International Workshop on
Visualization in Medicine and Life Sciences
- Current Challenges, State-of-the-art Approaches, and Future Directions -
July 19-21, 2006
1. Visualization in Medicine: Volume Rendering
Wednesday, 10:30am – 12:30pm.
Session Chair: Gerald Farin.
Speaker: Charles Hansen, University of Utah, USA.
Title:.Is Seeing Really Believing?
Computers have changed the way we live, work, and even recreate. Now, they are transforming how we think about science, engineering, and medicine. Direct volume rendering has proven to be an effective and flexible visualization method for interactive exploration and analysis of 3D scalar fields. At least a great deal of research has examined many aspects of volume rendering. However, while volume rendering is used in specific applications, why has it not been more widely adapted by the medical community? One reason is that volume rendering, by definition, takes the 3D data and generates an image as the result. This is powerful for exploring a dataset but is it sufficient for analysis? In this talk I will explore the idea of coupling volume rendering with segmentation for the analysis for 3D scalar fields with a particular emphasis on medical applications.
Speaker: Christof Rezk-Salama, University of Siegen, Germany.
Title: Medical Volume Visualization with Semantics.
While existing visualization techniques for medical image data are mature from a technical point of view, managing the complexity of visual parameters is still difficult for non-expert users. I will introduce an additional level of abstraction for parametric models of transfer functions. The proposed framework allows visualization experts to design high-level transfer function models which can be intuitively used by non-expert users. The results are user interfaces which provide semantic information for specialized visualization problems.
Speaker: Shigeo Takahashi, University of Tokyo, Japan.
Title: Selecting Optimal Feature Views for Medical Volume Datasets.
In this talk, we will discuss how to extend our approach to optimal viewpoint selection for medical volume datasets. A distinctive aspect of our approach is that it first extracts significant features from the given datasets and explicitly respects them in locating the corresponding optimal viewpoints. The most important issue here is to identify such features appropriately from medical datasets so that we can take advantage of high-level human visual perception when visualizing their underlying details. We will consider several prototype features characteristic of medical datasets, and demonstrate how these features are useful for selecting optimal feature views in a variety of medical applications.
Speaker: Peter Rautek, Technical University Vienna, Austria.
Title: Caricaturistic Visualization.
For many applications of medicine and life science, data is gathered or measured to find and to analyze the characteristics of the investigated object. Characteristics of a dataset can be expressed as the deviations from the norm. These deviations traditionally are found and classified using statistical methods. In many cases the statistical models do not appropriately describe the underlying phenomenon. They are therefore unsuitable for the data of interest. In this case visualization can replace the statistical methods. Expressive visualizations guide the user to find characteristics. Further the user is enabled to analyze the deviations of a given dataset. Caricaturistic visualization is an expressive method tailored to depict the deviations in an exaggerated way. It is guided by the idea of caricatures which exaggerate the outstanding features of an object. A method for caricaturistic visualization is presented and its power is shown on different examples. Caricaturistic visualization assumes the existence of a reference model. In many applications an explicit reference model is not available. To overcome this limitation different datasets are compared to each other. This results in the Caricature matrix, a 2D matrix of caricaturistic visualizations.
Speaker: Thomas Ertl, University of Stuttgart, Germany.
Title: VMLS activities at the University of Stuttgart.
This talk gives a short overview of the activities of the Stuttgart VIS/US group in the area of visualization in medicine and life sciences. I will report on our collaboration with the cognitive science group in Freiburg in the area of fMRI visualization, and our ongoing efforts together with the Neurocenter Erlangen in the area of web-based medical visualization. I will present some results of integrated simulation and visualization of deformable tetrahedreal volumes. Finally, I will show how our activities in point-based visualization extend to biomolecular simulations and DTI streamtubes.
2. Visualization in Medicine: Surfaces
Wednesday, 2:00pm – 4:00pm.
Session Chair: Charles Hansen.
Speaker: Steffen Oeltze, University of Magdeburg, Germany.
Title: Model-based Visualization of Tree-like Anatomic Structures.
For medical education as well as for many therapy planning tasks it is crucial to understand the branching pattern of tree-like anatomic structures, such as nerves, vascular and bronchial trees. For therapy planning, it is of paramount importance to recognize shape features and morphology of vascular structures as well as spatial relations between vascular and other relevant structures. For a convenient interpretation, the curvature, the depth relations, and the diminution of the diameter towards the periphery should be depicted correctly. In my presentation, I will describe methods to reconstruct, to render and to explore vascular trees based on medical volume data (CT or MR angiography). Traditional volume visualization methods, such as direct volume rendering, thresholdbased isosurface rendering, or maximum intensity projection are not well-suited for the above-mentioned goals. Due to the limited resolution of CT- and MR-scanners conventional visualizations show distracting aliasing artifacts in particular for small vessels where the diameter varies between slices due to partial volume effects and high-frequency noise. Also, the visual separation of contrast-enhanced vascular structures and other highintensity structures, such as bones, might be very difficult with conventional medical visualization techniques. Instead, vascular structures should be reconstructed based on the radiological data of a particular patient and some model assumptions as to the shape of vasculature.
