Automatic detection of microcalcification clusters using wavelet transformation and fuzzy logic

An estimated 215,990 new cases of breast cancer are expected to occur among women in the United States during 2004. Breast cancer ranks second among cancer deaths in women. Primary prevention seems impossible since the causes of this disease still remain unknown. Early detection of breast cancer is important. Computer-aided mammography diagnosis is an important and challenging task. An early sign of 30--50 percent of breast cancer cases detected mammographically is the appearance of clusters Mammographically detected clusters of fine, granular microcalcifications are an early sign 30--50 percent of breast cancer cases. We present a novel approach to microcalcification detection. We employ a fuzzy entropy principle and a fuzzy set theory to fuzzify the images. We use a wavelet transformation to enhance the fuzzified images. A thresholding method using both local and global thresholding is exploited to segment the microcalcifications. The local thresholding method is based on the local variances and means of two adaptive filter windows with different sizes. The global thresholding method is a p-tile scheme. We apply a denoising method based on a spatial relationship function to removeisolating spots. The clusters are detected and labeled. The free-response operating characteristic curve (FROC) is used to evaluate performance. Our method detects microcalcifications in very dense breasts. Compared to the results of existing algorithms on the same set of data, our algorithm achieves better results.