Abstract Background and aims Artificial intelligence (AI)–based clinical decision support systems (CDSS) are increasingly explored in acute stroke care. While high-level principles for trustworthy AI (TAI) exist, practical guidance for implementing these principles is limited. This hinders safe clinical adoption. We aimed to provide a transparent and reproducible framework to translate TAI principles into practice for CDSS in stroke care. Methods Within the EU Horizon project “VALIDATE”, we applied a multi-method approach to operationalize TAI in stroke CDSS development. We conducted participatory co-creation workshops, an external expert assessment using the Z-Inspection methodology®, semi-structured interviews with domain experts, and bi-annual internal audits with work package leads to elicit project-specific requirements. Results We identified key steps for the development and validation of TAI in stroke care. Early stakeholder engagement enables translation of ethical principles into concrete system specifications. Continuous internal governance and structured feedback mechanisms facilitate monitoring of ethical requirements and timely issue identification. Together, these processes improve transparency and align ethical governance with technical and clinical workflows. We further identified practical challenges - such as resource availability, hospital infrastructure, patient recruitment, and timing - along with recommendations for addressing them. We consolidated these findings into a comprehensive Standard Operating Procedure (SOP). Conclusions Our work provides a practical, process-oriented guide for TAI development and validation. The SOP will support future projects, helping to enhance trustworthiness, transparency and clinical safety of CDSS in stroke care. Conflict of interest All listed authors, main, presenting and other: Nothing to disclose.
Bürger et al. (Fri,) studied this question.