From Zero to Hero: Federated AI in Healthcare Systems — Cambridge, 24–25 August\nRafael Garcia-Dias, Senior AI Engineer on Federated Learning, represented the AI Centre at \nFrom Zero to Hero: Federated AI in Healthcare Systems\n, a two-day workshop held at the Old Divinity School, St John’s College, Cambridge, and sponsored by Flower Labs. Convened by organisers from the University of Cambridge, Queen Mary University of London, the University of Exeter, PharosAI, Munster Technological University and Flower Labs, the programme combined keynotes, research talks, live demonstrations and hands-on tutorials, with keynotes from André Dekker (Maastricht UMC+ and Maastro Clinic), Mark Durkee (Government Digital Service) and Carsten Maple (University of Warwick).\nRafael gave a live demonstration of FLIP, an open-source, reusable platform for federated learning in healthcare. FLIP lets multiple hospitals:\nDiscover cohorts across sites\nApprove projects locally\nPrepare and annotate data within each hospital’s secure enclave environment\nRun and monitor federated training from a central hub\nHe then joined the workshop’s closing panel, Scaling out: the limits of federated AI in healthcare today, and where it goes next, chaired by Nicholas Lane (Flower Labs) alongside Olaoluwa Ademolu (Barts Health NHS Trust / PharosAI), Carsten Maple (University of Warwick) and Bing Luo (Duke Kunshan University). The discussion returned repeatedly to a theme familiar from our own deployments: the hard part of federated learning in healthcare is rarely the algorithm, but the governance, infrastructure and institutional trust required to get a model running across several hospitals at once.\nView Rafael’s presentation slides via this link\n \nhere\n.

AI Centre Presents FLIP at Cambridge Federated AI Workshop 

At the AI Centre (AIC), we are partnering with the University of Cambridge, Cambridge University Hospitals, and NHS Trusts across the UK to harness the power of the newly launched Zenith AI Supercomputer. Our focus is on using this national-scale capability to accelerate the development and deployment of impactful healthcare AI solutions.\nThe launch of the Cambridge Zenith AI Supercomputer, unveiled during a visit by the Minister for Science, Innovation and Technology, marks a significant step forward for the UK’s AI infrastructure. As the country’s largest AI-for-science platform, Zenith provides the computational scale required to support advanced research and real-world clinical applications.\nThrough this collaboration, we are working to connect NHS data and expertise with high-performance computing infrastructure. This enables us to develop next-generation tools in predictive analytics, diagnostics, and personalised medicine. A key priority for the AI Centre is advancing cancer research through the MOSAIC (Multimodal Oncology Sovereign AI Collaboration) programme, where we are contributing to the development of NHS cancer foundation models.

AI Centre to harness newly launched Zenith AI Supercomputer to advance AI-driven healthcare innovation

On 14th April 2026, the AI Centre ran a workshop to showcase the progress of the AI Framework workstreams, with the aim to develop NHS system wide approach to AI adoption effectively and safely within Southeast London.\nThe team presented Successful AI use cases, processes for AI deployment, and a new AI Assessment tool. The tool, developed by Jessica Morley, aims to help NHS organisations assess whether a specific AI tool is appropriate for their specific context. It pairs 12 complexity dimensions (properties of the tool) with 12 readiness dimensions (capabilities of the deploying organisation) to produce a context-sensitive deployment recommendation. \nThis tool is aimed at Clinical safety officers, digital transformation leads, AI programme managers, procurement teams, and anyone involved in decisions about adopting AI tools in NHS settings. The assessment is designed to be completed collaboratively — it works best when clinical, technical, and operational perspectives are all represented. It is still in development and is being tested by organisations in primary, secondary, acute and mental health in Southeast London.\nThe workshop generated some great discussions with colleagues from across the system, including some honest reflections around adoption challenges, ethics concerns and patient views.  \nThank you to Jawad Chaudhry, Sigal Hachlili, Jessica Morley, Dijana Vilic, Matea Deliu, Findlay Macaskill, Gian-Marco Strappelli, Lisa Law, Nisha Wheeler and the London Institute for Healthcare Engineering. 

