PET/CT Image Analysis And Its Application In Assistant Diagnosis And Treatment Of Lung Cancer

Cancer is malignant tumor,which is hard to cure and have a high mortality rate.The mortality rate of lung cancer is the highest,which win the first place in the death rate of malignant tumors.It has become one of the greatest killers of human health.The diagnosis and treatment of lung cancer mainly depend on information from lung PET/C T images.Not only analyzing the information provided by PET/CT images,but also subjective experience analysis is used to diagnose lung cancer and design the treatment plans.The traditional method of fusion and segmentation has the limited which cannot afford doctors objective advice.In this paper,intensive study of the PET/C T image of lung cancer is done.The purpose is to make use of the image algorithm to assist the doctor to judge,to outline the target area objectively,and to design more reasonable treatment scheme for the patients.The main work of this thesis is as follows: PET/CT image fusion and PET/CT image combined segmentation.In order to assist doctors to judge whether they have malignant tumors and stage the tumor,we do fusion of medical images of two modes of PET/CT.Because the traditional sparse fusion algorithm calculates the sparse decomposition with redundant DC T dictionary,the fixed DCT dictionary cannot be applied to all kinds of images.In this paper,we use the Steering Kernel Regression function as the feature to block the original CT image.Then we get the smaller dictionary related to the image blocks after classification.Finally use the smaller dictionary to get the sparsity coefficient.The dictionaries trained according to the classified image blocks are more structured,so the algorithm can retain the unique information of CT and PET images better.The focus area is bright,the edge is clear,the algorithm performance is improved.It can be more accurate and reliable to assist doctors to diagnose the malignant tumor.In order to assist doctors to design better tumor treatment plan and target area,we studied the automatic segmentation of PET/C T images,regard it as an objective basis.The traditional Geodesic Active Contour model(GAC)has the advantage of rapid edge evolution and accurate segmentation for the weak boundary.However,the algorithm only uses one kind of image modal to segment.Our algorithm is improved base on traditional Geodesic Active Contour model.Redesign its edge function base on both gradient information of CT image and PET image.The algorithm does the combined segmentation using two modes of medical image information.Because two kinds of information are both combined in the edge function,the convergence speed of the algorithm has been improved,and the edge of the segmentation area is also more reasonable and the boundary is more accurate compared with segmentation method based on the single PET image.