Detection of Malaria Parasite Using Lightweight CNN Architecture and Smart Android Application

Malaria, a fatal disease, is brought about when a female anopheles mosquito bites. Death may occur from malaria if treatment is not received in its early stages. It is highly important to identify the malaria parasite as early as possible in order to proceed with the appropriate treatment. Thus, the paper proposes a Lightweight CNN model that is designed to detect malaria parasite by early stage from microscopic blood images. To reveal the usefulness of the proposed model, Grad-CAM was applied to detect the image part of most concern (which pixels draw more attention from the model related with the other pixels) by generating heatmap image. The Model's efficiency is calculated by accuracy, precision, recall, and F1-score. Comparative analyses are conducted against well-known transfer-learning (TL) models and state-of-the-art (SOTA) models, with the proposed method. The proposed model yielded 99.56% accuracy, 99.97% precision, 98.36% recall, 99% f1-score and 99.2% AUC score. As, the proposed CNN model is lightweight the prediction time for a single image is about 2 milliseconds which making it highly efficient for practical applications and have significant impacts on healthcare accessibility, efficiency and even cost savings in the resource limited settings.