Abstract
AbstractUnderstanding large-scale brain dynamics is a grand challenge in neuroscience. We propose functional connectome-based Hopfield Neural Networks (fcHNNs) as a model of macro-scale brain dynamics, arising from recurrent activity flow among brain regions. AnfcHNNis neither optimized to mimic certain brain characteristics, nor trained to solve specific tasks; its weights are simply initialized with empirical functional connectivity values. In thefcHNNframework, brain dynamics are understood in relation to so-called attractor states, i.e. neurobiologically meaningful low-energy activity configurations. Analyses of 7 distinct datasets demonstrate thatfcHNNs can accurately reconstruct and predict brain dynamics under a wide range of conditions, including resting and task states and brain disorders. By establishing a mechanistic link between connectivity and activity,fcHNNs offers a simple and interpretable computational alternative to conventional descriptive analyses of brain function. Being a generative framework,fcHNNs can yield mechanistic insights and hold potential to uncover novel treatment targets.Key PointsWe present a simple yet powerful computational model for large-scale brain dynamicsThe model uses a functional connectome-based Hopfield artificial neural network (fcHNN) architecture to compute recurrent “activity flow” through the functional brain connectomeFcHNNattractor dynamics provide a mechanistic account for gradient- and state-dynamics in the brain in resting conditionsFcHNNs conceptualize both task-induced and pathological changes in brain activity as a non-linear shift in these dynamicsOur approach is validated using large-scale neuroimaging data from seven studiesFcHNNs offers a simple and interpretable computational alternative to conventional descriptive analyses of brain functionProject website (with interactive manuscript):https://pni-lab.github.io/connattractor
Publisher
Cold Spring Harbor Laboratory