Developments in X-Ray Tomography XV

Understanding the biomechanics of the human middle ear remains a challenge, primarily due to the ossicle’s size, accessibility, and subtle movements. In this innovative study, dynamic synchrotron-based X-ray microtomography (dynamic tomography) has been used on acoustically stimulated intact human ears, allowing the three-dimensional visualization of moving eardrums and ossicular chains for the first time. The implementation of a dedicated analysis pipeline has demonstrated the ability to resolve fast micromotions at 128 Hz for two acoustic stimuli (110 and 120 dB SPL) on seven ex-vivo fresh-frozen human temporal bones. Achieving lower sound pressure levels and higher frequency stimulations is challenging: we will discuss in this presentation the current solutions under investigation.

Noise-induced hearing loss is characterized by a cumulative and irreversible loss of hair cells and auditory nerves in the cochlea. To investigate the effects of noise on the inner ear, the physiological information from hearing measurements must be analyzed together with the pathological changes of the inner ear. For this purpose, the cochleas of noise-exposed mice were examined using synchrotron radiation-based microtomography. The visualization of the 3D cochlear reconstruction enables in-depth quantitative analyses of the sensory cells of control and noise-exposed samples.

With the development of a multi-agent staining method for laboratory microCT imaging, the tissue-specific information of standard H&E (hematoxylin & eosin) histology can be transferred to the X-ray field with the advantages of tissue-friendly, non-destructive handling, 3D volume details and multiple use of samples; compared to standard histology and microscopy. By applying a washing step between the stains, the X-ray staining method promises individual information about cell nuclei and cytoplasm, which can point out alterations in tissue due to pathological changes. To test the staining modality in this study, endocrine glands of rats (wild-type and MENX mutated) were used as samples and measured in overview and high-resolution at a laboratory microCT (Zeiss VersaXRM-500 Xradia). The scans showed the tumour formation and nests, changes and destruction of tissue in the diseased samples and distinct anatomical regions in the wild-type glands.

Alzheimer's disease (AD) remains one of the foremost public health challenges of our time. Recently, attention has turned to the gut-brain axis, a complex network of communication between the gastrointestinal tract and the brain, as a potential player in the pathogenesis of AD. Here we exploited X-ray Phase Contrast Tomography to provide an in-depth analysis of the link between the gut condition and AD, exploring gut anatomy and structure in murine models. We conducted a comprehensive analysis by comparing the outcomes in various mouse models of cognitive impairment, including AD, frail mice, and frontotemporal dementia affected mice. We discovered an association between substantial changes in the gut structure and the presence of amyloid-beta (Aβ) in the brain. We found that the most important gut alterations are related to Aβ occurrence in the brain. In particular, we investigated the gut morphology, the distribution of enteric micro-processes and neurons in the ileum.

Computed laminography (CL) is developed and executed to achieve high spatial resolution in three-dimensional imaging of flat, laterally extended objects. This study explores the synergy of synchrotron laminography, X-ray fluorescence (XRF), and efficient data analysis software, aiming to enhance X-ray imaging capabilities at beamline 2-BM and 2-ID-E of the Advanced Photon Source (APS) in Argonne National Laboratory. Unlike the conventional computed tomography (CT) setup, our study leverages the pivotal innovation of laminography to streamline large sample scans by minimizing cutting procedures in this study.The investigation encompasses diverse subjects, including mammalian neural tissues and bones, plant tissues, and integrated circuits, investigated by synchrotron CL and synchrotron XRF techniques with a tilted and rotary stage along the beam direction. The integration of these techniques, facilitated by the advanced synchrotron source, holds the potential for substantial advancements in imaging technology.

Edge-illumination (EI) is an established x-ray phase-contrast imaging method that relies on gratings to obtain attenuation, differential phase and dark field contrast. Current EI setups however have a limited geometric flexibility. That is, the gratings are designed for a fixed magnification and the period and aperture size of the sample grating determine the resolution. To allow multi-resolution EI, a grating with an adaptive period is required. In this paper, we introduce a grating with an adaptive period that allows multiresolution EI, and prove this concept with Monte-carlo simulations of a resolution phantom at different magnifications.

We report on the development of a new x-ray computed tomography facility dedicated to academic and industrial users at University College London, as part of the NXCT (https://nxct.ac.uk/). The system provides x-rays at two different energies, according to the needs of the experiment, by using a Molybdenum or a Copper target. It comprises two imaging stations, one focused on high-resolution imaging (with a mm-sized field-of-view) and one focused on multi-contrast imaging (with a cm-sized field-of-view), although the resolution and contrast methods can be adapted on each side. We also provide several phase and dark-field contrast methods (edge-illumination, beam tracking, free-space propagation) which can be chosen to suit the experiment at hand. The system is designed with flexibility in mind, and it is supported through a free-at-the-point-of-access scheme.

The use of amplitude modulated beams in x-ray micro-CT systems is becoming increasingly popular, since such systems enable x-ray phase contrast imaging (XPCi) and “aperture-driven resolution”. The former provides superior contrast for weakly attenuating samples, while the latter allows increasing the resolution of an imaging system beyond the conventional limit set by the x-ray source and detector. This talk will review the current status of such systems, discuss the opportunities and challenges associated with them, and provide an outlook into what the future might hold, with a particular focus on leveraging and integrating machine learning into micro-CT workflows.

In this study, we explore interior tomography, a technique facilitating the observation of a region-of-interest (ROI) in computerized tomography (CT) through a strategically adjusted detector offset. By modifying the offset, we extend the field-of-view (FOV), consequently enlarging the ROI. Our innovative approach involves offsetting the detector to cover asymmetric regions during data acquisition, overcoming challenges faced by conventional reconstruction algorithms dealing with truncated projection data in interior tomography. To address these issues, we employ a deep learning (DL) network for interior tomography with a detector offset, comparing its performance with other reconstruction methods. Our DL network leverages the weighted filtered back projection (FBP) as input and incorporates the ROI map as additional information, enabling flexible ROI image acquisition within a single network. Trained on abdominal CT projection data, our network exhibits superior performance compared to existing methods. This methodology holds promise for advancing system fusion and miniaturization, particularly in omni-tomography, as it efficiently eliminates noise and artifacts in a shorter time

Interior photon-counting computed tomography (PCCT) scans are essential for obtaining high-resolution images at minimal radiation dose by focusing only on a specific region of interest. However, designing a deep learning model for denoising a PCCT interior scan is rather challenging. Recently, several studies explored deep reinforcement learning (RL)-based models with far fewer parameters than deep supervised and generative models to obtain a generalizable and interpretable model that can be optimized on a small dataset. In this work, we design a deep reinforcement learning model to perform multichannel PCCT scan denoising. Because a reliable reward function is crucial for optimizing the RL model, we focus on designing a small denoising autoencoder-based reward network to learn the latent representation of full-dose simulated PCCT data and use the reconstruction error to quantify the reward. We also use domain-specific batch normalization for unsupervised domain adaptation with limited multichannel PCCT data. Our results show that the proposed model achieves excellent denoising results, with a significant potential for clinical and preclinical PCCT denoising.

