SCIENTIFIC COMPUTING AND IMAGING INSTITUTE
at the University of Utah

Computer simulation and visualization are having a substantial impact on biomedicine and other areas of science and engineering. Advanced simulation and data acquisition techniques allow biomedical researchers to investigate increasingly sophisticated biological function and structure. A continuing trend in all computational science and engineering applications is the increasing size of resulting datasets. This trend is also evident in data acquisition, especially in image acquisition in biology and medical image databases.

For example, in a collaboration between neuroscientist Robert Marc and our research team at the University of Utah's Scientific Computing and Imaging (SCI) Institute (www.sci.utah.edu), we're creating datasets of brain electron microscopy (EM) mosaics that are 16 terabytes in size. However, while there's no foreseeable end to the increase in our ability to produce simulation data or record observational data, our ability to use this data in meaningful ways is inhibited by current data analysis capabilities, which already lag far behind. Indeed, as the NIH-NSF Visualization Research Challenges report notes, to effectively understand and make use of the vast amounts of data researchers are producing is one of the greatest scientific challenges of the 21st century.

Visual data analysis involves creating images that convey salient information about underlying data and processes, enabling the detection and validation of expected results while leading to unexpected discoveries in science. This allows for the validation of new theoretical models, provides comparison between models and datasets, enables quantitative and qualitative querying, improves interpretation of data, and facilitates decision making. Scientists can use visual data analysis systems to explore \"what if\" scenarios, define hypotheses, and examine data under multiple perspectives and assumptions. In addition, they can identify connections between numerous attributes and quantitatively assess the reliability of hypotheses. In essence, visual data analysis is an integral part of scientific problem solving and discovery.

As applied to biomedical systems, visualization plays a crucial role in our ability to comprehend large and complex data-data that, in two, three, or more dimensions, convey insight into many diverse biomedical applications, including understanding neural connectivity within the brain, interpreting bioelectric currents within the heart, characterizing white-matter tracts by diffusion tensor imaging, and understanding morphology differences among different genetic mice phenotypes.

Keywords: kaust

Neuroscientists are developing new imaging techniques and generating large volumes of data in an effort to understand the complex structure of the nervous system. The complexity and size of this data makes human interpretation a labor-intensive task. To aid in the analysis, new segmentation techniques for identifying neurons in these feature rich datasets are required. This paper presents a method for neuron boundary detection and nonbranching process segmentation in electron microscopy images and visualizing them in three dimensions. It combines both automated segmentation techniques with a graphical user interface for correction of mistakes in the automated process. The automated process first uses machine learning and image processing techniques to identify neuron membranes that deliniate the cells in each two-dimensional section. To segment nonbranching processes, the cell regions in each two-dimensional section are connected in 3D using correlation of regions between sections. The combination of this method with a graphical user interface specially designed for this purpose, enables users to quickly segment cellular processes in large volumes.

The National Alliance for Medical Image Computing (NA-MIC), is a multi-institutional, interdisciplinary community of researchers, who share the recognition that modern health care demands improved technologies to ease suffering and prolong productive life. Organized under the National Centers for Biomedical Computing 7 years ago, the mission of NA-MIC is to implement a robust and flexible open-source infrastructure for developing and applying advanced imaging technologies across a range of important biomedical research disciplines. A measure of its success, NA-MIC is now applying this technology to diseases that have immense impact on the duration and quality of life: cancer, heart disease, trauma, and degenerative genetic diseases. The targets of this technology range from group comparisons to subject-specific analysis.

Extracting isosurfaces represented as high quality meshes from three-dimensional scalar fields is needed for many important applications, particularly visualization and numerical simulations. One recent advance for extracting high quality meshes for isosurface computation is based on a dynamic particle system. Unfortunately, this state-of-the-art particle placement technique requires a significant amount of time to produce a satisfactory mesh. To address this issue, we study the parallelism property of the particle placement and make use of CUDA, a parallel programming technique on the GPU, to significantly improve the performance of particle placement. This paper describes the curvature dependent sampling method used to extract high quality meshes and describes its implementation using CUDA on the GPU.

