Skip to Main Content
Classification of airway shapes in chest X-ray images may be useful in computer-aided detection of lymphadenopathy associated with pediatric tuberculosis. This paper presents an interactive approach for airway segmentation from chest X-ray images that may be used in an airway shape classification algorithm. A local normalization filter is applied as a preprocessing step to enhance the visibility of the airways. Segmentation is then performed with the aid of active shape models (ASMs), which are warped to a set of manually defined control points on the image to be segmented, using an affine transformation. Two shape models are built, one of which consists of points on the airway edges only and the other consists of points on the airway edges as well as points on the ribs. The ASMs are built from a set of manually segmented images. The Hausdorff distance is used to compute the accuracy of the segmentations with reference to a manual segmentation.