A New Approach for Predicting Brain Functional Connectivity: A Cross-Modality DTI-fMRI Study

Diffusion tensor imaging (DTI) and functional magnetic resonance imaging (fMRI) provide complementary insights into brain connectivity, with DTI capturing structural connectivity (SC) and fMRI measuring functional connectivity (FC). Integrating these modalities offers the potential to deepen our understanding on the brain structural-functional coupling. In this study, a novel integrated DTI-fMRI approach was designed to explore the possibility to predict brain functional connections based on the underlying structural pathways.An asymmetric DTI-driven fMRI model was developed to investigate the role of brain structure, via a linear combination of DTI-derived features, in shaping FC within a normative framework. The model was trained on healthy young adults (n=12, 27.2 ± 0.8 years, 6M/6F) using a 4-fold cross-validation framework. The final model fitted on the training set was preliminarily tested on an independent set of healthy volunteers across a broader age range (n=14, 55.4 ± 18.2 years, 8M/6F). In the test set, any age-related variations in the model predictive performances were assessed.Cross-validation results showed that SC strength, path length, and physical distance between network nodes exhibited significant effects on FC prediction. When applied to the independent test set, the model's whole-brain similarity between actual and estimated FC was not associated with age or sex, suggesting stable structure-function coupling during adulthood and between males and females. In contrast, age-related differences were found at the link-level reconstruction error between actual and estimated FC, mainly in the connections of frontal areas.These findings underscore the potential for DTI-fMRI integration for investigating the complex relationships between brain structural architecture and dynamic functions both in physiology and in disease.