Detection of Microcalcifications in Digitized Mammograms using Discrete Wavelet Transform and Hybridized Algorithm

Breast Cancer, is a type of cancer that originates in the breast tissue. For the earlier detection of breast cancer, mammography is considered as the best modality by finding malignant (cancerous) lesions, masses and microcalcifications (MC's) in the breast tissue. MC's are the calcium deposits in the breast tissues with no regular patterns, shape and size. They are often located across non-homogeneous backgrounds; their intensity could be similar to that of noise. So, it is difficult to detect in the naked eyes even by the experienced radiologists. To assist them, a fully automated Computer Aided Diagnosis (CAD) system has been proposed. In the proposed work, mammogram images are enhanced using Discrete Wavelet Transform (DWT) for denoising and localization. The spatial location range High-High (HH) from DWT is considered and represented as data vector. It is, then, segmented using Possibilistic Fuzzy C-Means clustering (PFCM) algorithm. PFCM clustering algorithm segments the image into normal tissues and MC's suspected regions by finding atypicality values of each pixel of data vector. Finally, features of the segments are extracted using window-based feature extraction method and given as input to the Multi-Layer Perceptron (MLP) classifier to classify the tissue as normal tissue or malignant tissue. As per the experiments and results of the proposed system, the accuracy is calculated as 96.15%.