PET/MRI Medical Image Fusion Research

Registration and fusion technology of multimodality medical image have been widely used in clinical diagnosis and treatment because of the Multimodality images providing more information in detail. Magnetic resonance imaging-MRI is an anatomic imaging modality that has relatively more spatial texture information. Positron emission tomography-PET is a functional imaging modality which reveal human physiological and metabolic information, but the spatial resolution of PET is poor.This paper researches fusion technology of PET and MRI images based on DICOM standard. It focuses on DICOM structure analysis, PET/MRI medical image registration and fusion algorithm. The main work are summarized as follows:Firstly, the practical application of DICOM standard is discussed, and the main architecture and format of DICOM standard is described in detail. Then the parsing and conversion methods of attributes and pixels of DICOM file are given.Secondly, this thesis describes an approach of the conversion between of DICOM3.0Format and DIB Format and how to display them, and also investigates the registering problem of PET/MRI. It realizes the image pretreatment including gray degree transformation, image enhancement, geometric transformation and so on. And after that, it does the work of registration of PET/MRI by the mutual information.Finally, this paper analyzes the principle of the image fusion algorithm that respectively based on IHS color space and retina model in detail and points out the pros and cons of these two algorithms in the spectral and spatial texture effects. Based on the characteristics of these two algorithms, an improved IHS and retina model fusion technique is proposed to overcome the shortcomings of them to achieve coordinated color and spatial features. A series of Simulation experiments are conducted respectively based on retina enlightening model, HIS triangular model, HIS cylindrical mode, Brovey algorithm, Discrete Wavelet Transform, and aTrous wavelet transform. And spectral difference method, average gradient method and mutual information method are adopted as evaluation criteria to do overall merit of these fusion methods. The results show that the improved image fusion algorithm can retain features of the color and spatial texture better and it is superior to the other experimental methods. At last, the full text is summarized and the next step of the research work is expected.