Interpretable and Explainable AI for Trustworthy Clinical Decision Support

Artificial intelligence is transforming healthcare, supporting clinicians in diagnosis, prognosis, treatment planning, and monitoring. Yet clinical adoption remains limited by low transparency, insufficient trust, and uncertainty about the reliability of AI recommendations. Post-hoc methods such as SHAP, LIME, and Grad-CAM have improved interpretability, but attention is shifting toward intrinsically interpretable AI, where transparency is built into the model itself.This Research Topic aims to advance the next generation of trustworthy Clinical Decision Support Systems by bringing together work on interpretable machine learning, explainable AI, human-centered evaluation, and trustworthy AI. We welcome contributions that move beyond explaining predictions toward AI whose reasoning is inherently understandable, clinically meaningful, and aligned with clinicians’ workflows.We are especially interested in emerging interpretable methods (e.g. Generalized Additive Models, Explainable Boosting Machines, Kolmogorov–Arnold Networks, prototype- and concept-based learning, symbolic and modular architectures), as well as faithful and actionable post-hoc explanations. Counterfactual reasoning and explainability for time-series, monitoring, and longitudinal healthcare data are also central, given AI’s growing role in intensive care, wearables, and electronic health records.Beyond algorithmic transparency, the topic emphasizes human-centered XAI: how explanations shape calibrated trust, appropriate reliance, cognitive workload, decision-making, usability, and patient communication. We also welcome work on robustness, fairness, uncertainty, regulatory compliance, and the ethical deployment of interpretable AI in high-risk clinical settings.Topics of interest include (but are not limited to):- Intrinsically interpretable machine learning for clinical decision support.- Explainable AI methods (SHAP, LIME, Grad-CAM, Integrated Gradients, attention and feature attribution).- Hybrid approaches combining interpretable models with post-hoc explanations.- Interpretable neural components: concept bottlenecks, prototype learning, symbolic and modular AI.- Counterfactual and contrastive explanations, actionable recourse, and causal explainability.- Explainability for time-series, forecasting, wearables, ICU monitoring, and sequential clinical prediction.- Human-centered XAI evaluation: trust calibration, reliance, cognitive workload, and mental models.- Clinician- and patient-centered explanation interfaces, interactive explanations, and co-design.- Frameworks for explanation faithfulness, robustness, stability, and clinical utility.- Fairness, bias detection, uncertainty quantification, and trustworthy AI in healthcare.- Regulatory, ethical, and governance aspects of interpretable AI in high-stakes settings.- Applications to medical imaging, EHRs, oncology, cardiovascular and pulmonary medicine, intensive care, and remote monitoring. Topic editors: Danielle Sent, Aniek Markus Source: https://www.frontiersin.org/research-topics/83976/interpretable-and-explainable-ai-for-trustworthy-clinical-decision-support