Diffusion-Based Applications for Interactive Medical Image Segmentation

Volume segmentation is an important part of medical diagnostics, and fast solutions which nevertheless can handle large datasets are required. This paper presents a diffusion-based method that can be used for volume smoothing and segmentation purposes, and has been implemented on the GPU in order to achieve interactive performance. We describe the nonlinear isotropic and anisotropic diffusion methods, which – in combination with seeded region growing – are also used for segmentation purposes. The procedures are embedded in a highquality hardware based volume renderer and provide interactive data handling and parameter editing.