Medical Image Database System Based On Semantics

Medical image database is the kernel of medical information system, which is the basis of application of all kinds of image. It can be applied widely to PACS(Picture Archive and Communication System), telemedical system, medical archive system, and medical imaging system. In medical information system, DICOM (Digital Imaging and Communication in Medicine) standard is the technical guarantee that ensures all modalities device exchange image data compliantly. But the semantic image content was not included in standard, which makes it unsuitable for intelligent applications based on semantics in medical image system.In order to solve the limitation in DICOM, a medical image semantic model is proposed in this paper after discussing the information level composites in medical image. The semantic attributes and their relations are extended for DICOM, and the semantic extraction, storage of semantic information and retrieval of semantic content are explored. By using the visual presentative information, the visual content based retrieve experiments were made through typical image in domain of medical image. The abstractive content based retrieve experiments were made by the semantic matching procedure in constructed semantic network. In experiments we find that we can get the image category by simple visual detection in domain of medical image. Such results can be used as initial data set for further semantic analysis. The results also show that the method of semantic retrieve by semantic matching in semantic network is effective.The semantic extraction procedure is verified by experiment of extracting semantic content of microcalcifications in X-ray mammography by method of data mining. The image semantic attributes are defined by the private attributes in DICOM and are used to extent the semantic module in image database. For supporting the semantic research, a DICOM visual working platform is constructed.