Beyond the Algorithm: Scalable, Explainable, and Adaptive AI for High-Stakes Clinical Decision Support

Artificial Intelligence (AI) has shown strong potential in supporting clinical decision-making through Clinical Decision Support Systems (CDSSs). However, despite promising experimental results, most AI solutions remain confined to prototype stages and fail to achieve sustained adoption in real-world healthcare settings. This gap is largely due to unresolved engineering challenges, including integration with heterogeneous infrastructures such as electronic health records (EHRs) and telemedicine platforms, handling noisy and evolving data, ensuring robustness under operational conditions, and maintaining systems over time. Additionally, high-stakes clinical environments demand transparency, traceability, and compliance with regulatory standards, further complicating deployment. Recent advances in explainable AI, multimodal learning, and privacy-preserving techniques provide new opportunities, but a stronger focus on software engineering principles is required to transform these advances into reliable, production-ready systems.This Research Topic addresses the critical gap between the development of AI prototypes and their reliable deployment in real-world clinical decision support systems. While advances in machine learning have led to increasingly accurate models, their integration into operational healthcare environments remains limited due to challenges related to system design, scalability, robustness, lifecycle management, and validation in practice. The goal of this initiative is to foster contributions that explicitly tackle these software engineering challenges, promoting AI-driven CDSSs that are not only accurate but also deployable, maintainable, and trustworthy. We aim to encourage the development of engineering methodologies, system architectures, and evaluation strategies that enable continuous integration of AI into clinical workflows. Particular emphasis is placed on contributions that provide empirical evidence from real-world deployments, document engineering trade-offs, and report lessons learned from building and operating AI systems in high-stakes clinical contexts. This Research Topic focuses on the software engineering of AI systems for high-stakes clinical decision support, with particular attention to the full lifecycle of such systems, from design and development to deployment, validation, and maintenance in real-world settings. We welcome contributions that explicitly address engineering aspects such as system architectures for AI-enabled CDSSs, integration with clinical infrastructures, robustness and fault tolerance, monitoring and maintenance strategies, and lifecycle management of adaptive models. Submissions should clearly articulate their engineering contribution, including methods, artefacts, frameworks, or reproducible practices.Relevant topics include, but are not limited to:- system design and integration of AI within clinical workflows;- validation frameworks and benchmarking in real-world environments;- explainability techniques embedded into operational systems;- continuous, federated, or adaptive learning in deployed settings;- data engineering and multimodal data integration pipelines;- safety, robustness, and compliance in AI-driven clinical systems;- human-AI interaction and usability in decision support tools.We particularly encourage case studies, system papers, evaluation frameworks, and experience reports that provide empirical validation and insights from real deployments. Contributions should go beyond algorithmic performance and demonstrate practical impact, reproducibility, or transferable engineering knowledge. Purely theoretical or model-centric works without clear relevance to deployment and system-level considerations are out of scope. Topic editors: Martina Iammarino, Vittoria Nardone Source: https://www.frontiersin.org/research-topics/79573/beyond-the-algorithm-scalable-explainable-and-adaptive-ai-for-high-stakes-clinical-decision-support