A spatio-temporal fingerprints with hierarchical activation refinement for predicting task activation maps from resting-state fMRI

Predicting task-induced brain activation maps from resting-state fMRI (rs-fMRI) is an important challenge in computational neuroimaging, with potential applications for studying brain function without task-based scanning. However, most existing approaches rely primarily on functional connectivity representations, which often overlook the rich temporal dynamics embedded within rs-fMRI signals. To this end, we propose a novel hierarchical framework that uniquely combines spatio-temporal fingerprinting with residual refinement to predict individualized task-activation maps directly from rs-fMRI. The framework first captures complementary spatial and temporal characteristics of resting-state brain activity and subsequently employs hierarchical residual learning to progressively refine the predicted activation maps. Evaluation on the Human Connectome Project (HCP) dataset across seven task contrasts demonstrates strong agreement between predicted and ground-truth activation maps, with correlations reaching up to 0.884. Additional evaluation on the Chinese Human Connectome Project (CHCP) dataset further demonstrates the robustness and generalizability of the proposed framework. The results thus highlight the potential of integrating spatio-temporal representations with hierarchical residual learning for individualized prediction of task-related brain activation from resting-state fMRI.