A survey of explainable AI applications in infectious disease diagnosis.

Abstract

Artificial intelligence (AI), particularly deep learning, has shown significant promise in accelerating the diagnosis of infectious diseases. However, the inherent "black box" nature of deep learning is a major barrier to clinical adoption, preventing trust and accountability. Explainable AI (XAI) tries to address this problem by adding transparency to model decision-making. This paper presents a comprehensive scoping review of XAI applications in infectious disease diagnosis based on literature published between 2023 and 2025. We review studies of major diseases, including COVID-19, tuberculosis, malaria, monkeypox, and dengue, categorizing them by data modality (e.g., imaging, tabular) and the XAI techniques employed (e.g., SHAP, LIME, Grad-CAM). We conclude that XAI methods are effective in identifying clinically relevant features, such as key biomarkers from lab results and pathological regions in medical images, thereby aligning model reasoning with established clinical knowledge. Despite this progress, significant challenges persist, including poor generalizability from small, single-center datasets, the computational overhead of post-hoc methods, and a gap between technical explanations and clinical utility. We outline future directions that should be taken, including the need for standardized evaluation metrics, multi-center validation, and clinician-centered XAI design to facilitate the responsible integration of AI into clinical practice.Clinical trial numberNot applicable.