Keywords: CUDA, GPGPU, particle systems, volumetric data mesh extraction

Smoothness-increasing accuracy-conserving (SIAC) filtering has demonstrated its effectiveness in raising the convergence rate of discontinuous Galerkin solutions from order k + 1/2 to order 2k + 1 for specific types of translation invariant meshes (Cockburn et al. in Math. Comput. 72:577–606, 2003; Curtis et al. in SIAM J. Sci. Comput. 30(1):272– 289, 2007; Mirzaee et al. in SIAM J. Numer. Anal. 49:1899–1920, 2011). Additionally, it improves the weak continuity in the discontinuous Galerkin method to k - 1 continuity. Typically this improvement has a positive impact on the error quantity in the sense that it also reduces the absolute errors. However, not enough emphasis has been placed on the difference between superconvergent accuracy and improved errors. This distinction is particularly important when it comes to understanding the interplay introduced through meshing, between geometry and filtering. The underlying mesh over which the DG solution is built is important because the tool used in SIAC filtering—convolution—is scaled by the geometric mesh size. This heavily contributes to the effectiveness of the post-processor. In this paper, we present a study of this mesh scaling and how it factors into the theoretical errors. To accomplish the large volume of post-processing necessary for this study, commodity streaming multiprocessors were used; we demonstrate for structured meshes up to a 50× speed up in the computational time over traditional CPU implementations of the SIAC filter.

To make the most of current advanced computing technologies, experts in particular areas of science and engineering should be supported by sophisticated tools for carrying out computational experiments. The complexity of individual components of such tools should be hidden from them so they may concentrate on solving the specific problem within their field of expertise. One class of such tools are Problem Solving Environments (PSEs). The contribution of this paper refers to the idea of integration of an interactive computing framework applicable to different engineering applications into the SCIRun PSE in order to enable interactive real-time response of the computational model to user interaction even for large-scale problems. While the SCIRun PSE allows for real-time computational steering, we propose extending this functionality to a wider range of applications and larger scale problems. With only minor code modifications the proposed system allows each module scheduled for execution in a dataflow-based simulation to be automatically interrupted and re-scheduled. This rescheduling allows one to keep the relation between the user interaction and its immediate effect transparent independent of the problem size, thus, allowing for the intuitive and interactive exploration of simulation results.

Keywords: scirun

Bounding volume hierarchies (BVHs) are a popular acceleration structure choice for animated scenes rendered with ray tracing. This is due to the relative simplicity of refitting bounding volumes around moving geometry. However, the quality of such a refitted tree can degrade rapidly if objects in the scene deform or rearrange significantly as the animation progresses, resulting in dramatic increases in rendering times and a commensurate reduction in the frame rate. The BVH could be rebuilt on every frame, but this could take significant time. We present a method to efficiently extend refitting for animated scenes with tree rotations, a technique previously proposed for off-line improvement of BVH quality for static scenes. Tree rotations are local restructuring operations which can mitigate the effects that moving primitives have on BVH quality by rearranging nodes in the tree during each refit rather than triggering a full rebuild. The result is a fast, lightweight, incremental update algorithm that requires negligible memory, has minor update times, parallelizes easily, avoids significant degradation in tree quality or the need for rebuilding, and maintains fast rendering times. We show that our method approaches or exceeds the frame rates of other techniques and is consistently among the best options regardless of the animated scene.

Hierarchical, multiresolution data representations enable interactive analysis and visualization of large-scale simulations. One promising application of these techniques is to store high performance computing simulation output in a hierarchical Z (HZ) ordering that translates data from a Cartesian coordinate scheme to a one-dimensional array ordered by locality at different resolution levels. However, when the dimensions of the simulation data are not an even power of 2, parallel HZ ordering produces sparse memory and network access patterns that inhibit I/O performance. This work presents a new technique for parallel HZ ordering of simulation datasets that restructures simulation data into large (power of 2) blocks to facilitate efficient I/O aggregation. We perform both weak and strong scaling experiments using the S3D combustion application on both Cray-XE6 (65,536 cores) and IBM Blue Gene/P (131,072 cores) platforms. We demonstrate that data can be written in hierarchical, multiresolution format with performance competitive to that of native data-ordering methods.

