Breast cancer diagnosis from biopsy images by serial fusion of Random Subspace ensembles

Accurate and reliable classification of microscopic biopsy images is an important issue in computer assisted breast cancer diagnosis. In this paper, we investigated the effectiveness of a feature description approach by combining LBP texture analysis with Curvelet Transform and proposed a cascade Random Subspace ensembles with rejection options for microscopic biopsy images classification. While the LBP efficiently describes texture properties, the Curvelet Transform is particularly appropriate for the representation of piece-wise smooth images rich of edge information. A combined feature description can thus provide comprehensive image characteristics by taking advantages of their complementary strengths. The classification system is built as a serial fusion of two different Random Subspace classifier ensembles with rejection options to enhance the classification reliability. The first ensemble consists of SVM classifiers, with set of binary SVMs converting the original k-class classification problem into a K 2-class problems. The second ensemble, consisting of random subspace ensemble of MLPs, focus on the rejected samples from the first ensemble. For both of the ensembles, rejection option is implemented by relating the consensus degree from majority voting to confidence measure and abstaining to classify ambiguous samples if the consensus degree is lower than a threshold. Using a microscopic biopsy images dataset from Israel Institute of Technology, a high classification accuracy 97% is obtained with rejection rate 0.8% from the proposed system consisting of 30 base classifiers in each ensemble.