AI Framework workshop for Southeast London Integrated Care Board

\n“Demystifying artificial intelligence in health: What health policy-makers need to know”\n is a new publication written for those working in, or alongside, the health sector who are increasingly required to engage with artificial intelligence (AI). The book aims to demystify AI for health policy-makers, offering a clear-eyed view of its capabilities, limitations and implications. It equips leaders with the foundations for informed decision-making, helping them ask the right questions, understand trade-offs, and shape a future in which AI serves public health goals. Crucially, it argues that realising AI’s potential requires principled adoption, rigorous evaluation and inclusive governance arrangements, ensuring that AI remains a means to an end rather than an end in itself.\nAmong the case studies included is the \nOneLondon Secure Data Environment (SDE)\n and the work delivered within it by the AI Centre for Value Based Health Care. The case study explores how secure, well-governed data infrastructure can provide the foundations for responsible and scalable AI deployment within the NHS.\nThe OneLondon SDE provides a secure data environment that enables real-time access to linked health data to support system-wide decision-making across London. Applications include optimising care pathways for long-term conditions, identifying high-risk patients for targeted intervention, and generating population health insights to inform planning and service transformation. By embedding strong governance, transparency and oversight into its architecture, the SDE establishes the trusted conditions necessary for advanced analytics and AI.\nWithin this secure environment, The AI Centre has developed reusable and adaptable analytic pipelines and interoperable building blocks to reduce duplication, enable reproducibility and move beyond static, one-off datasets towards dynamic, real-time data flows. The infrastructure supports multimodal data analysis across hospital trusts through \nfederated learning approaches\n that allow AI model development without physically transferring sensitive patient data. In parallel, efforts to standardise data from proprietary hospital systems and to apply advanced natural language processing to transform unstructured clinical notes into structured, analysable information are expanding the scope and richness of the data available.\nPublic trust remains central to this work. Through public engagement initiatives, including citizen summits and collaboration with \nIntegrated Care Boards\n and patient committees, \nOneLondon\n has embedded patient and public voice into the governance of data use. This commitment to transparency and accountability reinforces the principle that responsible AI depends not only on technical capability, but on public trust.\nThe inclusion of this case study highlights an important message for health systems: sustainable AI cannot be built on isolated pilots or fragmented datasets. It requires strong foundations, secure environments, interoperable systems, robust governance and meaningful public engagement, aligned with policy priorities and the public interest.\nAs AI technologies move from research environments into health systems, the book provides a timely and practical guide for decision-makers navigating this complex landscape and seeking to translate technological capability into measurable value for patients and populations.\n

AI Centre for Value Based Health Care featured in new European Observatory on Health Systems and Policies book on AI and Health Policy

From 24 to 26 February 2026, we hosted the \nAI Radiology summit\n, bringing together selected industry-leading experts, academics, clinicians, and policy advocates for 3 days of intense collaborative discussions to shape the future of AI in healthcare.\nThe conversation around artificial intelligence in healthcare is no longer just about the algorithms themselves. It’s about the \nnetworks\n that will power them. This was the resounding theme of the panel featuring Dr. Joe Zhang, Joanna Wills, Sam Khashman, Henriettae Ståhlbrandt, and Bryce Travers, moderated by Dr. Christina Triantafyllou.\nThe room was filled with great expertise and a shared sense of purpose. The conversation laid bare a profound paradox: we have plenty of health data that could deliver immense value to patients and the health care system, but it’s scattered across thousands of disconnected silos, each with its own governance, incentives, and way of doing things.\nAs one Bryce Travers aptly put it, “My trusts have no real incentive to share data with each other.” Henriettae Ståhlbrandt highlighted the cultural hurdles, noting the need to “trust the private companies that they also have the greater good in mind,” underlining that these aren't purely technical problems; they are problems of trust, incentives, alignment, and infrastructure.

From Silos to Synapses: Why Federated Learning is the Network Healthcare AI Has Been Waiting For

London & Munich — February 26, 2026\n - The \nAI Centre for Value-Based Healthcare\n at \nKing’s College London\n and \nGuy’s and St Thomas’ NHS Foundation Trust\n, together with strategic partner \ndeepc\n, today announced the launch of \nFLIP\n, a fully open-source federated learning interoperability platform designed to support large-scale, collaborative development of healthcare AI while ensuring patient data remains local and under institutional control. Developed within the NHS ecosystem in collaboration with \nOneLondon\n and \nFlower Labs\n, FLIP is designed to enable secure, multi-institutional AI collaboration across health systems.\nDeveloped, governed, and maintained within the NHS, FLIP addresses a longstanding challenge in healthcare AI: the tension between the need for large, diverse data sets to train clinically robust models and the ethical, regulatory, and societal imperative to protect sensitive patient data. As modern AI increasingly relies on foundation models and multimodal learning approaches, this challenge has become a limiting factor for progress across clinical research and innovation.\nFLIP enables AI models to be trained across one or multiple institutions while data stays within secure local environments. By allowing models to learn from a wider range of data, it has the potential to improve patient care by increasing accuracy of diagnosis, support earlier detection of disease, and help bring new innovations into care more quickly - all while maintaining high standards of privacy and data protection.

