Biophysical Signal Prediction for Translation of Population Neuroimaging

The past decade has seen the emergence of population-level brain imaging.

For the first time, imaging can provide a rich, multi-faceted description of how an individual’s brain deviates from population norms, potentially informing about early pathology or susceptibility to disease. Studies with access to health records, like UK Biobank, enable that link to be explicitly identified.

However, the imaging in these new population-health resources, which use cutting-edge imaging technology, does not match hospital-based patient imaging in the real world, which incurs compromises in technologies and protocols.

We aim to address the deep challenges of information translation that currently prevent clinical imaging from taking advantage of population-level health data resources. This fellowship proposes to develop a biophysical approach to such translation. That is, linking the relevant biology to the physics of the imaging measurement.

Core to this approach is a biophysical framework for phenotype prediction (WP1), which will enable us to: (i) identify and validate the biological source of variations in UKB-derived imaging markers (WP2) (ii) design new imaging acquisitions that are maximally sensitive to the same biological information as UKB imaging (WP3) (iii) establish biophysically-valid harmonisation to enable direct comparison of newly-acquired data with UKB imaging phenotypes (WP4)