Implementation of Malaria Parasite Detection and Species Classification Using Dilated Convolutional Neural Network

Abstract

Malaria is an infectious disease caused by a bite of an Anopheles Mosquito which has caused a lot of death. Diagnosis of malaria is made by examining a red blood cell of an infected patient using a microscope, which takes time and requires a qualified laboratory expert to examine, read and interpret the results obtained. Convolutional Neural Network (CNN) has played important role in image classification; however, it has exhibited some problems in consuming computing resources which is one of the limitations of CNN. To reduce this problem, this paper presented a Dilated Convolution Neural Network for malaria parasites detection and species classification using blood smear images. A direct classification was carried out to detect infected and uninfected malaria parasites. Subsequently, species classification was carried out using 3 convolutional layers and Convolution2D for convolution operation while a dilation rate of 2 was used for the convolution layers. The model was trained with a publicly available dataset of 27699 images with a performance accuracy of 99.9% for parasite detection and species classification of 99.9% for falciparum, 64.6% for Malarie, 39.1% for Ovale and 37.3% for Vivax.