The performance of massively parallel applications is often heavily impacted by the cost of communication among compute nodes. However, determining how to best use the network is a formidable task, made challenging by the ever increasing size and complexity of modern supercomputers. This paper applies visualization techniques to aid parallel application developers in understanding the network activity by enabling a detailed exploration of the flow of packets through the hardware interconnect. In order to visualize this large and complex data, we employ two linked views of the hardware network. The first is a 2D view, that represents the network structure as one of several simplified planar projections. This view is designed to allow a user to easily identify trends and patterns in the network traffic. The second is a 3D view that augments the 2D view by preserving the physical network topology and providing a context that is familiar to the application developers. Using the massively parallel multi-physics code pF3D as a case study, we demonstrate that our tool provides valuable insight that we use to explain and optimize pF3D’s performance on an IBM Blue Gene/P system.

Despite recent large-scale profiling efforts, the best prognostic predictor of glioblastoma multiforme (GBM) remains the patient's age at diagnosis. We describe a global pattern of tumor-exclusive co-occurring copy-number alterations (CNAs) that is correlated, possibly coordinated with GBM patients' survival and response to chemotherapy. The pattern is revealed by GSVD comparison of patient-matched but probe-independent GBM and normal aCGH datasets from The Cancer Genome Atlas (TCGA). We find that, first, the GSVD, formulated as a framework for comparatively modeling two composite datasets, removes from the pattern copy-number variations (CNVs) that occur in the normal human genome (e.g., female-specific X chromosome amplification) and experimental variations (e.g., in tissue batch, genomic center, hybridization date and scanner), without a-priori knowledge of these variations. Second, the pattern includes most known GBM-associated changes in chromosome numbers and focal CNAs, as well as several previously unreported CNAs in greater than 3\% of the patients. These include the biochemically putative drug target, cell cycle-regulated serine/threonine kinase-encoding TLK2, the cyclin E1-encoding CCNE1, and the Rb-binding histone demethylase-encoding KDM5A. Third, the pattern provides a better prognostic predictor than the chromosome numbers or any one focal CNA that it identifies, suggesting that the GBM survival phenotype is an outcome of its global genotype. The pattern is independent of age, and combined with age, makes a better predictor than age alone. GSVD comparison of matched profiles of a larger set of TCGA patients, inclusive of the initial set, confirms the global pattern. GSVD classification of the GBM profiles of an independent set of patients validates the prognostic contribution of the pattern.

Analysis and visualization of complex vector fields remain major challenges when studying large scale simulation of physical phenomena. The primary reason is the gap between the concepts of smooth vector field theory and their computational realization. In practice, researchers must choose between either numerical techniques, with limited or no guarantees on how they preserve fundamental invariants, or discrete techniques which limit the precision at which the vector field can be represented. We propose a new representation of vector fields that combines the advantages of both approaches. In particular, we represent a subset of possible streamlines by storing their paths as they traverse the edges of a triangulation. Using only a finite set of streamlines creates a fully discrete version of a vector field that nevertheless approximates the smooth flow up to a user controlled error bound. The discrete nature of our representation enables us to directly compute and classify analogues of critical points, closed orbits, and other common topological structures. Further, by varying the number of divisions (quantizations) used per edge, we vary the resolution used to represent the field, allowing for controlled precision. This representation is compact in memory and supports standard vector field operations.

