Retrieval Driven Classification for Mammographic Masses

Accurate diagnosis is pivotal for successful treatment for breast cancer. High chances of survival are possible, if malignancy is detected at an early stage. Mammography is the most efficient and widely accepted modality for screening breast cancer. In this paper we propose a decision support system based on image retrieval which retrieves similar pathology based mammographic images to serve the physician in the diagnosis of breast cancer. The work explores how to use the retrieved similar cases as references to improve the classification performance. The rationale is that by incorporating the closeness information for decision making improves classifier performance rather than making decision from whole database. Experiments were carried out on DDSM database utilizing 4300 images of breast cancer. The results demonstrated the effectiveness of proposed system and show the vitality for clinical applications.