A hyper-spaced data model for content and semantic-based medical image retrieval

Images are usually retrieved on the basis of either low-level features, such as texture or color; context, such as date of acquisition or author; or high-level features, such as real-world objects and relations. We propose a data model for medical image retrieval. Our model integrates several types of image features and projects them to a set of spaces. Each image feature is then represented by several dimensions. Furthermore, the medical image has an evolutionary content that changes over time and according to several events. Our proposition allows medical users to easily describe the content evolution of many images. To achieve that, a spatial knowledge model is integrated into our approach that verifies the coherence between feature spaces and assists both system and user whenever precision is needed or specific ambiguity occurs. We briefly expose our Web-accessible implementation called MIMS (Medical Image Management System) developed with Java