This study was aimed at evaluating the effect of single photon emission computed tomography (SPECT) image denoising by using Noise2Void of deep learning method, which does not require prior learning and can recover images using only the target image. To demonstrate usefulness of Noise2Void for SPECT image denoising, simulation using a numerical brain phantom and measurement with physical brain phantom using a high-resolution SPECT system were performed. From projection image data processed by Noise2Void, SPECT images were reconstructed using ordered subset expectation maximization (OS-EM) method. In the simulation, Noise2Void processing improved peak signal-to-noise ratio (PSNR) from 6.69 dB to 20.44 dB and structural similarity (SSIM) from 0.443 to 0.836. Visual evaluation also demonstrated that the noise in the reconstructed image was reduced without blurring and the structure of the phantom became clearer. In an experiment using a physical phantom, Noise2Void also significantly suppressed the noise and clearly depicted the structure of the brain phantom.

A novel method for image restoration is introduced that uses a synthetic prior intermediate (SPI) which is passed through a forward imaging operator, creating a data pair well-structured for inverse operator optimization, of which network training is of particular interest. This technique is applied to a critical problem in X-ray reconstruction: noise and artefact removal. We discuss the creation of the SPI through state-of-the-art Deep Learning reconstruction, a spatially variant heuristic data-driven forward model for spectrally accurate noise and artefact modelling, and final image restoration via a convolutional neural network. Qualitative and quantitative performance is then benchmarked on a range of samples, comparing legacy reconstruction (FDK), state-of-the-art Deep Learning reconstruction, and SPI based reconstruction. SPI based reconstruction better recovers small features while also reducing residual sampling artefacts in large features. Quantitative analysis of SPI reconstruction showed a 40% throughput improvement relative to the state-of-the-art at a constant image quality.

This presentation introduces an X-ray scattering tensor tomography (XSTT) approach tailored to rapidly explore embedded micro-scale structures within composite materials on a centimeter scale. Thanks to advancements in rapid data acquisition and sophisticated reconstruction algorithm, this technique is extremely efficient for centimeter-scale studies of industrially significant fiber-reinforced composites (FRC). The integration of finite element method (FEM) simulations with XSTT data showcases its potential as an efficient tool for computer-aided engineering of FRCs. In addition to the time-steady characterization of FRCs, our pioneering work in tracking time-resolved deformations within viscous fluids containing micro-scale fibers also creates new opportunities for advancing rheological studies. These methodological advancements significantly impact material characterization, offering new perspectives and expanding possibilities in material science, engineering, and practical industrial applications.

X-ray micro/nano-tomography (multiscale CT) has been developed at SPring-8 BL47XU. The X-ray optics consists of a simple projection type imaging system for micro-tomography and a full-field x-ray microscope using a Fresnel zone plate (FZP) for nano-tomography, respectively. All components are mounted on high precision stages driven by stepper motor. It takes two or three minutes to change the mode from “micro” to “nano” or “nano” to “micro”. The spatial resolutions are 1 um and 100 nm for micro and nano tomography. The field of views are 800um and 60um. The measurement time is about 5 minutes for 1800 projections. The energy range is between 7 keV and 15keV.

The human brain contains 86 billion cells, but so far they could not be visualized in their entirety. The task corresponds to a dataset of petabyte size – analogous to plotting every star in the Milky Way. Quite recently, we demonstrated feasibility of cellular-resolution full-brain imaging for ethanol-immersed and paraffin-embedded human brain using the tomography setup at the beamline P07 (Petra III, DESY, Hamburg, Germany). The next challenge involves acquiring and disseminating a human brain atlas that will create a paradigm for investigating other human organs, high-performance engineering devices, and unique cultural heritage objects related to big data.

COSI is a NASA Small Explorer (SMEX) satellite mission for launch in 2027. COSI is a wide-field gamma-ray telescope designed to survey the entire sky at 0.2-5 MeV. It provides imaging, spectroscopy, and polarimetry of astrophysical sources, and its germanium detectors provide excellent energy resolution for emission line measurements. I will overview the COSI mission including the science, technical design, and status. The heart of the COSI instrument is an array of high-resolution cross strip germanium sensors. I will focus deeper on germanium detector developments in support of the COSI mission and ongoing work to optimize the in-flight detector performance.

We developed a detachable mouse table enabling temperature regulation during combined x-ray computed tomography and optical imaging sessions. The 3D-printed table allows seamless transportation of anesthetized mice between imaging modalities while maintaining consistent animal positioning and posture. An integrated carbon nanotube heating pad under the mouse sustainably warms the body to optimal temperature without interfering with x-ray imaging or introducing artifacts. This table optimizes multimodal preclinical imaging workflows and upholds animal welfare standards by facilitating an efficient operational environment and inhibiting hypothermia in anesthetized mice.

This study aims to investigate the potential of X-ray induced acoustic computed tomography (XACT) technology for vascular imaging, specifically in assessing oxygen content, blood flow, and velocity. XACT has shown promising capabilities in high-resolution imaging, and its application in vascular imaging could provide valuable information for diagnosing and monitoring vascular conditions. This research will focus on exploring the feasibility of using XACT to obtain quantitative measurements of oxygenation levels within blood vessels, as well as determining blood flow rates and velocities. By utilizing XACT, non-invasive and real-time assessment of vascular parameters could be achieved, offering a novel approach for studying vascular physiology and improving clinical outcomes. The findings of this study will contribute to the growing field of XACT and its potential applications in vascular imaging.

Combining Neutron and X-ray Tomography (NeXT) into existing neutron facilities is crucial for non-destructive testing and materials science research. By combining these two technologies, a comprehensive understanding of the internal structure of various materials can be achieved. This presentation will include the preliminary design for implementing a NeXT system at the HANARO/ENF facility, aiming to develop a data acquisition system and advanced data fusion algorithms essential for the seamless integration of neutron and X-ray datasets. For the development of the data fusion algorithm, data was acquired from both laboratory-based X-ray sources and the HANARO/ENF facility. The discussion will detail the fusion algorithm developed based on this data and the initial design concept of NeXT. Additionally, the potential and future possibilities of the facility will be explored, focusing on various imaging modes such as quantitative dual-mode imaging using monochromatic beams and quantitative dual-mode interferometry imaging using grating interferometers.

The conventional Talbot-Lau interferometer is capable of obtaining not only absorption images but phase and dark-field images through phase stepping methods. However, this approach requires a considerable amount of acquisition time. To overcome this limitation, a single absorption grating interferometer system has been proposed. Although this system has lower resolution compared to the conventional system, it significantly mitigate the complexity of the system by utilizing only one grating. This simplification alleviates the complexity of parameters and the burden of grating fabrication. By applying different imaging methods, it significantly reduces the acquisition time. To verify the functionality of the single grating interferometer system, simulations were conducted using McXtrace, and comparison analysis between real images and simulation data was performed with the neutron grating interferometer at HANARO. In addition, experiments has been conducted on a simple phantom to assess the feasibility of applying single absorption grating image to tomography. We apply algorithms for tomographic reconstruction and evaluate the image quality of the reconstructed tomography results.

