Wavelet-based Fractal Feature Extraction for Microcalcification Detection in Mammograms

A novel hybrid wavelet-based fractal feature extraction method and an Artificial Neural Networks (ANNs) classification system are proposed for the detection of microcalcification clusters (MCCs) in the digital mammograms. The hybrid wavelet-based fractal feature set consists of the surrounding region dependence based features and the newly proposed wavelet-based fractal features. Experiments demonstrated that the proposed hybrid feature has the best classification discriminating ability among three sets of features tested in the experiments. A satisfactory MCCs' detection rate and a good ratio of true positive fraction to false positive fraction (ROC curve) have been achieved. The proposed MCCs detection system provides an adequate framework for microcalcification detection in the mammograms.