Abstract: Knowledge of structure features of pulmonary fissures is useful in diagnosis of lung diseases.In CT images,pulmonary fissure segmentation is a formidable task due to various of factors such as intensity variability,clutters and pathological deformation.To overcome the problem,a computerized scheme based on vector partition model and 3 D skeletonization model are introduced for pulmonary fissure segmentation.Motivated by the fact that fissures appear as line-profiles in 2 D space,we adopt an oriented derivative of stick filter by merging the intensity and orientation information to construct fissure vector field,which can effectively discriminate between fissures and other tissues.Considering the fact that fissures appear as planar patches in 3 D space,an orientation partition scheme is adopted to separate fissure patches and other tissues in different orientation partition,thus resulting a good performance in clutters suppression.To achieve a complete fissure segmentation,a post-processing pipeline based on 3 D skeletonization model is introduced for clutters removal.When applying the computerized scheme to publicly available LOLA11 datasets,the median F1,FDR and FNR are 0.893,0.068 and 0.128,respectively.Which indicates that the computerized scheme can efficiently segment pulmonary fissures in CT images.