Edge Illumination (EI) X-ray phase contrast imaging (XPCI) is a promising technique that, next to conventional attenuation contrast, provides phase contrast and dark field contrast. Opposed to conventional X-ray imaging though, EI-XPCI is more vulnerable to vibrations, temperature changes and small variations of the source’s focal spot size, leading to intensity changes in the acquired projections. As a result, attenuation, phase, and dark field contrast parameters that are estimated from these projections also suffer from these variations. In this work, through intensive simulations using the CAD-ASTRA toolbox, we study the effect of a changing focal spot size on EI-XPCI parameter estimation. These variations are then modelled by computing eigen flat fields (EFFs). Subsequently the EFFs are used to normalize the flat fields for each phase step of an EI-XPCI acquisition. Results indicate that dynamic flat field correction (FFC) based on EFFs outperforms conventional FFC in EI-XPCI.

The Talbot-Lau grating interferometer represents a significant advancement in X-ray imaging technology, allowing phase contrast, dark-field imaging, and differential phase contrast imaging to be acquired from laboratory-based X-ray sources alongside conventional absorption images. In this study, we explore the ability of directional dark-field imaging (DDFI) to reveal microstructural details within a sample. Utilizing a Talbot-Lau grating interferometer setup, are measured and compared to other methods to demonstrate the practical implementation and efficiency of DDFI in visualizing the degree of anisotropy, orientation information, and microstructural information of various samples. By rotating the sample and analyzing the scattering direction information, we highlight the utility of DDFI in describing complex material features. Our findings highlight the potential of DDFI as a powerful tool in both practical and materials science applications, providing new insights into sample characterization and analysis.

The multi-scale CT developed at BL20XU of SPring-8 consists of three CT systems with different field of view (FOV)/resolutions for nondestructive, high-resolution imaging of internal hierarchical structures in samples. The first is a micro-CT (FOV 1mm / pixel size 0.5 um), which uses a simple projection optics. The second is a wide-field CT (FOV 6 mm / pixel size 3 um), in which a beam diffuser is inserted in the optical path to increase the beam irradiation area to the sample. The third is a nano-CT for high resolution (FOV 60 um, pixel size 30-40 nm), which employs a full-field x-ray microscope using a Fresnel zone plate as the x-ray objective. These measurement modes can be easily changed by users in 1 min without removing the sample from the stage. In order to measure samples with several millimeters in size, the nano-CT is capable of high energy regions of 15 keV - 37.7 keV. In addition, a working distance of approximately 100 mm is provided around the sample, allowing measurement with various test equipment brought in by the user on board, facilitating in-situ observation and operando measurement. The detail of the system and some results will be shown at the conference.

At SPring-8 BL28B2, Hoshino et al. developed a CT measurement technique using high-energy X-rays (200 keV) to observe internal structures of materials. Combining this with a sample exchange robot, high-def X-ray camera, and SP8-DC computer tech, they automated the process from measurement to image reconstruction. The system captures projection images with a maximum field of view of 48mm x 1.2mm, handling larger specimens by repeated scans and stacking CT images. Measuring a sample with the size of 48mm x 10mm takes about 1.5 hours with the effective pixel size of 3.72 um/pixel. Then, transferring the data to SP8-DC is completed within half an hour, and image reconstruction takes approximately 6 hours.

The High Energy Materials Science Beamline HEMS/P07 at PETRA III satisfies high energy X-ray diffraction and imaging techniques. It is tunable in the range 30 to 200 keV. Recently experiments has been commissioned which combinate full field imaging with diffraction contrast tomography. Within this poster the hardware requirements, especially the triggering of the PILATUS CdTe detector via the rotating stage, will be explained. First primary results will demonstrate the new experimental opportunities at the P07 beamline.

Nano Computed Tomography (nanoCT) is a powerful tool for non-destructive threedimensional specimen visualization and investigation at the submicron scale. Here, we present our lens-free lab-based nanoCT-setup, which combines a nanofocus X-ray source with a photon-counting detector. On the one hand, the X-ray source's coherence enables propagation-based phase-contrast which is beneficial for the investigation of low-absorbing specimens, e.g. in virtual 3D-histology of biopsies. On the other hand, the X-ray tube's acceleration voltage range allows for dual-energy computed tomography (CT), a useful tool in material science. We describe the set-up and discuss its optimization, focusing on acquisition speed and field-of-view increase. Furthermore, we show our latest results for dual-energy nanoCT and phase-contrast nanoCT.

The performance of pixelated semiconductor detectors, such as CdTe and CZT sensors, degrades because of multi-pixel events in medical imaging scenario. In this study, we investigate the characteristics of multi-pixel events in the context of XFCT imaging and implement corrections for the energy loss during charge sharing and the interaction location of fluorescence escape and re-capture bi-pixel events. The implemented corrections enhanced the contrast noise ratio and improved the minimum detection limit of XFCT, which is crucial for the preclinical investigation of precise tumor diagnosis and treatment using high atomic number element nanoparticles.

X-ray photon-counting detectors (PCDs) are gaining popularity in medical imaging, yet they face challenges like charge splitting and pulse pileup, which distort the energy spectrum. Based on a Wasserstein generative adversarial network (WGAN) with convolutional neural networks (CNNs) for PCD data correction, we evaluate various CNN architectures and training methods for effectiveness. Trained CNNs from scratch on our dataset outperforms the pre-trained networks. Additionally, we explore transformers for texture feature extraction, highlighting the importance of careful CNN selection and training in the WGAN framework for optimal photon counting CT data correction.

This study presents a novel approach in the construction of a system matrix in X-ray imaging models by combining machine learning with fundamental physical principles through the utilization of physics-informed machine learning (PIML). It incorporates physics-based constraints into the model to guarantee precision and flexibility. Thorough validation demonstrates the feasibility of this approach, offering potential transformative implications in the fields of electronics, medical imaging, and material inspection.

The rapid development of deep learning provides a powerful tool for CT image reconstruction from low-dose, limited-angle and sparse-angle projection data. It is very significant to design an efficient self-supervised deep learning method for CT image reconstruction because it is nearly impossible to obtain ground-truth images in practice. In this paper, we propose a self-supervised learning method for CT image reconstruction based on the so-called view-by-view backprojection tensor (Tao et al 2020), which is an important extension of the classic filtered backprojection (FBP) algorithm for CT image reconstruction. The proposed method can be applied to CT image reconstruction from low-dose, limited-angle and sparse-angle data. Experiments on various data including simulation data and real data demonstrate that the proposed method outperforms the state-of-art methods based on a self-superivised deep learning approach.

Compared to step-and-shoot X-ray scanning, continuous acquisition X-ray scanning is advantageous in terms of scanning speed and mechanical stability. Unfortunately, continuous motion during acquisition leads to blurry projections if conventional image reconstruction methods are applied. In this work, we propose Image Reconstruction for Continuous Acquisition (IRCA), in which the motion during X-ray exposure is modelled via a flexible blurring operator.

