UK Biobank is a prospective epidemiological study of over 500,000 individuals (40-69y when recruited) in the UK. Blood, urine and saliva samples were collected, samples for genetic analysis and physical measurements taken, and each volunteer answered an extensive questionnaire on aspects of health and lifestyle. Additionally, brain, heart and body imaging from 100,000 of the participants is now being acquired (17,000 to date). The brain imaging comprises 6 modalities (T1w, T2 FLAIR, resting fMRI, task fMRI, diffusion MRI and susceptibility weighted imaging) and has been highly optimised to run in 35 minutes.
Brain imaging data is not immediately usable for most research purposes in its raw form. It needs to be processed and analysed in a specific replicable manner for the outcome of an analysis to be valid and meaningful; the way in which such tools are applied is referred to as a “processing pipeline”. UK Biobank needs a pipeline that can process and integrate many modalities; it must be robustly automatable. Our pipeline also generates 4000 “imaging-derived phenotypes” (IDPs) aiming to identify biomarkers for early diagnosis. These include metrics such as subcortical structure volumes, white matter hyperintensities, microstructural measures in major tracts, and structural/functional connectivity metrics.
Finally, due to the huge number of subjects, manual analysis would be too costly, so we have developed an automated Quality Control tool using machine learning methods to identify images with problems either in their acquisition or in later processing steps.