Deep Learning-Based Convolutional Neural Networks for Automated Malaria Detection

Malaria is a potentially fatal illness that needs to be diagnosed as soon as possible to be effectively managed, especially in environments with limited resources. Using microscopic blood smear images, this study explores the use of deep learning-based Convolutional Neural Networks (CNNs) for automated malaria identification. Several CNN architectures were evaluated, including VGG16 (89.40%), DenseNet121 (90.2%), EfficientNetB0 (90.5%), and ResNet50 (83.67%), on a publicly available blood smear dataset. The Xception model achieved a notable accuracy of 91.5%. To further enhance diagnostic accuracy, an improved Xception-based model was developed, integrating advanced data augmentation, hyperparameter optimization, and regularization techniques, resulting in an accuracy of 96.73%. The results demonstrate the superiority of the enhanced model in distinguishing infected cells from uninfected ones. This work demonstrates how deep learning models, in particular optimized architectures, have the potential to transform malaria diagnostics by offering a scalable and effective automated detection solution that will aid medical professionals in the fight against the illness.