Austin Dibble.

Research

PhD in neuroscience, University of Glasgow, 2023 to 2027. Supervised by Dr Michele Svanera.

Thesis

Toward Precision Neuroimaging with Disease-Naïve Models for Brain Health Estimation

I built NeuroFM, a deep learning model for structural brain MRI. It learns a compressed representation of a T1-weighted scan that captures the overall health of the brain, rather than looking for any one disease. From that representation it estimates brain age, total brain volume, ventricular volume and sex.

The unusual part is the training data. NeuroFM was pretrained on 100,000 synthetic brain volumes and no real ones, so it isn't tied to any cohort's privacy or data-governance rules. We then tested it on more than 136,000 real MRI volumes from several cohorts, where its estimates pick up signs of future dementia years before diagnosis.

Read the preprint, get the weights on Hugging Face (small, medium and large), or see the code and project page.

Submission January 2027, viva around April 2027.

Talks and posters

Papers and preprints

Also on Google Scholar and ORCID.

First author

Co-author