全局最优混合活动轮廓模型肺组织三维重构

Three dimensional (3D) reconstruction of medical images is an important computer-aided visualization technology, which has an important impact on the medical diagnosis and adjuvant therapy. In computed tomography (CT) images, the densities of lung tissues are much lower than those of peripheral tissues, such as the chest wall and bones. As a result, the pulmonary parenchyma is easy to be sheltered by peripheral tissues while reconstructing 3D models of the lung tissues, which heavily affects the 3D observation and quantitative analysis. This paper proposes a non-sheltered 3D reconstruction method for lung tissues aiming at chest CT slices. Firstly, a global optimal hybrid active contour model is proposed to accurately segment the lung tissues from sequential chest CT images. Then, the Shear-Warp based maximum intensity projection volume rendering is applied to reconstruct 3D models of lung tissues. Experiments on CT data of 30 patients showed that the proposed hybrid active contour model could achieve accurate lung segmentation. The mean Dice similarity coefficient is 0.983, the mean bidirectional Hausdorff distance is 6.1 mm, and the average segmentation efficiency for each patient is 4.5 min. The non-sheltered 3D models of pulmonary parenchyma can be obtained by reconstruction using Shear-Warp volume rendering.