Stacked Ensemble Deep Learning Technique to Detect Malaria Parasite in Blood Smear

Abstract

Malaria Parasites are transferred from infected female mosquitos to humans which can lead to the death of the person. Malaria affects the majority of people each year, and most of these cases arise in remote areas. There has been lots of research in the field of Malaria Parasite detection using the automated technique, but these techniques require high computational power, and in remote areas, the availability of such systems is very unlikely. The proposed Ensemble Model can detect the presence of Malaria Parasite in the thick blood smear by taking the average of the output layers of ResNet50 and the custom CNN model. The models' performance has been evaluated and results reveal that it achieved 0.97 Specificity, 0.98 Sensitivity, 0.97 Precision, and 0.972 Accuracy with an image size of 64x64x3. The overall file size of the model is under 15Mb that also makes it portable.