Detection of microcalcifications in digital mammogram using wavelet analysis
Yashashri G. Garud, Neha G. Shahare · 2013
Clusters of microcalcifications in digital mammograms are important and early sign of breast cancer. This paper presents CAD system for detection of clusters of microcalcifications in digital mammograms. Microcalcifications are tiny deposits of calcium in breast tissue. Dense nature of breast tissue and poor contrast of mammograms prohibit effectiveness in detecting microcalcifications. Thus, to detect and differentiate the microcalcifications from normal tissue, proposed system uses wavelet analysis. Proposed system also makes use of extreme learning machine which has better generalization performance at extremely fast learning speed. ELM also avoids problems like local minima, improper learning rate. In this proposed system, raw mammographic image taken from MIAS database and it is morphologically preprocessed to remove labels and noise. Then, windowing function is applied to extract sub images of 32×32. Sub images are decomposed into 4 levels and wavelet features are computed. Whole process is supported with Extreme Learning Machine which is used as classifier.