A Triple-Source CT (TSCT) architecture featuring three imaging chains has been previously proposed for cardiac imaging with rapid temporal resolution and low radiation dose. However, due to the unconventional triangular sampling trajectory in TSCT, reconstructing high-quality images from the sparse and noisy data acquired in this TSCT is challenging. To overcome the obstacle, we propose a dual-domain network for deep reconstruction in TSCT. The network leverages information from both sinogram and image domains and actively promotes mutual learning between the two domains. Various sparse view sampling protocols were implemented by changing the view angle interval. A conjugate interpolation method was developed to address the unconventional sampling scheme in this TSCT. The interpolated raw data were reconstructed by the proposed dual-domain network method. Preliminary results from cardiac CT imaging of a realistic porcine heart demonstrate that the proposed method can effectively suppress sparse view artifacts and reduce image noise while preserving image details for cardiac diagnosis.

In our correlative characterization studies of biodegradable metal bone implants we have performed both synchrotron-radiation microtomography (SR-µCT) and histology on the same samples. Histological staining is still the gold standard for tissue visualization yet requires multiple lengthy sample preparation steps (fixing, embedding, sectioning then staining) before imaging is performed on individual slices, in contrast to the relatively non-destructive and 3D nature of tomography. In the process of correlating the corresponding data sets, we are able to combine the advantages of both modalities by using machine learning methods to generate artificially stained 3D data from SR-µCT datasets, a so-called ‘virtual histology’. For this we have developed an automated registration tool to find and fit the correct virtual tomographic plane to each histology slice. Preliminary results are promising after training a cycle generative adversarial network on our data, with two different histological stainings.

Imaging anatomical features of the human brain at micrometer resolution currently relies on series of physical sections with related slicing artefacts. This study aims to demonstrate the feasibility of imaging the entire human brain with micrometer resolution without the need for physical sectioning. Computed tomography of a formalin-fixated and ethanol-immersed human brain was performed at the P07 beamline, operated by Hereon, at DESY in Hamburg. The 16× extended field of view necessary to cover the 10 cm wide specimen with 2.54 µm voxels was realized acquiring eight rings, each with 48’000 projections along 360 degrees. The projections were stitched and superimposed before tomographic reconstruction. In a next step, several 10’000 slices, each several GB in size, must be tiled vertically to cover the entire brain. This synchrotron radiation-based study with 67 keV photons shows the feasibility of employing X-ray tomography to image the entire human brain using isotropic micrometer-sized voxels.

The human eye’s cornea is vital for visual clarity and quality of life. This study employs contrast-enhanced micro-CT to analyze the ultrastructure of fresh human corneas, using both phase-contrast and absorption micro-CT techniques with and without contrast enhancement. Human corneas, prepared in various hydrated states, underwent contrast enhancement with Lugol’s iodine and were imaged using micro-CT. The results allowed detailed ultrastructural analysis, revealing the spatial distribution of cells within corneal layers and allowing quantitative measurement of corneal hydration. The phase-contrast technique, enhanced with Lugol’s iodine, improved image detail. The method, utilizing Lugol’s iodine as a contrast agent, demonstrates the promising potential for virtual histology, offering detailed insights into corneal ultrastructure.

The anatomy of the human facial nerve (FN) is well described. The New Zealand rabbit is a broadly accepted animal model in translational neurosurgery, e.g. regeneration and grafting studies. Yet macro- and microanatomy are only insufficiently deciphered and limited to histological studies. Microtomography (µCT) provides complementary information through isotropic 3D characterization with true micrometer resolution. Four FN pairs of the New Zealand rabbit are embedded in 4% formaldehyde and ascending ethanol solution series. To increase soft tissue contrast, temporal bone is decalcified with 20% EDTA solution. Imaging is performed with nanotom®m (phoenix|x-ray, Waygate Technologies) and Xradia 620 Versa (Zeiss, Oberkochen, Germany), tomography setups at beamline P07, PETRA III, DESY (Hamburg, Germany) and at beamline ID19, ESRF (Grenoble, France). The study will shed light on controversially discussed 3D arrangement of FN fibers, provide comprehensive insight into connectivity of FN to other structures and will be valuable in investigating pathologies.

We present an anatomical description of the soft tissues of the cochlea of harbour porpoise (Phocoena phocoena). The samples from two individuals were analysed at high spatial resolution (2 μm voxel size) showing normal and altered anatomy, based on synchrotron phase-contrast micro-computed tomography (DESY/HEREON, Hamburg).

The Compton camera is emerging as an attractive approach for directly mapping the distribution of radiopharmaceuticals inside a patient without mechanical collimation. In this study, we design a high-efficiency and high-quality tomographic imaging system with a Compton camera for thyroid cancer imaging and develop a dedicated algorithm for Compton scattering-based SPECT imaging. Due to the inconvenience in using gamma ray sources, our system employs an x-ray source to emulate gamma rays since there is a substantial overlap between X-ray and gamma-ray spectra. Our system utilizes a Timepix-3-based Compton Camera, and metal samples are used to generate scattered photos from a laboratory X-ray source. Then, experimental data are reconstructed using analytic and deep learning methods respectively. The reconstruction results demonstrate that the proposed technology works satisfactorily, attaining results comparable to standard SPECT while significantly lowering the radiation dose. Further work is underway to capitalize its potential in the clinical translation.

This study examines the scope of non-destructive testing methods, specifically Scanning Acoustic Microscopy (SAM) and 3D X-ray imaging, for inspecting advanced 3D and 2.5D integrated circuit (IC) packaging. As the semiconductor industry evolves towards more complex packaging technologies for improved performance and cost reduction, the reliability of these structures is crucial. The paper reviews current non-destructive techniques for physical inspection, characterization, and failure analysis, focusing on SAM and 3D X-ray imaging amidst challenges of miniaturization and densely packed components. Through a case study, it evaluates these methods' effectiveness, comparing speed, accuracy, resolution, and analysis capabilities. It underscores the importance of non-destructive testing in enhancing semiconductor packaging and suggests future advancements in imaging technology and AI for better defect detection.

This paper introduces an automated, non-destructive metrology verification workflow using X-ray technology for inspecting fine-pitch interconnects and other features in advanced semiconductor packages. It includes high-resolution X-ray data acquisition, pre-processing for enhanced feature visibility, and the use of a machine learning model for feature segmentation. The model, trained on annotated datasets, can automatically recognize and segment critical IC package features like vias and micro-bumps. The approach culminates in metrology and layout validation against design specifications, offering a scalable and efficient solution for ensuring the integrity and precision of semiconductor fabrication processes.

Ensuring reliability of TSVs and micro-bumps in 2.5D and 3D presents automatically challenges due to current failure analysis constraints. This study utilizes fast scanning acoustic microscopy (SAM) spectra for automatic defect detection. High-resolution X-ray will help in annotating the SAM data, and a machine learning model will be developed for the automatic anomaly detection.

