ABSTRACT
Breast cancer is one of the few diseases that uses asymptomatic screening as a method for controlling the disease. Several countries in the world have national screening programs, and many others have recommendations for regular mammographic screening for asymptomatic women. Consequently, mammography is becoming a high-volume subspecialty. Furthermore, it is becoming one of the most common areas for radiological malpractice suits in the United States (1). Radiologists read screening mammograms in batches, sometimes 100 or more at a sitting. Because only about 0.5% of these cases will have breast cancer, it can be difficult to be ever vigilant to find the often subtle indications of malignancy on the mammogram. Consequently, between 5% and 30% of women, who have breast cancer and have a mammogram, are diagnosed as normal. Computerized detection of breast lesions can be used by radiologists as a “second opinion” and thereby reduce the chances that a cancer is missed. Therefore, automated analysis of mammograms is the most active area in computer-aided diagnosis research.
