EU Experience: CE-Marked Diagnostic AI
Article #8 in Medical ML for Ukrainian Doctors Series
![[Medical ML] EU Experience: CE-Marked Diagnostic AI](https://hub.stabilarity.com/wp-content/uploads/2026/02/photo-1576091160399-112ba8d25d1d.jpg)
By Oleh Ivchenko | Researcher, ONPU | Stabilarity Hub | February 8, 2026
Key Questions Addressed #
- How does the European regulatory framework for medical AI differ from the US FDA approach?
- What is the impact of the EU AI Act on medical device software, and how does dual regulation work?
- What lessons can Ukrainian healthcare draw from the EU’s evidence-focused approach?
Context: Why This Matters for Ukrainian Healthcare #
Ukraine’s regulatory trajectory aligns with the EU Medical Device Regulation (MDR) through ongoing European integration reforms. Understanding the European CE marking process—with its emphasis on clinical evidence and post-market surveillance—directly informs how Ukrainian hospitals should evaluate AI diagnostic tools.
A Different Regulatory Philosophy: EU vs US #
CE-Marked AI Devices: The Numbers #
~500-700
Total CE-marked AI devices
~53%
Radiology focus
~60%
Class IIa devices
~35%
Class IIb/III (high-risk)
The EU AI Act: A New Layer of Regulation #
The World’s First Comprehensive AI Law #
The EU AI Act (Regulation EU 2024/1689), adopted March 2024, represents the world’s first comprehensive legal framework for artificial intelligence. For medical devices, this creates a dual regulatory burden.
“`mermaid
graph TD
A[Medical AI Device] –> B{MDR Classification?}
B –>|Class I| C[Not High-Risk AI]
B –>|Class IIa+| D[High-Risk AI]
D –> E[MDR Compliance
+ AI Act Compliance]
E –> F[Risk Management
Art. 9]
E –> G[Data Governance
Art. 10]
E –> H[Technical Documentation
Art. 11]
E –> I[Transparency
Art. 13]
E –> J[Human Oversight
Art. 14]
style D fill:#dc3545,color:#fff
style E fill:#ffc107,color:#000
“`
AI Act Risk Classification #
Timeline for AI Act Compliance #
“`mermaid
timeline
title EU AI Act Compliance Milestones
August 2024 : AI Act enters into force
May 2025 : General-purpose AI obligations
August 2026 : Most high-risk AI obligations
August 2027 : Medical device AI obligations (36-month transition)
“`
The Dual Regulatory Framework: MDR + AI Act #
Integration, Not Duplication #
Article 11(2) of the AI Act allows manufacturers to maintain a single technical documentation file that combines MDR and AI Act requirements.
Requirements for High-Risk Medical AI #
European Market Leaders and Notable Devices #
Top Companies in European Medical AI #
Notable CE-Marked AI Applications #
Breast Cancer #
Transpara (ScreenPoint)
Mammography triage, Class IIb
🫁 Chest X-ray #
qXR (Qure.ai)
TB/pneumonia detection, Class IIa
Stroke Detection #
Viz LVO (Viz.ai)
Large vessel occlusion alert
Clinical Evidence: The European Approach #
MDR’s Clinical Evidence Requirements #
Post-Market Clinical Follow-up (PMCF) #
A distinctive EU requirement is ongoing PMCF studies after market entry:
- PMCF Plan: Proactive collection of clinical data post-CE mark
- PMCF Studies: Prospective/retrospective studies in real-world use
- PMCF Report: Regular updates, part of technical documentation
- Periodic Safety Update Report: Mandatory for higher-risk devices
European AI in Action: Case Studies #
Case Study 1: AI in European Breast Cancer Screening #
Case Study 2: The Colonoscopy Deskilling Warning #
(!)️ Critical Finding #
A pivotal 2025 Lancet study on AI-assisted colonoscopy revealed:
- Initial AI assistance: Improved adenoma detection
- After 6 months: Physician detection rates FELL when AI was withdrawn
- Implication: Over-reliance creates skill degradation
“Researchers found that over six months of using AI, clinicians’ ability to spot concerning features in colonoscopies declined when AI was not used during specific procedures.”
— Time, January 2025
Challenges Facing European Medical AI #
“`mermaid
mindmap
root((EU Medical AI
Challenges))
Regulatory Complexity
Dual MDR + AI Act
Two sets of documentation
Definition ambiguity
Notified Body Bottleneck
Only ~40 MDR bodies active
AI Act accreditation emerging
Limited dual-accredited
Market Fragmentation
National reimbursement
Language requirements
HTA variation
SME Impact
High documentation costs
Disproportionate burden
Innovation barrier
“`
Practical Implications for Ukrainian Healthcare #
What the EU Experience Teaches #
Recommendations for Ukrainian Hospitals #
- Prioritize CE-marked devices: EU certification provides stronger clinical evidence than 510(k)-only clearance
- Demand PMCF data: Ask vendors for post-market clinical follow-up results
- Plan for dual compliance: Expect both MDR-style and AI Act-style requirements
- Build monitoring infrastructure: Implement local performance tracking
- Train for collaboration: Prevent deskilling through proper training
Conclusions: Original Insights #
Global Future #
The EU’s dual regulatory framework represents the global future—other jurisdictions will likely adopt similar models
Yes Quality Filter #
Stricter clinical evidence requirements may result in fewer but better-validated tools
(!)️ Deskilling Risk #
European studies show AI assistance can degrade physician skills when over-relied upon
PMCF Advantage #
Devices with robust post-market data will increasingly differentiate themselves
Questions Answered #
Yes How does the EU regulatory framework differ from the US FDA approach?
The EU requires stronger clinical evidence through MDR conformity assessment, mandatory Notified Body review, and ongoing Post-Market Clinical Follow-up. The AI Act adds a second regulatory layer.
Yes What is the impact of the EU AI Act on medical device software?
Medical AI devices (Class IIa and above) are automatically classified as “high-risk AI,” requiring dual compliance by August 2027.
Yes What lessons can Ukraine draw from the EU approach?
The EU model emphasizes evidence-based approval, ongoing monitoring, and human-AI collaboration—principles Ukraine should adopt as it aligns with EU standards.
Next in Series: Article #9 – UK NHS AI Lab: Lessons Learned
Series: Medical ML for Ukrainian Doctors | Stabilarity Hub Research Initiative
Author: Oleh Ivchenko | ONPU Researcher | Stabilarity Hub
References (1) #
- Stabilarity Research Hub. [Medical ML] EU Experience: CE-Marked Diagnostic AI. doi.org. dtil