DWT and RT-based approach for feature extraction and classification of mammograms with SVM

A new methodology to automatically extract features from mammograms for classification is presented. The approach consists of combining the discrete wavelet transform (DWT) and the Radon transform (RT). First, the DWT is employed to obtain the mammogram´s high-high (HH) sub-band image. Next, the RT is applied to the latter with four different orientations to obtain four RT signals whose energies and entropies are computed. Then, these statistics are fed to a support vector machine (SVM) with polynomial kernel to distinguish between normal mammograms and those showing malign microcalcifications tumours. The approach was tested on a database of one hundred mammograms and shows improved classification accuracy in comparison to using the DWT or RT alone.