Study On Content-based Medical Image Retrieval

A mass of medical image data are generated everyday in the clinic when the medical digital image equipments such as CT, MRI, and PET-CT are used in the clinic works more and more. However, it becomes a significant problem which cries for solved that how to manage this large number of data and then apply them to the clinic in the process of diagnoses. The traditional archive management systems adopt image database with simple label-based or even the entire manual way to administer the image data, which is gradually unable to meet the needs of searches with the cosmic medical image database and is affecting the function of the image used in the diagnoses severely. Thus, the technique of Content-based Medical Image Retrieval emerges as the times require.Content-based Medical Image Retrieval is the application of content-based image retrieval in the field of the medicine. The goal of this paper is to organize the image retrieval with the medical images and provide a convenience and precise way to search the image for the physicians and the assistant suggestion to the diagnoses.Based on the technique of the content-based image retrieval and principle of the medical image and from the demands of clinic diagnoses, an experiment system of content-based medical image retrieval is developed after the requirement and the application of technique of CBIR used in medical are analyzed. From the example of intracranial hemorrhage, the search methods of mass intracranial CT images based on local color and spatial features, shape feature of Fourier descriptors, region of interest, combination features are all achieved.The results indicated that the search method based on local color and spatial features has shown excellent precision when the gray level of the organ and the pathological part in the images differs from each other. At the same time, the search method based on shape feature of Fourier descriptors has preferable inflexibility of translation and scale and rotation. In the method based on region of interest, the focus distribution information of the region of interest in the medical image can be reflect exactly by using the feature vector of the cumulation histogram so that the preciseness, robustness, all-sidedness, and efficiency of the search method are improved.