Plasmodium Detection from Blood-cell Images using MobileNet and Firefly Algorithm
V. Rajinikanth
Abstract
Malaria is a most prevalent mosquito-borne disease that can result in mild to severe health issues. The main cause of malaria is blood infection due to the plasmodium parasite, which enters the bloodstream by the bite of an infected Anopheles mosquito. Blood samples taken from infected individuals are often examined under a microscope to facilitate clinical level diagnosis of malaria. To plan and execute the treatment, an accurate diagnosis of the red-blood cell (RBC) infection is required. To automate the RBC examination task, this research proposes a Deep-Learning (DL) technique employing the MobileNet (MN). Proposed DL-tool consist following phases; collecting and resizing the microscopic images, deep-feature extraction with a chosen MN-variant, reducing features using Firefly-Algorithm (FA), creating fused-deep-features (FDF), and classifying data using 3-fold cross validation. In this work, the suggested MN-tool uses the individual and FDF with SoftMax to classify the selected microscopic images into healthy RBC/Plasmodium. The performance of proposed DL-tool is verified using two forms of FDFs: (a) FDF1 generated with deep features after 50% dropout, and (b) FDF2 generated with deep features optimized using FA. The results of this study validate that the suggested DL-tool contributes to >99% accuracy when FDF2 based classification is used.