Navigating the EU AI Act for stroke treatment pathways
As European regulatory frameworks evolve at the forefront of the international stage, our project aims to contribute to a ‘future-ready’ ecosystem designed to minimize door-to-reperfusion times while ensuring absolute patient safety and data privacy.
The landscape of digital health is undergoing a seismic shift. With the finalization of the EU AI Act, Europe has introduced the world’s first comprehensive horizontal legal framework for Artificial Intelligence. While our initial focus lies in implementing and assessing standard technological tools such as telemedicine, the core vision of MDR in AIS is ‘future-readiness’, contributing to shape stroke care processes that are not only efficient today but are specifically built to accommodate AI and emerging digital tools, allowing healthcare systems to adapt to innovation in real-time. For projects like MDR in AIS, which are at the intersection of Neuroradiology and technological innovation in treatment workflows, this is not just a regulatory hurdle – it is a roadmap for building trust in life-saving technology.
The High-Risk Challenge in Neuroradiology
Under the EU AI Act, AI systems used in healthcare – specifically those falling under the scope of the Medical Device Regulation (MDR 2017/745) – are largely classified as “High-Risk” (Annex III). For AI-driven diagnostic tools in stroke management, the stakes are exceptionally high.
As “Time is Brain” AI algorithms that assist in detecting Large Vessel Occlusions (LVO) or calculating Intracranial Hemorrhage (ICH) volumes must be fast and reliable offering immense opportunities for acute management of Stroke. The challenge lies in aligning clinical evidence of safety and performance to the many layers and requirements regarding data governance, transparency, and human oversight (Article 14 ).
How MDRinAIS Bridges the Gap
The MDR in AIS project, co-funded by the THCS (Transforming Health and Care Systems) partnership, acts as an invested contributor in regulatory science. Rather than viewing compliance as a retrospective “check-the-box” exercise, MDR in AIS integrates regulatory thinking into the development lifecycle.
Our methodology focuses on:
- Standardized Validation: Developing pathways that simplify how clinical centers – like our host unit in Siena – can validate AI tools, ensuring that “innovation” does not come at the cost of “uncertainty.”
- Interdisciplinary Stakeholder Integration: Recognizing that improving stroke management is a multi-faceted challenge, we bring together medical professionals, regulatory experts, and industry innovators to foster an ecosystem where clinical needs, legal safety, and technical feasibility are addressed in unison.
Building Ecosystems, Not Just Algorithms
A core tenet of the MDR in AIS mission – supported by our preliminary findings – is the conviction that healthcare technologies, including AI, cannot succeed in isolation. Beyond evaluating clinical outcomes, our researchers are investigating the broader healthcare environment to assess its readiness for integrated ecosystems, from the local stroke network to the European Health Data Space (EHDS).
These ecosystems should:
- Reduce Hurdles: By creating pre-validated pathways and standardized data-sharing protocols, we lower the barrier to entry for innovative SMEs and research centers.
- Ensure Patient Safety: Through rigorous human-in-the-loop requirements, ensuring that AI remains a “second opinion” that empowers, rather than replaces, the neuroradiologist.
- Protect Privacy: Implementing “Privacy by Design” to comply with both GDPR and the AI Act’s strictures on high-quality training datasets, ensuring that sensitive patient data remains secure within the consortium’s infrastructure.
The Path Forward
The future of stroke care depends on a shared language between clinicians, developers, and regulators. In the coming months, MDRinAIS will intensify its interdisciplinary efforts, adapting our research to the developing regulatory frameworks including AI Act ,into actionable protocols for stroke networks. Our upcoming workshop in March 2026 will serve as a critical forum for these stakeholders to align their objectives.
Author: Martin Vitiello, Siena University Hospital
Sources: EU AI Act Official Site, https://artificialintelligenceact.eu