Predicting response speed and age from task-evoked effective connectivity.

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Computational Neuroscience

Background:

• Task-evoked functional connectivity (FC) shows promise in predicting individual traits.
• The predictive power of task-evoked effective connectivity (EC) for individual differences remains largely unexplored.

Purpose of the Study:

• To investigate the predictive capacity of intrinsic EC (I-EC) and task-modulated EC (M-EC) for individual reaction time (RT) and age.
• To compare the performance of EC against task-evoked FC in predicting these traits.
• To evaluate the impact of different data processing and modeling choices on prediction accuracy.

Main Methods:

• Dynamic causal modeling (DCM) was used to calculate I-EC and M-EC from fMRI data during a stimulus-response compatibility task.
• General linear model (GLM) designs (event-related vs. block-based), Bayesian model reduction, and cross-validation schemes were varied.
• Machine learning models were employed for prediction of RT and age.

Main Results:

• M-EC demonstrated superior prediction of RT compared to I-EC and task-evoked FC.
• All connectivity types performed similarly in predicting age.
• Event-related GLM and DCM designs yielded better predictions than block-based designs.
• Significant differences were observed in predicting RT and age between I-EC and M-EC.

Conclusions:

• Task-evoked EC, particularly M-EC, holds significant potential for predicting behavioral traits like RT.
• The choice of GLM and DCM design critically influences the accuracy of EC-based predictions.
• Findings contribute to understanding how neuroimaging analysis choices impact predictive modeling of individual differences.