Spatial and Spectral Features for Early Detection of Microcalcifications in Mammograms

In this paper a technique for early detection of breast cancer from digital mammograms is proposed. The proposed technique focuses on the detection of microcalcifications using multilayer back propagation (MLBP) neural network. A set of spatial domain features based on discrete Radon transform (DRT) and spectral domain features based on the discrete cosine transform (DCT) of mammograms are extracted from the X-ray image. The extracted features by the proposed methods are exploited to classify regions of interest (ROI’s) into positive ROI’s containing clustered microcalcifications and negative ROI’s containing normal tissues. The true positive rate (TPR) of the proposed system is 95.07 % with false positive rate (FPR) equals to 1.36 %. Compared with previous work in this area, the proposed system is superior in classification accuracy.