morphological atlases are an important tool in organismal studies, and modern high-throughput Computed Tomography (CT) facilities can produce hundreds of full-body high-resolution volumetric images of organisms. However, creating an atlas from these volumes requires accurate organ segmentation. In the last decade, machine learning approaches have achieved incredible results in image segmentation tasks, but they require large amounts of annotated data for training. In this paper, we propose a self-training framework for multi-organ segmentation in tomographic images of Medaka fish. We utilize the pseudo-labeled data from a pretrained Teacher model and adopt a Quality Classifier to refine the pseudo-labeled data. Then, we introduce a pixel-wise knowledge distillation method to prevent overfitting to the pseudo-labeled data and improve the segmentation performance. The experimental results demonstrate that our method improves mean Intersection over Union (IoU) by 5.9% on the full dataset and enables keeping the quality while using three times less markup.

cs.CV↗

arXiv2023

Self-Supervised Learning for Biological Sample Localization in 3D Tomographic Images

In synchrotron-based Computed Tomography (CT) there is a trade-off between spatial resolution, field of view and speed of positioning and alignment of samples. The problem is even more prominent for high-throughput tomography--an automated setup, capable of scanning large batches of samples without human interaction. As a result, in many applications, only 20-30% of the reconstructed volume contains the actual sample. Such data redundancy clutters the storage and increases processing time. Hence, an automated sample localization becomes an important practical problem. In this work, we describe two self-supervised losses designed for biological CT. We further demonstrate how to employ the uncertainty estimation for sample localization. This approach shows the ability to localize a sample with less than 1.5\% relative error and reduce the used storage by a factor of four. We also show that one of the proposed losses works reasonably well as a pre-training task for the semantic segmentation.

cs.CV↗

arXiv2022

Using the Order of Tomographic Slices as a Prior for Neural Networks Pre-Training

The technical advances in Computed Tomography (CT) allow to obtain immense amounts of 3D data. For such datasets it is very costly and time-consuming to obtain the accurate 3D segmentation markup to train neural networks. The annotation is typically done for a limited number of 2D slices, followed by an interpolation. In this work, we propose a pre-training method SortingLoss. It performs pre-training on slices instead of volumes, so that a model could be fine-tuned on a sparse set of slices, without the interpolation step. Unlike general methods (e.g. SimCLR or Barlow Twins), the task specific methods (e.g. Transferable Visual Words) trade broad applicability for quality benefits by imposing stronger assumptions on the input data. We propose a relatively mild assumption -- if we take several slices along some axis of a volume, structure of the sample presented on those slices, should give a strong clue to reconstruct the correct order of those slices along the axis. Many biomedical datasets fulfill this requirement due to the specific anatomy of a sample and pre-defined alignment of the imaging setup. We examine the proposed method on two datasets: medical CT of lungs affected by COVID-19 disease, and high-resolution synchrotron-based full-body CT of model organisms (Medaka fish). We show that the proposed method performs on par with SimCLR, while working 2x faster and requiring 1.5x less memory. In addition, we present the benefits in terms of practical scenarios, especially the applicability to the pre-training of large models and the ability to localize samples within volumes in an unsupervised setup.

eess.IV↗

arXiv2022

Exploiting of flux shadowing effect on In$_{x}$Ga$_{1-x}$As asymmetric shell growth for strain and bending engineering in GaAs - In$_{x}$Ga$_{1-x}$As core - shell NW arrays

Here we report on non-uniform shell growth of In(x)Ga(1-x)As onto GaAs nanowire (NW) core by molecular beam epitaxy (MBE). The growth was realized on pre-patterned silicon substrates with pitch size (p) ranging from 0.1 um to 10 um. Considering the preferable bending direction with respect to the MBE cells as well as the layout of the substrate pattern, we are able to modify the strain distribution along the NW growth axis and the subsequent bending profile. For NW arrays with high number density, the obtained bending profile of the NWs is composed of straight (barely-strained) and bent (strained) segments with different lengths which depend on the pitch size. A precise control of the bent and straight NW segment length provides a recipe to design NW based devices with length selective strain distribution.

physics.ins-det↗

arXiv2021

In-situ X-ray analysis of misfit strain and curvature of bent polytypic GaAs-In(x)Ga(1-x)As core-shell nanowires

Misfit strain in core-shell nanowires can be elastically released by nanowire bending in case of asymmetric shell growth around the nanowire core. In this work, we investigate the bending of GaAs nanowires during the asymmetric overgrowth by an In(x)Ga(1-x)As shell caused by avoiding substrate rotation. We observe that the nanowire bending direction depends on the nature of the substrate's oxide layer, demonstrated by Si substrates covered by native and thermal oxide layers. Further, we follow the bending evolution by time-resolved in-situ X-ray diffraction measurements during the deposition of the asymmetric shell. The XRD measurements give insight into the temporal development of the strain as well as the bending evolution in the core-shell nanowire.

