Understanding AI Decisions in Vector-Borne Disease Detection: A Systematic Review of Methods, Practical Challenges, and Clinical Relevance

Abstract

Vector-borne diseases such as dengue, malaria, and Zika remain major public health challenges, particularly in resource-constrained regions with limited diagnostic infrastructure. Artificial intelligence (AI) has demonstrated significant potential for early disease detection using medical imaging, clinical symptoms, and environmental data; however, the lack of model interpretability limits clinical adoption. This systematic review, conducted following PRISMA guidelines, critically examines AI and explainable AI (XAI) techniques for vector-borne disease detection through a structured search of four major databases using predefined inclusion and exclusion criteria. Machine learning, deep learning, federated learning, and edge computing approaches are comparatively evaluated based on reported performance and clinical applicability. The review further compares SHAP and LIME in terms of interpretability, computational complexity, and diagnostic suitability. A conceptual multimodal framework integrating imaging, clinical, and environmental data is proposed, together with a validation roadmap for future clinical evaluation. Finally, key research gaps are identified, emphasizing that combining high diagnostic accuracy with transparency and usability is essential for the reliable clinical deployment of AI-based vector-borne disease detection systems.