Population Graph GNNs for Brain Age Prediction

摘要: Many common neurological and neurodegenerative disorders, such as Alzheimer9s disease, dementia multiple sclerosis, have been associated with abnormal patterns of apparent ageing the brain. Discrepancies between estimated brain age actual chronological (brain gaps) can be used to understand biological pathways behind process, assess an individual9s risk for various disorders identify new personalised treatment strategies. By flexibly integrating minimally preprocessed neuroimaging non-imaging modalities into a population graph data structure, we train two types neural network (GNN) architectures predict in clinically relevant fashion well investigate their robustness noisy inputs sparsity. The multimodal approach has potential learn from entire cohort healthy affected subjects both sexes at once, capturing wide range confounding effects detecting variations trends different sub-populations subjects.