AI Centre and deepc Launch FLIP, an Open-Source Platform to Enable Health AI Innovation at Scale, in Partnerships with OneLondon and Flower Labs

During London Life Sciences week (17-21 November 2025) NHS London teams showcased that London is open for business through strong partnerships across the capital's NHS, universities and businesses. 
\n
\nThe London NHS team announced that the London Secure Data Environment is scaling to benefit all Londoners and launched three new driver projects that are addressing challenges in children and young people's mental health, cardiovascular disease, cancer outcomes and antimicrobial resistance. 

\nLondon Secure Data Environment Driver Projects - OneLondon\n\nWe also heard how the development of the London Secure Data Environment has been shaped by several large-scale deliberations with the public, and how it is already supporting groundbreaking research in collaboration with the AI Centre. \nThis is part of a shared vision to make London the healthiest global city, and the best place to partner for research and innovation in MedTech, drug discovery, clinical trials etc. – that will improve patient outcomes.

It was fantastic to hear from the minister, Dr Zubir Ahmed, and some of the founding CEOs of Health Data for London that as a leadership community we intend to pick up the mantle and 'collaborate to compete' to make London and the UK the world's most innovative place for life sciences and health improvement.

Health Data for London event as part of London Life Sciences week

The AI Centre presented a series of technical developments at \nISPOR Europe 2025\n, showcasing progress in federated learning for imaging and clinical data, large language model–based information extraction in Oncology, and new methodologies to support population health research across the NHS.\nFLIP: Federated Learning Interoperability Platform\nThe team introduced updates to FLIP (Federated Learning Interoperability Platform), an open-source infrastructure designed to enable secure, hospital-retained analytics across NHS sites.\nFLIP provides capabilities for federated cohort discovery, multi-modal data engineering — spanning structured data, natural language processing, and radiology — and AI model training without transferring sensitive patient data outside hospital boundaries. The platform supports deep phenotyping using un-curated, real-world datasets, addressing limitations of traditional, highly curated research environments.\nDevelopment of a Unified Medication Adherence Index\nResearchers also reported findings on a proposed Medication Adherence Index, addressing the current absence of a standardised, robust measure of long-term medication adherence.\nUsing GP electronic health records, the AI Centre evaluated the index across three major classes of long-term medications and validated it against both subjective and objective adherence indicators. Results demonstrated strong predictive performance, underscoring the index’s potential to support medicines optimisation, risk stratification, and population health management.\nNational Deployment of OncoLlama\nThe AI Centre additionally highlighted progress on OncoLlama, an NHS-developed large language model designed to extract detailed Oncology information from unstructured clinical text.\nPoor availability of structured cancer data has long constrained research, innovation, and clinical trial readiness. Through a national collaboration, the programme will deploy secure infrastructure within NHS Trusts, implement advanced LLM processing pipelines, and generate high-resolution Oncology datasets. The initiative aims to enhance real-world evidence generation, improve cancer registry data quality, and support clinical trial recruitment.

AI Centre Team Presents Advances in Federated Learning, Oncology Data Extraction, and Population Health at ISPOR Europe 2025

Dr James Teo, Consultant Neurologist (\nKCH\n) and Chief Medical Officer at the AI Centre, has been appointed to the Health Systems Working Group of the National Commission into the Regulation of AI in Healthcare, established by the Medicines and Healthcare products Regulatory Agency (MHRA).\nLaunched in September 2025, the National Commission brings together global experts in artificial intelligence, clinical practice, and health regulation to help shape a new framework governing AI use in healthcare. Chaired by Professor Alastair Denniston, with Professor Henrietta Hughes as Deputy Chair, the Commission will publish its final recommendations in 2026. Its work underpins commitments set out in the government’s 10-Year Health Plan and Life Sciences Sector Plan, aimed at transforming NHS services and accelerating growth within the UK’s life sciences industry.\nAs part of the Health Systems Working Group, Dr Teo will provide clinical and operational insights to guide development of regulatory approaches that ensure AI tools are deployed safely, effectively, and for maximum patient benefit. The group’s agenda covers critical topics such as post-market surveillance, pre-market approval processes, professional practice, and governance responsibilities for healthcare providers.\nThe Commission’s programme will run from November 2025 to May 2026, examining key areas including the continuous oversight of AI systems, model update controls, and the balance between developer autonomy and regulatory assurance.\nDr Teo’s appointment highlights the importance of bringing frontline clinical experience into national policy discussions to ensure that emerging AI regulation supports both innovation and patient safety.

