Research And Application Of The Semantic Based Image Retrieval Technology In Medical Image System

Medical image is important for the doctors' diagnosis. Currently, digital medical image data increase rapidly, along with the development of computer technology. It's become an urgent need to find the way to get certain medical image from the vast information effectively. Analysis have been made on the current image retrieval methods, then, the semantic-based image retrieval technology has been applied to the medical field, and a deep study has been made with brain CT image as an example.Based on a thorough investigation, most of the current medical image retrieval systems are based on the attributes of image file or low-level features including color, texture, shape, etc. but, medical image is a kind of image which has particular domain features. Based on the analysis of doctor's diagnosis, the pathological feature has been taken as medical image's semantic feature, which develop the scope of retrieval.Combine the Medical imageology, the semantic understanding of text and image processing technology, a semantic-based image retrieval model has been proposed. First of all, semantic features are extracted automatically using image segmentation, mathematical morphology and other image processing technology; Second, the semantic features are represent by vector, and the diagnostic report can be generated by then; Finally, using the VSM (vector space model), vector weights are revised with the knowledge of the medical item relations, and the similarity algorithm is improved to support the similarity merging of multi-features.According to the retrieval model and retrieval algorithm above, a medical image retrieval system is designed and implemented. The experiment prove that the semantic based retrieval has been realized.