Automated detection of masses in digital mammograms based on pyramid
He Wang, Lin-Lin Huang, Xiaojie Zhao · 2007
In this paper, a new approach to detect masses in digital mammograms is presented. Firstly, preprocessing procedure is carried out to enhance the contrast between mass and surrounding tissue based on the algorithm of exponential transformation, which maps a narrow range of high grey-level values into a wider range of output levels. After preprocessing, pyramid segmentation algorithm is used to segment suspicious regions. Six features extracted from the suspicious region are used as the input of classifier to discriminate masses from normal tissue regions. The binary decision tree is adopted as classifier due to its conceptual simplicity and computational efficiency. A set of 50 mammograms with 52 masses have been tested. We achieved a sensitivity rate of 86.5% at a 0.6 FP per image. The results show that the proposed method is quite effective.