Leveraging Deep Learning for Malaria Detection: Multi-Model Comparative Analysis of CNN Architectures with Future Directions for AI-Driven Healthcare

Malaria is a serious, life-threatening disease that necessitates early and precise detection for effective treatment. This study assessed the performance of five deep learning models—EfficientNetV2B3, DenseNet201, MobileNetV3Large, ResNet152V2, and VGG19—in classifying malaria-infected and uninfected blood cell images. The models were trained and evaluated using the Malaria Cell Images Dataset, with an 80:20 split for training and testing. MobileNetV3Large emerged as the top-performing model, achieving an accuracy of 97%, followed by EfficientNetV2B3 and VGG19, which achieved accuracy of 96% and 93%, respectively. DenseNet201 achieved a moderate accuracy of 88%, while ResNet152V2 had the lowest accuracy at 66%, highlighting the need for further refinement. The research evaluates different CNN models to determine their effectiveness in detecting malaria through medical images. Future research will focus on enhancing accuracy through ensemble learning techniques, incorporating attention mechanisms, and employing advanced data augmentation methods. Additionally, deploying these models on mobile devices and edge computing systems could enable real-time malaria detection in resource-limited settings, facilitating faster diagnosis and improving healthcare access.