Integrating Explainable AI for Enhanced Dengue Detection from Blood Smears

Dengue fever, a mosquito-borne viral infection caused by the DENV virus, has been spreading rapidly across tropical and subtropical regions, posing a major public health concern and contributing to significant global mortality. Accurate identification of dengue cases remains challenging, making it essential to adopt reliable and efficient diagnostic approaches. This study explores an automated method for detecting dengue from peripheral blood smear (PBS) images using deep learning (DL) techniques, which have gained prominence in computer-assisted diagnosis of various medical conditions. The research utilizes the “Dengue Disease using Blood Smears†dataset and applies multiple classification models, including ResNet50, MobileNets, MobileNetV3-Small, MobileNetV3-Large, and a hybrid ensemble of Xception and MobileNet. For object detection, members of the YOLO family—YOLOV5x6, YOLOV5s6, YOLOV8n, and YOLOV9n—are employed. Experimental results reveal that the Xception–MobileNet ensemble yields the highest classification accuracy, while YOLO-based models effectively localize abnormalities in PBS images. These findings highlight the potential of DL-based approaches to enhance dengue diagnosis, offering a promising direction for improving public health surveillance and outcomes.