Automatic Segmentation for Pulmonary Vessels in Plain Thoracic CT Scans

As pulmonary CT has a lot of noise produced by thoracic imaging and partial volume effect, it is difficult that the computer segments out the correct blood vessels for plain thoracic CT. Therefore, after a deep investigation into the enhancement, segmentation methods and the upgrading ability of fractional differential operation, the paper proposes an automatic segmentation method for pulmonary vessel in the plain thoracic CT scans. The method steps are: pulmonary CT images have been enhanced with fractional differential operator firstly, and the image regions have been divided into sub-regions based on two control indexes, then a local optimal threshold is used for the vessels segmentation in each sub-region. The experiment results illustrate: the proposed method can segment out more complete low-level vessels, and the obtained vessel networks has more correct detail information. Compared with traditional methods, it has more accurate segmentation ability for pulmonary vessels in plain CT scans.