A GPU-based caching strategy for multi-material linear elastic FEM on regular grids.

In this study, we present a novel strategy to the method of finite elements (FEM) of linear elastic problems of very high resolution on graphic processing units (GPU). The approach exploits regularities in the system matrix that occur in regular hexahedral grids to achieve cache-friendly matrix-free FEM. The node-by-node method lies in the class of block-iterative Gauss-Seidel multigrid solvers. Our method significantly improves convergence times in cases where an ordered distribution of distinct materials is present in the dataset. The method was evaluated on three real world datasets: An aluminum-silicon (AlSi) alloy and a dual phase steel material sample, both captured by scanning electron tomography, and a clinical computed tomography (CT) scan of a tibia. The caching scheme leads to a speed-up factor of ×2-×4 compared to the same code without the caching scheme. Additionally, it facilitates the computation of high-resolution problems that cannot be computed otherwise due to memory consumption.

Correlative Fluorescence- and Electron Microscopy of Whole Breast Cancer Cells Reveals Different Distribution of ErbB2 Dependent on Underlying Actin.

Epidermal growth factor receptor 2 (ErbB2) is found overexpressed in several cancers, such as gastric, and breast cancer, and is, therefore, an important therapeutic target. ErbB2 plays a central role in cancer cell invasiveness, and is associated with cytoskeletal reorganization. In order to study the spatial correlation of single ErbB2 proteins and actin filaments, we applied correlative fluorescence microscopy (FM), and scanning transmission electron microscopy (STEM) to image specifically labeled SKBR3 breast cancer cells. The breast cancer cells were grown on microchips, transformed to express an actin-green fluorescent protein (GFP) fusion protein, and labeled with quantum dot (QD) nanoparticles attached to specific anti-ErbB2 Affibodies. FM was performed to identify cellular regions with spatially correlated actin and ErbB2 expression. For STEM of the intact plasma membrane of whole cells, the cells were fixed and covered with graphene. Spatial distribution patterns of ErbB2 in the actin rich ruffled membrane regions were examined, and compared to adjacent actin-low regions of the same cell, revealing an association of putative signaling active ErbB2 homodimers with actin-rich regions. ErbB2 homodimers were found absent from actin-low membrane regions, as well as after treatment of cells with Cytochalasin D, which breaks up larger actin filaments. In both latter data sets, a significant inter-label distance of 36 nm was identified, possibly indicating an indirect attachment to helical actin filaments via the formation of heterodimers of ErbB2 with epidermal growth factor receptor (EGFR). The possible attachment to actin filaments was further explored by identifying linear QD-chains in actin-rich regions, which also showed an inter-label distance of 36 nm.

Sparse Scanning Electron Microscopy Data Acquisition and Deep Neural Networks for Automated Segmentation in Connectomics.

With the growing importance of three-dimensional and very large field of view imaging, acquisition time becomes a serious bottleneck. Additionally, dose reduction is of importance when imaging material like biological tissue that is sensitive to electron radiation. Random sparse scanning can be used in the combination with image reconstruction techniques to reduce the acquisition time or electron dose in scanning electron microscopy. In this study, we demonstrate a workflow that includes data acquisition on a scanning electron microscope, followed by a sparse image reconstruction based on compressive sensing or alternatively using neural networks. Neuron structures are automatically segmented from the reconstructed images using deep learning techniques. We show that the average dwell time per pixel can be reduced by a factor of 2-3, thereby providing a real-life confirmation of previous results on simulated data in one of the key segmentation applications in connectomics and thus demonstrating the feasibility and benefit of random sparse scanning techniques for a specific real-world scenario.

Linear Chains of HER2 Receptors Found in the Plasma Membrane Using Liquid-Phase Electron Microscopy.

The spatial distribution of the human epidermal growth factor 2 (HER2) receptor in the plasma membrane of SKBR3 and HCC1954 breast cancer cells was studied. The receptor was labeled with quantum dot nanoparticles, and fixed whole cells were imaged in their native liquid state with environmental scanning electron microscopy using scanning transmission electron microscopy detection. The locations of individual HER2 positions were determined in a total plasma membrane area of 991 µm2 for several SKBR3 cells and 1062 µm2 for HCC1954 cells. Some of the HER2 receptors were arranged in a linear chain with interlabel distances of 40 ± 7 and 32 ± 10 nm in SKBR3 and HCC1954 cells, respectively. The finding was tested against randomly occurring linear chains of six or more positions, from which it was concluded that the experimental finding is significant and did not arise from random label distributions. Because the measured interlabel distance in the HER2 chains is similar to the 36-nm helix-repetition distance of actin filaments, it is proposed that a linking mechanism between HER2 and actin filaments leads to linearly aligned oligomers.

