Postdoctoral Researcher in advanced image analysis and machine learning for brain imaging Full-time

Postdoctoral Researcher in Advanced Image Analysis and Machine Learning for Brain Imaging

Nuffield Department of Clinical Neurosciences (NDCN), John Radcliffe Hospital, Headington, Oxford, OX3 9DU

This role will be to develop and apply image processing algorithms and software, to process two-timepoint (longitudinal) data from UK Biobank brain imaging.

This will involve development of image analysis and machine learning (AI) methods and software. All of this will be specific to longitudinal imaging, i.e., where we have more than one timepoint imaging from each participant.

Background:

The OxCIN (FMRIB) Analysis Group in Oxford is a leading research group in brain image analysis, and produced FSL (FMRIB Software Library), the world's most widely used research software that spans all the major structural and functional brain imaging modalities. FMRIB members Smith and Miller are the scientific leads of the brain imaging component of UK Biobank.

FMRIB Oxford developed FBP (FMRIB Biobank Pipeline,) v1 for processing brain imaging data on behalf of UKB. To date, we have processed ~95,000 participants' timepoint1 datasets for release by UKB. This processing includes: semi-automated QC, image processing and artefact removal, and estimation of over 4,000 IDPs – image-derived phenotypes, distinct measures of brain structure and function, ranging from simple estimates of tissue volume to measures of functional connectivity between different brain regions. All seven brain imaging modalities are processed: T1, T2FLAIR and swMRI structural data, diffusion structural connectivity and tissue microstructure, resting and task functional data, and ASL blood flow. All enhancements to FBP v1 have been “backwards compatible” expansions, in order to maintain consistency with already-released subjects' IDPs, as new subjects are added. There are many aspects of the pipeline that can be improved on, and we are now starting to design and create FBP v2, revisiting all processing and algorithmic decisions to take advantage of advances in the field. This new version will not be backwards compatible with v1, to allow for unconstrained optimal design.

None of the above includes modelling of longitudinal data. The timepoint2 datasets (around 20,000 to date and rising to 60,000 over the next few years) have been processed as if they were from new participants.

Our proposal here is to develop longitudinal modelling as part of pipeline v2, which fits well with the timing of the timepoint2 scanning, and pipeline v2 development.

Using both the existing and new pipeline components, we will process all timepoint2 data. We will then utilize these developments to study aging and later-life disease.

A major goal of our group is to extend the reach of imaging neuroscience to take full advantage of the richness of the new big data imaging resources, such as: major advances in the quality of data; complementarity of multi-modal imaging; large subject numbers; and linked information about health outcomes, genetics, and risk factors. This requires new research across multiple domains, including: integrated cross-modal analysis, particularized biologically-interpretable markers, machine learning (advanced AI methods), and modelling of variation in populations to the characterisation of individual subjects.

Successful applicants will also have the opportunity to contribute to the next generation of FSL tools, which is the most widely used software toolbox that combines structural, functional, and diffusion MRI analysis. FSL is used in more than 1000 universities and hospitals worldwide (160,000 downloads in the past 5 years).

We are looking for two postdocs with advanced experience in both AI (machine learning) and image analysis, including both new methods development and software. Applicants will be expected to hold a BSc (or equivalent) degree in a relevant area (i.e. Physics, Engineering, Computer Sciences, Neurosciences), and a PhD in a relevant computational domain. Also deep knowledge of statistics and Bayesian learning, and necessary mathematical and computational skills to put these into practice. Skills in computer programming (Python or C++) are essential.

This post is full time and Fixed-term until 30 November 2028. Only applications received before midday 12:00 on Friday 16th October 2026 will be considered. Interviews will be held as soon as possible thereafter.