dentification and characterization of cancer subtypes are important areas of research that are based on the integrated analysis of multiple heterogeneous genomics datasets. Since there are no tools supporting this process, much of this work is done using ad-hoc scripts and static plots, which is inefficient and limits visual exploration of the data. To address this, we have developed StratomeX, an integrative visualization tool that allows investigators to explore the relationships of candidate subtypes across multiple genomic data types such as gene expression, DNA methylation, or copy number data. StratomeX represents datasets as columns and subtypes as bricks in these columns. Ribbons between the columns connect bricks to show subtype relationships across datasets. Drill-down features enable detailed exploration. StratomeX provides insights into the functional and clinical implications of candidate subtypes by employing small multiples, which allow investigators to assess the effect of subtypes on molecular pathways or outcomes such as patient survival. As the configuration of viewing parameters in such a multi-dataset, multi-view scenario is complex, we propose a meta visualization and configuration interface for dataset dependencies and data-view relationships. StratomeX is developed in close collaboration with domain experts. We describe case studies that illustrate how investigators used the tool to explore subtypes in large datasets and demonstrate how they efficiently replicated findings from the literature and gained new insights into the data.

Computing failure probability is a fundamental task in many important practical problems. The computation, its numerical challenges aside, naturally requires knowledge of the probability distribution of the underlying random inputs. On the other hand, for many complex systems it is often not possible to have complete information about the probability distributions. In such cases the uncertainty is often referred to as epistemic uncertainty, and straightforward computation of the failure probability is not available. In this paper we develop a method to estimate both the upper bound and the lower bound of the failure probability subject to epistemic uncertainty. The bounds are rigorously derived using the variational formulas for relative entropy. We examine in detail the properties of the bounds and present numerical algorithms to efficiently compute them.

Keywords: failure probability, uncertainty quantification, epistemic uncertainty, relative entropy

Automated segmentation of electron microscopy (EM) images is a challenging problem. In this paper, we present a novel method that utilizes a hierarchical structure and boundary classification for 2D neuron segmentation. With a membrane detection probability map, a watershed merge tree is built for the representation of hierarchical region merging from the watershed algorithm. A boundary classifier is learned with non-local image features to predict each potential merge in the tree, upon which merge decisions are made with consistency constraints to acquire the final segmentation. Independent of classifiers and decision strategies, our approach proposes a general framework for efficient hierarchical segmentation with statistical learning. We demonstrate that our method leads to a substantial improvement in segmentation accuracy.

Representing uncertainty when creating visualizations is becoming more indispensable to understand and analyze scientific data. Uncertainty may come from different sources, such as, ensembles of experiments or unavoidable information loss when performing data reduction. One natural model to represent uncertainty is to assume that each position in space instead of a single value may take on a distribution of values. In this paper we present a new volume rendering method using per voxel Gaussian mixture models (GMMs) as the input data representation. GMMs are an elegant and compact way to drastically reduce the amount of data stored while still enabling realtime data access and rendering on the GPU. Our renderer offers efficient sampling of the data distribution, generating renderings of the data that flicker at each frame to indicate high variance. We can accumulate samples as well to generate still frames of the data, which preserve additional details in the data as compared to either traditional scalar indicators (such as a mean or a single nearest neighbor down sample) or to fitting the data with only a single Gaussian per voxel. We demonstrate the effectiveness of our method using ensembles of climate simulations and MRI scans as well as the down sampling of large scalar fields as examples.

Keywords: Uncertainty Visualization, Volume Rendering, Gaussian Mixture Model, Ensemble Visualization

Identifying functional networks from resting-state functional MRI is a challenging task, especially for multiple subjects. Most current studies estimate the networks in a sequential approach, i.e., they identify each individual subject's network independently to other subjects, and then estimate the group network from the subjects networks. This one-way flow of information prevents one subject's network estimation benefiting from other subjects. We propose a hierarchical Markov Random Field model, which takes into account both the within-subject spatial coherence and between-subject consistency of the network label map. Both population and subject network maps are estimated simultaneously using a Gibbs sampling approach in a Monte Carlo Expectation Maximization framework. We compare our approach to two alternative groupwise fMRI clustering methods, based on K-means and Normalized Cuts, using both synthetic and real fMRI data.We show that our method is able to estimate more consistent subject label maps, as well as a stable group label map.

