A new combined live wire and active surface approach for volume-of-interest segmentation

Volume-of-interest (VOI) segmentation is an important prerequisite for quantitative analysis of structure and function in multimodal tomographic images (e.g., MRI, CT, PET, SPECT images) of humans and small animals. A suitable VOI segmentation method should allow user interaction to position the VOI where desired and be automated to support the user as much as possible. Thus, semiautomatic methods are best for this kind of requirements. The new combined live wire (2D) and active surface approach (3D) for semiautomatic VOI segmentation was developed further and optimized. The method was evaluated based on software phantoms of different size and image quality (noise level). Compared to a simple polygon-based initialization of the active surface model, the live wire-based initialization yields better results with less user interaction. By optimization, the whole procedure could be improved in segmentation quality and computation time. Segmentations of a set of MRI images from different small animal and human brain studies of different image quality show the excellent performance of the method in segmenting various kinds of VOIs.