The Research And Application Of Medical Image Retrival Technology Based On Texture Feature

Accompanied by the popularization of medical information, the importance ofmedical image information is more and more obvious. image information relatedprovides a great deal of support and help for disease diagnosis and treatment, academicexchanges, medical teaching and scientific research. In order to achieve the sharing ofmedical information, medical images need to be managed scientifically and effectivelyin order to retrieve useful information quickly and accurately from the mass medicalimages. According to the retrieval result scientific clinical judgment is made, and theneeds stimulate further research interest in the field of image retrieval.At present, mainstream retrieval method uses keywords and key phrases whichcontains information to complete the retrieval, namely: comments are added manuallyas keywords to retrieve images and search conducts in massive image library. In thisway the retrieval of the image is convenient, but a large number of comments need to beadded manually, which causes the workload and time-consuming increased. Subjectivedifference, non-standardized or imprecise of terminology exist when comments areadded, and bottom features and visual features are described inappropriate. To solve thisproblem, new methods and technologies are needed to improve and refine imageretrieval. Thus retrieval techniques based on content emerges. The so-called content-based retrieval technology is based on the visual characteristics of the bottom of theimage changes, and the combination of these changes of the underlying visual featuresto retrieve images.Various low-level visual features of medical images are analyzed deeply, in particular,the characteristics of the underlying texture described in medical images have specialstatus and wide range of applications. Medical image retrieval method based on texturefeatures is mainly used to study human cardiac CT images and MR brain images. Firstlythe appropriate segmentation algorithm is selected for the object of study, and the imagetexture features of the algorithm is studied to find that Tamura algorithm and GLCMalgorithm has its particularity; Then texture feature extraction processing conductedusing the Tamura algorithm or GLCM algorithm after the image segmentation. Finally,medical image retrieval system based on texture characteristics is implemented. Themain contents is divided into six parts: image feature database and repository are built; DICOM images and BMP images can be browsed and displayed; Image preprocessingand segmentation; Tamura algorithm and GLCM algorithm are used to extractsegmentation image texture feature values; The weighted Mahalanobis distance is usedto complete the process to retrieve target image which match the retrieving imagesimilarly from the feature repository; image database is retrieved and the search resultscan be displayed.This thesis is based on developing platform of Microsoft Visual Studio2005andSQL Server2005, using C++language programming with the help of OpenCV and ITKtoolkit and based on C/S structure, the medical image retrieval system is completedbased on texture features. The experimental results show that the algorithm significantlyimproves the medical image retrieval recall and precision rate and also has somepractical value.