Deep Neural Network Based Malaria Infected Cells Classification

This paper proposes a methodology that automates the classification of Red Blood Cells (RBCs) infected with malaria by using Convolutional Neural Networks (CNNs). Timely and precise identification of patients who are infected by malaria disease with blood smear images is essential for prompt treatment and management of the. Although blood samples can be analyzed manually by laboratory technicians, it is timeconsuming and highly susceptible to mistakes. To solve these problems, the diagnosis method of using a CNN model has been developed, which is trained on a specific data set with labeled blood smear images that allow it to identify healthy and malaria-infected cells with relative ease. To enhance the accuracy of model and predictive power, a significant amount of both classification and data augmentation was done. In addition, using previously trained models also boosted the precision of the classification. Such models, compared to standard machine learning approaches, are very effective and can be relied on greatly especially where resources are limited. Further developments are going to be on the application of explanation-based AI for model understanding, clinical validation of the solution, and reduction of the latency for near real-time malaria diagnosis.