Spectral Graph Wavelets for the Classification of Microcalcifications in Mammograms

Among the various carcinoma occurrences, breast cancer remains the most female malignancy in the world. The existence of MicroCalcifications (MCs) is a primary sign of breast cancer and their diagnosis process is still a complex problem. Nowadays, digital mammography technique is used as the most common and effective tool in screening mammography. In this study, an automated MCs classification system is proposed based on Spectral Graph Wavelet Theory (SGWT) and K-Nearest Neighbour (KNN) classifier. The decomposed mammogram at various resolution levels by SGWT provides more information than spatial domain. The energy of each coefficient in different sub-bands is computed and all sub-bands are summed together to form the feature vector and classification is achieved by KNN classifier. Results prove that the MCs classification system provides accurate results at 3rd level SGWT level with 100% accuracy.