WHAT WE FOCUS ON IN OUR SPECIALIST RESEARCH

Our research spans today’s most pressing challenges and tomorrow’s emerging opportunities in AI and digital health. Some of these technologies are already transforming care, while others are still being developed. Across all our work, we aim to ensure that AI is safe, effective, equitable and inclusive, sustainable, technically robust, and affordable for health systems.

Research Projects

These key projects form the backbone of CERSI-AI’s implementation phase and represent our most critical areas of delivery and focus. Developed in collaboration with our UK-wide network of partners, they tackle the priority challenges identified by innovators, regulators, and health systems, ensuring that AI and digital health technologies are safe, effective, and trusted. From building transparency through AI medical device databases and post-market surveillance, to addressing algorithmic bias and secure data environments, these projects are designed to deliver real-world impact and position the UK as a global leader in regulatory science and innovation. Explore the projects below and see how we’re driving smarter regulation for better innovation.

Optimising SDEs for AI Health Technologies

This project explores how Secure Data Environments can be optimised to support AI health technology development. Led by UHB and the University of Birmingham, it balances innovation, privacy and governance, proposing technical and policy improvements to enable safe, trusted NHS data access.

Qualification and Classification of Large Language Models

This project clarifies when large language models should qualify as medical devices in healthcare. Led by the University of Birmingham and UHB, it defines LLM use cases, risk classification and regulatory guidance to support safe, accountable adoption and inform UK and international regulation.

The AI Readiness Checklist

The AI Readiness Checklist is a practical self-assessment tool helping UK healthcare organisations prepare for safe, effective and equitable AI adoption. Funded by NHS England and the Health Foundation, it evaluates data, governance, workforce and ethics readiness to support deployment.

Hardian Regulatory Intelligence Platform (HaRi)

The Hardian Regulatory Intelligence Platform (HaRi) brings together global medical device regulatory and safety data into one connected system. Developed by Hardian Health, it supports faster, safer, and more transparent decision-making for regulators, healthcare professionals, researchers, and patients worldwide.

Borderline Manual for Software as a Medical Device (SaMD)/ AI as a Medical Device (AIaMD)

The Borderline Manual for SaMD and AIaMD provides practical UK guidance on classifying AI health technologies as medical devices. Developed by the University of Birmingham and Hardian Health, it clarifies regulatory status, risk classification and compliance pathways to support safe, accelerated innovation.

Projects Funded by Cersi-AI

These projects are funded through the CERSI-AI Project Support Fund and reviewed by our Project Prioritisation Committee. Delivered by our CERSI-AI network partners, they represent our shared commitment to smarter regulation for better innovation, ensuring the UK leads in safe, effective, and equitable adoption of AI and digital health technologies. Each project tackles key priorities such as AI-enabled medical device regulation, large language model qualification, post-market surveillance, and algorithmic performance and bias. Learn more about each project below.

Qualifying Large Language Models for Oversight in Patient Safety Governance

This University of York project evaluates how large language models can support NHS patient safety governance. It compares LLM reviews with expert oversight, identifies risks, and develops qualification frameworks to inform safe, transparent and scalable regulatory use in non-diagnostic settings.

Model stress testing using causal generative AI

This Imperial College London project uses causal generative AI to stress test medical imaging models with realistic synthetic data. It identifies performance weaknesses across populations and conditions, supporting safer clinical deployment and informing MHRA guidance on AI evaluation.

ARIES GenAI tool producing radiology-style reports in 3 imaging modalities

ARIES-MED evaluates a generative AI tool producing radiology-style reports across CT trauma, CT head and chest X-ray imaging. Led by the University of Glasgow, the project assesses clinical accuracy, safety and acceptability to support safe NHS adoption and regulatory guidance.

Autonomous Reporting of Chest X-Rays- side by side comparison of two CXR tools in Glasgow

This project evaluates two AI tools—qXR and Annalise Enterprise CXR—using 110,000 chest X-rays from NHS Greater Glasgow and Clyde. It assesses whether these tools can accurately identify normal scans without missing clinically important findings, potentially reducing reporting delays caused by radiologist shortages.

CLIO – extension to current Newton’s Tree UK project monitoring the risk of ambient scribes

To implement and evaluate a federated monitoring platform for AI scribe tools across NHS outpatient settings, enabling monitoring of input quality, AI output drift, and clinician over-reliance.