Airway Segmentation for Low-Contrast CT Images from Combined PET/CT Scanners Based on Airway Modeling and Seed Prediction

Abstract

The combination of positron emission tomography (PET) and computed tomography (CT) scanning provides a superior access to both functional information and anatomical structures of the airway tree. However, airway tree segmentation from such low-dose and low-contrast CT images is a challenging task due to the limitation of the image resolutions. Complex anatomical structure of airway tree and partial volume effect pose other difficulties in airway segmentation. Conventional airway segmentation algorithms often produce less than satisfying results. In this paper, we propose a novel method for fully automatic airway tree segmentation for CT images from combined PET/CT scanners. In our method, airway modeling is used in seed extraction and prediction, and a new strategy is devised for identifying potential airway branches that are not detectable by conventional 3D region growing. In comparison with traditional 3D region growing segmentation algorithm, our method outperforms with not only retrieving considerably larger number of branches, but also providing more accurate geometric information.