Kernel Granger causality mapping effective connectivity on FMRI data

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Machine Learning

Background:

• Linear Granger causality is established for functional magnetic resonance imaging (fMRI) effective connectivity.
• Detecting nonlinear connectivity in fMRI remains a challenge.

Purpose of the Study:

• To introduce and evaluate kernel Granger causality (KGC) for uncovering nonlinear effective connectivity in fMRI.
• To demonstrate KGC's ability to identify couplings missed by linear methods.

Main Methods:

• Kernel Granger causality (KGC) is based on reproducing kernel Hilbert spaces.
• KGC extends linear Granger causality into a feature space to handle arbitrary nonlinearity.
• The method was tested using simulation studies and real fMRI data from a motor imagery task.

Main Results:

• KGC successfully identified effective couplings not detectable by linear Granger causality.
• The study generated effective connectivity networks using KGC, with the supplementary motor area (SMA) as a seed.
• KGC demonstrated superior performance in capturing complex brain interactions.

Conclusions:

• Kernel Granger causality (KGC) offers a powerful approach for detecting nonlinear effective connectivity in fMRI.
• KGC enhances the analysis of brain networks by revealing previously undetectable couplings.
• This method has significant implications for understanding complex brain dynamics.