This abstract presents a novel approach to addressing the challenges in failure analysis and quality control for heterogeneously integrated devices. The novelty lies in the justification of use through the introduction of case studies showcasing the benefits of this analysis and of two methodologies, early fusion, and decision-level fusion.

In high-resolution imaging of soft tissue, refraction of X-rays provides significantly more signal than attenuation. With a grating interferometer, the refraction of X-rays generated with a conventional X-ray tube can be detected, making the technique applicable in clinical diagnostics. We pursue its application to the imaging of the breast: a soft-tissue organ imaged at high resolution. We developed two grating-interferometry CT systems. We used the first, an ex-vivo scanner, to image five fresh tumorectomy specimens of human breast tissue up to 10 cm in size. The subjective image quality was assessed in a reader study. The second is a recently commissioned prototype of a rotating-gantry clinical breast CT. We characterised its performance, for both attenuation and refraction, in terms of MTF, TTF, NPS and the dose. The results show that the refraction signal in grating-interferometry CT enhances delineation of the soft tissue in the breast and that the technique can be translated into a medical device.

X-ray phase-contrast tomography (XPCT) offers a highly sensitive 3D imaging approach to investigate different disease-relevant networks from the single cell to the whole organ. We present here a concomitant study of the evolution of tissue damage and inflammation in potential target organs of the disease in the murine model of multiple sclerosis. XPCT identifies and monitors structural and cellular alterations throughout the central nervous system, but also in the gut and eye, of mice induced to develop multiple sclerosis-like disease and sacrificed at pre-symptomatic and symptomatic time points. This approach rests on a multiscale analysis to detect early appearance of imaging indicators potentially acting as biomarkers predictive of the disease. The longitudinal data permit an original evaluation of the sequential evolution of multi-organ damage in the mouse model, shedding light on the role of the gut-brain axis in the disease initiation and progression, of relevance for the human case.

Micro-CT, widely used for analyzing hard tissues like bone, faces challenges when inspecting soft tissues due to low contrast. Here, we introduce a method to enhance the contrast between adjacent soft and hard tissues by reducing hard tissue attenuation through decalcification and by increasing soft tissue attenuation through contrast enhancement. By exploring contrast-enhanced micro-CT utilizing Lugol’s iodine, iodine, and phosphotungstic acid (PTA) in ethanol for cellular components in bone on 28-day-old B6 mice femur samples and murine mandibles, Lugol’s iodine displayed high specificity to cellular structures. This resulted in outstanding contrasts of cellular components due to low binding to prior decalcified bone or cartilage. This approach, employing both phase-contrast and absorption micro-CT techniques, holds promise for detailed investigations into the cellular details of bone tissues, offering potential advancements in its understanding of bone and its soft tissue components.

Aqueous zinc metal batteries (ZMBs) have great potential for grid energy storage applications. However, the cycle life is closely connected to the reversibility of Zn-metal plating and stripping and it is central to follow these processes in real time to advance the technology. We demonstrate how operando synchrotron X-Ray Tomographic Microscopy (XTM) can be used to visualize zinc deposition in 3D during cell operation. From the 3D renderings reveals different structures of zinc deposits and their evolution during cycling. In addition, we can quantitatively follow the volume of zinc deposits which can be directly correlated to the electrochemical data. The results of this work provide insight into real time processes in the cell and the method is applicable also to other battery chemistries.

The liquid-metal-jet microfocus source presently provides a 5-25 µm x-ray spot with up to 1 kW e-beam power. This source enables several biomedical imaging applications where high spatial resolution, high contrast and short exposure time are critical. Examples: Cellular-resolution phase-contrast imaging for virtual x-ray histology as well as for clinical resection margin assessment, potentially with rapid intraoperative feedback. Preclinical in-vivo phase-contrast small-animal lung imaging and a path towards clinical phase-contrast medical imaging. Finally, x-ray fluorescence tomography for high-resolution molecular detection of tumors in live mice.

Inline X-ray phase tomography has proven to be one of the most suitable imaging techniques for the three-dimensional examination of soft tissue at the microscopic level. Due to the specific requirements, such as the coherence of the beam, the method was limited to synchrotron radiation. The latest developments in detectors (optical magnification) and X-ray sources (source size, liquid metal source) make it possible to transfer the technology to the laboratory environment. For investigations of the coroid plexus, we visualized entire mouse brains in ethanol by Polaris (Exciscope, Kista, Sweden), equipped with a MetalJet X-ray source from Exillum. For the investigation of the human brain on the sub-cellular level with voxel sizes of 300 nm, we used human brain punches of 1 mm embedded in paraffin by Xradia 610 Versa (Zeiss, Oberkochen, Germany).

In tomography, the ratio between field-of-view and pixel size is limited. For many applications it would be beneficial to increase the field-of-view while maintaining the high resolution, which can be done by tiling projections. This has been demonstrated on synchrotron data acquired in parallel-beam geometry. To our knowledge, fully automated methods for projection tiling have not yet been demonstrated in a laboratory instrument with cone-beam geometry. In this presentation, we will walk through our ongoing implementation of projection tiling at the Exciscope Polaris. This will include both acquisition strategies and our cloud-based data pipeline, where large resources are available on demand.

A rotating-anode x-ray source and custom-built sCMOS-based detector have been integrated into a lab-based micro-CT system capable of CT acquisition in as little as 0.13 s. This has been used to examine in 4D the expansion of a polymer foam with a temporal resolution of 2 Hz. The system is easily adapted to carry out fast phase-sensitive multi-contrast CT with sub-10 s CT acquisition times. This is made possible through the beam-tracking technique, which is capable of multi-contrast CT using only a single-shot per projection angle, while also being compatible with incoherent sources. This opens the door to dynamic, phase-sensitive, multi-contrast micro-CT in the home laboratory.

High-resolution synchrotron X-ray CT in situ pull-out tests with stepwise increased loading were performed to investigate the force transfer between a shape memory alloy wire and the surrounding epoxy polymer matrix. The advancing interfacial failure was observed. The stochastic surface structure of the SMA wire was utilized to determine the axial and radial strains introduced into the SMA wire during the test by performing digital volume correlation on the reconstructed surface.

Outsourcing semiconductor and PCB fabrication is a key problem, necessitating the establishment of quality assurance systems for interconnects and semiconductors supplied globally. Automatic bill of materials (BOM) generation and netlist extraction are standard PCB assurance methods. In this study, our main focus includes netlist extraction and addressing the challenges faced during the X-ray CT process that can be mitigated with the addition of optical features. In addition, we plan to evaluate state-of-the-art multi- modal machine learning techniques and feature correlations across imaging modalities to determine their usefulness for PCB assurance purposes.

