Related Experiment Video
Updated: Feb 17, 2026
11:28
Concurrent EEG and Functional MRI Recording and Integration Analysis for Dynamic Cortical Activity Imaging
Published on: June 30, 2018
12.3K
Spatio Temporal EEG Source Imaging with the Hierarchical Bayesian Elastic Net and Elitist Lasso Models
Deirel Paz-Linares1,2, Mayrim Vega-Hernández2, Pedro A Rojas-López1,2
1The Clinical Hospital of Chengdu Brain Science Institute, MOE Key Lab for Neuroinformation, University of Electronic Science and Technology of China, Chengdu, China.
Frontiers in Neuroscience
|December 5, 2017
Summary
This study introduces a new Bayesian framework for Electrophysiology Source Imaging, improving the accuracy of estimating EEG sources using Structured Sparse Bayesian Learning for Elastic Net and Elitist Lasso models.
Area of Science:
- Neuroscience
- Computational Neuroscience
- Biomedical Engineering
Background:
- Estimating electroencephalography (EEG) sources is a challenging Inverse Problem (IP) in neuroscience due to solution non-uniqueness.
- Structured sparsity priors, like Elastic Net (ENET) and Elitist Lasso (ELASSO), offer regularization but traditional penalized regression methods yield suboptimal, computationally expensive solutions.
- Existing Bayesian formulations for ENET are computationally intensive, and ELASSO has not been explored in a Bayesian context.
Purpose of the Study:
- To develop and validate a novel Bayesian framework for solving the EEG IP using ENET and ELASSO models.
- To propose a Structured Sparse Bayesian Learning algorithm for accurate parameter and hyperparameter estimation in EEG source imaging.
- To enhance the interpretability and accuracy of neurophysiological patterns derived from EEG data.
Main Methods:
- A Structured Sparse Bayesian Learning algorithm combining Empirical Bayes and iterative coordinate descent was developed.
- The algorithm estimates parameters and hyperparameters for ENET and ELASSO models within a Bayesian framework.
- Realistic simulations and real EEG data from a visual attention experiment were used for validation.
Main Results:
- The proposed Bayesian methods accurately recover complex EEG source configurations, outperforming classical LORETA, ENET, and LASSO Fusion.
- Hyperparameter estimation demonstrated robustness across various sparsity scenarios.
- Analysis of visual attention experiment data revealed more interpretable neurophysiological patterns compared to existing methods.
- Freely available Matlab code facilitates reproducibility and application.
Conclusions:
- The developed Bayesian framework for ENET and ELASSO models provides a more accurate and robust solution for the EEG Inverse Problem.
- This approach enhances the estimation of neural sources and the interpretation of neurophysiological data.
- The availability of the code promotes further research and application in the field of Electrophysiology Source Imaging.
