Breast tissue classification based on unbiased linear fusion of neural networks with normalized weighted average algorithm

The diagnosis of breast cancer is performed based on informed interpretation of representative histological tissue sections. Tissue distribution detected from cytologic examinations is useful for tumor staging and appropriate treatment. In this paper, we propose a normalized weighted average (Normwave) algorithm for the unbiased linear fusion, and also construct the multiple classifier system that includes a group of Radial Basis Function (RBF) neural classifiers for the classification of breast tissue samples. The empirical results show that the proposed Normwave algorithm may improve the performance of the RBF-based multiple classifier system, and also reliably outperforms some widely used fusion methods, in particular the simple average and adaptive mixture of experts.