ABSTRACT
Current computed tomography (CT) technology allows for high resolution imaging of the complete chest. These scans have become indispensable in thoracic radiography, but have also increased substantially the amount of data to be analyzed by radiologists. Automating the analysis of such data is a growing necessity. In CT images, inside the lungs, main pulmonary structures (such as airways and vessels) can be clearly observed. This paper presents an algorithm to segment the pulmonary vessel tree in CT images. Our method starts separating the lungs from the rest of the image. Then, the segmentation of the vessel tree inside the lung’s area is performed by a region growing technique combined with hysteresis thresholding. An experimental evaluation based on real CT images demonstrated the outcome of the proposed method is generally consistent with a visual segmentation.
