Leveraging population-level data for precision image-derived phenotyping in mental health

Project Description

Despite the emergence of sophisticated tools over the last decade, from genomic profiling and high-resolution neuroimaging to AI models, the search for objective diagnostic tests and personalised treatments has so far come short for nearly every psychiatric disorder. These efforts have been fundamentally limited by a lack of understanding of how symptoms in mental illnesses map onto disrupted brain circuits. Magnetic Resonance Imaging (MRI) allows unique opportunities to provide insight into this question.

However, a key limitation in these explorations is routed in the indirect nature of MRI measurements. In particular, MRI-derived features reflect a range of nuisance factors that are not linked to biological variability. Hardware/software differences between MRI scanners can yield between-scanner variability for the same individual, which can be as large as, or in some cases exceed, between-subject variability and biological effects of interest. This lack of harmonisation between MRI scanners limits robustness and reproducibility of imaging-derived findings.

In this project we will develop a novel solution to this challenge, using unique pre-existing data that our team has built from a leading travelling-heads study. This will allow us to link population-level epidemiological data from the UK Biobank (comprising of 100,000 participants, recruited over ten years and each scanned with 5 imaging modalities), with targeted studies (that are smaller in cohort size, e.g. less than 100 participants, but likely considerably richer with respect to clinical phenotyping). We will subsequently leverage these capabilities to explore brain-symptom associations at scale in mental health cohorts by linking multiple imaging studies.

The project will provide hands-on experience and interdisciplinary, transferable skills relevant to a wide range of research topics, including: acquisition/processing of multi-modal neuroimaging data, handling large datasets, harmonising MRI data obtained from different clinical/research scanners, scripting/programming for scientific computing, data science and analytics for biomarker discovery, neuroimaging in mental illness. Healthcare professionals will be ultimately empowered to apply such research and analysis skills in practice and translate data into patient benefit.