Segment Segmentation¶

To allow for regional analysis of lung parenchyma, CIRRUS Lung includes an automatic approximation of the pulmonary segments. The segmentation of the pulmonary segments is based on manual annotations of segment locations in 500 chest CT scans. The segmentation of the pulmonary segments is initialized by lobar segmentation, after which a classification into pulmonary segments is performed for each lobe separately using a Linear Discriminant Classifier trained  using the 500 manually annotated scans and employing features based on the fissures and the relative position of voxels in a lobe. The computation of the segmentation of the pulmonary segments is performed offline. The resulting segmentation can be inspected in CIRRUS Lung as a 3-dimensional overlay on the original CT data.

Technical publications about segment segmentation algorithms included in CIRRUS Lung¶

E.M. van Rikxoort, B. de Hoop, S. van de Vorst, M. Prokop and B. van Ginneken. "Automatic segmentation of pulmonary segments from volumetric chest CT scans", IEEE Transactions on Medical Imaging 2009;28:621-630.Abstract/PDF DOI PMID