科研之友
微信
新浪微博
Facebook
分享链接
摘要
This paper presents an automated approach for extracting vessels from computed tomography (CT) images that is based on the multiscale analysis of the vessel structure in the image. According to the scale responses to local line-like structures, the feature vessel image is built to reduce noise and enhance vessel structures. The multiscale linkage model is used to build a parent-child relationship between feature images with adjacent scales. Fuzzy distance measures are developed to describe the fuzzy similarity of vessel structures in multiscale images. The fuzzy similarity is embedded into conventional fuzzy clustering framework to build new multiscale vessel segmentation algorithm. The proposed multiscale similarity linking fuzzy C-means (MSLFCM) algorithm optimizes segmentation on all scales of the feature images. The inter- and intra-scale constraints are combined to automatically extract vessels. The segmentation experiments are performed on synthetic and CT pulmonary vessel images. The experiment results demonstrate the satisfactory performance of the proposed approach.
