My master's thesis is about illustrative visualization of volumetric medical data, with an emphasis on GPGPU-based methods. The basic problem concerns the fast and spectacular displaying 3d datasets resulting from CT, MRI or PET scans in a medical context. To ensure that the viewer can see exactly what he/she wants to, I am exploring non-photorealistic ways to display the data. Since we are dealing with huge datasets and possibly complex ways of processing, I am using GPGPU methods - shaders written in Cg - for the majority of calculations. I have chosen the most general and flexible framework for displaying: raycasting through the volumetric data. Each voxel has a color and a transparency, with these being appropriately accumulated along each ray casted. The dependency of color and opacity on voxel values is controllable in real-time. They can also depend on location, for example, only showing some region of interest. In the current version, a plane can be specified, and the data is clipped to a "layer" around it - of course, the plane's location is controllable as well. Most of the controls are keyboard-based, the exception being the drag-and-drop mechanism of "data cube" rotation.