Pulmonary Airway Tree Segmentation Algorithm In Three-Dimensional CT Images

As the most sensitive chest imaging mode, CT image is widely used in the diagnosis of pulmonary diseases. Among them, the rapid development of the Multi-slice computed tomography technology makes the computer aided diagnosis and quantitative evaluation of lung disease based on medical images become possible.As an important functional tissue in the human respiratory system, the tubular tree structure of pulmonary airway also provide the important reference for the interval division and location of lesion. Segmentation of pulmonary airway tree in CT images is the basis of lung disease analysis, pathological parameters measurement and subsequent image configuration, and is also an important part of the application of virtualbronchoscopy. However, due to the impact of the noise, motion artifacts and partial volume effects in CT images, the gray distribution of the pulmonary airway is very uneven, the local tube wall rupture is prone to occur at the terminal branches, and easily diffuse into the lung parenchyma caused leakage. In addition, the impact of the complex three-dimensional structure,uneven shape changes and mucous of airway tree make the problem becomes more complicated. In view of the above difficulties, we research from the aspects of preprocessing chest CT image, segmentation of pulmonary airway tree by the shape,density and fuzzy connectedness, etc. Specific research includes:1. In order to avoid the interference of independent tissues in the original CT image, we propose a simple lung region segmentation algorithm. By using threshold segmentation, connectivity analysis and morphological operation, the preprocessing of the chest CT images is accomplished, and the region of interest is defined to reduce the amount of the subsequent calculations.2. Considering the specific problem of the shape and density variety between large bronchi and small bronchi, we propose a method for pulmonary airway tree segmentation by combining region growing and morphological grayscale reconstruction algorithms. On the one hand, the trachea and large bronchi are segmented by using an improved region growing method on basis of an iterative hysteresis threshold, and a local volume explosion index is adopted to suppress the lateral leakage; On the other hand, the smaller bronchi are segmented using a morphological grayscale reconstruction algorithm, and the tube shape descriptor isdefined to remove the pseudo-tracheal regions; Finally, a complete airway tree is obtained by integrating segmentation results from the above two steps.3. Considering the problem of fuzzy in CT image, we propose a method for pulmonary airway tree segmentation by combining region growing and fuzzy connectedness algorithms. First, the trachea and large bronchi are segmented by using an improved region growing method on basis of an iterative hysteresis threshold, and the front voxel point of the segmentation results as the new seed point set; Then, the fuzzy connectedness between the seed and any point was calculated by constructing affinity function according to intensity homogeneity and tubular structure characteristics of the trachea. Besides, the fuzzy connectedness was segmented by threshold segmentation and the connected component analysis to obtain airway tree;Finally, a complete airway tree is obtained by integrating segmentation results from the above two steps.In order to verify the effectiveness of the algorithm, we set up a comprehensive evaluation system. Using the 20 publicly available test CT cases of the EXACT'09challenge for qualitative and quantitative evaluation, and the quantitative evaluation is conducted with the four kinds of evaluation criteria(bifurcation number, branch number, branch ratio and average operation time). Compared with that of other considered methods, the performance indices of the proposed methods are moderately better. Although the algorithm appears to be relatively simple and has low complexity,the proposed methods effectively prevent leakage.