The Non-Clinical Tomography Users Research Network (NoCTURN) is working to advance Findability, Accessibility, Interoperability, and Reuse (FAIR) and Open Science (OS) practices in the domain of tomographic imaging. By leveraging input from a broad community of tomography educators, researchers, and industry stakeholders, NoCTURN is engaging the international scientific tomographic community to stimulate improvements for tomographic imaging standards that focus on FAIR and OS principles. We aim to reduce the barriers to entry that isolate individuals and research labs, and we anticipate that developing community standards and improving methodological reporting will enable long-term, systemic changes necessary to aid those at all levels of experience in furthering their access to and use of tomographic imaging.

Shark vertebral bodies (centra) possess remarkable resistance to millions of cycles of large in vivo strains exceeding 4-8%. These strains are enormous for a mineralized tissue, and it appears that the centra evolved to achieve this performance through a hierarchy of structures spanning dimensions from centimeters to nanometers. At the 1 µm scale, blocks cut from centra and imaged with synchrotron microCT demonstrate that the centra tissue consists of closely spaced, mineralized trabeculae. An outstanding question is: How do these trabeculae deform to accommodate these large strains. This talk presents recently obtained microCT results on in situ loading of blocks of shark centra and examines the deformation modes of the interconnected array of trabeculae.

X-ray computed laminographic tomography (CLT) is a viable tool for high-throughput volumetric imaging of large, planar samples. Compared to classic X-ray computed tomography (CT) systems, CLT scans objects from a limited angular range, enabling significantly higher imaging throughput for applications such as rapid inline wafer-level 3D packaging inspection for integrated circuits (IC). In this work, we present a self-supervised deep image restoration workflow to produce noise-free, artifact-free volumetric reconstructions. The core of our pipeline is an unbiased progressive artifact removal algorithm that effectively reduces laminographic artifacts. We demonstrate our CLT method on a variety of samples scanned with an in-house prototype system, including test samples provided by Imec R&D, which feature various arrays of Through-Silicon Vias (TSVs) with different population densities and deliberately programmed voids of different extents, as well as several commercially available stacked memory chips. Our proposed method notably outperforms classic reconstruction methods, providing a feasible solution for both IC inline inspection and rapid failure analysis.

Our research pioneers a novel application of deep learning in X-ray tomography, addressing the challenge of detecting nanometer-scale defects within large volumes. We focus on solid-state batteries, where the trade-off between volume and resolution is critical. By training deep neural networks to reconstruct large volume datasets at 400 nm resolution, we reveal sub-micron defects previously obscured by lower resolution scans. These discoveries are meticulously validated using a femtosecond laser for sample cutting and subsequent SEM imaging. Our work not only advances defect analysis but also offers a non-destructive path to estimate crucial material properties, impacting diverse applications in materials science and engineering.

Quantitative comparison of mineral in rodent incisors requires segmentation of relevant tissues, which can be accomplished quickly and accurately with convolutional neural networks. Here we describe a protocol to adapt base networks to new data, thereby creating segmentation tools broadly useful for the diverse datasets necessary to compare rodent models. Specifically, we used μCT images collected for 18 mouse lines from both synchrotron and laboratory X-ray sources with a variety of voxel dimensions and image artifacts. We demonstrate the success of our adapted networks and show the results of a subsequent data processing pipeline for quantitatively comparing stages of mineral formation in the incisor. We envision enamel researchers adapting base networks with modest amounts of their own data, following the protocol shown here, to suit their needs and rapidly extract quantitative measures. The insights gained from these methods will greatly contribute to our understanding of pathologies in dental tissues.

gVirtualXray (gVXR) is an open-source framework that relies on the Beer–Lambert law to simulate X-ray images in realtime on a graphics processor unit (GPU) using triangular meshes. A wide range of programming languages is supported (C/C++, Python, R, Ruby, Tcl, C#, Java, and GNU Octave). Simulations generated with gVXR have been benchmarked with clinically realistic phantoms (i.e. complex phantoms) using Monte Carlo simulations, real radiographs and real digitally reconstructed radiographs (DRRs). It has been used in a wide range of applications, including real-time medical simulators, proposing a new densitometric radiographic modality in clinical imaging, studying noise removal techniques in fluoroscopy, teaching particle physics and X-ray imaging to undergraduate students in engineering, and X-ray CT to masters students, predicting image quality and artefacts in material science, etc. gVXR has also be used to produce a high number of realistic simulated images in optimisation problems and to train machine learning algorithms.

Phase-contrast computed tomography enables the visualization of weakly-absorbing samples with high contrast. Speckle-based imaging (SBI) is a phase-sensitive X-ray imaging technique that requires the use of a wavefront marker (e.g., sandpaper) to retrieve multi-modal information: absorption, refraction and scattering. These quantities are derived by analyzing the distortions in a reference pattern generated when the sample is inserted into the beam. Unified Modulated Pattern Analysis (UMPA) is a speckle-tracking method capable of processing such datasets. While high-resolution tomographic reconstructions can be achieved at the synchrotron, there is usually a trade-off with sample dimensions. Here, we use UMPA with a multi-frame approach for signal retrieval, enabling the expansion of the reconstructed FOV by moving the sample instead of the diffuser transversely to the beam. We demonstrate this technique on a human placenta tissue sample.

This work proposes the application of Deep Image Prior (DIP) to reduce undersampling reconstruction artifacts arising from sparse-view and restricted opening angle acquisition in X-ray Computed Tomography (CT) systems tailored for large-scale cargo scanning in harbors. Specifically targeting a static, rectangular multi-source X-ray CT system, DIP leverages a Convolutional Neural Network (CNN) architecture, to produce high-quality reconstructions without the need for a training dataset. Comparative analysis against FDK, SIRT, and BB-TV on pixel-wise metrics (PSNR, SSIM) and perceptual image quality (FSIM) demonstrates DIP’s superiority, promising enhanced image reconstruction quality in sparse-view cargo scanning.

The Helmholtz-Zentrum Hereon is operating imaging beamlines for X-ray tomography (P05 IBL, P07 HEMS) for academic and industrial users at the synchrotron radiation source PETRA III at DESY in Hamburg, Germany. The high flux density and coherence of synchrotron radiation enables high-resolution in situ/operando/vivo tomography experiments and provides phase contrast, respectively. Large amounts of 3D/4D data are collected that are difficult to process and analyze. Here, we report on the application of machine learning for the reconstruction and analysis of synchrotron-radiation tomography data. Applications include instance and semantic segmentation, a guided and interactive segmentation framework with a human in the loop, virtual histo-tomography for multimodal data analysis, denoising, digital volume correlation, and self-supervised learning for phase retrieval.

The Helmholtz-Zentrum Hereon, Germany, is operating the user experiments for microtomography at the beamlines P05 and P07 using synchrotron radiation produced in the storage ring PETRA III at DESY, Hamburg, Germany. Attenuation-contrast and phase-contrast techniques were established to provide an imaging tool for applications in biology, medical science and materials science. In the recent years we rebuilt the data preprocessing pipeline before reconstruction to provide different scanning techniques to investigate samples larger than the field of view of the X-ray beam. Within this talk the hardware requirements and calibration used at the imaging stations will be given. Furthermore, we adjust the preprocessing pipeline to deal with different mechanical accuracies of the translation / rotation stages used for addressing the full volume of the sample. Several examples using low photon energies at P05 and high photon energies at P07 will demonstrate the new stitching pipeline.

