Colloquy Cycle SS 2001

• Colloquy Cycle WS00/01
• Colloquy Cycle SS00
• Colloquy Cycle WS99/00
• Colloquy Cycle SS99
• Colloquy Cycle WS98/99
• Colloquy Cycle SS98
• Colloquy Cycle WS97/98
• Colloquy Cycle SS97
• Colloquy Cycle WS96/97

Segmentbestimmung im Thorax CT zur Planung von Lungenresektionen

Stefan Krass, MeVis, University Bremen, Germany

Das Bronchialkarzinom ist die Krebserkrankung mit der höchsten Mortalitätsrate. Diagnostik und Planung der chirurgischen Therapie erfordern eine exakte Lokalisierung des Tumors und eine möglichst genaue Prognose der postoperativen Lungenfunktion. Bei der Erfüllung dieser morphologischen und funktionellen Anforderungen würde die präoperative Bestimmung und Visualisierung der Lungenlappensegmente einen wichtigen Beitrag liefern. Der Vortrag beschreibt eine Methode zur Segmentbestimmung auf Grundlage computertomographischer (CT) Daten.

Der Bronchialbaum wird durch ein spezielles Bereichswachstumsverfahren segmentiert. Nach Skelettierung des Segmentierungsergebnisses und Überführung des Skeletts in eine Graphenrepräsentation können die Unterbäume der Lungenlappen und der Lappensegmente identifiziert werden. Ein auf Wachstumsmodellen basierender Algorithmus approximiert aus den identifizierten Unterbäumen und der ebenfalls segmentierten Parenchymgrenze die Grenzen der Lappensegmente.

In einer Machbarkeitsstudie wurde die Methode auf klinische Einzeilen- und Mehrzeilen-Spiral-CTs angewandt. Die Validierung erfolgte in-vitro anhand von Präparaten der menschlichen Lunge.

In klinischen CT-Daten war eine sichere Segmentierung des Bronchialbaums bis zur 3. Ordnung (Einzeilen-CT) bzw. bis zur 5. Ordnung (Mehrzeilen-CT) möglich. Die Validierung ergab eine Genauigkeit der Segmentapproximation von 70 % (Einzeilen-CT) bzw. 80 % (Mehrzeilen-CT).

Mit der vorgestellten Methode wird die Beurteilung der Lagebeziehung von Tumoren und Segmenten verbessert. Weiterhin ist eine verbesserte Abschätzung der postoperativen Lungenfunktion zu erwarten.

SCCH - Linking Science to Industry

Klaus Pirklbauer, Werner Winiwarter, Software Competence Center Hagenberg

This talk offers an overview of the research activities at the Software Competence Center Hagenberg (SCCH). After a general presentation of the SCCH, we provide a summary of the main research topics at the SCCH. In the second part of the talk we focus on current projects in the knowledge-based area of the SCCH. Finally, we desribe multilevel data mining methods in detail and present results for their application to image segmentation.

Towards High-Quality, Real-Time Volume Rendering

Torsten Möller, Simon Fraser University, Vancouver, Canada

Volume Rendering is a subfield of graphics that deals with the exploration, communication, and presentation of medical or scientific data. The presentation on a computer screen reduces the 3D nature of the data by one dimension. The 3D understanding of these data sets can be enhanced using so called motion parallax, i.e. the real-time interaction with the 2D display. Hence real-time rendering algorithms are crucial for the visualization of complex volumetric data.

In this talk I will survey typical volume rendering techniques and the current status of such algorithms. I will include the premise of high-quality visualization, since for many applications, especially medically based, reliability of the representation plays an important role. I will survey current software and hardware developments. Especially I will talk about several results for the improvement of splatting - one specific volume rendering method. I will argue for splatting to be one of the most promising volume rendering algorithms, since it can achieve high frame rates as well as high quality images. I hope that some conclusion I have to offer will stimulate some debate.

Visualization of Geological Data in Virtual Environments

In this talk we describe tools and techniques for the exploration of geo-scientific data from the oil and gas domain in stereoscopic virtual environments. The two main sources of data in the exploration task are seismic volumes and multivariate well logs of physical properties down a bore hole. We have developed a props-based interaction device called the cubic mouse to allow more direct and intuitive interaction with a cubic seismic volume. The device consists of a cube-shaped box with three perpendicular rods passing through the center and buttons on the top for additional control. The rods represent the X, Y, and Z axes of a given coordinate system. Pushing and pulling the rods specifies constrained motion along the corresponding axes. Twisting the rods typically rotations around the corresponding axes. Embedded within the device is a six degree of freedom tracking sensor, which allows the rods to be continually aligned with a coordinate system located in a virtual world. This device effectively places the seismic cube in the user's hand. We have also integrated the device with other visualization systems for crash engineers and flow simulations. In these systems the Cubic Mouse controls the position and orientation of a virtual model and the rods move three orthogonal cutting or slicing planes through the model. We have also developed a 3D texture based multi-resolution approach for handling massive volumetric data sets common in the oil and gas industry, the medical domain, and as a result from computer simulations. Due to the restricted texture memory available and the limited bandwidth into the texture memory, these data sets cannot be rendered at full resolution. Our approach uses a two-level hierarchical paging technique to guarantee a given frame rate. This technique displays lower resolutions of the data when a slice or volume rendering is moved fast through the data set, and fills in the high resolution, when the user slows down or stops. This behaviour correlates really well with motion blur.

Virtual Humans: Simulation of Individuals, Groups, and Crowds

Simulation, VR, and Entertainment applications (games, films) needs to have Virtual Humans able to move in a flexible and elegant way. They should more and more react to the other characters and to the user. Moreover, animating crowds is challenging both in character animation and a virtual city modeling. The problem is basically to be able to generate variety among a finite set of motion requests and then to apply it to either an individual or a member of a crowd. A single autonomous agent and a member of the crowd present the same kind of 'individuality'. The only difference is at the level of the modules that control the main set of actions.

Biography

Daniel Thalmann is a pioneer in research on Virtual Humans. His current research interests include Real-time Virtual Humans in Virtual Reality, Networked Virtual Environments, Artificial Life, and Multimedia. He is coeditor-in-chief of the Journal of Visualization and Computer Animation, member of the editorial board of the Visual Computer, the CADDM Journal (China Engineering Society) and Computer Graphics (Russia). He is cochair of the EUROGRAPHICS Working Group on Computer Simulation and Animation and member of the Executive Board of the Computer Graphics Society. Daniel Thalmann was member of numerous Program Committees, Program Chair of several conferences and chair of the Computer Graphics International '93, Pacific Graphics '95, ACM VRST '97, and MMM '98 conferences. He is Program Cochair of IEEE VR 2000. He has also organized 4 courses at SIGGRAPH on human animation. He has published more than 250 papers in Graphics, Animation, and Virtual Reality. He is coeditor of 25 books, and coauthor of several books including the recent book on "Avatars in Networked Virtual Environments", published by John Wiley and Sons. He was also codirector of several computer-generated films with synthetic actors including a synthetic Marilyn shown on numerous TV channels all over the world.