A Causal Validation augmented Temporal Convolutional Framework for Brain Effective Connectivity Networks Estimation.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Network Science

Background:

• Effective connectivity networks (ECNs) are crucial for understanding brain function and disease.
• Neural networks (NNs) offer powerful ECN estimation but often retain redundant temporal data and lack causal validation.
• Existing methods struggle with temporal redundancy and rigorous causal inference in ECN analysis.

Purpose of the Study:

• To introduce a novel end-to-end framework, the Causal Validation augmented Temporal Convolutional Framework (CVTCF), for robust ECN estimation.
• To address limitations of existing NN-based ECN methods, specifically redundant temporal information and insufficient causal validation.
• To enhance the reliability and accuracy of ECN estimation for neurological applications.

Main Methods:

• Utilized Temporal Convolutional Networks (TCNs) with Least Absolute Shrinkage and Selection Operator (Lasso) regression for ECN estimation.
• Proposed a convolutional Hierarchical Group Lasso (cHGL) to detect Granger Causality (GC) inputs and eliminate temporal redundancy.
• Incorporated permutation importance validation with the Wilcoxon signed-rank test for enhanced GC detection reliability.

Main Results:

• The CVTCF framework demonstrated superior performance compared to state-of-the-art methods in simulations using the Lorenz-96 model and a BOLD dataset.
• Successfully identified causal interactions within the cerebral cortex, revealing complex relationships in neurological functioning.
• Provided detailed analysis of causal interactions relevant to neurodegenerative conditions like Alzheimer's Disease and Parkinson's Disease.

Conclusions:

• The CVTCF framework offers a significant advancement in ECN estimation, effectively handling temporal redundancy and providing causal validation.
• ECN estimation using CVTCF shows potential as reliable biomarkers for neurodegenerative diseases.
• This study lays the groundwork for future diagnostic and therapeutic strategies targeting neurological disorders.