Segmentation of the pulmonary vascular tree

A method to semi-automatically segment the pulmonary vascular tree from computed axial tomography (CAT) images is presented. The main goal is to aid the diagnosis and treatment of acute respiratory distress syndrome and pulmonary embolism. The proposed methodology is based on a variational region growing method and a multi-scale vessel enhancement filter. Preliminary studies were made with a 20 CAT image dataset that included healthy and pathological lung scans. The results were satisfactory, although in some cases vessels were not correctly distinguished from other thin structures such as mucus-filled bronchi, nodules and airway walls connected to vessels.