Font Size: a A A

Research On Medical Image Registration Technology

Posted on:2014-01-16Degree:MasterType:Thesis
Country:ChinaCandidate:X H XueFull Text:PDF
GTID:2248330395487190Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
As a important research field in Medical Image Processing,Medical Image Registrationhas been widely used in surgery navigation,lesions tracking,therapeutic review and otherclinical diagnosis and treatment. With difference of applications,there are many technicalproblems to be solved in accuracy, speed, and robustness of medical image registration.Basedon the review of various theories and technologies in Medical Image Registration,the paperstudies the methods to inhibit Local Extremum in the process of registration and investigateshow to increase the accuracy for large deformation image registration.The main work are asfollows:(1)The dissertation makes a study of the influence of different window functions ofinterpolation to Mutual Strictly Concave Function Measure,and designs a rigid registrationbased on this measure,which could inhibit Local Extremum effectively.Mutual StrictlyConcave Function Measure performs more sensitive to symmetry of PV interpolation windowneighborhood,when the window neighborhood is not symmetrical,Local Extremumphenomenon would be aggravated in the registration based on the measure.The methodsproposed in the paper is based on Mutual Strictly Concave Function Measure for thesimilarity measure,PV or3orders B-spline GPVE for interpolation function to rigid registermedical image.Experiments shows that the methods proposed could inhibit Local Extremumand get more accuracy without increasing cost.(2)The dissertation proposes G&MAD non-rigid algorithm,which considers theinfluence of deformation to both gray and gradient comprehensively.To decide the rightparameter of driving force,the algorithm takes account of the influence of deformation to bothgray and gradient at the same time in processing the problems of large deformed imageregistration.The result of experiment shows that G&MAD non-rigid algorithm proposed inthe paper could acquire more accurate effect for large deformed image registration thanMAD.(3)The gray of different medical images,which are acquired in different manners,displays differently,according to which we designed a non-rigid registration method fit to theregistration of non-consistent gray image based on G&MAD algorithm.At first the papertakes use of Exact Histogram Regulations technology to transform the gray of medicalimages to be registered,whose gray is non-consistent,then takes advantage of G&MADalgorithm to register the images have been preprocessd before.The experiments followsshows that the method suggested in the paper performs more suitable for graynon-consistency of images to be registrated.
Keywords/Search Tags:Medical Image Registration, Local Extremum, Mutual Strictly ConcaveFunction Measure, Exact Histogram Regulations
PDF Full Text Request
Related items