A model based approach to object detection in digital mammography

Breast cancer has the highest incidence level of any cancer among women living in developed countries. Early detection of masses, and microcalcification clusters, is presently the best means whereby breast cancer mortality rates may be reduced. To this end, high resolution digital mammograms, generally at the 50 μm resolution, are used with a view to automatically detecting any potential anomalies. This paper presents an original technique for the detection of such microcalcifications. This is done by specifying a parametric model and using this to gain estimates of object positions and sizes directly.