Professor David Brown
Professor in Interactive Systems for Social Inclusion at Nottingham Trent University’s School of Science and TechnologySpeaking at the UKAI and Curia Healthcare and Life Sciences Parliamentary Showcase, Notting Trent University’s Professor David Brown highlighted how privacy preserving data sharing, synthetic data and explainable AI could help the NHS identify lung cancer patients at greatest risk of disease earlier.
Professor in Interactive Systems for Social Inclusion at Nottingham Trent University’s School of Science and Technology, Professor David Brown used his presentation at the UKAI and Curia Healthcare and Life Sciences Parliamentary Showcase in Parliament to set out how trusted artificial intelligence could help enable earlier detection of lung cancer.
Speaking after the keynote from Lawrence Tallon, Chief Executive of the Medicines and Healthcare products Regulatory Agency, Professor Brown focused on one of the central questions facing the future of AI in healthcare: how can the NHS use data at scale while protecting privacy, maintaining trust and ensuring that AI systems can be properly explained?
Using AI to identify risk earlier
Professor Brown’s central use case was lung cancer detection. He explained that artificial intelligence has transformative potential for the NHS, particularly in early diagnosis, risk prediction, and personalised care. His work uses data from the Clinical Practice Research Datalink to identify patients who may be at greatest risk of lung cancer before symptoms appear.
This matters because lung cancer remains one of the UK’s most serious healthcare challenges. Professor Brown noted that lung cancer is the leading cause of cancer death in the UK and is often diagnosed late, when survival rates are much lower. By contrast, earlier diagnosis can dramatically improve survival prospects. His argument was clear: if AI can help identify high risk patients earlier, it could improve outcomes for individuals while also reducing pressure on the NHS.
However, he also stressed that the NHS cannot simply screen everyone. Low dose computed tomography is highly accurate, but expensive. The challenge is to identify intelligently who should be prioritised for screening. AI models, if developed responsibly, could help the NHS move patients at greatest risk more quickly into the right diagnostic pathway.
The challenge of fragmented health data
A major barrier, Professor Brown argued, is that health data is often fragmented across different systems, including primary care, community services, and mental health. Different coding systems, disconnected records, and organisational boundaries make it harder to build integrated models that support prediction and early intervention.
His answer is not to lower standards on privacy, but to design privacy into the system from the beginning. Professor Brown described work where sensitive health data is transformed into privacy-preserving representations including synthetic data, while techniques such as differential privacy ensure that individual patient identities cannot be re-identified. This allows large datasets for areas such as lung cancer, prostate cancer, ischaemic stroke, and Alzheimer’s disease to be used by AI developers without exposing identifiable patient data.
Privacy preserving AI and federated learning
Professor Brown also highlighted the PHASE IV AI project (https://www.phase4ai-project.eu/ ), a Horizon Europe project in which he is the UK lead partner and principal investigator. The project supports the use of synthetic health data, federated learning, models as a service and multi-party computation to help researchers and innovators develop AI safely.
Federated learning was a central part of his presentation. Rather than moving sensitive health data between organisations or countries, local models are trained on local data. Only the model weights and parameters are shared to update a centrally held model. This means researchers can learn from data across different health systems without directly exposing the underlying patient information.
For the NHS, this approach could help address one of the major barriers to AI adoption: the need to collaborate at scale without compromising privacy, trust, or legal compliance.
Explainability and public trust
Professor Brown was also clear that AI must be explainable. He said that, where black box models are used, their outputs must still be explained in terms of which features contributed to the prediction. For lung cancer risk, relevant factors include smoking history, family and personal cancer history, ethnicity, geographic location, age, and environmental exposures.
This is an important point for public trust. If AI is being used to help determine someone’s risk of disease, patients and clinicians need to understand why a model has reached a particular conclusion. Professor Brown described the use of techniques such as SHAP values to help explain how opaque models produce their outputs.
Nottingham’s role in healthcare innovation
Professor Brown closed by highlighting Nottingham Trent University’s Medical Technologies Innovation Facility, a £23 million dual site facility supporting medtech development, healthy ageing, diagnostics and point of care solutions. He also pointed to ambitions for a digital health city of Nottingham, bringing together regional partners to support health, wellbeing, and business growth.
His presentation showed how regional innovation ecosystems can contribute to national NHS transformation. The message was not simply that AI can detect disease earlier. It was that trusted AI depends on the right foundations: secure data sharing, privacy preserving infrastructure, explainable models, regional collaboration, and a clear focus on patient benefit.