Malaria Cell Images Classification with Deep Ensemble Learning

Malaria is a deadly infectious disease and a major threat to global health. Efficient and accurate malaria detection is crucial for timely identification of patients and subsequent treatment. Most traditional classification models use a single network to extract features, but each single model can only extract a limited number of image features for classification. To overcome this limitation and to improve the classification performance, this paper proposes a classification study of malaria cell pathology images based on a deep ensemble learning model. Transfer learning technique is employed to pretrain multiple networks, transferring the learned knowledge to the target dataset. The pretrained networks with optimal performance are selected as the basic classifiers for the ensemble model. A weighted voting strategy is used to integrate multiple pretrained networks for the final classification of malaria cell images. To validate the effectiveness of the proposed model, the classification performance is evaluated on a publicly available malaria cell images dataset. Experimental results demonstrate that the proposed deep ensemble model achieves excellent classification performance on the target dataset, with a classification accuracy of 98.49%, outperforming the classification performance of a single CNN model using transfer learning. The proposed deep ensemble learning technique proves to be feasible for classifying malaria cell pathology images.