Over summer, the Elizabeth Blackwell (EBI) and Jean Golding (JGI) Institutes together with the Faculty of Health and Life Sciences strategic research support fund, and University Hospitals Bristol and Weston NHS Foundation Trust, ran a pump-priming funding call to support innovative applications of AI in health or biomedical research. This funding call came with the expectation that the funded activities would provide a basis for developing and submitting external bids for future research programmes and projects that use or address AI in health and biomedical research contexts.
We are excited to announce 10 projects involving more than 30 researchers supported by this funding. Check out the successful awardees and their projects below.
AI-assisted personalisation of neurostimulation
- Petra Fischer, School of Physiology, Pharmacology and Neuroscience
- Conor Houghton, School of Engineering Mathematics and Technology
Dystonia is a heterogenous neurological disorder, which causes involuntary muscle contractions, often resulting in pain and severely restricted movement, affecting millions of people worldwide.
A key challenge in neuroscience is understanding how brain networks process sensory input to control movement. Neural synchronisation plays a vital role in organising this activity, occurring both locally and across distant regions. Excessive synchronisation is linked to disorders like dystonia, Parkinson’s, and schizophrenia, and targeted modulation has emerged as a promising therapy.
Dr Fischer’s lab uses non-invasive, phase-specific vibrotactile stimulation to selectively enhance or disrupt synchronisation in dystonia with the aim to improve symptoms. Currently it is still unclear whether local or interregional modulation drives therapeutic effects. The interdisciplinary team will use existing dystonia data and AI-assisted analysis to:
- Map brain-wide effects of stimulation, and
- Predict outcomes and effective stimulation parameters based on neural data to replace a trial-and-error based search procedure
Findings will support an MRC funding application to develop a clinical stimulation tool, with potential extension to other techniques like transcranial electrical stimulation for direct cortical targeting.
AI-Organoid: A Smart Predictive Platform for Advanced Neurological Modelling
- James Armstrong, Bristol Medical School
- Qiang Liu, Engineering Mathematics & Technology
This interdisciplinary project brings together biomedicine and AI to develop AI-Organoid, a predictive, interpretable platform for tracking and forecasting the development of organoids—lab-grown cell models that mimic human organs. Organoids are vital for reducing animal testing and studying human-specific diseases, but their inconsistent growth limits pharmaceutical applications.
Building on promising pilot data (74% accuracy in predicting brain organoid outcomes), the project will refine AI-Organoid to improve reproducibility and provide mechanistic insights into organoid development. Beyond forecasting organoid outcomes, the platform will identify when developmental trajectories diverge and provide mechanistic insights into the underlying biology of brain organoid growth.
The EBI-JGI grant will support data collection, model training, and dissemination, enabling future applications in disease modelling (e.g., neurodevelopmental disorders) and expansion to other organoid types (e.g., ovarian, intestinal, liver). Outputs will include publications, open-source tools, and a foundation for commercialisation and further funding bids.
An AI-integrated lung-on-a-chip platform for the rapid screening and optimisation of mesenchymal stem cell secretome therapeutics
- Wael Kafienah, School of Biochemistry and Cellular and Molecular Medicine
- Lucia Marucci, School of Engineering Mathematics and Technology
- Darryl Hill, School of Biochemistry and Cellular and Molecular Medicine
Inflammatory lung diseases, such as acute respiratory distress syndrome (ARDS), are devastating conditions with high mortality and no effective drug treatments. A promising new therapy involves using the cocktail of healing molecules secreted by mesenchymal stem cells (MSCs). These cells can be manipulated to optimise the secretome composition towards a specific therapeutic target. However, identifying the most potent secretome composition is a major bottleneck, relying on slow, laborious methods and animal models that poorly predict human responses.
This project aims to develop an AI-integrated Lung-on-a-Chip (LoC) platform to accelerate discovery of effective MSC therapies for inflammatory lung diseases like ARDS, which currently lack treatments. By mimicking lung inflammation and analysing cell responses in real time, the AI will identify optimal MSC secretome compositions more efficiently than current methods.
The pilot will deliver proof-of-concept data, including a high-accuracy AI model and a rich imaging and gene expression dataset, laying the foundation for reducing reliance on animal models and enabling rapid development of regenerative therapies with commercial potential.
