It is thought that brain structure is the primary determinant of functions of brain regions. For example, cortical areas with functional differences also have different structural connectivity (SC) patterns. We used SCs derived from diffusion tensor imaging (DTI) data in 100 healthy adults included in the Human Connectome Project (HCP) to successfully predict cortical activation responses across distinct cognitive tasks and found that predictive performance varied among tasks. We also observed that predictive performance could be used to characterize task load in both relational reasoning and N-back working memory tasks and was significantly positively associated with behavioral performance. Moreover, we found that the default mode network (DMN) played a more dominant role in both activation prediction and behavioral performance than was found for other functional networks. These results support our hypothesis that individuals who performed tasks better might exhibit a more accurate predicted activation pattern as task-evoked activities are more inclined to flow over inherent structural networks than over more flexible paths. In the high difficulty condition, the decreased correlation between predicted and empirical activation may be associated with the more random brain activity in these conditions/participants due to the lack of engagement. Together, our findings highlight the feasibility of using SCs to estimate various cognitive task activations and thus further facilitate the exploration of the relationship between the brain and behavior by providing strong evidence for the relevance of structure to function in the human brain.
Keywords: Default mode network; Predictive performance; Structural connectivity; Task load; Task-functional magnetic resonance imaging.