A software framework for preprocessing and level set segmentation of medical image data

In this work a software platform for semiautomatic segmentation of medical images based on geometric deformable models will be presented. Including filters for image preprocessing, image segmentation and 3D visualization this toolkit offers the possibility of creating highly effective segmentation pipelines by combining classic segmentation techniques like seeded region growing and manual se-mentation with modern level set segmentation algorithms. By individually combining input and output of different segmentation methods, specific and at the same time easy to use segmentation pipelines can be created. Using open source libraries for the implementation of a number of frequently used preprocessing and segmentation algorithms allowed effective programming by at the same time providing stable and highly effective algorithms. The usage of modern programming standards and developing cross-platform algorithm classes guarantees extensibility and flexible implementation in different hard- and software settings. Segmentation results, created in different research projects will be presented and the efficient usage of this framework will be demonstrated. The implementation of parts of the framework in a clinical setting is in progress and currently we are working on the embedding of statistical models and prior knowledge in the seamentation framework.