Double Sparse Dictionary-Based Electroencephalography Channel Selection for Depression Analysis.

Area of Science:

• Neuroscience
• Signal Processing
• Computational Biology

Background:

• High-density electroencephalography (EEG) analysis for depression (DP) faces challenges with channel redundancy and computational complexity.
• Identifying core brain functional networks (BFNs) in DP patients is crucial for understanding disease mechanisms.

Purpose of the Study:

• To propose an elastic net-based double sparse dictionary channel selection (EN-DSDCS) method for efficient EEG analysis in depression.
• To identify key EEG channels and analyze topological changes in the BFN of DP patients.
• To reduce signal reconstruction error and channel sparsity compared to existing methods.

Main Methods:

• An improved coarse-graining method reconstructs EEG signals and calculates multi-scale permutation entropy (MSPE).
• A double sparse dictionary structure is created using DCT matrices and optimized via sparse K-SVD.
• Elastic net regularization is applied for joint optimization of the dictionary and sparse coefficients, enabling channel selection.

Main Results:

• The EN-DSDCS method reduced signal reconstruction error by 3x10^-4.
• Channel sparsity was decreased by 3.93% compared to Lasso-based methods (L-DSDCS).
• Selected channels were predominantly in the frontal and temporal lobes, revealing differential connectivity and a left-hemispheric bias in Hub node distribution in DP patients.

Conclusions:

• The EN-DSDCS method effectively identifies core EEG channels and analyzes BFN alterations in depression.
• The findings highlight specific topological changes and connectivity patterns in the brains of DP patients.
• This approach offers a more computationally efficient and accurate method for EEG-based depression research.