A convolutional attention mapping deep neural network for classification and localization of cardiomegaly on chest X-rays.

Building a reliable and precise model for disease classification and identifying abnormal sites can provide physicians assistance in their decision-making process. Deep learning based image analysis is a promising technique for enriching the decision making process, and accordingly strengthening patient care. This work presents a convolutional attention mapping deep learning model, Cardio-XAttentionNet, to classify and localize cardiomegaly effectively. We revisit the global average pooling (GAP) system and add a weighting term to develop a light and effective Attention Mapping Mechanism (AMM). The model enables the classification of cardiomegaly from chest X-rays through image-level classification and pixel-level localization only from image-level labels. We leverage some of the advanced ConvNet architectures as a backbone-model of the proposed attention mapping network to build Cardio-XAttentionNet. The proposed model is trained on ChestX-Ray14, which is a publicly accessible chest X-ray dataset. The best single model achieves an overall precision, recall, F-1 measure and area under curve (AUC) scores of 0.87, 0.85, 0.86 and 0.89, respectively, for the classification of the cardiomegaly. The results also demonstrate that the Cardio-XAttentionNet model well captures the cardiomegaly class information at image-level as well as localization at pixel-level on chest x-rays. A comparative analysis between the proposed AMM and existing GAP based models shows that the proposed model achieves a state-of-the-art performance on this dataset for cardiomegaly detection using a single model.

Towards More Structure: Comparing TNM Staging Completeness and Processing Time of Text-Based Reports versus Fully Segmented and Annotated PET/CT Data of Non-Small-Cell Lung Cancer.

Results of PET/CT examinations are communicated as text-based reports which are frequently not fully structured. Incomplete or missing staging information can be a significant source of staging and treatment errors. We compared standard text-based reports to a manual full 3D-segmentation-based approach with respect to TNM completeness and processing time. TNM information was extracted retrospectively from 395 reports. Moreover, the RIS time stamps of these reports were analyzed. 2995 lesions using a set of 41 classification labels (TNM features + location) were manually segmented on the corresponding image data. Information content and processing time of reports and segmentations were compared using descriptive statistics and modelling. The TNM/UICC stage was mentioned explicitly in only 6% (n=22) of the text-based reports. In 22% (n=86), information was incomplete, most frequently affecting T stage (19%, n=74), followed by N stage (6%, n=22) and M stage (2%, n=9). Full NSCLC-lesion segmentation required a median time of 13.3 min, while the median of the shortest estimator of the text-based reporting time (R1) was 18.1 min (p=0.01). Tumor stage (UICC I/II: 5.2 min, UICC III/IV: 20.3 min, p < 0.001), lesion size (p < 0.001), and lesion count (n=1: 4.4 min, n=12: 37.2 min, p < 0.001) correlated significantly with the segmentation time, but not with the estimators of text-based reporting time. Numerous text-based reports are lacking staging information. A segmentation-based reporting approach tailored to the staging task improves report quality with manageable processing time and helps to avoid erroneous therapy decisions based on incomplete reports. Furthermore, segmented data may be used for multimedia enhancement and automatization.

Phased Array Ultrasound System for Planar Flow Mapping in Liquid Metals.

Controllable magnetic fields can be used to optimize flows in technical and industrial processes involving liquid metals in order to improve quality and yield. However, experimental studies in magnetohydrodynamics often involve complex, turbulent flows and require planar, two-component (2c) velocity measurements through only one acoustical access. We present the phased array ultrasound Doppler velocimeter as a modular research platform for flow mapping in liquid metals. It combines the pulse wave Doppler method with the phased array technique to adaptively focus the ultrasound beam. This makes it possible to resolve smaller flow structures in planar measurements compared with fixed-beam sensors and enables 2c flow mapping with only one acoustical access via the cross beam technique. From simultaneously measured 2-D velocity fields, quantities for turbulence characterization can be derived. The capabilities of this measurement system are demonstrated through measurements in the alloy gallium-indium-tin at room temperature. The 2-D, 2c velocity measurements of a flow in a cubic vessel driven by a rotating magnetic field (RMF) with a spatial resolution of up to 2.2 mm are presented. The measurement results are in good agreement with a semianalytical simulation. As a highlight, two-point correlation functions of the velocity field for different magnitudes of the RMF are presented.

Amelioration of Metabolic Syndrome-Associated Cognitive Impairments in Mice via a Reduction in Dietary Fat Content or Infusion of Non-Diabetic Plasma.

