Computer-aided Detection for Screening Mammography

摘要

Computer-aided detection of breast cancer in screening mammography is just now becoming a clinically useful tool for radiologists. Systems are being implemented to act as second readers of mammograms. These systems automatically detect suspicious regions in digitized films, or directly acquired digital images, then prompt the radiologist to examine the identified regions more carefully. The purpose of using a computerized system as a second reader is to reduce the chances of overlooking visible signs of malignancies and hence improve the breast cancer detection rates in screening programs. This paper presents an overview of a mammogram image analysis system developed at Intelligent Systems, M.D., Inc. for identifying all primary signs of cancer in mammogram images. The system has been tested on a set of 88 mammography cases of four views each. The cases contain a total of 96 malignant abnormalities. The system currently achieves 91% sensitivity at an average of just over one false prompt per image. Ongoing system refinement and updates are expected to further improve performance.