A Mobile Application Framework for Malaria Parasite and Species Classification Using Dilated Convolutional Neural Networks

Malaria remains one of the most pressing global health concerns, particularly in sub-Saharan Africa and Southeast Asia, where accurate and timely diagnosis is essential for effective treatment.Traditional microscopic examination, though considered the gold standard, is time-consuming, expertise-dependent, and unsuitable for large-scale or point-of-care deployment.This study presents a novel mobile-based framework for automated malaria parasite detection and species classification using a Dilated Convolutional Neural Network (D-CNN) integrated with multi-scale feature extraction.The proposed system was trained and validated on a dataset of over 27,000 microscopic images for parasite and species, incorporating data augmentation and SMOTE to address class imbalance most especially for the species.Experimental results demonstrate strong performance, achieving 96.26% accuracy in parasite detection and 99.63% accuracy in species classification across four Plasmodium species (P.falciparum, P. vivax, P. malariae, and P. ovale), with AUC values above 98%.The trained model was optimized and deployed on mobile device via PyTorch Mobile, enabling real-time.The mobile application provides a practical diagnostic tool for field use in resource-limited settings, demonstrating robustness against image variability and morphological similarities between parasite species.This research highlights the potential of D-CNN-based mobile solutions in advancing malaria diagnosis, supporting clinical decision-making, and enhancing access to life-saving healthcare.