Malaria diagnosis using microscopic imaging

Abstract

Malaria, a dangerous disease caused by Plasmodium, which is spread by being bitten by infected mosquitoes (Female Anopheles). It is crucial to diagnose malaria pathogens quickly and accurately at the right time. Traditional microscopy is commonly used in developing countries to diagnose malaria parasites, where pathologists examine the slide under a light microscope. However, in the case of traditional microscopy, requires more time and careful attention. Here, we have proposed a method to diagnose malaria based on computer vision. As a pre-processing stage, SAMF (Systematically Applied Mean Filter) algorithm is proposed that removes impulse noise from the corrupted malaria-infected images. Otsu method is used to obtain the binary version of images for cropping blood cells from complete image. 17 texture and color features were extracted from these cropped cells and these features were used to train Cubic SVM (Support Vector Machine) classifier. Hence a precise malaria diagnosis system was developed for detecting Plasmodium parasites, identifying their life stages and species using images of thin blood smears. A total of 348 images from CDC (Centre for Disease Control and Prevention) database were used to train and test the performance of the system.