Automatic segmentation of lung lobes in CT images based on fissures, vessels, and bronchi

Lobewise analysis of the pulmonary parenchyma is of clinical relevance for diagnosing and monitoring pathologies. In this work, a fully automatic lobe segmentation approach is presented, which is based on a previously proposed watershed transformation approach. The proposed extension explicitly considers the pulmonary fissures by including them in the cost image for the watershed segmentation. The fissure structures are found through a tailored feature analysis of the Hessian matrix. The method is evaluated using 42 data sets, and a comparison with manual segmentations yields an average volumetric agreement of 96.8%. In comparison to the previously proposed approach, this method increases segmentation accuracy where the fissures are visible.