Abstract

This systematic literature review, conducted under the PRISMA framework, investigates the use of Explainable Artificial Intelligence (XAI) techniques for biomarker identification in lung cancer detection. By analyzing 15 peer-reviewed studies published between 2018 and 2025, we identify dominant XAI methods—particularly gradient-based approaches such as SHAP, DeepLIFT, and Integrated Gradients—and their association with specific model architectures and biomarker types. The review highlights critical limitations, including small and demographically narrow datasets, insufficient multimodal integration, and limited evaluation of clinical usability. A significant gap is the complete absence of studies using Latin American datasets, revealing a lack of regional inclusiveness in AI-driven healthcare research. This work contributes a structured synthesis of methodological trends, practical barriers, and future directions, aiming to inform the development of clinically relevant, trustworthy, and regionally inclusive explainable AI tools for lung cancer diagnos