Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Incomplete Depression Feature Selection with Missing EEG Channels

Zhijian Gong, Wenjia Dong, Xueyuan Xu, Fulin Wei, Chunyu Liu, Li Zhuo

TL;DR

This work introduces IDFS-MEC, a robust feature selection framework for EEG-based depression detection under incomplete channel data. By integrating a missing-channel indicator, adaptive channel weighting, weighted orthogonal regression, and global redundancy minimization, the method jointly optimizes feature representation and channel importance, using an alternating optimization scheme that combines Generalized Power Iteration with ALM and quadratic programming. The approach demonstrates superior performance over 10 baselines across 3-, 64-, and 128-channel EEG datasets and exhibits strong robustness to missing data, with analyses highlighting the contributions of each component and the method’s practical computational efficiency. The findings suggest IDFS-MEC’s potential for reliable depression identification in real-world EEG deployments that frequently suffer from channel loss and noise.

Abstract

As a critical mental health disorder, depression has severe effects on both human physical and mental well-being. Recent developments in EEG-based depression analysis have shown promise in improving depression detection accuracies. However, EEG features often contain redundant, irrelevant, and noisy information. Additionally, real-world EEG data acquisition frequently faces challenges, such as data loss from electrode detachment and heavy noise interference. To tackle the challenges, we propose a novel feature selection approach for robust depression analysis, called Incomplete Depression Feature Selection with Missing EEG Channels (IDFS-MEC). IDFS-MEC integrates missing-channel indicator information and adaptive channel weighting learning into orthogonal regression to lessen the effects of incomplete channels on model construction, and then utilizes global redundancy minimization learning to reduce redundant information among selected feature subsets. Extensive experiments conducted on MODMA and PRED-d003 datasets reveal that the EEG feature subsets chosen by IDFS-MEC have superior performance than 10 popular feature selection methods among 3-, 64-, and 128-channel settings.

Incomplete Depression Feature Selection with Missing EEG Channels

TL;DR

This work introduces IDFS-MEC, a robust feature selection framework for EEG-based depression detection under incomplete channel data. By integrating a missing-channel indicator, adaptive channel weighting, weighted orthogonal regression, and global redundancy minimization, the method jointly optimizes feature representation and channel importance, using an alternating optimization scheme that combines Generalized Power Iteration with ALM and quadratic programming. The approach demonstrates superior performance over 10 baselines across 3-, 64-, and 128-channel EEG datasets and exhibits strong robustness to missing data, with analyses highlighting the contributions of each component and the method’s practical computational efficiency. The findings suggest IDFS-MEC’s potential for reliable depression identification in real-world EEG deployments that frequently suffer from channel loss and noise.

Abstract

As a critical mental health disorder, depression has severe effects on both human physical and mental well-being. Recent developments in EEG-based depression analysis have shown promise in improving depression detection accuracies. However, EEG features often contain redundant, irrelevant, and noisy information. Additionally, real-world EEG data acquisition frequently faces challenges, such as data loss from electrode detachment and heavy noise interference. To tackle the challenges, we propose a novel feature selection approach for robust depression analysis, called Incomplete Depression Feature Selection with Missing EEG Channels (IDFS-MEC). IDFS-MEC integrates missing-channel indicator information and adaptive channel weighting learning into orthogonal regression to lessen the effects of incomplete channels on model construction, and then utilizes global redundancy minimization learning to reduce redundant information among selected feature subsets. Extensive experiments conducted on MODMA and PRED-d003 datasets reveal that the EEG feature subsets chosen by IDFS-MEC have superior performance than 10 popular feature selection methods among 3-, 64-, and 128-channel settings.

Paper Structure

This paper contains 17 sections, 15 equations, 3 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. THE PROPOSED FRAMEWORK
  3. Global Redundancy Matrix
  4. Objective Function
  5. OPTIMIZATION STRATEGY
  6. Update $W^{\left( v \right)}$ by fixing $\Theta^{\left( v \right)}$, $\alpha^{\left( v \right)}$
  7. Update $\Theta^{\left( v \right)}$ by fixing $W^{\left( v \right)}$, $\alpha^{\left( v \right)}$
  8. Update $\alpha^{\left( v \right)}$ by fixing $W^{\left( v \right)}$, $\Theta^{\left( v \right)}$
  9. EXPERIMENTAL RESULTS AND DISCUSSION
  10. Experimental Details
  11. Performance Analysis and Discussion
  12. Feature Selection Porformance
  13. Ablation Experiment
  14. Parameter Sensitivity Analysis
  15. Computational Costs
  16. ...and 2 more sections

Figures (3)

  • Figure 1: Overall framework of IDFS-MEC. IDFS-MEC comprises three modules: Weighted orthogonal regression, adaptive channel weighting learning, and global redundancy minimization learning. Missing-channel indicator information is introduced as the weighted matrix for orthogonal regression. IDFS-MEC integrates weighted orthogonal regression and adaptive channel weighting learning to lessen the effects of incomplete channels on model construction, and then utilizes global redundancy minimization learning to reduce redundant information among selected EEG feature subsets.
  • Figure 2: Classification accuracies on the 3-, 64-, and 128-channel EEG datasets.
  • Figure 3: Average classification accuracies for different parameters on the 3-, 64-, and 128-channel EEG datasets.