An Integrated AI and machine learning platform to enable high throughput, precision oncology driven drug testing
- Deepali Pal, School of Biochemistry and Cellular and Molecular Medicine
- Colin Campbell, Mathematics, Engineering and Technology
- Stephen Cross, Wolfson Bioimaging Unit, Faculty of Life Sciences
- Rihuan Ke, School of Mathematics
250 children in the UK die from cancer each year. Leukaemia, affecting the human bone marrow, is the commonest children’s cancer. Yet it is a rare disease, which makes studying new treatments in clinical trials challenging. Therefore preclinical prioritisation is key, requiring predictive patient-derived models. However, hospital samples are difficult to cultivate, and where complex tissue-like biomimetic models to allow patient-sample cultivation have been engineered, these are hard to read, making output data inaccessible.
This project will develop an AI-powered bioimaging analysis tool to accurately detect leukaemia cells within complex bone marrow microenvironments, enabling predictive personalised drug screening.
The integrated machine deep learning and AI tool will analyse 3D bioprinted organoids to distinguish leukaemia from morphologically similar bone marrow cells, overcoming limitations of marker-based identification. The interdisciplinary team will apply image processing, expert annotation, and algorithm development to validate the tool.
Impact includes a proof-of-concept precision oncology organoid platform that generates high-throughput, interpretable drug screening data within clinically relevant timeframes. The project also offers commercial potential in the growing organoid market including applicability across other diseases.
Rational AI Driven Target Acquisition from Genomes (RAIDTAG)
- Darryl Hill, School of Cellular and Molecular Medicine
- Sean Davis, School of Chemistry
This pilot study uses AI to accelerate the discovery of highly repetitive DNA sequences for rapid microorganism identification—critical for healthcare, agriculture, and food production. These ‘repetitive signatures’, absent in closely related species, offer precise, cost-effective diagnostic markers.
Building on proof-of-concept using gold nanoparticle detection, the project will deliver a minimal viable product (MVP): an AI pipeline that automates marker discovery and validates candidates in the lab. This interdisciplinary effort, combining expertise in AI and computer science with nanomaterials and diagnostics, will provide the foundation for future external funding applications and translational research.
Predicting PD-L1 status from H&E slides using AI
- Tom Dudding, Bristol Dental School
- Qiang Liu, School of Engineering Mathematics and Technology
- Sarah Hargreaves, Bristol Dental School
- Miranda Pring, Bristol Dental School
There are approximately 377,700 newly diagnosed mouth cancers each year worldwide. Despite treatment, survival rates remain low, and side effects are life changing. Some people with early-stage cancer unexpectedly experience poor outcomes, such as recurrence or early death, and these high-risk patients are often hard to identify. New ways to detect and manage these high-risk cancers are needed.
PD-L1 is a marker found on many cells, including mouth cancer cells. It helps cancers hide from the immune system and may contribute to poorer outcomes. This marker can however be targeted using drugs like Pembrolizumab, which block PD-L1 and help the patient’s immune system fight the cancer. Because of this, PD-L1 is now an important marker used in clinical practice, to guide treatment decisions for people presenting with late-stage head and neck cancers.
This project aims to develop an AI tool to predict PD-L1 expression in mouth cancer directly from digital histology slides, bypassing costly and limited lab tests.
Using the HN5000 cohort—750 digitised slides with linked biospecimens and long-term follow-up—the pilot will create a proof-of-concept AI model. This will support future funding bids, improve diagnostic equity, and expand access to immunotherapy in both NHS and global settings. Deliverables include a validated AI tool for PD-L1 detection, benchmarking against immunohistochemistry to establish reliability, and preliminary analysis to underpin external bids enabling translation of AI-enabled PD-L1 testing into multi-centre validation and ultimately routine clinical practice.
Genetic doppelgangers: using AI to reveal the true face of streptococcal disease
- Alice Halliday, Biochemistry and Cellular & Molecular Medicine
- Colin Campbell, Engineering Mathematics and Technology
- Rachel Bromell, Biochemistry and Cellular & Molecular Medicine
- Anu Goenka, Bristol Medical School
- Sion Bayliss, Bristol Veterinary School
This project aims to use AI tools to evaluate and develop diagnostics to distinguish between Streptococcus pyogenes (GAS) and related bacterium, Streptococcus dysgalactiae subspecies equisimilis (SDSE). GAS and SDSE are very similar genetically, such that they are akin to ‘genetic doppelgangers’. Modern DNA-based tests (qPCR) have not been well evaluated for SDSE detection and struggle with closely-related bacteria. This diagnostic confusion impacts our understanding of responsible pathogens and clinical consequences, with increasing evidence that SDSE’s disease burden is significantly underestimated.
Using genetic material from bacteria isolated from clinical throat swabs, this new interdisciplinary team will build on Bristol AI expertise to develop a machine learning classification tool for distinguishing these bacterial species, based on genome ‘k-mers’.