Effective detection of and response to infectious disease outbreaks depend on the ability to capture and analyze information and on how public health officials respond to this information. Researchers have developed various surveillance systems to automate data collection, analysis, and alert generation, yet the massive amount of collected data often leads to information overload. To improve decision-making in outbreak detection and response, it's important to understand how outbreak investigators seek relevant information. Studying their information-search strategies can provide insight into their cognitive biases and heuristics. Identifying the presence of such biases will enable the development of tools that counter balance them and help users develop alternative scenarios.

We implemented a large-scale high-fidelity simulation of scripted infectious-disease outbreaks to help us study public health practitioners' information- search strategies. We also developed Epinome, an integrated visual-analytics investigation system. Epinome caters to users' needs by providing a variety of investigation tools. It facilitates user studies by recording which tools they used, when, and how. (See the video demonstration of Epinome at www.sci.utah.edu/gallery2/v/ software/epinome.) Epinome provides a dynamic environment that seamlessly evolves and adapts to user tasks and needs. It introduces four userinteraction paradigms in public health:

• an evolving visual display,
• seamless integration between disparate views,
• loosely coordinated multiple views, and
• direct interaction with data items.

Using Epinome, users can replay simulation scenarios, investigate an unfolding outbreak using a variety of visualization tools, and steer the simulation by implementing different public health policies at predefined decision points. Epinome records user actions, such as tool selection, interactions with each tool, and policy changes, and stores them in a database for postanalysis. A psychology team can then use that information to study users' search strategies.

This paper describes an adaptive implementation of a high order Discontinuous Galerkin (DG) method for the solution of elastohydrodynamic lubrication (EHL) point contact problems. These problems arise when modelling the thin lubricating film between contacts which are under sufficiently high pressure that the elastic deformation of the contacting elements cannot be neglected. The governing equations are highly nonlinear and include a second order partial differential equation that is derived via the thin-film approximation. Furthermore, the problem features a free boundary, which models where cavitation occurs, and this is automatically captured as part of the solution process. The need for spatial adaptivity stems from the highly variable length scales that are present in typical solutions. Results are presented which demonstrate both the effectiveness and the limitations of the proposed adaptive algorithm.

Keywords: Elastohydrodynamic lubrication, Discontinuous Galerkin, High polynomial degree, h-adaptivity, Nonlinear systems

Background: Enlargement of the lateral ventricles is thought to originate from abnormal prenatal brain development and is associated with neurodevelopmental disorders. Fetal isolated mild ventriculomegaly (MVM) is associated with the enlargement of lateral ventricle volumes in the neonatal period and developmental delays in early childhood. However, little is known about postnatal brain development in these children.

Methods: Twenty-eight children with fetal isolated MVM and 56 matched controls were followed at ages 1 and 2 years with structural imaging on a 3T Siemens scanner and assessment of cognitive development with the Mullen Scales of Early Learning. Lateral ventricle, total gray and white matter volumes, and Mullen cognitive composite scores and subscale scores were compared between groups.

Results: Compared to controls, children with prenatal isolated MVM had significantly larger lateral ventricle volumes at ages 1 and 2 years. Lateral ventricle volume at 1 and 2 years of age was significantly correlated with prenatal ventricle size. Enlargement of the lateral ventricles was associated with increased intracranial volumes and increased gray and white matter volumes. Children with MVM had Mullen composite scores similar to controls, although there was evidence of delay in fine motor and expressive language skills.

Conclusions: Children with prenatal MVM have persistent enlargement of the lateral ventricles through the age of 2 years; this enlargement is associated with increased gray and white matter volumes and some evidence of delay in fine motor and expressive language development. Further study is needed to determine if enlarged lateral ventricles are associated with increased risk for neurodevelopmental disorders.