H. Osta, R. Qahwaji, and S. Ipson (UK)
mammography, feature extraction, wavelet transform, support vector machine
In this paper, we investigate wavelet-based feature extraction from mammogram images and efficient dimensionality reduction techniques. The aim of this study is to compare the classification accuracy performance of two sets of features using radial-basis-functions neural network (RBFNN) and support vector machines (SVM). The two sets of features used in this study are extracted from region of interest. The first set is extracted from wavelet decomposition and the second is a reduced obtained by applying the minimal-redundancy-maximal relevance criterion (mRMR) to the first set. It is found that SVM performs much better than RBFNN and achieves outstanding results reaching an accuracy of 89.3%.
Important Links:
Go Back
