高分辨率肺部CT图像纹理分割与配准算法研究-生物医学工程专业论文.docxVIP

ABSTRACTABSTRACT ABSTRACT ABSTRACT To keep pace谢t11 the development of modem medical imaging techniques， research to realize Compmer Aided Diagnosis or Computer Intelligent Diagnosis has become a hot topic and general trend in the field of medical imaging processing and analysis．In order to successfully detect pathology on medical image，many of the image processing techniques，such as registration，segmentation and construction of digital probabilistic atlas have to be researched．In this case，a lung’S atlas—based Computer-Aided Diagnosis(CAD)system is presented． For most cases of lung image data,texture trees of lung，contain airway,artery and vein，and usually reveal vital functional and anatomical messages．These messages may have certain relationship with some lung lesions(tracheal stenosis，tumor etc)，SO that Can be used for diagnosing and locating the lesions，even for establishing of anatomical atlas．By using Adaptive Region Growing Approach and segmentation algorithm based on SMDC-Connection Cost，we segment airway and vessel orderly． Then according to the anatomical feature between each vessel and bronchus，we separate the lung arteries from the vein． Airway skeleton is very important for building lung airway parameters mathematical model and picking up the landmarks．Firstly,the two extracted airways ale skeletonized by an efficient voxel—coding based method．Then，the points wim distinct topological features in the two breath—hold skeletons are picked as the 1andmarks and the correspondence anatomical relationship between the landmarks are described by Huffman-coding． The rigid registration is a key process in the CAD system．Through registration process，we can get anatomical messages which are very major steps for building statistical atlas and realiszing atlas-based pathology automatic detection．We deform the expiratory image volume by using the affine transformation．Then the landmarks of two thoracic are picked and matched by using the above process．Finally,a coarse