Research Of Indexing Technology In Medical Image Retrieval Platform

Nowadays,X-ray, Ultrasound, Nuclear magnetic resonance and other techniques has been used in the medical field more and more widely, content-based retrieval method for massive medical image is becoming an increasingly important area of research, and the efficiency of high-dimensional indexing method for content-based multimedia information retrieval plays a vital role. However, the efficiency of high-dimensional indexing method is often not ideal, in particular for the data whose dimensions exceeds a certain threshold value, there's almost no indexing method can completely solve the "dimension disaster" problem. So to find an efficient high-dimensional indexing method is the real medical image retrieval technology resources to the health system will be used in the course request. So finding an efficient method for high-dimensional indexing is the inevitable requirement to apply the retrieval technology of Medical Imaging Resource in real health care systemAs real data in high-dimensional space is not evenly distributed, Clustering method has been chosen to divide the entire data space into subintervals, and then the dividing of data points are no longer blind but has the Clustering features. Then calculate the approximate vector for original features of each interval using the approximate vector indexing method, and then obtain simplified approximate vector. Using approximate vector in searching can reduce I/O costs, but also reduce the computational complexity of CPU. And such methods are also having the inherent defects, such as low query precision after the first filter because of losing information caused by data compression and quantify. So when choosing VA-File method for high-dimensional indexing technology, optimize its storage structure to speed up the search speed, and improve its Matching Algorithm into Two-step query to increase query precision. Experiments show that when using the KNN-Search, the new index method has higher searching efficiency under the same dimension than the original approximate vector indexing method; this also means that the new algorithm can serve for higher-dimensional vector data.Based on requirements of Medical Image Resource retrieval platform, take an architecture design method based on search strategies. System decide flexible which algorithm to use for each step such as preprocessing, feature extraction or matching calculation based on the current search strategy. According to the Feature characteristics of feature information for medical imaging, realize the using of high-dimensional indexing techniques in Retrieval of medical imaging platforms by combination of approximate vector indexing method and clustering method. In the windows environment, encoded using java language to get a simple and efficient systems to store and retrieval the resources of medical imaging.