Segmentation of lung lobes in volumetric CT images for surgical planning of treating lung cancer

Study has shown that three-dimensional (3D) visualization of lung cavities has distinct advantages over traditional computed tomographic (CT) images for surgical planning. A crucial step for achieving 3D visualization of lung cavities is the segmentation of lung lobes by identifying lobar fissures in volumetric CT images. Current segmentation algorithms for lung lobes rely on manually placed markers to identify the fissures. This paper presents an autonomous algorithm that effectively segments the lung lobes without user intervention. This algorithm applies a two-stage approach: (a) adaptive fissure sweeping to coarsely define fissure regions of lobar fissures; and (b) watershed transform to refine the location and curvature of fissures within the fissure regions. We have tested this algorithm on 4 CT data sets. Comparing with visual inspection, the algorithm provides an accuracy of 85.5 - 95.0% and 88.2 - 92.3% for lobar fissures in the left and right lungs, respectively. This work proves the feasibility of developing an automatic algorithm for segmenting lung lobes.