High-order interactions to track brain aging: A deep-learning approach

Brisbane Convention & Exhibition Centre 

Brain function changes during our lifespan. So far, only pairwise functional interactions between brain regions have been employed to predict brain age. However, high-order interactions (HOI) can capture complex non-linear associations between pairwise regions and the rest of the brain, detecting key synergistic and redundant couplings. To bridge this gap, we employed cutting-edge deep learning techniques to analyze HOI and predict brain age from functional MRI (fMRI) and electroencephalography (EEG) data, encompassing over 5,000 participants from 15 countries. Our brain-age gap models revealed a progressive acceleration of brain aging from healthy individuals to those with neurocognitive disorders. We observed increasing brain-age gaps from healthy controls to individuals with mild cognitive impairment (MCI), and further increases in those with Alzheimer’s disease (AD) and frontotemporal dementia (FTD). We found that individual clinical conditions, country-level socioeconomic factors, gender disparities, and environmental exposomes influence brain-age gaps. Individuals in countries with more significant socioeconomic inequalities, higher pollution levels, and limited healthcare access exhibit older brain ages. Our research advances our understanding of brain aging. It highlights the urgent need for global health policies that address the social and environmental factors contributing to accelerated brain aging, particularly in regions facing greater socioeconomic challenges.