Lung Nodule Segmentation Using Pleural Wall Shape

Abstract

A lung nodule segmentation method is proposed to deal with juxta-pleural nodules in CT scans by smartly wrapping pleural wall shape into segmentation. The global pleural wall shape model is estimated by components analysis from adjacent CT slices to capture its invariant features of anatomical structure. In order to grasp more refined pleura features in each slice, the estimated global shape model is combined with local intensity and morphological features adaptively to produce final pleural wall segmentation through a level set based propagation algorithm. With the extracted pleural wall, the lung nodules can be segmented well by a level set method based on intensity contrast between nodule and background even when the nodules are attached to the pleural wall. The experimental results show that the proposed method can achieve DSC value of 0.70 on 175 juxta-pleural nodules from LIDC-IDRI database, outperforming the state of art method on this kind of nodules.