The acquisition of large tomography volumes, exceeding the typical detector field of view, requires advanced acquisition techniques. Current approaches are the tiling of local reconstructed volumes or the tiling in projection space, also known as mosaic tomography. In this work we propose a third, hybrid approach, to profit from the speed and dose efficiency of projection tiling, but relaxing the requirements for mechanical and beam stability. The volume to be imaged is covered by overlapping cylinders, each corresponding to the reconstructed volume of one mosaic tomogram. The image registration approach for stitching of three-dimensional tiles was modified to force a consistent solution. We demonstrate hybrid tiling for a 2 cm section from the brain stem at 0.65 µm voxelsize. The results of the hybrid acquisition in seven tiles with four rings each are compared to a pure projection tiling approach with eight rings and to local regions representing reconstruction tiling.

Solving the inverse tomographic problem in X-ray CT on the tetrahedra of an adaptive volume mesh enables a memory efficient multi-resolution representation of an object. Furthermore, in contrast to voxel-based reconstruction methods, tetrahedral reconstruction has the potential to avoid partial volume effects. In this work, we focus on methods to optimize vertex positioning and mesh quality while reconstructing attenuation values with the SIRT algorithm, GPU accelerated by CAD-ASTRA (Paramonov et al. Optics Express 32(3), 2024).

On-the-fly 3D data reconstruction is a challenging need in synchrotron micro-tomography facilities. This presentation describes the data workflow and infrastructure implemented at the FaXToR beamline of the ALBA Spanish synchrotron to follow dynamics inside the samples. The beamline is expected to present a high data throughput making use of state-of-the art CMOS fast detectors. Therefore, a particular care is required in order to cope with the computing requirements. The IT infrastructure will accomplish 3D data reconstruction, redundant distributed processing and PB data storage.

Understanding the fluid dynamics of the central nervous system (CNS) is important for developing therapies for neurological disorders which alter cerebrospinal fluid (CSF) dynamics. Based on unique time-resolved in vivo microtomography, we investigated the movement of CSF spaces in the mouse CNS. Anesthetized mice were placed in a heated holder to minimize head motion and maintain body temperature. Contrast agent was injected into the ventricle to enhance visibility of the CSF space. Projections were retrospectively sorted according to simultaneously acquired ECG recording. The 3D+time image sequences were automatically analyzed for local, short motion patterns. Periodic motion of 16 µm was observed in the anterior nasopharynx region. So far, local ventricle motion was not detected at 6.45 µm voxel resolution. Future higher-resolution local scans will be used to investigate the limits of in vivo ventricle motion magnitudes in mice.

In this study, we explore the capabilities of an innovative technique for three-dimensional micrometer-scale analysis across biomedical and materials science applications, leveraging Talbot Array Illuminators (TAI) as an optical element in combination with Unified Modulated Pattern Analysis (UMPA) as a phase-retrieval algorithm. This approach enables high-precision phase retrieval and exhibits high sensitivity without prior assumptions on the sample composition, making it versatile across a broad spectrum of sample types with diverse material properties. We demonstrate the application of this methodology to various types of specimens at the PETRA III beamline P07, operated by Hereon, at DESY, Hamburg. Additionally, we present methods for improving image quality using a hanging-axis approach combined with an ethanol bath in which the sample is immersed. Our findings underscore the technique's potential in advancing imaging and analysis on the micrometer scale.

Hierarchical Phase-Contrast Tomography (HiP-CT), utilizing the European Synchrotron Radiation Facility's Extremely Brilliant Source, is a novel synchrotron X-ray computed tomography technique enabling non-destructive, high-resolution imaging of entire human organs down to cellular scale. This method can scan whole human organs with a resolution of 20 μm per voxel, and subsequent zooms at 1um in local areas. It bridges the gap between global clinical imaging and detailed histology, without the need for physical sectioning. HiP-CT has successfully imaged various organs such as lungs, hearts, and brains, contributing to disease research, notably on COVID-19 lung damage and cardiac pathologies. This data is freely available on the Human Organ Atlas website.

The APS Upgrade project (APS-U) is set to revolutionize hard X-ray research, providing significantly improved coherence and brightness (500-fold increase) in X-ray beams. Managed by the Imaging Group at APS, three specialized beamlines—2-BM, 7-BM, and 32-ID—are dedicated to full field imaging and ultra-high-speed applications. This range of instruments allows the group to cover three orders of magnitude in spatial and temporal resolution. Following the APS-U project, the Imaging Group at APS will provide a diverse array of instruments and techniques to the user community, facilitating versatile multi-scale and multi-modal approaches. An overview of the expected capabilities and instrumentation will be presented, ensuring a comprehensive understanding of the opportunities available for researchers.

Simultaneous neutron and X-ray tomography (NeXT) leverages the complementarity in contrast of the two modalities for improved material identification, especially in dynamic systems. A challenge of this multimodal method is the long acquisition times required due to the low brightness of neutron sources. The acquisition time for a 10 µm spatial resolution neutron image can be 60 seconds or longer. This challenge presents an opportunity for novel scan methodologies for the X-ray system such as dual-energy, Ross bandpass filter, and single-shot region-of-interest (ROI). This talk will discuss instrument upgrades to the NIST-NeXT system that will facilitate these new X-ray scan schemes, what the benefits of combining these new X-ray scan scheme with neutron tomography are, and what the data storage, reconstruction, and segmentation challenges are for working with higher dimensionality tomography data and large, 62 or 102 megapixel, camera that would push towards 1 TB sized datasets.

Measuring the oxygenation distribution at a high spatial resolution in bone marrow is important because it helps us in understanding its roles on stem cell proliferation and differentiation inside the bone marrow. Current technologies have limitations in imaging the oxygenation of deep targets. To overcome these limitations, X-ray Luminescence Computed Tomography (XLCT) emerges as a promising method to image the oxygenation of bone marrow at a spatial resolution close to the focused x-ray beam size, which is better than 50 micrometers. We have built a four channel XLCT imaging system in our lab to verify the proposed method.

This study evaluates the impact of charge summing correction (CSC) on a Cadmium Telluride (CdTe) Photon Counting Detector in breast CT. A laboratory benchtop system that consists of a 0.1 mm pixel pitch CdTe detector and a tungsten anode X-ray source. Images were acquired at 55 kVp with 2 mm Al external filtration under three different tube currents: 25, 100, and 200 mA. Performance was assessed using contrast to noise ratio (CNR), modulation transfer function (MTF), noise power spectrum (NPS), and iodine quantification. Anticoincidence (AC) and single pixel (SP) modes were compared, both with signal-to-thickness calibration and FDK reconstruction. AC mode showed enhanced low-energy contrast and accurate iodine quantification, while SP mode had better CNR at low-energy. High X-ray fluence reduced AC mode uniformity, but not SP. Results suggest that CSC in breast CT improves iodine quantification but at the cost of increased noise in low-energy images. These effects are dependent on the studied system and operational parameters.

