Extraction Method Of LungTexture Tree From HRCT Images

With the fast development of medical imaging technique, the large numbers of high resolution digital imagesare produced. On the one hand, these images can offer the clearer anatomic or functional structures for diagnosis. On the other hand, the reading burden is so heavy that misses and errors increase along. In the case, the researches about computer aided diagnose (CAD) expand a lot. Computed tomography is the primary imaging modality for investigation of lung function and lung diseases. There are lots of report about lung function measurement and disease diagnosis at home or abroad,the better result of these research based on better segmentatiion and registration.lung texture tree(airway,artery,vein) contains important anatomical message about lung and has anatomical relation of lung lesions(tracheal stenosis,tumor el),and provides important information for CAD system design and research of virtual bronchoscopy and The establishment of anatomical atlas.Because of the importance of lung texture, this thesis put forward a method of segmentation of airway and vasclular based on the feature of them,then remove them from the roughly segmented lungs,what left is a pure lung parenchyma .The next we apply a skeleton extraction method based on two distance filed to the airway segmented result,then we can obtain the anatomical point of lung airway,at last we store the skeleton information and anatomical point in a xml file for further use.The idea of this thesis is as follows.Firstly,we uses a adaptive segmentation parameter airway extraction method to segment airway from orginal volume HRCT images.Secondly,we segment the lung parenchyma roughly,then uses SMDC-connection cost to segment lung vascular.At last,we apply a skeleton extraction method based on two distance filed to the airway segmented result to get a one pixel width and connected skeleton,and also obtain a the anatomical point of lung airway, then we store the skeleton information and anatomical point in a xml file for further use.There are mainly three aspects in the research:the automated segmentation of airway,automated vascular segmentation of lung,extraction of airway skeleton.In order to segment lung airway and vascular,we first roughly segment the lung parenchyma which reserves the detail of airway branch and vascular .We proposed a method of segmenting airway tree branch by branch according to the anatomic structure knowledge of lung airway,for each branch,we uses a method like wavefront,from the initial front take in the adjacent pixel that meet the condition to form a new wavefront iteratedly,for each new wavefront we detect the connectedness for the branch split,then stop.when the wavefront stop,we analysis the result,if it is bad ,we adjust the parameter to repeat.For segmentation of lung vascular ,because of the character of SMDC-connection cost operator that can fill holes according to their morphology size,we segment the vascular as follow:first we chose a small size parameter to "fill in" the noise in the image ,second we chose a bigger one to "fill in" the vascular structure in the image ,then we subtract the above two result and chose a gray threshold ,we can achieve the segmented lung vascular.At last ,we compare the result with the result segmented by fuzzy connectedness method. At last,we extract the skeleton from the segmented airway by a method using two distance filed,the extracted skeleton represents the topology structure of airway which is connected and one pixel width,and located in the center of branch.We store the skeleton information in a xml file. We can use the skeleton information to find automate anatomic landmark to use in the registration in lung HRCT image.In this research, we realize the automated segmentation of airway and lung vascular,and extraction of skeleton and automate find the anatomical point in airway.Therefore, it is valuable to some extent.This research was sponsored by the key project program of The Education Department of Anhui Province (project No. 2006KJ097A) and Nature and Science Foundation of China (project No. 60771007).