Predicting imaging phenotypes from multi-dimensional spectral measurements of MRI-relevant tissue properties

The past decade has seen the emergence of population-level magnetic resonance imaging (MRI) studies like the UK Biobank, which is scanning an unprecedented 100,000 individuals. This imaging has enormous potential to inform about early pathology or susceptibility to disease. However, to translate insights from population-level health data resources into the clinic, we require approaches to translating, or ‘harmonising’, between datasets acquired under very different conditions.

A newly funded collaboration between Oxford and the NIH aims to deliver a novel harmonisation approach by linking relevant tissue biology to the physics of the imaging measurement. Core to this ‘biophysical’ approach is a framework for predicting imaging phenotypes from multi-dimensional spectral measurements of MRI-relevant tissue properties.

This DPhil project will deliver the multi-spectral measurements at the heart of this prediction framework. The student will work within our collaborative team to:

• Year 1: implement multi-spectral acquisition protocols and associated analysis pipelines for use in a large cohort;
• Years 2-3: develop novel biophysical modelling that enables us to characterise and restrict the number of modelled tissue compartments, enabling fewer measurements for clinical scanners;
• Year 4: demonstrate the ability to predict imaging phenotypes based on these measurements in order to harmonise measurements from multiple protocols.

This project would be jointly supervised by the neuroimaging experts in Oxford who are leading brain MRI in UK Biobank (Miller) and physics experts at NIH who have pioneered these multi-spectral measurements (Benjamini).