Speaker: Frits Post, TU Delft, Netherlands.
Title: Advances in Virtual Colonoscopy and Multi-Field Data Visualization.
In this presentation we will show recent results from two different projects in medical visualization at TU Delft: virtual colonoscopy, and visualization of multi-field medical data. In virtual colonoscopy, visualization and semi-automatic detection of colonic polyps are necessary for the early diagnosis of colon cancer. Computer-aided diagnosis (CAD) is a helpful addition to visual inspection for preselection of suspected colonic polyps, especially in mass screening of low-incidence populations. The colon surface shape can be characterized and visualized using lines of curvature, or streamlines that follow principal curvature directions. We show that the patterns of such curves are a good indicator of polyps. Features strongly correlated with true positive detections are calculated on lines of curvature and used for the polyp candidate selection. Visualization of shaded colonic surfaces is also enhanced with lines of curvature. In the second project we present interactive visualization techniques based on multi-field data, combining diffusion tensor imaging (DTI), functional MRI, and structural MRI data to assist in planning brain tumor surgery. The functional impairment caused by the operation is minimized by analysing predicted damage to the affected area, and also to fiber bundles that connect these areas with other parts of the brain. We can extract the fiber bundles that pass through a region around the tumor. These bundles can then be explored by filtering on distance to the tumor, or by selecting a specific functional area. This approach enables the surgeon to combine all this information in a highly interactive environment in order to explore the pre-operative situation.
Speaker: Guido Brunnett, TU Chemnitz, Germany.
Title: Visualization of Cross Sectional Data for Morphogenetic Studies.
We report on a visualization system that has been implemented to study cellular events which regulate morphogenesis in large organs or whole embryos. For data acquisition, the high resolution scanning system "Huge Image" is used. Huge Image integrates single fields of view of the camera into one high resolution image of an entire histological section. These images are loaded in our visualization system for semi-automatic contour extraction and cytological diagnosis. Based on the contours, a 3d reconstruction of the object is created and visualized. Since several iterations of data acquisition and reconstruction may be necessary to create high quality models, the time needed for model creation is a critical factor. The talk focuses on our approaches to reduce time and storage requirements for contour creation and 3d reconstruction.
Speaker: Stefan Gumhold, TU Dresden, Germany.
Title: Higher Order Surface Primitives for Interactive Visualization.
Several visualization techniques for medical applications and applications in life sciences map data types such as diffusion tensors, stream lines or molecular datasets to a geometric representation that is based on higher order surface primitives. Typical surface primitives are spheres, ellipsoids, cylinders, generalized cylinders or more complex surfaces that are composed of implicit surface patches such as dynamic skin surfaces.Current graphics hardware does not support direct rendering of those primitives. Therefore, one typically tessellates the surfaces primitives resulting in increased processing times and a significantly augmented size of the processed data. This hinders interactive visualization of large data sets. This talk introduces a technique called incremental raytracing, which exploits programmable graphics hardware to allow for direct rendering of ellipsoids, generalized cylinders and clipped quadric patches without the need for tessellation. In incremental raytracing the graphics hardware is programmed to compute ray-primitive intersections and fed for each primitive with a bundle of rays that completely covers the primitive. To achieve maximum rendering performance, the work is optimally split among the processing units of the graphics hardware.
Speaker: David C. Banks, Florida State University, USA.
Title: Global Illumination Applied for Brain Imaging.
We demonstrate how global illumination can be incorporated into a rendering tool for displaying geometry acquired from MRI and Diffusion MRI of the brain.
3. Visualization in Life Sciences: “Visualizomics”.
Thursday, 10:00am – 12:00pm.
Session Chair: Thomas Ertl.
Speaker: Gunther Weber, University of California, Davis, USA.
Title: PointCloudXplore: A Visualization Tool to Examine Quantitative Relationships Between 3D Gene Expression Pattern.
