Research On And Implementation Of Unsupervised Segmentation Method For Liver CT Images

Cancer is one of the most serious diseases which harm to people’s health,and has been become the main reason leading to people’s death.However,among many types of cancer,liver cancer top in the list.In clinical practice,the most common treatments of liver cancer include liver transplantation and liver resection surgery.Whatever the treatment,the extraction of liver parenchyma always plays a crucial role in it.In addition,liver segmentation is the basis of the computer assisted diagnosis(CAD)and the 3d reconstruction of the segmentation results can provide doctors with visual diagnostic data.As a result,the study on liver CT image segmentation algorithm is of great clinical significance.Liver segmentation is a difficult subject in the field of medical image segmentation.The reason lies in many aspects,such as many tissues in the abdomen having similar density to liver,the adhesion with similar-density tissues,and the difference of liver shape between different CT images etc.In recent years,there appear a lot of segmentation methods at home and abroad including supervised segmentation method and unsupervised segmentation method.Due to involving human participation,the supervised method is relatively mature,while the unsupervised method is still an open problem and some of the existing algorithms cannot achieve the effective liver segmentation result.Aiming at the shortage of existing methods,this thesis employs different schemes and proposes a method named liver CT image unsupervised segmentation method.Region growing algorithm is one of the typical methods in the image segmentation and has advantages of being simple,feasible and faster in image segmentation.The segmentation method suggested in this thesis is based on spatial region growing and makes full use of the anatomical relationship between abdominal tissues.Compared with traditional region growing algorithm,it mainly has the following advantages.Firstly,the thesis presents a separation algorithm which can achieve the separation of liver and heart and applies to abdominal coronal CT slices to avoid the over-segmentation into heart during liver segmentation.Secondly,this thesis proposes a method named the separation of liver and muscle with the result of avoiding the over-segmentation into muscles during liver segmentation.Thirdly,according to the theory of confidence interval estimation,this thesis improves the traditional region growing criteria to enhance its adaptivity.Besides,based on the characteristic of thin joint between liver and stomach,the thesis adds a restricted condition to growing criterion avoiding the over-segmentation into stomach during liver segmentation.Fourthly,taking advantage of the spatial characteristic,the phenomenon that the growing of one liver lobe cannot access to another leading to leaked-segmentation will not happen during the spatial region growing of liver.Fifth,the whole procedure of liver CT image segmentation does not need people to participate achieving the unsupervised liver CT segmentation and improving the efficiency of segmentation.In addition,the 3D construction of liver is presented in the last chapter of the thesis.After a large number of experiments,it has been verified that the method proposed in this thesis can realize an accurate segmentation result for liver CT images and have realistic clinical significance.