Automatic Cardiomyopathy Diagnosis with a Cost-sensitive Ensemble Classifier

This paper proposed a cost-sensitive ensemble classifier for automatic cardiomyopathy diagnosis using features extracted from cardiac magnetic resonance images. However, with numerous features extracted from images, it is hard for a single classifier to achieve accurate prediction. In contrast, an ensemble classifier combines multiple weak classifiers which could benefit from each others and improve the performance. Therefore, we proposed a cost-sensitive ensemble classifier assembling five heterogeneous classifiers: logistic regression (LR), Gaussian naive bayes (GNB), support vector machine (SVM), multi-layer perception(MLP), and convolutional neural network(CNN). The weight of each classifier was determined according to the special cost-sensitive function. In the experiment, the proposed method was evaluated on a publicly available Automated Cardiac Diagnosis Challenge (ACDC) dataset [1], where the proposed ensemble classifier achieves a considerable improvement.