Research And Implementation Of Multi-modality Medical Image Registration And Fusion

Since the last century, medical imaging technology has experienced a development process which is from static to dynamatic,from formal to functional, from flat to three-dimensional. With the development of medical imaging technology, it produces many human formal and functional information for the medical diagnosis. A variety of medical imaging technologies have their own characteristics and provide information on different characteristics of the human body. Different medical imaging technologies make up for each other.Therefore,single-mode image can not reflect the information of the human body completely.In order to get more complete and more abundant shape and functional information, it is necessary for multi-modality medical image fusion.It contains more comprehensive and reliable information to help medical diagnosis. Because of different characteristics, it needs to integrate different information which is provided by multi-modality medical images. First, different images must match in the spacein order to complete the unification of the spatial location and anatomical structures. It is the medical image registration. Medical image registration is crucial for medical image fusion,it has a very important significance for the medical diagnose.This thesis has researched and implemented multi-modal medical image registration and fusion technology.The main contents are taken as follows:1.Medical image registration technologies are researched and analyzed. The research purpose, research background and significance have been clear.The state-of-the-art research at home and abroad are analyzed.2.The medical image registration method and its classification are introduced in detail. And medical image registration is analyzed space transformation method, image interpolation, optimization strategies and similarity measures.The steps and evaluation criteria of medical image fusion are introduced, and its methods and classification are also described. And medical image fusion method and classification are described.3.A registration algorithm of multiresolution medical images that uses a modified artificial fish-swarm algorithm combined with the Powell method is proposed to solve the local maximum problem of mutual information-based image registration.Fuzzy weighted normalized mutual information and normalized partial energy-weighted matching are proposed as new similarity measures; Multi-resolution strategy and HPV interpolation are also used, and global optimization is performed using the hybrid algorithm that combines a modified artificial fish-swarm algorithm and Powell's method. Experimental results and comparisons with some existing classical registration methods in multi-mode medical image registration show that the proposed algorithm improves the accuracy and performance of registration.4.In order to combine CT and MRI images to be better, this thesis presents a new CT and MRI medical image fusion method which use adaptive exponentially weighted fuzzy entropy fusion rules and improved regional information PCNN fusion rules. It is in the framework of multi-wavelet for CT and MRI medical image fusion. Firstly, the images are decomposed based on multi-wavelets.Then different image fusion algorithms are used for different frequency components. Experiments show that the proposed algorithm is superior to other fusion algorithms.It improves the clarity of the image and reserves the details of the image in a large extent, with the advantages of prominent edge information, high brightness and contrast.