During animal development, complex patterns of gene expression form that provide positional information within an embryo and are used to organize cells. To allow a more rigorous understanding of the underlying gene regulatory networks, the Berkeley Drosophila Transcription Network Project (BDTNP) has developed a suite of methods that support quantitative, computational analysis of three- dimensional (3D) gene expression in early Drosophila embryos with cellular resolution. We report on first components of a visualization tool, PointCloudXplore that allows exploring relationships between different genes’ expression in these data sets visually. Two aspects of gene expression are of particular interest to the BDTNP: (i) gene expression patterns defined by spatial locations of cells expressing a gene, and (ii) relationships between expression levels of multiple genes. We present a visualization tool that takes both these aspects into account by providing two corresponding types of views on PointCloud data: (i) embryo views show gene expression on a model of the physical embryo, and (ii) abstract expression space views, such as 3D scatter plots, discard spatial information and plot expression levels of multiple genes with respect to each other. To support more complex analyses, we use brushing to define subsets of embryo cells and emphasize associated data within a view. Furthermore, we utilize linking to show in additional views the expression data for a group of cells that have first been highlighted as a brush in a single view, allowing further data subset properties to be determined.
Speaker: Dieter Fellner, Graz University of Technology, Austria.
Title: BioBrowser – Visualization of and Access to Macro-Molecular Structures.
This interdisciplinary research project BioBrowser addresses topics from biology and computer science and aims at innovative access methods to the biological knowledge about molecules, which is directly 'embedded' into the interactive visualization of the macro molecules (proteins, DNA, RNA). The BioBrowser makes this possible on standard (e.g. in no way specialized) computers even for very complex molecules. The immediate linking of knowledge items in bio-science and the visualization of the molecules offers of a new way of interaction in the biological research: the 3D-structure of a molecule as the central access tool to relevant information in biology itself as well as in other related disciplines. Or, phrased differently in the context of a digital library for biological information, the 3D model of the molecule becomes the central document type in the daily biological work.
Speaker: Vijay Natarajan, University of California, Davis, USA.
Title: Segmenting Protein Surfaces.
We describe an algorithm for segmenting a protein surface into segments that identify characteristic features of the molecule. Topological analysis of a scalar function defined on the surface and its associated gradient field reveals the relationship between the features of interest and critical points of the scalar function. Feature-driven simplification of the scalar function merges segments together resulting in a hierarchical segmentation of the surface. We use the segmentation to identify rigid components of protein molecules and to study the role of cavities and protrusions in protein-protein interactions.
Speaker: Julia Löcherbach, Ernst-Moritz-Arndt-University Greifswald, Germany.
Title: Visual Analysis for Gel-free Proteomics.
In proteomics protein mixtures are analyzed for qualitative and quantitative purposes. To support this analysis we developed a visual exploration system for gel-free approaches. The most widely used method involves liquid chromatography (LC) where the peptides (enzymatically digested proteins) are separated and mass spectrometry (MS) where spectra of the peptides are aquired. The most abundant peptides are further analyzed in a second mass spectrometry step (tandem MS or MS/MS). LC-MS data has the properties of being non-equidistantly distributed in the time dimension (measured by LC) and scattered in the mass-to-charge ratio dimension (measured by MS). We describe a hierarchical data representation and visualization method for large LC/MS data. Based on this visualization we have developed a tool that supports various data analysis steps. Our visual tool provides a global understanding of the data, intuitive detection and classifcation of experimental errors, and extensions to LC/MS/MS, LC/LC/MS, and LC/LC/MS/MS data analysis. We have developed methods to unite intensity peaks that refer to the same peptide. These peaks occur due to the presence of isotopes and possibly multiply charged ions. This deisotoping and deconvolution step is visually documented by our system, such that misclassifcation can be detected intuitively. For differential protein expression analysis we compute and visualize the quantities of up- and down-regulation of test vs. control experiments. In order to compute the differential expression, the experiments need to be registered. For registration we perform a non-rigid warping step based on landmarks. The landmarks can be assigned automatically using protein identifcation methods.
4. Visualization in Medicine: Data Processing.
Friday, 8:00am – 10:00am.
Session Chair: Frits Post.
Speaker: Jörg Meyer, University of California, Irvine, USA.
Title: Large-scale Volume Rendering of Multimodal Image Sets.
Arising from the clinical need for multimodal imaging, automated large-scale image processing, registration and volume rendering techniques for simultaneous processing of image data from multiple sources have been developed. The algorithms satisfy real-time data processing constraints, as required in a clinal routine. We present an integrated pipeline for multimodal diagnostics comprising of multiple-source image acquisition; efficient, wavelet-based data storage; automated image registration based on mutual information and histogram transformations; and texture-based volume rendering for interactive rendering on multiple scales. We will discuss how multimodal images can be stored and processed efficiently after histogram transformation and registration. We will also discuss how the variable resolution issue when using different modalities can be resolved efficiently by using a wavelet-based storage pattern.
