Benchmarking CNN Models for Malaria Cell Detection

Malaria is a life-threatening disease prevalent in tropical and subtropical regions, caused by Plasmodium parasites and transmitted by female Anopheles mosquitoes. Traditional diagnostic methods, such as manual microscopy and rapid diagnostic tests, face limitations in terms of accuracy and efficiency, necessitating more advanced diagnostic techniques. This study aims to evaluate and compare the performance of seven state-of-the-art convolutional neural network (CNN) architectures-VGG16, VGG19, DenseNet121, ResNet50, Xception, MobileNetV3, and EfficientNetB3-for the detection of malaria parasites in digitized blood smear images. The dataset used comprises thousands of annotated blood smear images, categorized into infected and uninfected cells. Each model's performance was assessed based on key metrics including accuracy, precision, recall, and Fl-score. EfficientNetB3 demonstrated superior performance with a diagnostic accuracy of 97.09<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>, significantly outperforming the other models. This study highlights the potential of deep learning models to enhance the diagnostic accuracy and efficiency of malaria detection, facilitating quicker and more reliable treatment interventions. The findings support the further development and deployment of these technologies, especially in resource-limited, malaria-endemic regions, thus potentially improving healthcare outcomes and reducing the burden of malaria globally.