physics.ins-det↗

arXiv2019

Megahertz X-ray microscopy at X-ray Free-Electron Laser and Synchrotron sources

We demonstrate X-ray phase contrast microscopy performed at the European X-ray Free-Electron Laser sampled at 1.128 MHz rate. We have applied this method to image stochastic processes induced by an optical laser incident on water-filled capillaries with micrometer scale spatial resolution. The generated high speed water jet, cavitation formation and annihilation in water and glass, as well as glass explosions are observed. The comparison between XFEL and previous synchrotron MHz microscopy shows the superior contrast and spatial resolution at the XFEL over the synchrotron. This work opens up new possibilities for the characterization of dynamic stochastic systems on nanosecond to microsecond time scales at megahertz rate with object velocities up to few kilometers per second using X-ray Free-Electron Laser sources.

physics.ins-det↗

arXiv2018

A Shack-Hartmann sensor for single-shot multi-contrast imaging with hard X-rays

An array of compound refractive X-ray lenses (CRL) with 20x20 lenslets, a focal distance of 20 cm and a visibility of 0.93 is presented. It can be used as a Shack-Hartmann sensor for hard X-rays (SHARX) for wavefront sensing and permits for true single-shot multi-contrast imaging the dynamics of materials with a spatial resolution in the micrometer range, sensitivity on nanosized structures and temporal resolution on the microsecond scale. The object's absorption and its induced wavefront shift can be assessed simultaneously together with information from diffraction channels. This enables the imaging of hierarchical materials. In contrast to the established Hartmann sensors the SHARX has an increased flux efficiency through focusing of the beam rather than blocking parts of it. We investigated the spatiotemporal behavior of a cavitation bubble induced by laser pulses. Furthermore, we validated the SHARX by measuring refraction angles of a single diamond CRL, where we obtained an angular resolution better than 4 microrad.

physics.app-ph↗

arXiv2012

Improved success rate and stability for phase retrieval by including randomized overrelaxation in the hybrid input output algorithm

In this paper, we study the success rate of the reconstruction of objects of finite extent given the magnitude of its Fourier transform and its geometrical shape. We demonstrate that the commonly used combination of the hybrid input output and error reduction algorithm is significantly outperformed by an extension of this algorithm based on randomized overrelaxation. In most cases, this extension tremendously enhances the success rate of reconstructions for a fixed number of iterations as compared to reconstructions solely based on the traditional algorithm. The good scaling properties in terms of computational time and memory requirements of the original algorithm are not influenced by this extension.

physics.comp-ph↗

arXiv2012

Charge-density waves in deconfining SU(2) Yang-Mills thermodynamics

At one-loop accuracy we compute, characterize, and discuss the dispersion laws for the three low-momentum branches of propagating longitudinal, electric U(1) fields in the effective theory for the deconfining phase of pure SU(2) Yang-Mills thermodynamics. With an electric-magnetically dual interpretation of SU(2)$_{\tiny CMB}$ we argue that upon a breaking of plasma isotropy and homogeneity, introduced e.g. by a temperature gradient, the longitudinal modes could conspire to provide magnetic seed fields for magneto-hydrodynamical dynamos inside structures of galaxy, galaxy-cluster, and cosmological scales. Such a scenario ultimately links structure with seed-field formation. As judged from the present cosmological epoch, the maximally available coherent field strength of 10$^{-8}$\,Gauss from SU(2)$_{\tiny CMB}$ matches with the upper bound for cosmological present-day field strength derived from small-angle anisotropies of the cosmic microwave background.

hep-th↗

arXiv2010

Improved ground-state estimate by thermal resummation

For the deconfining phase of SU(2) Yang-Mills thermodynamics and for high temperatures we point out that a linear dependence on temperature of a one-loop selfconsistently resummed thermal correction to the pressure and the energy density takes place despite a quartic dependence arising from an unsummed two-loop correction. This linearity is hierarchically smaller than the one belonging to the tree-level estimate of the thermal ground-state. We discuss and interprete this result.

hep-th↗

arXiv2010

Modification of black-body radiance at low temperatures and frequencies

In contrast to earlier reports, where the spectrum of the {\sl energy density} of photonic black-body radiation modified by SU(2) effects was discussed, we discuss the low-frequency spectrum of the {\sl radiance} at temperatures ranging between 5 and 20 Kelvin. We conclude that compared to the conventional theory the only observable effect is associated with the spectral gap (total screening). We also discuss how a low-temperature black body cavity under the influence of a sufficiently strong static electric field is forced to emit according to Planck's radiation law (pure U(1) theory) even at low frequencies and how this effect can be utilized to measure SU(2) induced deviations.

hep-ph↗