Speaker: Jennis Meyer-Spradow, University of Münster, Germany.
Title: Supporting Depth and Motion Perception in Medical Volume Data.
Understanding spatio-temporal relations from a single still image is a challenging task. However, there are many application areas where dynamic visualization cannot be used and the user can only view a still image. In this talk we introduce two sets of visualization techniques. One for supporting the user in perceiving depth information in 3D angiography images and one for depicting motion inherent in time-varying medical volume datasets. In both cases no dynamic visualization is required since the proposed techniques have been developed to be applicable in a single static image.
Speaker: Hamish Carr, University College Dublin, Ireland.
Title: Applying Topological Analysis to Medical and Biochemical Applications.
Topological analysis in the form of contour trees and Morse-Smale complexes has developed rapidly over the past ten years in visualization circles. For biomedical and clinical tasks, however, the principal difficulty is defining or selecting suitable problems to which to apply these methods, and developing working relationships with collaborators in these disciplines. I will discuss the directions being taken at University College Dublin to bring together researchers in diagnostic imaging, visualisation and biochemistry, both to apply topological analysis to application problems, and to use application problems to derive further insight into topological techniques.
Speaker: Han-Wei Shen, Ohio State University, USA.
Title: Navigating Multiresolution Volume Data.
Having the ability to navigate through large scale multiresolution volume data allows the user to flexibly inspect features of various scales and trade image quality for speed. One important factor for the user to select a suitable data resolution is the resulting image quality and the computation cost. In this talk, I describe an image-based level-of-detail selection algorithm and an intuitive user interface for interactive navigation of multiresolution volumetric data. The design of our quality metric is based on an efficient method that evaluates the contribution and distortion of multiresolution volume blocks. We developed a user interface called LOD map – a representation of LOD quality and a visual interface for navigating multiresolution data exploration. Our measure for LOD quality is based on the formulation of entropy from the information theory. A LOD map is generated through the mapping of key LOD ingredients to a treemap representation. This visual interface not only indicates the quality of LODs in an intuitive way, but also provides immediate suggestions for possible LOD improvement through visually-striking features. It also allows us to compare different views and perform rendering budget control.
Speaker: Georges-Pierre Bonneau, IMAG, France.
Title: Substructure Topology Preserving Simplification of Tetrahedral Meshes.
Interdisciplinary efforts in modeling and simulating phenomena have led to complex multi-physics models involving different physical properties and materials in the same system. Within a 3d domain, substructures of lower dimensions appear at the interface between different materials. Correspondingly, an unstructured tetrahedral mesh used for such a simulation includes 2d and 1d substructures embedded in the vertices, edges and faces of the mesh. The simplification of such tetrahedral meshes must preserve (1) the geometry and the topology of the 3d domain, (2) the simulated data and (3) the geometry and topology of the embedded substructures. Although intensive research has been conducted on the first two goals, the third objective has received little attention. This talk focuses on the preservation of the topology of 1d and 2d substructures embedded in an unstructured tetrahedral mesh, during edge collapse simplification. We define these substructures as simplicial sub-complexes of the mesh, which is modeled as an extended simplicial complex. We derive a robust algorithm, based on combinatorial topology results, in order to determine if an edge can be collapsed without changing the topology of both the mesh and all embedded substructures. Based on this algorithm we have developed a system for simplifying scientific datasets defined on irregular tetrahedral meshes with substructures. We demonstrate the power of our system with real world scientific datasets from electromagnetism simulations.
5. Visualization in Life Sciences: Data Processing.
Friday, 10:30am – 12:30pm.
Session Chair: Wolfgang Straßer.
Speaker: Robert van Liere, CWI, Netherlands.
Title: Visualization of living cells.
During the past 6 years our visualization group has closely collaborated with cell biologist researchers at the Swammerdam Institute of Life Sciences. The organisation of the living cell is the general problem that biologists are interested in. Throughout the years, we have developed a number of visualization methods and systems that helped the users to gain insight in these problems. In this short presentation, I will reflect on this work and identify three tough problems in which I believe should be put on the visualization agenda: 1. high level motion analysis: extract and render 'meaningful' motion 2. comparitive visualization: fuse and compare 'wet' and 'dry' data 3. multi-scale visualization: 10^4 levels of spatial and temporal scales. I will give some examples of how we have approached these issues.
Speaker: David Duke, University of Leeds, UK.
Title: Grid-enabling Visualization: Coarse versus Fine-grained Approaches.
