The Study Of Medical Image Retrieval Method Based On Hausdoeff Distance

With the application of network technology in the medical field,medical images data,such as X-ray images,B-ultrasound images,CT images,MRI images has been increasing explosively,which play an important role in clinic,teaching and scientific research.Medical image database stores a large number of medical records.It can provide reference to doctors in correctly identify the condition identifying pathologic condition correctly,reducing the rate of missed diagnosis and misdiagnosis.Efficient storage as well as obtaining useful medical images for doctors has become a difficult problem in the medical field.Therefore,it is of great significance for clinic medical to study efficient and accurate medical image retrieval method.Compared with the traditional text-based medical image retrieval technology,content-based medical image retrieval technology has been paid more and more attention by researchers.In this paper,two key factors of medical image of features extraction and similarity metrics in content-based image retrieval technology are studied to improve the superiority of medical image retrieval.Primarily,this paper mainly studies medical image characteristics and similarity measures.In the process of medical image retrieval,due to the imaging characteristics of medical images,the majority of images are gray scale images.Analyzing the texture features of grayscale images which is the main basis for doctors to judge and diagnose pathology.Hence,in this paper,we will focus on the texture features to describe medical images.Currently,distance measurement method usually used as similarity measure of medical image retrieval.Among different distance measurement methods,we choose Hausdorff distance as the similarity measure function,which is better than the Euclidean distance in the principle and widely studied and applied.Then,via selecting appropriate texture features to describe medical image content.By fusing texture features to improve the superiority of medical image retrieval.To improve improving the accuracy of medical retrieval,this paper focusedon the medical image retrieval approach by texture features fusion based on Hausdorff distance.We constructed and implemented a method of texture features serial fusion based on Hausdorff distance,that is,to extract Tamura texture features and wavelet texture features of brain MRI medical images and lung CT medical images,respectively.Serial fusion texture features as a new feature vector.The Hausdorff distance is used to measure the similarties of the new feature vectors,and to verify the validity of this retrieval method.Finally,to consider the method of features fusion in medical image,we constructed and implemented a method of texture features with equal weighted fusion based on Hausdorff distance.The Hausdorff distance is used to measure the similarties of Tamura texture features and wavelet texture features respectively.The similarities are equally weighted according to the weight value of 1:1.The obtained result is taken as the total similarity measure result.On the basis,this medical image retrieval method of texture features with adaptive weighted fusion is proposed,using Echo State Network(ESN)to train medical images,to set the adaptive weight of different images.Finally,weight value is assigned to the corresponding texture feature,and the obtained value is taken as the total similarity distance.To verify the availability of two retrieval methods.Experimental results show,the proposed Hausdorff distance has better medical image retrieval performance.For medical image retrieval,it can improve the retrieval accuracy by using Echo State Network to set the adaptive weight to fusion texture features.