Detecting micro-calcifications in mammograms by using an intelligent computer-aided detection algorithm

It is well known that clustered micro-calcifications (MCC) on mammograms are important hints of breast cancer. Nevertheless, it is a complex and difficult task for radiologists to detect the clustered MCC from the tissue background of mammograms only by the naked eye. This paper describes a method for computer-aided detection of MCC in digital mammograms. The detection algorithm mainly consists of two different methods. The first one, based on the difference-image technique, recognizes high-frequency signals and very high-frequency noise. The second one is able to extract high-frequency signals by exploiting a wavelet based noise suppression and artificial neural network (ANN) classification. In the false-positive reduction step, false signals are separated from MCC by means of an AND operation on signals from the two methods. The algorithm was tested with a series of clinical mammograms. The results were compared with the judgment of radiological experts, and a sensitivity of 78% was obtained at a relatively low false-positive (FP) detection per image.