Combining traditional microbiology, novel DNA-based assays and cutting-edge ML analysis of genome sequence data, the team aim to evaluate and design diagnostic tools that accurately identify both pathogens. This could lead to improved diagnostic capabilities, enhanced disease surveillance, better outbreak investigations, and improved patient outcomes.
Bristol Respiratory Infection Dashboard (BRID Project)
- Andrew Dowsey, Bristol Veterinary School
- Raul Santos-Rodriguez, Engineering Maths and Technology;
- Maha Albur, Consultant Microbiologist
- Peter Muir, Consultant Clinical Scientist
- Paul North, Business Support Manager and Data Analytics, Severn Pathology, North Bristol NHS Trust and UKHSA
- Amy Carson, Academic Clinical Fellow
- Gavin Deas, Doctor in Training
- Marceli Wac, Engineering Maths and Technology
- Jack Stanley, Academic Clinical Fellow, Severn Pathology, North Bristol NHS Trust and University of Bristol
This project will develop the AI-powered Bristol Respiratory Infection Dashboard (BRID) at Severn Pathology, serving the South West region of the UK. By integrating real-time data from pathology and care sources, BRID will enable early detection of respiratory infection trends and support targeted interventions.
Respiratory Tract Infections (RTIs) are a major cause of hospital admissions, with over 400,000 cases in 2024 and winter surges up to 80%. Despite available treatments, disparities in vaccine uptake and care access persist. Real-time surveillance is essential to guide equitable, effective responses and improve outcomes.
This project supports NHS England’s strategy for managing acute respiratory infections (ARIs) through integrated care and digital innovation. Backed by medical directors from both merging trusts, the project will compare AI modelling using local Electronic Patient Record (EPR) data versus the South West Secure Data Environment (SWSDE), evaluating data quality, linkage, and implementation.
A co-designed dashboard with clinicians will guide funding bids to scale the platform, aiming to reduce admissions, optimise resources, and improve public health.
Automated image analysis to facilitate the incorporation of quality assurance measures into surgical RCTs
- Natalie Blencowe, Bristol Medical School
- Michael Wray, School of Computer Science
- Anni King, Bristol Medical School
- Sheraz Marker, GOLF study, University of Oxford
- Nainika Menon, GOLF study, University of Oxford
This project aims to develop an AI model to streamline quality assurance (QA) in surgical randomised controlled trials (RCTs), addressing bias caused by variability in surgical technique and skill. Using annotated videos from the GOLF trial, the AI will assess key operative steps based on anatomical visibility as a proxy for quality.
This project provides valuable pilot work to further the application of AI to surgical videos, enabling QA processes to be efficiently embedded into surgical RCTs, meaning they can be adopted more widely. In turn this will improve RCT quality, ultimately improving patient outcomes. There is also potential for these methods, powered by AI, to be used in ‘real time’ during operations to alert surgeons if a key step has not been fully completed, immediately improving surgical quality. Both these applications have wider implications outside of research studies. In routine clinical practice, they could shorten surgeons’ learning curves through the provision of bespoke, real-time feedback. This could transform the way surgeons learn, as well as optimising patient care.
Explainable AI for early categorisation of child deaths: real-time insights for prevention
- Karen Luyt, Bristol Medical School
- Edwin Simpson, School of Engineering Mathematics and Technology
- Brian Hoy, Bristol Medical School
- James Gopsill, School of Electrical, Electronic and Mechanical Engineering
- David Odd, School of Medicine, Cardiff University
This project brings National Child Mortality Database analysts from the Faculty of Health and Life Sciences and AI experts from the Faculty of Science and ngineering together to develop an explainable high-confidence early categorisation system for the cause of child deaths.
The Child Mortality Analysis Unit (CMAU) at the University of Bristol is the national hub for analysing statutory child death data in England. CMAU links over 25,000 notifications and 19,500 completed reviews with national datasets across health, education, social care, policing, and child safety. This enables the identification of patterns, causes, and risk factors in child mortality, informing preventative action and national policy.
Now, in partnership with analysts from the National Child Mortality Database and leading AI experts, CMAU is developing a pioneering early categorisation system to identify suspected causes of child deaths in real time. This innovation will enhance national surveillance and accelerate public health responses.
Future work will expand to unstructured data from documents such as clinical notes, unlocking insights and spotting patterns that are currently difficult to detect manually. This will mark a major leap forward in understanding and preventing child deaths.
An explainable, high-confidence early categorisation system could be a game-changer, revolutionising how services across sectors monitor, respond to, and ultimately prevent child mortality.
The project team will be sharing the findings via this website, LinkedIn, academic publications and industry events. Are you a national health provider who would like to do something similar? Please reach out and contact us to learn more.