Predictive Brain Development Modeling using Temporal GNNs

Gaining an understanding of the changes in the brain between early childhood and late adulthood is a challenging task, especially when most studies are limited to observing the brains of individuals in one age group or making predictions about the brain based on limited observations of brain scans from various age groups. However, by using this model and process, it is possible to gain a better understanding of the brain's changes by learning a timeline that represents how the brain’s organizational structure changes over time. Furthermore, by embedding the brain scans of individual subjects into this timeline, it is possible to understand brain changes over time even for life stages that are underrepresented in the HCP dataset (such as those between 30 and 50 years of age). Thus, each of these brain scans can be matched with a point along the timeline, allowing each individual to have their own profile representing the development of their brain over time. The model will also estimate structural changes and shifts. Initial study suggests that this approach is more sensitive to subtle differences in brain aging and can offer a stronger foundation for early detection of both developmental delays and age-related decline.