Second-order instantaneous causal analysis of spontaneous MEG

Area of Science:

• Neuroscience
• Computational Neuroscience
• Brain Imaging Analysis

Background:

• Identifying instantaneous causal relationships in observational brain imaging data (e.g., MEG, fMRI) is a persistent challenge.
• Existing methods like Granger Causality and Non-Gaussian Structural Equation Models (SEM) have limitations in handling instantaneous effects or data non-Gaussianity.

Purpose of the Study:

• To propose a novel model capable of discovering instantaneous causal relationships in temporally dependent variables, common in neuroimaging.
• To develop an efficient and simple method for estimating causal directions from observational brain data.

Main Methods:

• Developed a model incorporating instantaneous causality for time-dependent variables.
• Proposed a likelihood ratio-based method, related to mutual information, to estimate causal directions.
• Constructed a simple decision criterion for instantaneous causal discovery in time-series data.

Main Results:

• The proposed method demonstrates strong performance on simulated data, even with limited sample sizes.
• Applied to MEG data, the method yields consistent intra- and inter-subject causal directionalities.
• Outperforms Granger Causality and non-Gaussian SEM in a brain age prediction task.

Conclusions:

• The novel method provides a computationally simple and effective approach for instantaneous causal discovery in neuroimaging.
• The findings suggest potential utility in analyzing causal brain connectomes from functional brain imaging data.
• The method offers an improvement over existing techniques for understanding brain connectivity.