Approved research

Assessing the link between physiological measurements and structural brain images using machine learning

Lay summary

The overarching goal of the present proposal is to characterize structural brain changes in healthy individuals in a wide age range. Recent studies attempted to characterize the relationship between brain structure and age by predicting an individual's age using a structural MRI image. Advancement in the field of machine learning (ML) has enabled remarkable improvement in this domain, with several studies reporting a mean error of ~5 years in predicting age using neuroimaging as input. Linking brain anatomy with age enables to estimate the difference between subjects' brain-age as predicted by the ML model and the subject's real biological age or the delta brain age (DBA). We intend to further improve those existing tools, specifically create a more interpretable result, reduce the need for image preprocessing and examine how DBA is influenced by health-related risk or resilience factors. We plan to examine whether genetic markers of rapid or slow aging can be identified. We expect that the project would last 2-3 years, after which the results, the trained model and code would be openly available.