Brought to you by
ECR 2013 / C-2474
An automatic system for segmentation, matching, anatomical labeling and measurement of airways from CT images
Congress: ECR 2013
Poster No.: C-2474
Type: Scientific Exhibit
Keywords: Chronic obstructive airways disease, Computer Applications-3D, CT, Respiratory system
Authors: J. Petersen1, A. Feragen1, M. Owen2, P. Lo3, M. M. W. Wille4, L. H. Thomsen5, A. Dirksen5, M. De Bruijne6; 1Copenhagen/DK, 2Waterloo, ON/CA, 3Los Angeles, CA/US, 4Hellerup/DK, 5Gentofte/DK, 6Rotterdam/NL
DOI:10.1594/ecr2013/C-2474

Methods and Materials

Airway tree segmentation

The fully automatic segmentation process begins by detecting a seed point within the trachea as a black circular region within the top slice. From this seed point the airway centreline tree is then iteratively extended by searching for locally optimal paths that most resemble the airway centrelines based on a statistical model derived from a training set. An initial segmentation of the airway lumen is then grown in a tubular fashion around the centreline. The method is described in detail in Lo et al. (2009). See fig. 1 for an illustration of the process.

 

Airway wall segmentation

The interior and exterior wall surfaces are found using a graph built from the initial airway tree segmentation. The graph directs the search for the surfaces along flow lines inward and outward from the initial segmentation and is able to simultaneously optimize the position of both, taking image gradient, surface smoothness and separation into account. Flow lines do not intersect and so the found surfaces are guaranteed not to self intersect. This results in good segmentations even in the otherwise very difficult to segment bifurcation regions. The approach is described in detail in Petersen et al. (2011). See fig. 2 for an illustration.

 

Airway branch matching

Using deformable image registration, corresponding points within two or more images of the same subject are found. Mass-preserving image registration as described by Gorbunova et al. (2012) is especially well suited to lung CT images because the model allows intensity variation, which is common due to differences in inspiration level, under the assumption of preservation of lung mass. See fig. 3 for an illustration of a registration result. Once corresponding points are found the centrelines are transferred to a common coordinate system in which the airway branches are matched based on distance and angle. Branches not found in all available images of the same subject are removed to correct possible segmentation errors and ensure corresponding measurements. The matching process is described in Petersen, Gorbunova et al. (2011) and fig. 4 illustrates the process.

 

Airway branch labeling

The anatomical names of all branches down to and including the segmental level are assigned based on distances to a training set of expert labeled trees. Distances are measured in a geometric tree-space, a mathematical construction, incorporating both topology and centreline shape differences. The method is described in Feragen et al. (2012).

POSTER ACTIONS Add bookmark Contact presenter Send to a friend Download pdf
SHARE THIS POSTER
2 clicks for more privacy: On the first click the button will be activated and you can then share the poster with a second click.

This website uses cookies. Learn more