Content-Based Medical Imaging Image Retrieval

With the emergence of more and more modalities in hospitals, especially 256 rows CT and high-field Magnetic Resonance,Efficient methods of image retrieval and management are needed because of the complexity and importance of medical image.The traditional information retrieval techniques,which are based on text comparison,are not satisfied to retrieve large scale medical image database.So the content based image retrieval techniques have emerged as the times require.CBIR is on the basis of images' content itself which is extracted by computers,such as color,texture and shape.The system retrieves the feature vectors consisting of the above information from the large database to satisfy the users' requirement.CBIR aims at searching the similar images using low level features,and medical image retrieval needs it for the retrieval of similar images.Medical images contain not only a lot of content data,but also a lot of semantic information.So,the paper presents an approach by combining high level features(DICOM semantics)and low level features(texture and shape) to perform retrieval on medical image databases.At the first step,characteristics of medical imaging image and the establishment of image database are introduced,considering characteristics of CT/MRI and doctor' s image retrieval habits.The semantic information is extracted from DICOM header of images for the pre-filtering of the images.Then the extracted methods of texture and shape feature are discussed.In the experimental CBIR system,we choose DT-CWT for extracting texture features and Canny for extracting shape features.According to characteristics of the sample image,the proper DICOM headed information is chosed to pre-filter images.The DT-CWT features and Canny features of pre-filtered images and example images are extracted and compared to retrieval similar images.The method of image retrieval in PACS is based on the text,while the method of image retrieval in CBIR is based on the content.Radiologists require the organic combination of CBIR and PACS.So,at the end of the paper,an integration model is presented that relies on standard protocols and maintains the autonomy of both PACS and CBIR.Experimental results show that by combining the high level semantics (DICOM features) and low level content features(texture and shape) the retrieval time is reduced and the performance of medical image retrieval is increased.