||Segmentation plays an important role in image based medical applications to extract different kinds of information. However, segmenting (volume) data is still a time-consuming, often manual slice-by-slice process that becomes more and more an issue with the growing number of datasets acquired during medical examinations nowadays.
In collaboration with the Department of Radiology, Uppsala University Hospital, who provided a number of (manually segmented) MRI datasets of the human abdomen, we looked for a faster way of determining the liver volume in combination with using commodity graphics hardware. Based on the seeded region growing approach, some preprocessing and the actual region growing algorithms has been implemented on the graphics hardware. This allows for an easy visualization of the segmentation's progress "on the fly", and thus enables the user to directly interact with the process, e.g. to control the segmentation in ambiguous regions. In addition, the overall performance of those non-graphical algorithms can be highly improved by exploiting the processing power of modern graphics hardware.