Malaria Parasite Detection using Deep Learning in Thick Blood Smears

Detection of malaria disease is done by finding the presence of malaria parasite or plasmodium in blood smear. Here malaria parasites are detected in thick blood smears. This paper proposes a version to detect the presence of malaria parasite(plasmodium) in thick blood smears automatically with the help of deep learning and not using microscopy examinations and chemical tests. This detection will be done using two steps, that is, intensity-based screening which is the preprocessing step, the first step, that extracts candidates for processing, and next is customized convolutional neural network (CNN), the processing step, which takes the preprocessed images and detects whether malaria parasite is present or not. Hold-out(3:1) technique is used for evaluation of the model. The model has achieved an accuracy reaching 91%. The two preprocessing and processing steps improves object detection of the system. Malaria is usually detected using chemical tests and microscopy examinations. This process requires a lot of resources mainly laboratories. Parasitologists who are experienced are sometimes difficult to find, so manually counting the malaria parasites can be prone to major errors. Due to which the cost for testing and even time for malaria diagnosis increases drastically. Since the traditional process of malaria parasite in blood smears detection has many drawbacks it needs a sophisticated, accurate diagnosing equipment or system which has low cost. This system can be used in regions and areas where there are constraints on resources, time of people and cost which they can afford. This system provides many advantages to rural diagnostic centres where the supplies are limited and not easily accessible.