Detection of Malaria on an Erythrocyte Using Deep Residual Network

Abstract

In the recent years, many researches have evolved in the field of CAD particularly in malaria diagnosis. Malaria diagnosis encompasses many domains, such as labelling of blood samples, detection of infected cells and furthermore determining the development stage and parasite type. Various techniques were implemented, among them, Deep learning has recently caught the attention of researchers. In this research project, we have implement ResNet-50, a variant of Deep Residual Learning to detect the presence of malaria on an erythrocyte. For this experiment, a dataset of 27,560 Red Blood Cells from NIH has been used. Samples were equally distributed. The Model achieved a training, validation and testing accuracy respectively of 96.50%, 96.78% and 97%. The whole process took 1 hour, and training-validation lasted for 50 epochs.