How should a fixed budget of dwell time be spent in scanning electron microscopy to optimize image quality?

In scanning electron microscopy, the achievable image quality is often limited by a maximum feasible acquisition time per dataset. Particularly with regard to three-dimensional or large field-of-view imaging, a compromise must be found between a high amount of shot noise, which leads to a low signal-to-noise ratio, and excessive acquisition times. Assuming a fixed acquisition time per frame, we compared three different strategies for algorithm-assisted image acquisition in scanning electron microscopy. We evaluated (1) raster scanning with a reduced dwell time per pixel followed by a state-of-the-art Denoising algorithm, (2) raster scanning with a decreased resolution in conjunction with a state-of-the-art Super Resolution algorithm, and (3) a sparse scanning approach where a fixed percentage of pixels is visited by the beam in combination with state-of-the-art inpainting algorithms. Additionally, we considered increased beam currents for each of the strategies. The experiments showed that sparse scanning using an appropriate reconstruction technique was superior to the other strategies.

Exemplar-based inpainting as a solution to the missing wedge problem in electron tomography.

A new method for dealing with incomplete projection sets in electron tomography is proposed. The approach is inspired by exemplar-based inpainting techniques in image processing and heuristically generates data for missing projection directions. The method has been extended to work on three dimensional data. In general, electron tomography reconstructions suffer from elongation artifacts along the beam direction. These artifacts can be seen in the corresponding Fourier domain as a missing wedge. The new method synthetically generates projections for these missing directions with the help of a dictionary based approach that is able to convey both structure and texture at the same time. It constitutes a preprocessing step that can be combined with any tomographic reconstruction algorithm. The new algorithm was applied to phantom data, to a real electron tomography data set taken from a catalyst, as well as to a real dataset containing solely colloidal gold particles. Visually, the synthetic projections, reconstructions, and corresponding Fourier power spectra showed a decrease of the typical missing wedge artifacts. Quantitatively, the inpainting method is capable to reduce missing wedge artifacts and improves tomogram quality with respect to full width half maximum measurements.

Spherically symmetric volume elements as basis functions for image reconstructions in computed laminography.

BACKGROUND: Laminography is a tomographic technique that allows three-dimensional imaging of flat and elongated objects that stretch beyond the extent of a reconstruction volume. Laminography images can be reconstructed using iterative algorithms based on the Kaczmarz method. OBJECTIVE: This study aims to develop and demonstrate a new reconstruction algorithm that may provide superior image reconstruction quality for this challenged imaging application. METHODS: The images are initially represented using the coefficients over basis functions, which are typically piecewise constant functions (voxels). By replacing voxels with spherically symmetric volume elements (blobs) based on the generalized Kaiser-Bessel window functions, the images are reconstructed using this new adapted version of the algebraic image reconstruction technique. RESULTS: Band-limiting properties of blob functions are beneficial particular in the case of noisy projections and with only a limited number of available projections. Study showed that using blob basis functions improved full-width-at-half-maximum resolution from 10.2±1.0 to 9.9±0.9 (p < 0.001). Signal-to-noise ratio also improved from 16.1 to 31.0. The increased computational demand per iteration was compensated by using a faster convergence rate, such that the overall performance is approximately identical for blobs and voxels. CONCLUSIONS: Despite the higher complexity, tomographic reconstruction from computed laminography data should be implemented using blob basis functions, especially if noisy data is expected.

Feature Adaptive Sampling for Scanning Electron Microscopy.

A new method for the image acquisition in scanning electron microscopy (SEM) was introduced. The method used adaptively increased pixel-dwell times to improve the signal-to-noise ratio (SNR) in areas of high detail. In areas of low detail, the electron dose was reduced on a per pixel basis, and a-posteriori image processing techniques were applied to remove the resulting noise. The technique was realized by scanning the sample twice. The first, quick scan used small pixel-dwell times to generate a first, noisy image using a low electron dose. This image was analyzed automatically, and a software algorithm generated a sparse pattern of regions of the image that require additional sampling. A second scan generated a sparse image of only these regions, but using a highly increased electron dose. By applying a selective low-pass filter and combining both datasets, a single image was generated. The resulting image exhibited a factor of ≈3 better SNR than an image acquired with uniform sampling on a Cartesian grid and the same total acquisition time. This result implies that the required electron dose (or acquisition time) for the adaptive scanning method is a factor of ten lower than for uniform scanning.

Combining miRNA and mRNA Expression Profiles in Wilms Tumor Subtypes.

