Cascading one-class kernel subspace ensembles for reliable biopsy image classification

Reliable classification of microscopic biopsy images is an important issue in computer assisted breast cancer diagnosis. In this paper, a new cascade scheme with reject options is proposed for microscopic biopsy image classification. The classification system is built as a serial fusion of two different classifier ensembles with reject options to enhance the classification reliability. The first ensemble consists of a set of Kernel Principle Component Analysis (KPCA) one-class classifiers trained for each image class with different image features. The second ensemble consists of a Random Subspace Support Vector Machine (SVM) ensemble, that focuses on the rejected samples from the first ensemble. For both of the ensembles, the reject option is implemented so that an ensemble abstains from classifying ambiguous samples if the consensus degree is lower than some threshold. Using a benchmark microscopic biopsy image dataset obtained from the Israel Institute of Technology, a high classification reliability of 99.46% was obtained (with a rejection rate of 1.86%) using the proposed system.