Extractiong Method Of Airway Tree From Lung CT Image

In recent years,with the improvement and development of imaging equipment,medical imaging technology has been popularized aboardly. In clinical diagnosis,electronic computer X-ray tomography （CT）is the most popular for its highresolution and comprehensive functions. Precise and accurate analysis of CT imagescan improve the diagnostic accuracy for related diseases, but the higher resolution ofthe CT equipment the more CT slices will be produced, and analysing a largenumber of CT slices is a burden for image-analysing physician. In addition, the simpleand boring work is easy to lead miscarriage of justice. To change the situation, acomputer-aided analysis system for medical images was exploited.In the check of chest, chest CT images provide the information of anatomicalstructure for lung airway tree and lung vessel tree. From the information of lungairway tree, physicians can judge and locate lung diseases (such as：emphysema,pulmonary embolism, partial tracheal stenosis and lung cancer), further more the3Dlung airway tree image is used to guid the planning and the implementation forminimally invasive surgical intervention. Therefore, airway tree segmentation fromchest CT is an important research field in computer-aided CT image analysis systems.Because of the niose in CT image, there is breakage phenomenon happened when thelung bronchial extend into lung parenchyma. For the reasons that tiny bronchial wallis thin and the grey level of bronchial cavity is similar to its surrounding tissue, it iseasy to prone to leakage when segement airway tree based on gray level regulation. Itis a key point for realizing computer aided analysing system that whether airway treecan be segementated successfully without leakage, which is the basis of chest CTimage analysis. In this paper,the main research work are as follow:1) Parenchymal segmentation is an essential part of chest CT image processing. Tosolve the promblem caused by neglecting the relationship among slices in twodimensional parenchymal segmentation, the paper proposed a three-dimentionalparenchymal extracting method，which is based on morphology and3-D regionregrowing. The result shows that there is no iner-slice intermittent phenomenonusing new method.2) To overcome leakage and breakage problem happened in tiny bronchisegmentation, based on traditional method, this paper put forward morphologicalgradient based3D region growing algorithm. From the3D results, we know thatnew method improved the airway tree segmentation accurcy1-2order.3) To reduce the long time caused by large amount of3D CT image data, amulti-scaled analysising based C-V level set model has been proposed. In theprocess of chest CT image, wavelet is introduced into build multi-scale frame.Take C-V model as an example, under the framework of multi-scale the CV modelsegmentation Algorithm can greatly shorten the processing time, meanwhilereduce the impact of noise on the results.Finally, the authors especially thank the provided support for this work from NationalNatural Science Foundation (81000651) and Jiangsu Provincial Natural ScienceFoundation (BK2010236).