FIST: fast interactive segmentation of tumors

摘要: Automatic segmentation methods for tumors are typically only suitable a specific type of tumor in imaging modality and sometimes lack accuracy whereas manual achieves the desired results but is very time consuming. Interactive however speeds up process while still being able to maintain segmentation. This paper presents novel method fast interactive (called FIST) from medical images, which all somewhat spherical any 3d modality. The user clicks center belief propagation based iterative adaption initiated, thereby considering image gradients as well local smoothness priors surface. During that process, instant visual feedback given, enabling intervene by sketching parts contour cross section. The approach has successfully been applied liver CT datasets. Satisfactory could be achieved 15.20875 seconds on average. Further trials oropharynx tumors, prostate MR images lymph nodes bladder volumes demonstrate generality presented approach.