Content-Based Retrieval of Medical Images

We consider the requirements for the design and implementation of Image DataBase (IDB) systems which support the retrieval of medical images by content. Attention is focused on a methodology for the efficient representation and retrieval of medical images based on spatial information. The content of medical images is represented by Attributed Relational Graphs (ARGs) holding features of objects or regions and relationships between such objects or regions. The method can answer queries by example, such as “find all image examinations similar to Smith’s examination”. The performance of the method has been evaluated using a dataset of 13,500 simulated, but realistic computed tomography and magnetic resonance images. Key-Words: image database, medical image, attributed relational graph, query by example.