Task-Aware Effective Connectivity Modeling for Cognitive Function Prediction.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Brain Imaging

Background:

• Resting-state functional Magnetic Resonance Imaging (rs fMRI) is crucial for understanding brain function.
• Current effective connectivity (EC) methods often require individual retraining and overlook population-level data.
• Existing EC approaches are typically independent of specific cognitive tasks, limiting their ability to capture task-dependent variations.

Purpose of the Study:

• To develop a flexible Task-Aware Effective Connectivity (TAEC) model for constructing individualized, task-aware, and nonlinear causal brain networks.
• To enable end-to-end prediction and identify task-dependent EC patterns without subject-specific retraining.
• To improve the generalizability and task-specific sensitivity of EC analysis.

Main Methods:

• A Causal Discovery Module (CDM) employing a spatial-temporal attention mechanism estimates individual EC.
• A Task-Aware Graph Neural Network (GNN) Predictor is utilized for end-to-end prediction with a task-aware penalty.
• The model is validated on twelve cognitive tasks from the Human Connectome Project (HCP) dataset.

Main Results:

• The TAEC model achieves state-of-the-art performance in task-aware effective connectivity modeling.
• The framework successfully identifies discriminative and task-specific EC patterns.
• The proposed method demonstrates effectiveness in capturing individualized and nonlinear causal brain networks.

Conclusions:

• The TAEC model offers a novel and efficient approach to analyzing brain connectivity.
• This framework enhances our understanding of cognitive functions by revealing task-specific neural network dynamics.
• The TAEC model provides valuable insights into brain function across various cognitive tasks.