Dr James Teo Appointed to MHRA Working Group on AI Regulation in Healthcare

Artificial Intelligence (AI) is rapidly becoming a crucial component in the future of the NHS and Integrated Care Systems (ICS) across the UK. Across the NHS the implementation of AI technologies is showing promising results in improving healthcare services and patient experiences. From improving diagnosis rates to faster automation of patient registrations and more accurate patient referral support tools.\nAI Framework Launch\nThe new \nAI Framework\n launched on 3rd March will help the NHS in London to deliver safe and ethical pathways to use these tools effectively. The framework covers the agreed way of introducing, using and monitoring Artificial Intelligence (AI) products across the NHS.\nWe see significant levels of demand and recognise that the NHS needs to change to be able to best support people and communities in our area. For example, patients can find it hard to get appointments and can face a long wait for treatment. AI tools could improve this if used effectively and responsibly; one product has already reduced the average wait time to register
with a GP practice from days to hours.\nWe want to harness the potential power of AI products in a safe, effective and efficient way to improve health and care services for people and communities. To do this, we have driven the development of the Framework for our area and for London.

NHS AI Framework launched to support safe and ethical use of AI in healthcare across London

AWS Public Sector AI Symposium London — 9 September\nAlexandre Triay Bagur, Senior AI Engineer, presented NHS Federated Learning: Training AI Without Moving Patient Data in the healthcare track of the \nAWS Public Sector AI Symposium\n at Convene 133 Houndsditch, in partnership with OneLondon.\nThe first half of the session set out the scale of the underlying data engineering problem. A single London NHS Trust runs eight or more structured clinical systems alongside separate stores of narrative documents, letters and imaging — together holding 20 years of real-world data on 5.2 million patients, including some 10 billion rows of clinical concepts and 300 million documents. Aggregated across the capital, this becomes a living data fabric covering 11 million patients, built on shared code and common data models rather than one-off extracts.\nMuch of that clinical detail is locked in free text, and the session briefly introduced MESA (Medical-concept Extraction with Schema Alignment), the AI Centre’s open-source toolkit for fine-tuning clinical language models on AWS, now in active deployment across five cancer centres. MESA is available at \ngithub.com/londonaicentre/MESA\n.\nRadiology is the harder case: 17.2 million imaging studies spanning two decades, at petabyte scale and impractical to centralise on any realistic timeline. This is what FLIP was built for. Rather than moving images, FLIP places a node alongside a secure enclave inside each Trust and coordinates training across those nodes from a central hub. Apache 2.0 licensed and built on AWS with NVIDIA FLARE, Flower and deepc, it is deployed across Guy’s and St Thomas’, King’s College Hospital, King’s College London, the London Secure Data Environment and Bangkok Dusit Medical Services in Thailand.\nThe session walked through a worked example: federated fine-tuning of the Ark+ chest X-ray foundation model across the UK and Thailand. With the pre-trained backbone frozen, only a small classification head was trained — a useful illustration of how little needs to cross a network boundary to do something clinically meaningful.\nAlex closed on security and data sovereignty, which FLIP enforces structurally rather than by policy alone. Each Trust approves every project separately; only aggregate statistics, filtered model updates and task status ever leave the hospital; and Trusts poll outbound only, so the hub has no network route in. The National Data Opt-Out is applied where the data is prepared, so it propagates to every query, project and model. FLIP is a Health Research Authority REC-approved Research Database overseen by a Data Allocation Committee, and has been assessed against Cyber Essentials, the NHS Data Security and Protection Toolkit, the NCSC Cyber Assessment Framework and the SATRE specification for federated analytics.\nView Alex's presentation slides via this link \nhere\n.

AI Centre for Value Based Healthcare

News

FLIP demonstrated at AWS Public Sector AI Symposium

AI Centre Presents FLIP at Cambridge Federated AI Workshop

AI Centre to harness newly launched Zenith AI Supercomputer to advance…

AI Framework workshop for Southeast London Integrated Care Board

AI Centre for Value Based Health Care featured in new European Observatory…

From Silos to Synapses: Why Federated Learning is the Network Healthcare…

AI Centre and deepc Launch FLIP, an Open-Source Platform to Enable Health…

Health Data for London event as part of London Life Sciences week

AI Centre Team Presents Advances in Federated Learning, Oncology Data…

Dr James Teo Appointed to MHRA Working Group on AI Regulation in Healthcare

NHS AI Framework launched to support safe and ethical use of AI in…