Segmentation of lung vessels using radon transform

Abstract

Lung cancer is the most common cause of cancer- related deaths in the world. For the detection of lung cancer, pulmonary CT (computed tomography) image processing is required. Discrimination of vessels from other organs in the lung is required for the detection of lung nodules. This is an important step in the diagnosis of lung cancer since the size and shape of the nodules indicate the severity of lung cancer in the patient. Therefore, vessel segmentation has received a large amount of interest, and a variety of techniques have been proposed for this task. Vessel detection techniques have been widely used in brain MRI images and retinal images, and have been shown to apply to segmentation of lung vessels as well. Radon transform is a technique which transforms line-containing images to a radon transform domain where every line is represented as a peak or a valley. This helps to understand the intensity and position of the lines in the image. This technique has shown success in the segmentation of blood vessels in retinal images, and in this paper, we use it for the segmentation of lung vessels. Pre-processed lung CT images are partitioned and further processed. The Radon transform is then applied. Vessel mask parameters are defined and analyzed to obtain a final vessel map highlighting the blood vessels in the source lung CT image.