Obesity, metabolic syndrome (MetS) and type 2 diabetes (T2D) are associated with decreased cognitive function. While weight loss and T2D remission result in improvements in metabolism and vascular function, it is less clear if these benefits extend to cognitive performance. Here, we highlight the malleable nature of MetS-associated cognitive dysfunction using a mouse model of high fat diet (HFD)-induced MetS. While learning and memory was generally unaffected in mice with type 1 diabetes (T1D), multiple cognitive impairments were associated with MetS, including deficits in novel object recognition, cued fear memory, and spatial learning and memory. However, a brief reduction in dietary fat content in chronic HFD-fed mice led to a complete rescue of cognitive function. Cerebral blood volume (CBV), a measure of vascular perfusion, was decreased during MetS, was associated with long term memory, and recovered following the intervention. Finally, repeated infusion of plasma collected from age-matched, low fat diet-fed mice improved memory in HFD mice, and was associated with a distinct metabolic profile. Thus, the cognitive dysfunction accompanying MetS appears to be amenable to treatment, related to cerebrovascular function, and mitigated by systemic factors.

Assessment of the 3 D Pore Structure and Individual Components of Preshaped Catalyst Bodies by X-Ray Imaging.

Porosity in catalyst particles is essential because it enables reactants to reach the active sites and it enables products to leave the catalyst. The engineering of composite-particle catalysts through the tuning of pore-size distribution and connectivity is hampered by the inability to visualize structure and porosity at critical-length scales. Herein, it is shown that the combination of phase-contrast X-ray microtomography and high-resolution ptychographic X-ray tomography allows the visualization and characterization of the interparticle pores at micro- and nanometer-length scales. Furthermore, individual components in preshaped catalyst bodies used in fluid catalytic cracking, one of the most used catalysts, could be visualized and identified. The distribution of pore sizes, as well as enclosed pores, which cannot be probed by traditional methods, such as nitrogen physisorption and isotherm analysis, were determined.

High-throughput phenotyping and genetic linkage of cortical bone microstructure in the mouse.

BACKGROUND: Understanding cellular structure and organization, which plays an important role in biological systems ranging from mechanosensation to neural organization, is a complicated multifactorial problem depending on genetics, environmental factors, and stochastic processes. Isolating these factors necessitates the measurement and sensitive quantification of many samples in a reliable, high-throughput, unbiased manner. In this manuscript we present a pipelined approach using a fully automated framework based on Synchrotron-based X-ray Tomographic Microscopy (SRXTM) for performing a full 3D characterization of millions of substructures. RESULTS: We demonstrate the framework on a genetic study on the femur bones of in-bred mice. We measured 1300 femurs from a F2 cross experiment in mice without the growth hormone (which can confound many of the smaller structural differences between strains) and characterized more than 50 million osteocyte lacunae (cell-sized hollows in the bone). The results were then correlated with genetic markers in a process called quantitative trait localization (QTL). Our findings provide a mapping between regions of the genome (all 19 autosomes) and observable phenotypes which could explain between 8-40 % of the variance using between 2-10 loci for each trait. This map shows 4 areas of overlap with previous studies looking at bone strength and 3 areas not previously associated with bone. CONCLUSIONS: The mapping of microstructural phenotypes provides a starting point for both structure-function and genetic studies on murine bone structure and the specific loci can be investigated in more detail to identify single gene candidates which can then be translated to human investigations. The flexible infrastructure offers a full spectrum of shape, distribution, and connectivity metrics for cellular networks and can be adapted to a wide variety of materials ranging from plant roots to lung tissue in studies requiring high sample counts and sensitive metrics such as the drug-gene interactions and high-throughput screening.

Osteocyte lacunar properties in rat cortical bone: Differences between lamellar and central bone.

Recently, the roles of osteocytes in bone maintenance have gained increasing attention. Osteocytes reside in lacunae that are interconnected by canaliculi resulting in a vast cellular network within the mineralized bone matrix. As the structure of the lacuno-canalicular network is highly connected to osteocyte function, osteocyte lacunar properties such as volume, shape, orientation, and density are now frequently reported in studies investigating osteocyte activity. Despite this increasing interest in lacunar morphometrics, many studies show a large spread in such values, suggesting a large inter-species but also inter-site variation in lacunar properties. Here, osteocyte lacunae in rat cortical bone have been studied using synchrotron radiation micro computed tomography (SR µCT) and backscattered electron (BE) microscopy. Quantitative lacunar geometric characteristics are reported based on the synchrotron radiation data, differentiating between circumferential lamellar bone and a central, more disordered bone type. From these studies, no significant differences were found in lacunar volumes between lamellar and central bone, whereas significant differences in lacunar orientation, shape and density values were observed. The 3D nature of the SR µCT data sets furthermore revealed that lacunae in central bone, which appear to be poorly aligned in transverse 2D cross sections, are in fact highly aligned along the bone long axis. These results demonstrate the importance of using 3D methods to investigate anisotropic biological materials such as bone and that the appropriate choice of subregions for high resolution imaging is not trivial.

Synchrotron X-ray micro-tomography imaging and analysis of wood degraded by Physisporinus vitreus and Xylaria longipes.

