An indexed atlas for content-based retrieval and analysis of mammograms

We describe the development of an indexed atlas of digital mammograms to facilitate content-based retrieval and comparative analysis of mammograms for computer-aided diagnosis (CAD) of breast cancer. Specifically, the requirements and the design of the components of the indexed archival and retrieval system are examined. In order to facilitate search by categories, the mammograms in the atlas are indexed by case number, year of acquisition, category (normal, benign disease, and malignant disease that could be screen-detected or interval cancer), and the presence of signs of disease such as masses, calcifications, bilateral asymmetry, and architectural distortion. In the initial phase of the project, mammograms with masses have been indexed with objective diagnostic features related to their shape, edge definition, and texture. Interfaces to the atlas provide tools for selection and retrieval of cases by text-based or content-based indices. The system should assist radiologists and clinical specialists in CAD of breast cancer.