Recently, we developed a full-field multimodal imaging technique of X-ray transmission, fluorescence and scattering tomography with a polychromatic X-ray source. A new three-mode CT imaging system was built for transmission, fluorescence, and scattering, which consists of a conventional polychromatic X-ray tube, a pinhole collimator, a spectral photon-counting detector with high energy resolution (~1keV@60keV) for scattering and fluorescence tomography and an energy-integrating flat-panel detector with high spatial resolution for transmission tomography. We called it full X-ray particle information CT (PI-CT) technique. A new algorithm was proposed to simultaneously reconstruct the full-field linear attenuation coefficients with high spatial resolution at different energies and the quantitative images of both High-Z elements concentrations and electron density distributions. Using the multifunctional medicines loaded with nanoparticles for diagnosis and treatment, a visualized precise radiation enhanced therapy scheme of keV X-ray beam can be developed. Experimental results on PI-CT will be presented in the applications of multimodal CT imaging and tumor diagnosis and treatment.

When an intense laser interacts with a gas of atoms, high-order harmonics are generated. In the time domain, this radiation forms a train of extremely short light pulses, of the order of 100 attoseconds. Attosecond pulses allow the study of the dynamics of electrons in atoms and molecules, using pump-probe techniques. This presentation will highlight some of the key steps of the field of attosecond science.

Integrating physiology and structure at the neuronal circuit scale can provide a mechanistic understanding on how that circuit works. The glomerular columns in the mouse olfactory bulb contain the first synapse of the olfactory sensory pathway, through a circuit that is compact, modular and accessible to optophysiology setups. A correlative multimodal imaging pipeline that combines in vivo 2-photon microscopy and synchrotron X-ray computed tomography with propagation-based phase contrast provides a robust and versatile approach to identify all neurons imaged in vivo in a multi-mm3 resin-embedded sample of brain tissue. Follow-up targeted imaging is possible with either X-ray nanoholotomography or volume EM, and doing so becomes simplified by milling the sample using a femtosecond laser. Altogether, this approach enables harnessing the resolving power of multiphoton, hard X-ray and volume electron microscopy technologies to create detailed multiscale, multimodal maps of brain circuits.

Imaging natural history collections is becoming a classic conservation tool that also serves research purposes. So far, microtomography is limited to rare and important specimens such as types or reference specimens, but the data generated takes conservation to another level. This is because not only coarse surface features, but also very fine texture and internal structures become digitally available, depicting the object in almost all its complexity and dimension. Generating this kind of data provides first-hand scientific data and limits further handling of sometimes fragile specimens. Fossil species are often the best preserved and are witnesses of past life on our planet. They provide information on how today's biodiversity has evolved; they preserve in their physical structure morphological parameters related to past environmental conditions and are thus a good indicator for the reconstruction of past climates. In addition, they often fascinate a wide audience and are therefore good ambassadors for communicating scientific findings. Recording them with the help of X-ray microtomography should therefore be a general goal, which we illustrate here with examples.

This paper summarises recent progress in the field of X-ray-based three-dimensional imaging related to clinical, laboratory-based and synchrotron radiation-based instrumentation, algorithms for big data, image reconstruction and artefact removal, and the wide variety of applications including materials and plant sciences, pathology, and anthropology. Multi-modal imaging, which includes synergistic and reciprocal information, is now common and requires interdisciplinary approaches of researchers and users from medicine/dentistry, biology, earth and materials science, crystallography, solid-state and soft-matter physics, chemistry, computer science, engineering, and applied mathematics.

Born in 1928, the German physicist Ulrich Bonse had a challenging time as pupil during and after World War II. He had to serve as Luftwaffenhelfer (anti-aircraft auxiliary). Starting in 1949, he studied physics at the University of Münster, Germany. Under the supervision of Eugen Kappler, Ulrich Bonse developed an X-ray-based method to experimentally determine strain fields of defects in silicon and germanium single crystalline materials - a timely research topic closely related to the invention of the transistor. His awarded PhD-thesis was internationally recognised. Thus, he was invited to US in 1961 and became guest professor at Cornell from 1963 to 1965, where he developed together Michael Hard the first X-ray interferometer, a development awarded by the German Physical Society. So it is no surprise that in 1970 Ulrich Bonse became Professor in Physics and the Founding Director of X-ray and neutron imaging research on the Angstrom scale at the University of Dortmund, Germany.

Ulrich Bonse had an enormous impact during the early days of x-ray microComputed Tomography (microCT), both with absorption contrast and with phase contrast. This talk focuses on absorption contrast microCT and the early trajectory of the field, in particular, the use of area detectors and others’ work occurring at about the same time. In addition to technical advances, Bonse and coworkers produced some of the first microCT studies of several different types of engineering materials and anatomical structures. Bonse also helped foster the growth of the field of non-clinical tomography by establishing the present conference and chairing the first five meetings. Summaries of early conferences are presented as well as a perspective on the subsequent ten meetings.

The development of full-field imaging at synchrotron radiation sources in Europe was driven by the scientific research of Ulrich Bonse together with many students from the Department of Experimental Physics I at the University of Dortmund, Germany. Using bending magnets at the storage ring DORIS III at DESY, the first full-field tomograms were conducted in parasitic mode. With the dedicated operation of DORIS for synchrotron radiation user experiments and the integration of a bypass to DORIS to integrate wiggler insertion devices the development attenuation- and phase-contrast SRµCT as a research tool for 3D investigation of biological and materials science samples became possible. Furthermore, the group of Ulrich Bonse used this knowledge as first non-official user of the ESRF to acquire tomograms of trabecular bone at the beamline ID6 during test phase for insertion devices. Within this talk the experimental developments form hardware to software for running full-field imaging from DORIS via ESRF to PETRA III based on the work of Ulrich Bonse will be elucidated.

Ulrich Bonse's contributions to the X-ray imaging field are instrumental in several aspects. The talk starts with his original work on X-ray interferometry with Michael Hart. This breakthrough enabled the subsequent development of X-ray phase-contrast imaging methods. An emphasis is placed on tomographic algorithms which go from classic methods either analytic or iterative to deep learning-based methods for phase-retrieval and image reconstruction. Future directions are also discussed.

This presentation is personal appreciation of Professor Bonse as an individual. In 1997.he emailed me to ask me to join a new effort to explore the design and evaluate novel micro-CT systems. At his invitation I joined him and his co-workers at the (to be) first SPIE conference on X-ray Tomography. Although his academic background was at the basic physics level and my background was at the macroscopic level of interaction of organ structure to function. He mentored many upcoming investigators and championed broadening the application of x-ray. There is no doubt that he was in charge but he was very sociable and hospitable as illustrated by the evening get-togethers after the end of the day’s session. He planned coming events with great care in order to not be blind-sided. He was an avid supporter of the Dortmund football team. Although obsessed with minor details he could admit to a joke on himself. He spent fruitless hours explaining to me the intricacies of the German language.