Research And Application Of Intelligent Medical Image Retrieval Technology

Medical images are very useful and indispensable in diagnosing diseases,and with time accumulation and the increasing number of medical imaging equipment and medical image types,hospitals already have a large number of medical image data.There are different kinds of medical image data,and different medical image data have different formats,meanwhile,there always exist many noises.So,the effective organization and retrieval of the medical images has been the focus of the study.The intelligent retrieval technology of medical images is the key to the construction of digital integrated medical system and intelligent diagnosis of diseases.Considering unique features of medical images,based on the technology of intelligent image retrieval,this paper mainly improves some technologies in intelligent medical image retrieval.At the beginning,some global features are selected for retrieval based on the global features of medical images.According to Characteristics of medical images and the retrieval results based on different features,this paper choose LBP texture feature as the feature of medical images.Based on the texture feature of LBP,an improved LBP algorithm is proposed,which divide medical image into non-uniform block and has an improved encoded mode.Experiments prove that this improved algorithm can increase the function of the retrieval system.The medical image is divided into blocks,which can add the spatial position information to the feature,improve the dimension of feature,and improve the performance of uniform LBP.At the same time,the improved encoding LBP operator increases the stability compared with the common LBP feature.This paper proposed two kinds of improved algorithms based on medical image ret rieval technology to solve the problem of word disorder in the visual feature of SIFT.The retrieval based on SIFT and global LBP feature of medical image is proposed,which fuse the SIFT feature vector and the global LBP features.This algorithm solves the problem of no word order and enhances the retrieval function.Meanwhile,due to the introduction of irrelevant interference factors,the retrieval result will be unstable sometimes.The retrieval based on SIFT-LBP feature of medical image is proposed,which combine the SIFT feature vector and the local LBP feature.The algorithm eliminate the influence of noise,at the same time it can enhance the retrieval function inthe retrieval system.In the relevance feedback module of the retrieval system,this paper uses the idea of ensemble learning technology,and select the Bagging ensemble learning model.During the learning process,the proportion of positive and negative samples in the training data is adjusted to make full use of the unlabeled image data.At the same time,the model makes up the problem of over fitting and variance of the SVM classifier and improves the generalization ability,which is applicable to medical image data sets.