Design of Deep Convolutional Neural Network for Efficient Classification of Malaria Parasite

Abstract

The standard method used for the testing of malaria is the blood stream examined under the parasite-infected red blood cells by qualified technicians. The test is carried out under a microscope. The inefficiency of this traditional method has been improved with advanced computer vision and deep learning techniques to automatically identify the object as infected or not infected through the use of microscope. The work proposed employs a deep convolutional neural network (DCNN) to identify the parasites of malaria from images of human bladder cells. The focus of the work is primarily on the comparison of validation and precision by setting the hyperparameters so that images are classified into parasitized and uninfected cells. The expertise in finding the parasitized and non-parasitized and quality of smear play an important role in system accuracy; the proposed deep learning technique provides better solution with end-to-end extraction and classification of feature. The datasets consist of 26,188 images which contain 13,105 parasitized images and 13,083 uninfected images. The kappa coefficient (KC) and Matthews correlation coefficient (MCC) are evaluated for the proposed network.