Browsing by Author "Rößler, Friedemann Andreas"

Now showing 1 - 1 of 1

Open Access
Bridging the gap between volume visualization and medical applications
(2009) Rößler, Friedemann Andreas; Ertl, Thomas (Prof. Dr.)
Direct volume visualization has been established as a common visualization technique for tomographic volume datasets in many medical application fields. In particular, the introduction of volume visualization techniques that exploit the computing power of modern graphics hardware has expanded the application capabilities enormously. However, the employment of programmable graphics processing units (GPUs) usually requires an individual adaption of the algorithms for each different medical visualization task. Thus, only few sophisticated volume visualization algorithms have yet found the way into daily medical practice. In this thesis several new techniques for medical volume visualization are presented that aid to bridge this gap between volume visualization and medical applications. Thereby, the problem of medical volume visualization is addressed on three different levels of abstraction, which build upon each other. On the lowest level a flexible framework for the simultaneous rendering of multiple volume datasets is introduced. This is needed when multiple volumes, which may be acquired with different imaging modalities or at different points in time, should be combined into a single image. Therefore, a render graph was developed that allows the definition of complex visualization rules for arbitrary multi-volume scenes. From this graph GPU programs for optimized rendering are generated automatically. The second level comprises interactive volume visualization applications for different medical tasks. Several tools and techniques are presented that demonstrate the flexibility of the multi-volume rendering framework. Specifically, a visualization tool was developed that permits the direct configuration of the render graph via a graphical user interface. Another application focuses on the simultaneous visualization of functional and anatomical brain images, as they are acquired in studies for cognitive neuroscience. Moreover, an algorithm for direct volume deformation is presented, which can be applied for surgical simulation. On the third level the automation of visualization processes is considered. This can be applied for standard visualization taks to support medical doctors in their daily work. First, 3D object movies are proposed for the representation of automatically generated visualizations. These allow intuitive navigation along precomputed views of an object. Then, a visualization service is presented that delegates the costly computation of video sequences and object movies of a volume dataset to a GPU-cluster. In conclusion, a processing model for the development of medical volume visualization solutions is proposed. Beginning from the initial request for the application of volume-visualization techniques for a certain medical task, this covers the whole life cycle of such a solution from a prototype to an automated service. Thereby, it is shown how the techniques that where developed for this thesis support the generation of the visualization solutions on the different stages.