Abstract

Breast cancer detection and diagnosis is a critical and complex procedure that demands high degree of accuracy. In computer aided diagnostic systems, the breast cancer detection is a two stage procedure. First, to classify the malignant and benign mammograms, while in second stage, the type of abnormality is detected. The classifier ensemble optimization is a method that can be applied to increase the classification accuracy at both stages. In this paper, we have proposed a novel technique to enhance the classification of malignant and benign mammograms using multi-classification of malignant mammograms into six abnormality classes. DWT (discrete wavelet transformation) features are extracted from preprocessed images and passed through different classifiers. To improve accuracy, results generated by various classifiers are ensembled. Mammograms declared as malignant by ensemble classifiers are divided into six classes. The ensemble classifiers are further used for multi-classification using one against all technique for classification. Output of all ensemble classifiers is combined by product, median and mean rule. It has been observed that the accuracy of classification of abnormalities is more than 97% in case of mean rule. Mammographic Institute Society Analysis [MIAS] dataset is used for experimentation.

Key words: Ensemble classifier, mammography, multi-classification.