一种人体气道树快速自动提取算法的设计

为了从胸部CT图像中提取气道树，提出一种人体气道树的快速自动提取算法，其中包括三维区域生长、三维波传递和形态学优化3个步骤，通过多次迭代完成最终提取.区域生长的作用是提取气道树主体，三维波传递和形态学优化用于提取外围细支气管.在波传递中采用模糊逻辑判据，并利用形态学特征防止泄漏.将算法应用于28例数据，结果表明，利用该算法均能成功提取第5级或第6级支气管，且未泄漏；此外，该算法运行速度特别快，提取一个气管树耗时一般小于2s.

Algorithm Design for Fast Automatic Segmentation of Airway Tree

To extract the tracheobronchial tree from lung CT images, an automatic rapid algorithm was proposed, which consisted of three steps including 3D region growing, 3D wave propagation and morphological optimization. The final segmentation results were obtained through iterations of the three steps. The basic airway tree was segmented during region growing, and then the bronchial trees were obtained through 3D wave propagation and morphological optimization. A fuzzy logic criterion was adopted to end the wave propagation and the morphological features were used to prevent leakage. The proposed algorithm was applied to 28 cases and it was shown that the airway tree with the 5th or 6th bronchia could be extracted successfully while no leakage occurred. Besides, the proposed algorithm worked very fast, taking less than 2 seconds to extract one airway tree.