Detection of vivax and falciparum malaria using deep learning techniques

Abstract

Although malaria is uncommon in temperate climates, malaria is still common in tropical and subtropical countries. Every year, approximately 290 million people are infected with malaria, and more than 400,000 people die from this disease.

In this research, the dataset was downloaded from the Kaggle website, which consists of 27,558 images of infected people and healthy people, divided into: 13,779 images of healthy people, 6890 images of people infected with falciparum malaria, and 6889 images of people infected with vivax malaria. Initially, the images were pre-processed (fogification - noise removal - morphological operations), after which the convolutional neural network (CNN) was built, trained and tested, where the data was divided into two groups, 80% training and 20% testing, and then comparison was made between several pre-trained models. The proposed model achieved the best evaluation accuracy among the pre-trained models, and the results gave an accuracy rate of 96.5%, a sensitivity rate of 95%, and a specificity rate of 97.6%. Thus, the results demonstrate the effectiveness of the proposed algorithm as an assistive model for doctors in diagnosing malaria.