Labeling of the Airway Tree¶

To allow for automatic localization of airway locations, CIRRUS Lung includes anatomical labeling of the segmented airway tree. After airway tree segmentation, the tree is divided into airway branches based on detected bi- and trifurcations. The labeling algorithm is based on training data acquired by manually labeling the airways upto segmental level in a set of scans. In this training data, a large set of characteristics of each labeled brach were collected and orientation, average radius, and angle with the parent branch were found most discriminative. Assuming independent normal distributions for these characteristics over the branches, we determine a probability that a particular label should be assigned to a branch in the scan that is being labeled. To assign the actual label, the probability of the branch is considered conditioned on those of its children and grand children. Labels are assigned in a recursive process, starting at the trachea, which is known from the initial seed point of the airway segmentation algorithm. At every step all unlabeled branches with a labeled parent are considered and the most probable labels are assigned. 

Technical publications about airway tree labeling algorithms included in CIRRUS Lung¶

E. van Rikxoort, W. Baggerman and B. van Ginneken. "Automatic segmentation of the airway tree from thoracic CT scans using a multi-threshold approach", in: The Second International Workshop on Pulmonary Image Analysis, 2009, pages 341-349. Abstract/PDF

B. van Ginneken, W. Baggerman and E.M. van Rikxoort. "Robust segmentation and anatomical labeling of the airway tree from thoracic CT scans", in: Medical Image Computing and Computer-Assisted Intervention, volume 5241 of Lecture Notes in Computer Science, 2008, pages 219-226. Abstract/PDF DOI PMID