Incubation of Norway spruce with Physisporinus vitreus and sycamore with Xylaria longipes results in reduction in density of these wood species that are traditionally used for the top and bottom plate of a violin, which follows by enhanced acoustic properties. We used Synchrotron X-ray micro-tomography, to study the three-dimensional structure of wood at the micro-scale level and the alterations of the density distribution after incubation with two white-rot fungi. Micro-tomography data from wood treated at different incubation periods are analyzed and compared with untreated (control) specimens to determine the wood density map and changes at the cell-wall level. Differences between the density of early- and latewood, xylem ray and around bordered pits in both Norway spruce and sycamore are studied. Three-dimensional hyphal networks of the P.vitreus and Xylaria longipes hyphae are visualized inside the cell lumina and their significance on the density of the early- and latewood cells after different incubation periods are discussed. The study illustrates the utility of X-ray micro-tomography for both qualitative and quantitative studies of a wide variety of biological systems and due to its high sensitivity, small structural changes can be quantified.

Coherent X-ray imaging: bridging the gap between atomic and micro-scale investigations.

We present a review of state-of-the art X-ray imaging techniques based on partially coherent synchrotron radiation. Full-field X-ray tomography, X-ray ptychography, scanning small-angle X-ray scattering, and scanning transmission X-ray microscopy are imaging techniques that gather structural information at spatial resolution ranging from several microns to a few tens of nanometers in both real- and reciprocal space. These methods exploit contrast mechanisms based on absorption, phase, and spectroscopic signals. We provide examples of how these techniques can be applied to address scientific questions ranging from imaging of biological samples, to foam rheology, and cement composition.

Coherent X-ray Imaging: Bridging the Gap between Atomic and Micro-scale Investigations.

We present a review of state-of-the art X-ray imaging techniques based on partially coherent synchrotron radiation. Full-field X-ray tomography, X-ray ptychography, scanning small-angle X-ray scattering, and scanning transmission X-ray microscopy are imaging techniques that gather structural information at spatial resolution ranging from several microns to a few tens of nanometers in both real- and reciprocal space. These methods exploit contrast mechanisms based on absorption, phase, and spectroscopic signals. We provide examples of how these techniques can be applied to address scientific questions ranging from imaging of biological samples, to foam rheology, and cement composition.

A quantitative framework for the 3D characterization of the osteocyte lacunar system.

Assessing the role of osteocyte lacunae and the ways in which they communicate with one another is important for determining the function and viability of bone tissue. Osteocytes are able to play a significant role in bone development and remodeling because they can receive nourishment from, interact with, and communicate with other cells. In this sense the immediate environment of an osteocyte is crucial for understanding its function. Modern imaging techniques, ranging from synchrotron radiation-based computed tomography (SR CT) to confocal laser scanning microscopy, produce large volumes of high-quality imaging data of bone tissue on the micrometer scale in rapidly shortening times. These images often contain tens of thousands of osteocytes and their lacunae, void spaces which enclose the osteocytes. While theoretically possible, quantitative analysis of the osteocyte lacunar system is too time consuming to be practical without highly automated tools. Moreover, quantitative morphometry of the osteocyte lacunar system necessitates clearly defined, robust, and three-dimensional (3D) measures. Here, we introduce a framework for the quantitative characterization of millions of osteocyte lacunae and their spatial relationships in 3D. The metrics complement and expand previous works looking at shape and number density while providing novel measures for quantifying spatial distribution and alignment. We developed model, in silico systems to visualize and validate the metrics and provide a concrete example of the attribute being classified with each metric. We then illustrate the applicability to biological samples in a first study comparing two strains of mice and the effect of growth hormone. We found significant differences in shape and distribution between strains for alignment. The proposed quantitative framework can be used in future studies examining differences and treatment effects in bone microstructure at the cell scale. Furthermore, the proposed strategy for quantitative bone cell morphometry will allow investigating structure-function relationships in bone tissue, for example by linking cellular morphometry to bone remodeling.

High-throughput full-automatic synchrotron-based tomographic microscopy.

At the TOMCAT (TOmographic Microscopy and Coherent rAdiology experimenTs) beamline of the Swiss Light Source with an energy range of 8-45 keV and voxel size from 0.37 µm to 7.4 µm, full tomographic datasets are typically acquired in 5 to 10 min. To exploit the speed of the system and enable high-throughput studies to be performed in a fully automatic manner, a package of automation tools has been developed. The samples are automatically exchanged, aligned, moved to the correct region of interest, and scanned. This task is accomplished through the coordination of Python scripts, a robot-based sample-exchange system, sample positioning motors and a CCD camera. The tools are suited for any samples that can be mounted on a standard SEM stub, and require no specific environmental conditions. Up to 60 samples can be analyzed at a time without user intervention. The throughput of the system is dependent on resolution, energy and sample size, but rates of four samples per hour have been achieved with 0.74 µm voxel size at 17.5 keV. The maximum intervention-free scanning time is theoretically unlimited, and in practice experiments have been running unattended as long as 53 h (the average beam time allocation at TOMCAT is 48 h per user). The system is the first fully automated high-throughput tomography station: mounting samples, finding regions of interest, scanning and reconstructing can be performed without user intervention. The system also includes many features which accelerate and simplify the process of tomographic microscopy.