Wilms tumor (WT) is the most common childhood renal cancer. Recent findings of mutations in microRNA (miRNA) processing proteins suggest a pivotal role of miRNAs in WT genesis. We performed miRNA expression profiling of 36 WTs of different subtypes and four normal kidney tissues using microarrays. Additionally, we determined the gene expression profile of 28 of these tumors to identify potentially correlated target genes and affected pathways. We identified 85 miRNAs and 2107 messenger RNAs (mRNA) differentially expressed in blastemal WT, and 266 miRNAs and 1267 mRNAs differentially expressed in regressive subtype. The hierarchical clustering of the samples, using either the miRNA or mRNA profile, showed the clear separation of WT from normal kidney samples, but the miRNA pattern yielded better separation of WT subtypes. A correlation analysis of the deregulated miRNA and mRNAs identified 13,026 miRNA/mRNA pairs with inversely correlated expression, of which 2844 are potential interactions of miRNA and their predicted mRNA targets. We found significant upregulation of miRNAs-183, -301a/b and -335 for the blastemal subtype, and miRNAs-181b, -223 and -630 for the regressive subtype. We found marked deregulation of miRNAs regulating epithelial to mesenchymal transition, especially in the blastemal subtype, and miRNAs influencing chemosensitivity, especially in regressive subtypes. Further research is needed to assess the influence of preoperative chemotherapy and tumor infiltrating lymphocytes on the miRNA and mRNA patterns in WT.

Multi-omics enrichment analysis using the GeneTrail2 web service.

MOTIVATION: Gene set analysis has revolutionized the interpretation of high-throughput transcriptomic data. Nowadays, with comprehensive studies that measure multiple -omics from the same sample, powerful tools for the integrative analysis of multi-omics datasets are required. RESULTS: Here, we present GeneTrail2, a web service allowing the integrated analysis of transcriptomic, miRNomic, genomic and proteomic datasets. It offers multiple statistical tests, a large number of predefined reference sets, as well as a comprehensive collection of biological categories and enables direct comparisons between the computed results. We used GeneTrail2 to explore pathogenic mechanisms of Wilms tumors. We not only succeeded in revealing signaling cascades that may contribute to the malignancy of blastemal subtype tumors but also identified potential biomarkers for nephroblastoma with adverse prognosis. The presented use-case demonstrates that GeneTrail2 is well equipped for the integrative analysis of comprehensive -omics data and may help to shed light on complex pathogenic mechanisms in cancer and other diseases. AVAILABILITY AND IMPLEMENTATION: GeneTrail2 can be freely accessed under https://genetrail2.bioinf.uni-sb.de CONTACT: : dstoeckel@bioinf.uni-sb.de SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

The Ettention software package.

We present a novel software package for the problem "reconstruction from projections" in electron microscopy. The Ettention framework consists of a set of modular building-blocks for tomographic reconstruction algorithms. The well-known block iterative reconstruction method based on Kaczmarz algorithm is implemented using these building-blocks, including adaptations specific to electron tomography. Ettention simultaneously features (1) a modular, object-oriented software design, (2) optimized access to high-performance computing (HPC) platforms such as graphic processing units (GPU) or many-core architectures like Xeon Phi, and (3) accessibility to microscopy end-users via integration in the IMOD package and eTomo user interface. We also provide developers with a clean and well-structured application programming interface (API) that allows for extending the software easily and thus makes it an ideal platform for algorithmic research while hiding most of the technical details of high-performance computing.

Marker Detection in Electron Tomography: A Comparative Study.

We conducted a comparative study of three widely used algorithms for the detection of fiducial markers in electron microscopy images. The algorithms were applied to four datasets from different sources. For the purpose of obtaining comparable results, we introduced figures of merit and implemented all three algorithms in a unified code base to exclude software-specific differences. The application of the algorithms revealed that none of the three algorithms is superior to the others in all cases. This leads to the conclusion that the choice of a marker detection algorithm highly depends on the properties of the dataset to be analyzed, even within the narrowed domain of electron tomography.

CausalTrail: Testing hypothesis using causal Bayesian networks.

Summary Causal Bayesian Networks are a special class of Bayesian networks in which the hierarchy directly encodes the causal relationships between the variables. This allows to compute the effect of interventions, which are external changes to the system, caused by e.g. gene knockouts or an administered drug. Whereas numerous packages for constructing causal Bayesian networks are available, hardly any program targeted at downstream analysis exists. In this paper we present CausalTrail, a tool for performing reasoning on causal Bayesian networks using the do-calculus. CausalTrail's features include multiple data import methods, a flexible query language for formulating hypotheses, as well as an intuitive graphical user interface. The program is able to account for missing data and thus can be readily applied in multi-omics settings where it is common that not all measurements are performed for all samples. Availability and Implementation CausalTrail is implemented in C++ using the Boost and Qt5 libraries. It can be obtained from https://github.com/dstoeckel/causaltrail.