Combining Neural Learners with the Naive Bayes Fusion Rule for Breast Tissue Classification
Yunfeng Wu, Sin-Chun Ng · 2007
Early detection of suspicious breast lesions is commonly performed by analysis of breast profiles detected by effective modalities. Tissue distribution in each modality can provide important information about the elastic characteristics of breast which is useful for computer-aided diagnosis. In this paper, the naive Bayes (NB) fusion rule is utilized to combine a group of radial basis function (RBF) neural learners in a multiple classifier system for classification of breast tissues. The empirical results show the NB fusion rule may effectively diminish the mean-squared errors, and also improve approximately 15% classification accuracy, which is significantly better than the component RBF neural learners. Moreover, the NB fusion rule also outperforms the widely used simple average and majority voting fusion rules.