The UK Biobank Imaging Data Symposium (20th November 2025) provided a unique opportunity to explore one of the most ambitious imaging projects ever undertaken.
With the landmark v20 data release, researchers now have access to imaging data from 100,000 participants, alongside cutting-edge tools to support their next discoveries.
This free half-day virtual event examined the new data available, showcased new research from the scientific community, and highlighted tools and support available through UK Biobank's Research Analysis Platform (UKB-RAP).
Watch now
Watch the full event here:
Opening Remarks & Data Overview
Adam Lewandowski (Deputy Chief Scientist, UK Biobank) welcomed attendees and celebrated the milestone of reaching 100,000 imaging participants, a feat that required four dedicated imaging centres and years of effort. He outlined the scope of the imaging study, which included brain, cardiac, body MRI, DXA, ultrasound, ECG, and OCT.
He also shared details of the repeat imaging project, a new longitudinal project targeting 60,000 participants, already at 20,000 scans.
Oliver Gray (Product Manager, UK Biobank) then gave an overview of the v20 release, shared below with corresponding Showcase links.
- Cardiac imaging IDPs for an additional 40,000 participants (total ~81,000) available in category 157.
- NEW Biventricular cardiac mesh models for 55,000 participants, category 538.
- NEW FreeSurfer longitudinal metrics, available in categories 530 to 537.
- Brain imaging pipeline update, including updated brain imaging outputs for 20,000+ more participants (total ~90,000), plus NEW pre-processed brain imaging data for 82,000. Category 508.
- NEW Skeletal measures derived from DXA images for 38,000 in category 522.
- DXA-derived IDPs for an additional 30,000 (total ~70,000) in categories 124 and 125.
- NEW Retinal features derived from colour fundus eye images for ~35,000, available in category 521.
- Clinical grading of retinal images in category 1080.
Case Studies
Three leading researchers showcased how imaging data made available previously is already driving discovery:
Louise Thomas: Abdominal MRI
Louise Thomas (Professor of Metabolic Imaging, University of Westminster) showcased her work on metabolic imaging at scale. Using automated segmentation of abdominal MRI scans, her team mapped visceral fat, muscle quality, and organ composition across thousands of participants.
This approach revealed striking insights, such as the high prevalence of fatty liver disease, affecting nearly 30% of the cohort, and its genetic associations.
Beyond fat and muscle, Louise highlighted advanced phenotyping techniques that capture organ shape and fat distribution, as well as opportunistic findings like spleen iron levels, opening new avenues for metabolic health research.
Steve Smith: PANDORA and the Future of Brain Imaging
Steve Smith (Professor of Biomedical Engineering, University of Oxford) introduced PANDORA (Population Archive of Neuroimaging Data Organised for Rapid Analysis), a transformative resource for brain imaging analysis.
PANDORA reorganises brain imaging data into massive matrices, enabling voxel-level and supervoxel-level analysis for over 82,000 participants. Steve explained how this structure allows researchers to run efficient regressions against phenotypes and genetic data, uncovering correlations with traits such as smoking, neuroticism, and PTSD.
By leveraging ICA-based supervoxels, PANDORA dramatically reduces computational complexity while preserving rich spatial detail, making large-scale neuroimaging studies more accessible than ever.
Read more:
- PANDORA FSL-GLM on UKB-RAP (GitHub)
- Imaging and Machine Learning Analyses on UKB-RAP (GitHub)
Alistair Young: Cardiac Mesh Models
Alistair Young (Professor of Cardiovascular Data Analytics and Artificial Intelligence, King’s College London) presented the latest advances in cardiac imaging through mesh models now available in UK Biobank. These models capture the three-dimensional shape and motion of the heart, providing a more nuanced view than traditional metrics like ejection fraction.
Alistair demonstrated how mesh-based analysis improves disease prediction and supports genetic discovery related to cardiac remodelling. He also introduced digital twin simulations, which use these models to estimate hidden physiological parameters such as conduction velocity, offering a glimpse into the future of personalised cardiovascular medicine.
Tool Spotlight: MONAI on UKB-RAP
Ben Busby (Global Alliances Manager, NVIDIA) introduced MONAI, an AI framework for medical imaging now integrated into UK Biobank's Research Analysis Platform (UKB-RAP). MONAI is able to support researchers with:
- AI-Assisted Annotation: Accelerates segmentation for large-scale datasets.
- Model Zoo & Pretrained Bundles: Ready-to-use models for brain and cardiac imaging.
- Interactive Workflows: 3D Slicer integration for visualisation and labelling.
Read more:
- Using MONAI on UKB-RAP
- Imaging tools on UKB-RAP
- Imaging and Machine Learning Analyses on UKB-RAP (GitHub)
Inside UKB-RAP
Lea-Maria Khoueiry (Senior Data Analyst, UK Biobank) walked through practical steps for researchers to get the most out of imaging data in UKB-RAP.
The session explored practical aspects for a selection of tools available for visualisation and analysis, including 3D Slicer for exploring scans and MONAI Label for AI-assisted segmentation and annotation.
Lea-Maria also showed how researchers can build their own workflows using Docker images and applets, with GitHub resources available to support reproducibility and collaboration.
Read more:
- Viewing images with 3D Slicer
- Imaging tools on UKB-RAP
- JupyterLab with FSL and FreeSurfer
- GitHub notebooks
Guidance for Early Career Researchers
Oliver Gray returned to address some common challenges for ECRs.
- Access & compliance: Training modules and AMS project setup.
- Funding support: AWS credits, transition grants, and global access funds.
- Learning resources: UKB Community Forum, GitHub tutorials, and video guides.
- Publishing best practices: Acknowledgements, anonymisation, and image use guidelines.
Read more:
- Financial support and credit programmes
- UKB-RAP getting started guides
- UK Biobank GitHub
- DNAnexus GitHub
- UKB-RAP documentation
- UK Biobank YouTube
- DNAnexus YouTube
Q&A
A selection of questions asked in the text section during the session:
What about Diffusion MRI and fMRI?
These are part of the brain protocol. Please see the following links on Showcase for information on our resting fMRI, task fMRI and diffusion data.
Do we have access to brain MRI raw files (voxelwise), or do we have to use pre-processed IDPs?
You can access the raw DICOM data for all scans, aside from the T1 and T2 structural scans - these images are provided in a defaced, NIFTI format to protect participant anonymity
Are the raw DICOM images available for the DXA data, or only the derived measures?
You can access the raw DXA data from field 20158.
Are there images from colon MRI?
We provide abdominal MRI scans to researchers. More information available in category 105 on Showcase.
Do you have lung MRI / lung CT imaging datasets?
UK Biobank doesn't provide CT scans. You can find more information on the imaging measurements in Category 100003 on Showcase.
Is it possible to share any info about electrophysiological recordings? I can see EEG datasets mentioned in a few projects, but EEG as a modality is not listed under "types of data".
We don't provide EEG data in terms of electrophysiological measures in general. We do offer ECG data, which you can read more about here: Electrocardiogram Data.
How complex/simple is filtering out sub-samples with the single PANDORA file?
Very easy: you just create one file with your regression design matrix (for any subset of subjects) and a second file with the corresponding list of subjects, and fsl_glm will do the rest for you.
Is there a way to download and reuse your derived results that you have generated within UKB-RAP outside the UK Biobank or will it stay exclusively in UK Biobank?
Researchers aren't permitted to download any individual data from UKB-RAP. However, you are able to save your derived results within your project, and reuse any of your derived results within this environment. If you wish to do so, you can download any summary data that you have derived (e.g. figures).
Comments
0 comments
Please sign in to leave a comment.