Ensemble of Multiple Classifiers for Automatic Multimodal Brain Tumor Segmentation

Ensembles of classifiers can improve the performance of individual classifiers on several classification tasks. In this paper, we investigate the employment of ensemble methods for improving the accuracy of multimodal brain tumor segmentation. Four different ensemble methods are evaluated: Adaboost, bagging, stacking, and voting, on MICCAI BRATS 2016 challenge’s MRI dataset. Our experimental results confirm the performance improvement produced by the ensemble methods over those of 20 different individual classifiers. Majority voting based ensemble method performed the best among the four ensemble methods.