Modern medical imaging tools (CT, MRI) are able to capture high resolution volumetric data. The resolution of a high quality image can consist of more than 515x512x512 voxels. To process easily and receive the expected results, this huge amount of data must be presented in the right form to the user. For this purpose, the presentation by slices is not always the best way, because the 3D information can be lost. However, the real-time volumetric visualization is still a difficult task, even using modern CPUs. Therefore, it can be very useful to utilize the huge computational capacity of the Graphics Processing Unit (GPU). With CUDA, one can easily write programs for the GPU at high level, using a C like syntax.
Visualisation is just a part of the task. The user must be able to extract relevant regions from the data. This task can be performed with segmentation. However, multiple factors make this task hard. The first one is the size of the data. The running time of segmentation methods largely depends on the amount of data to be processed. The second one is the data itself. It contains only local density values, with noise from the measurement, and these values do not significantly differ between the different tissue types. So performing segmentation is not an easy task and it is still an active field of research. Using GPU can be useful in this task too, because its computational power can be utilized to help handling the huge amount of data.