Automatic Classification of Plasmodium for Malaria Diagnosis based on Ensemble Neural Network

Malaria is one of the important public health issues of global concern. It is a kind of infectious disease caused by Plasmodium which can endanger human life and health. The examination of Plasmodium blood smear is the main method to diagnose malaria. Applying machine learning method to automatically analyze malaria smear images is very important for rapid diagnosis and surveillance of malaria. However, the existing machine learning methods need further improvement in feature extraction and generalization ability. For this reason, this paper introduces an ensemble neural network to automatically learn more accurate image features and achieve automatic classification of malaria images. In addition, we propose an adaptive threshold control method for cell segmentation, and then put the cell as center to extract images to obtain the training dataset, which solves the noise problem caused by sliding window cutting. And the idea of transfer learning is applied to solve the problem of shortage of training data. We additionally apply the proposed method to malaria image recognition and achieve 94.58% accuracy. The experimental results show that the model has good robustness and generalization ability and provides a valid data processing method for clinical malaria rapid diagnosis application.