Dynamic brain effective connectivity analysis based on low-rank canonical polyadic decomposition: application to...

Area of Science:

• Neuroscience
• Signal Processing
• Computational Biology

Background:

• Epilepsy is characterized by abnormal brain activity and altered functional connectivity.
• Tracking dynamic changes in brain effective connectivity networks is crucial for understanding epilepsy.
• Current methods may not fully capture the complex, time-varying nature of these networks.

Purpose of the Study:

• To propose and validate a new method for tracking brain effective connectivity networks in epilepsy using electroencephalographic (iEEG) data.
• To identify dominant directed graph structures underlying iEEG signals within specific time windows.
• To analyze the temporal and spectral evolution of these networks across multiple seizures.

Main Methods:

• Combines partial directed coherence (PDC) with constrained low-rank canonical polyadic tensor decomposition.
• Infers dominant directed graph structures from iEEG signals within time windows.
• Decomposes the PDC-based tensor into space, time, and frequency signatures.

Main Results:

• The method successfully infers dominant brain network structures from simulated and real iEEG data.
• Time and frequency signatures allow tracking network evolution and identifying operating frequency bands.
• Analysis across multiple seizures revealed insights consistent with clinical expertise.

Conclusions:

• The proposed method offers a robust way to track brain effective connectivity in epilepsy.
• It enables detailed investigation of network dynamics over time, frequency, and across seizures.
• This approach enhances understanding of the neural underpinnings of epileptic activity.