Visualization applications in the biological sciences can involve both high computational costs and/or integration of data from heterogeneous sources. They are thus strong candidates for building on grid technologies. However, grid computing is far from routine in visualization. This talk looks at one of the problems, the kind of abstractions from which applications are assembled. Two ongoing projects at Leeds illustrate alternative approaches: coarse-grained components composed within a web/grid services model, and recent work on pipelines defined from fine-grained functions.
Speaker: Jos Roerdink, University of Groningen, Netherlands.
Title: Visualization of biological networks in interactive environments.
Visualization is becoming ever more important in the life sciences, as it generically contributes to the interpretation of data which are of high dimension and/or large size. Techniques from graph visualization are increasingly applied to represent, retrieve, display, and explore biological networks, such as phylogenetic trees, metabolic pathways or regulatory genetic networks, either as traditional two-dimensional images or in three dimensions, using interactive displays and virtual environments. Emphasis is put on interactive manipulation of visualized structures by providing users with tools to search, reorganize, control the level of detail (pan and zoom), interrogate, and derive new useful information. I will discuss some general requirements for such systems and briefly mention a number of systems which have recently appeared. Almost invariably current systems are limited in terms of interactivity, adaptiveness of views, possibilities for collaborative work, knowledge representation, allowable model dynamics, and literature coverage. I will briefly outline some recent more generic approaches based on graph visualization frameworks, and discuss plans to extend the current tools to techniques for simultaneous visualization of many cellular processes as modules which can be combined in whole "virtual cell" visualizations. This is joint work with the Department of Molecular Genetics, University of Groningen.
Speaker: Christian Heine.
Title: Visualization of Barrier Tree Sequences.
The increasing complexity of models for prediction of the native spatial structure of RNA molecules requires visualization methods that help to both analyze and understand the models and their predictions. This presentation introduces a visualization method for a sequence of barrier trees. The barrier trees of these sequences are rough topological simplifications of folding landscapes -- energy landscapes in which kinetic folding takes place. The folding landscapes themselves are generated for RNA molecules where the number of nucleotides increases. Successive landscapes are thus correlated and so are the corresponding barrier trees. The landscape sequence is visualized by an animation of a barrier tree that changes with time. The animation is created by an adaption of the foresight layout with tolerance algorithm for dynamic graph layout problems. Since it is very general, the main ideas for the adaption are presented: how to construct a supergraph of the trees, how to lay out that supergraph, and how to build the final animation from its layout. The presentation is concluded by an example animation, techniques that further emphasize relevant information in the animation, and a discussion of the value of the visualization to RNA secondary structure prediction.
6. Visualization in Medicine: Modeling.
Friday, 2:00pm – 3:00pm.
Session Chair: Lars Linsen.
Speaker: Min Chen, University of Swansea, UK.
Title: Visualizing the Unknown.
This presentation examines various techniques used for facial aging simulation, which is traditionally studied under craniofacial measurement and analysis, or image processing and computer vision. Many previous attempts were made in formulating a global age progression model, and had so far limited success in obtaining a practically usable technique. In this presentation, the speaker will describe a data-driven approach where the age progression model is considered unknown for any specific person, and instead, such a model is obtained dynamically by learning from existing facial progression data using genetic programming. This example represents a shift from the traditional 'measurement + modeling' approach to a 'data + learning + visualization' approach. It is believed that such an approach will help broaden the scope of research in medical and life science.
Speaker: Gerd Reis, University of Kaiserslautern, Germany.
Title: Towards a virtual ultrasound tutoring system.
Nowadays ultrasound is a very important visual examination tool in medicine. Ultrasound is a save and non-invasive imaging modality and it does not put patients to great inconvenience during examination. However, ultrasound has some limitations (i.e. resolution, anisotropy, artefacts) that call for a very intensive education. At least in Germany the education is done during the clinical practice. This is of advantage in that students can learn under qualified supervision (e.g. real pathologies, discussion of special cases during an examination, etc). On the other hand there is a bunch of drawbacks, e.g. there is always need for a supervisor, the education cannot be performed at any time and of course there is need for real patients with respective pathologies. Indeed, studies have shown that during a standard education of approximately one year students can learn about 80% of the important pathologies, only. To remedy the situation we propose a tutoring system, which is in the scope of recent research at the DFKI GmbH. Based on data from a standardized finding system and several ontologies, geometric models of a pathologic heart are created and transformed into tissue descriptions. Now students can examine the data set which is blended into a mannequin using a pose tracked virtual transducer. Their findings are recorded using a second instance of the finding module such that measurements and other examination details as well as the diagnosis as such can be compared automatically. Current research focuses on the automatic generation of pathologic heart models and the real time generation of artificial ultrasound images, the latter being in the special scope of this presentation.