ECARMI: An Ensemble Based Approach for Medical Images Based Disease Classification and ROIs Recognition

In this paper, we proposed a novel medical images based computer aided diagnosis method named ECARMI. It combines the cost-sensitive learning with selective ensemble techniques to improve the medical images based diagnosis performance. At first, selective cost-sensitive SVM ensemble is utilized to perform the classification of medical images. Then, the Regions of Interest (ROIs) in positively identified image are identified by using a selective ensemble of cost-sensitive Fuzzy C-Means models. The real dataset based experiments show that, the ECARMI approach not only improved generalization ability but also achieved a satisfactory result in both the accuracy of classification and correctly labeling the ROIs.