Study Of Medical Image Registration And Fusion Based On Brain Glioma

Brain tumor is one of the most serious tumor diseases.At present,more and more attention has been paid to the research related to brain tumor images.The diagnosis of brain glioma is mainly based on the CT and MRI slice scan images,and the fusion results are helpful for the diagnosis of brain tumors.However,the coordinate system between CT and MRI is not corresponding,and the existing registration algorithms for this problem are time-consuming and low accuracy.At the same time,the existing fusion algorithms can not well reflect the characteristics of CT and MRI images.At present,there is no related medical analysis and diagnosis software to provide services.Therefore,the above related content is studied in this thesis.In this thesis,CT images and MRI T1 images of patients' brain are used as samples to study the image preprocessing,image retrieval,image registration and fusion,glioma segmentation and so on.The main contents include:(1)The contrast enhancement of medical images increase the visual experience of images.And the removal of check beds from CT images can eliminate the interference.(2)From the point of view of image features,different methods are used to extract features and compare the degree of approximation between the features,so as to find out the most similar images in the two modes.(3)The principal component analysis(PCA)and parameter estimation are used to improve the image registration algorithm,and it is applied to multi-scale registration,which improves the performance and precision of the registration method.(4)The image fusion methods in frequency domain and spatial domain are compared,and the fusion effect is evaluated by using relevant fusion image evaluation criteria.CT value and bone window image are fused with corresponding MRI images to eliminate the interference of irrelevant pixels in CT images.(5)The GMM model is used to model the gray distribution of the image,and the optimal segmentation threshold is determined to realize the automatic selection of the segmentation threshold.The method of deep learning is used to segment the image,so that the segmentation is not the gray analysis of the image,but the use of higher-level features and segmentation through learning,which makes the algorithm more simple and intelligent.(6)In order to observe the brain gliomas from three-dimensional angle,the segmentation results of brain gliomas are visualized in this paper.In this paper,the algorithm is implemented and integrated into CT and MRI registration fusion analysis system.The experiments show that the improved multi-scale contrast enhancement algorithm has better visual experience,and the automatic similar image retrieval algorithm has better retrieval accuracy.Compared with the original method,the image registration algorithm improves the accuracy and efficiency.The GMM model and the segmentation algorithm based on deep learning have good adaptability and intelligence level.