CLASSIFICATION AND QUANTIFICATION OF MALARIA PARASITES USING CONVOLUTIONAL NEURAL NETWORKS.

Abstract Malaria is currently one of the most deadly diseases in the world. While there are different treatment methods for the disease, the search for new drugs against malaria is still a very important area of research. One of the main challenges in manufacturing drugs against malaria is efficiently evaluating the performance of the drugs on the parasites since it requires, amongst others, precise measurements of the parasite growth-stages as well as their counts in blood smear images. The current gold-standard for making such detail diagnosis is manual microscopy which is tedious. This research showed that convolutional neural networks can be used to identify the different growth-cycle stages of Plasmodium parasites, even in situations where there is little data. Employing a variety of data augmentation techniques and transfer learning, a semantic segmentation model was built to discriminate between trophozoites, gametocytes and normal red blood cells with an accuracy of 85.86% in 353 Giemsa-stained thin blood smears. The results showed that it is possible to perform dense predictions on Giemsa-stained thin blood smears using convolutional neural networks