The Research Of CT Pulmonary Vascular Segmentation Based On Level Set Method

Segmentation of pulmonary vascular in CT Image is an important basic part of computer-aided diagnosis system,the segmentation results of pulmonary vessels can provide important basis for the diagnosis and localization of common pulmonary diseases.Traditional segmentation methods,such as threshold method,region growing method and clustering method,are very sensitive to noise.They fail to segment lung blood vessels with changeable structure,and many smaller blood vessels are easily missed.The level set method can deal with the topology change of image better,and can complete segmentation better.The research has theoretical significance and application value.According to the characteristics of CT images of pulmonary vessels,two different level set algorithms are used to segment pulmonary vessels,which has strong pertinence.Moreover,P-M filter is added to bi-directional level set evolution,which can make the segmentation of pulmonary vessels more accurate.The main research work can be summarized into five aspects.(1)Reserve the theoretical knowledge and propose the method for segmentation.Firstly,read the literature,study the lung CT images,and summarize their features and differences of the lung and pulmonary vascular.For its advantages and disadvantages,learning level set method mainly.Proposed algorithm scheme,using two level set segmentation algorithm,complete pulmonary vascular segmentation in two steps.(2)Lung parenchyma was extracted by distance regularization level set.According to the characteristics of one-way evolution of distance regularization level set and the special structure of lung,two initial positions are selected to evolve the chest CT images,extract lung parenchyma,remove most interference information and highlight the target lung blood vessels to be segmented.(3)Strengthen the pulmonary vessels.Removing the boundary of lung parenchyma by corrosion and operation to reduce its effection.Through the index operation to enhance image,and the pulmonary vascular,makes the segmentation area more explicit.(4)Segmentation of pulmonary vessels based on P-M filter bi-directional level setalgorithm.Bi-directional level set evolution segmentation of lung blood vessels,is not affected by the initial position,to improve its energy,add P-M filter,enhance the boundary,make the segmentation boundary information enhancement,is conducive to boundary recognition,improve the segmentation accuracy.(5)The algorithm implementation and experiment.Four groups of thoracic CT images were used for the experiment,and the results were obtained respectively.Selection of supervised and unsupervised two kinds of evaluation methods to evaluate,validate the feasibility and effectiveness of the algorithm.There are two innovations in this study:(1)Level set algorithm is used twice.One-way evolution distance regularization level set algorithm is used to extract lung parenchyma,according to the special structure of lung,two lung lobes are segmented at two initial positions to ensure the integrity of lung parenchyma information.(2)In view of the complex topology of pulmonary vessels,bi-directional level set algorithm is used to segment the lung vessels,which solves the segmentation difficulties caused by its special structure.P-M filter is added to bi-directional level set algorithm to enhance the boundary,improve the recognition degree of the boundary,make the evolution target more significant and improve the segmentation accuracy.The shortcomings of the research are:(1)The selection of initial position of one-way level set evolution has a great influence on the results,the location selection has a great influence on the segmentation accuracy,the segmentation accuracy of the algorithm results has been improved,but the excess rate has also been improved,which is the disadvantage of the algorithm in this paper.(2)In the program to set the initial contour and position,need to constantly modify in the experiment,the degree of automation is low,the speed is slow.