Christoph Poliwoda

VIRIM: a massively parallel processor for real-time volume visualization in medicine

Architecture and applications of a massively parallel processor are described. Volumes of 256×256×128 voxels can be visualized at a frame rate of 10 Hz using volume oriented visualization algorithms. A prototype of the scalable and modular system is currently set up. 3D rotation around an arbilrary rotation axis, perspective, zooming, and arbilrary gray value mapping are provided in real-time. Multiuser access over high-speed networks is possible. A volume oriented visualization algorithm is used that is tailored to the requirements in medicine [5]. With this algorithm, small structures of a size down to the pixel resolution, and structures without defined surfaces can be visualized as well as semitransparent objects. One application of the system is therapy planning in heart surgery.

Reconstruction of 3D Catheter Paths from 2D X-Ray Projections

The diagnosis and therapy of intensive care patients requires the usage of several catheters inside the patients chest. The information about the position and path of the catheters inside the patients body is important for the doctor, but is nowadays not part of the clinical routine. One possible source of this information are CT or NMR scans, which lead to an organizational overhead and additional stress for the intensive care patient. To minimize the overhead we implemented an algorithm to extract the 3D path of catheters in the body of the patient from two or more standard X-ray images. The approach is based on only few assumptions, runs completely in three dimensions, and uses the X- ray images only as a guideline for the path reconstruction process. It shows an inherently robust behaviour against misleading structures in the X-ray images, like loops and intersections. The algorithm has been tested with a selection of test images, including images from the clinical routine.

Virim: A massively parallel processor for real-time volume visualization in medicine

Abstract Architecture and applications of a massively parallel system currently developed are described, which allows real-time visualization using volume oriented visualization algorithms. Volumes of 256 × 256 × 128 voxels can be visualized with a frame rate of 10 Hz. The system is scalable and modular, and will allow a multi-user access over high-speed networks. 3D-rotation around arbitrary rotation axis, perspective, zooming and arbitrary grey value mapping are provided in real-time. A volume oriented algorithm is used that is tailored to the requirements in medicine [H.-P. Meinzer et al., IEEE Comp. Graph. Appl. , 34 (Nov. 1991)]. With this algorithm, small structures without defined surfaces, e.g. , tumours, can be visualized as well as semi-transparent objects. One planned application of the system is heart surgery.

Evaluation of a real-time direct volume rendering system

VIRIM, a real-time direct volume rendering system is evaluated for medical applications. Experiences concerning the hardware architecture are discussed. The issues are the flexibility of VIRIM, the restriction to two gradient components only, the duplication of the volume data sets on different modules, the size of the volume data set, the gray-value segmentation tool, and the support of algorithmic improvements like space-leaping, early ray-termination and others. It turned out that flexibility is the main benefit and absolutely necessary for VIRIM. Given this flexibility the application areas of real-time rendering systems increase dramatically: Most of the user requirements focus now not on visualization but on general volume data processing. The most serious bottleneck of VIRIM is the limited volume memory that is integrated on the first prototype. The most frequently used tool of VIRIM is gray-value segmentation. It is highly useful if original, i.e. unsegmented data have to be dealt with, and if pre-segmented data have to be investigated. All other benefits and architectural shortcomings are not critical for the application areas of VIRIM, i.e. operation simulation and control in head surgery.

Multiresolution Data Handling for Visualization of Very Large Data Sets

We discuss some features of an experimental system for visualization of large (medical) volume data sets. Input voxel data sets are subdivided into blocks first. Then each block is decomposed into a multiresolution data representation by applying a reversible 3D integer Haar wavelet transform (S-transform). The resulting transform coefficients are encoded using a Golomb-Rice algorithm. For volume visualization, we selectively load and decompress blocks to a proper resolution. Then they are rendered into a common view. From our experiments, we learned that large volume data sets should preferably be stored using a multiresolution data representation. Depending on the size of the volume data set and the rendering mode, we also found that a block-based data representation can provide some advantages, but it may not always be the best choice.