The Research Of Medical Image Retrieval Based On Joint Index

At present in the field of medicine,medical imaging technology has become the most trusted auxiliary technology when a doctor conducts the medical diagnosis.Every day,the major hospitals generate medical images of different formats and sizes using CT,nuclear magnetic resonance and other imaging techniques.These massive medical images contain a large number of pathological information from different patients,so using the modern information science and technology to unearth the pathologists that can not be found by doctors has become a valuable research topic.Content based image retrieval technology can recognize images that are visually similar to the retrieved images using modern computer technology.This technique can help doctors find the most valuable medical images in a large number of images,this technique can provide valuable historical data support for medical clinical diagnosis,medical teaching and scientific research by exploring more cases with the same imaging features.Therefore,the research on content-based medical image retrieval are important for clinical diagnosis and medical research.This paper has done the following researching:The method of extracting and calculating the fusion features of medical image proposed in this paper,fusion features of medical image contain three parts: Global features of EVLAD images based on SIFT local invariant sub features,gray feature based on local sensitive hash,image texture feature based on gabor wavelet transform.Experiments show that the algorithm proposed in this paper can achieve better accuracy.In the clinical diagnosis,teaching and scientific research,it is often necessary to carry out the accurate research on the specific focus.Therefore,we propose a candidate set sorting algorithm for a specific lesion region based on region segmentation algorithm.We split the candidate region of the candidate set with the help of seed points specified by the doctor.Then we extracted the features of the lesion area and reorder candidate according to the features.By the above method,we realized the accurate retrieval of suspicious lesions and the candidate set reordering.In this paper,we also propose a joint index structure.First of all,the three features of the fusion medical characteristics were clustered separately.Then we combine the clustering centers from different layers and we get the index nodes.We only need to compute a small number of cluster centers to get multiple index nodes in this way.In this paper,we use weights to calculate the similarity between index nodes and images,so we can improve the retrieval efficiency and ensure the retrieval accuracy.In order to solve the problem of huge amount of medical image data,we design a distributed computing architecture based on the retrieval algorithm.We divide the large search tasks into several small tasks and execute them on different hardware.It greatly improves the speed of algorithm feature extraction,index construction and online query.In this way,we can solve the problem of huge amount of computation.