Identification and Classification of Types of Malarial Parasites Using Customized CNN Model

Over the past decade, medical imaging has achieved many remarkable milestones in medical science. However, the incidence of infectious diseases like malaria continues to rise rapidly. Malaria a serious infection caused by the Plasmodium parasite, is a life-threatening paradox in global health. Even though conventional methods could identify the malaria parasite, they still failed to present accurate results and are time-consuming manual processes. Early detection and analysis of the complex morphological structure of blood cells are challenging tasks and prevalent using morphological detection tools. In our work, our primary goal is to develop a customized CNN model to pick malaria species and classify the different variants of malaria species in microscopic thin blood smear images. This specialized CNN model leverages deep learning abilities and fine-tunes the model capacity employing dropout techniques. The model can successfully validate quantitative analyses of different variants i.e. Plasmodium falciparum (or P. falciparum), Plasmodium malariae (or P. malariae), Plasmodium vivax (or P. vivax), Plasmodium ovale (or P. ovale) improving patient outcomes and presents overall classification results with an accuracy of 93.30 %, specificity of 97.6%, F measure of 96.5%, and AUC of 95.21 %. Through careful hyperparameter tuning, the customized CNN model has achieved superior performance results as compared to other state-of-the-art methods.