Evaluation of Deep Learning Models for Malaria Detection: A Comparative Study on Accuracy

Abstract

The detection of malaria is a critical task in healthcare, especially in regions where the disease is prevalent. In this study, the accuracy of several deep learning models in classifying malaria-infected and uninfected cell images was evaluated. The dataset used for training consisted of 416 images, with 220 labeled as "Parasite" and 196 as "Uninfected," while the testing set included 134 images (91 "Parasite" and 43 "Uninfected"). The performance of five prominent models VGG16, VGG19, MobileNetV2, InceptionV3, and CNN was assessed by measuring the accuracy on the test set. It was found that the highest accuracy was achieved by VGG19, with a value of 96%, followed by VGG16, which achieved an accuracy of 94%. The MobileNetV2 model, while slightly less accurate, attained a reasonable accuracy of 92%. Comparable results were observed for the InceptionV3 and CNN models, which achieved accuracies of 93%. These results demonstrate that deep learning models, particularly VGG16 and VGG19, can be effectively utilized for malaria detection with high accuracy. The potential of convolutional neural networks (CNNs) to provide reliable, automated diagnostic tools for malaria was highlighted, especially in resource-limited settings. The findings underscore the promise of deep learning techniques in enhancing malaria detection accuracy and improving healthcare outcomes.