Deep Learning-Based Malaria Detection: A Comparative Study of Algorithm Performance

Malaria is caused by Plasmodium parasites and remains a serious threat to health worldwide, with high morbidity and mortality rates. However, effective treatment depends on the accurate and fast diagnosis of malaria, which traditionally has been through microscopy and rapid diagnostic tests that have limitations, including requiring expert interpretation, limited sensitivity in the early stages of infection, and variable accuracy. By automatically detecting malaria from microscopic blood smear images using deep learning models, our study has overcome the above constraints. Four sophisticated deep learning models-DenseNet201, NasNet, EfficientNetB7, and AlexNet were tested against performance metrics of accuracy, precision, recall, F1 score, and computational efficiency. With 99% accuracy, NasNet was the most successful model among them and is hence very dependable for the diagnosis of clinical malaria. In addition, AlexN et was faster in inference speed and had moderate accuracy compared to DenseNet201 and EfficientNetB7. Such results are a significant step forward in malaria detection technology, and the potential of NasNet for automated, scalable diagnostic systems is promising.