Segmentation of airway trees from multislice CT using fuzzy logic

The segmentation and reconstruction of the human airway tree from volumetric computed tomography (CT) images facilitates many clinical applications and physiological investigations. The main problem with standard automated region-growing segmentation algorithms is leakage into the extra-luminal regions due to thinness of the airway wall during the process of segmentation. This phenomenon causes regions of lung parenchyma to be wrongly identified as airways. Main previous solutions to this problem include region of interest modification-based techniques, morphology-based method and fuzzy connectivity based method in which early leaks are detected and avoided. In this paper, an airway segmentation focusing on 2D line profile based evaluation of the degree of existence of airway wall using fuzzy logic is presented. New features are proposed and the usefulness of the features are evaluated. Comparison with a commonly used region-growing segmentation algorithm shows that the proposed method retrieves a significantly higher count of airway branches and less leaks. Our algorithm provides a way for fast realization of the major 3D airway trees. The algorithm succeeds in segmenting airways that have moderate to obvious airway walls in 2D slices. It provides a structure for follow-up branch growing algorithm.