Related Experiment Video
Updated: Jun 25, 2025
09:57
Author Spotlight: Advancing Pediatric Epilepsy Surgery in Children Through Novel Biomarkers and Enhanced Localization
Published on: September 20, 2024
2.6K
Selecting EEG channels and features using multi-objective optimization for accurate MCI detection: validation using
Majid Aljalal1, Saeed A Aldosari2, Marta Molinas3
1Department of Electrical Engineering, College of Engineering, King Saud University, Riyadh, Saudi Arabia. maljalal@ksu.edu.sa.
Scientific Reports
|May 30, 2024
Summary
This study introduces a multi-objective optimization approach using NSGA-II to select Electroencephalogram (EEG) channels and features for detecting mild cognitive impairment (MCI). Optimized selection significantly improves diagnostic accuracy, paving the way for clinical application.
Area of Science:
- Neuroscience and Biomedical Engineering
- Machine Learning in Healthcare
Background:
- Timely detection of mild cognitive impairment (MCI) is crucial for effective dementia management.
- Electroencephalogram (EEG) signals offer a potential non-invasive method for MCI detection.
- Optimizing EEG channel and feature selection is key to improving diagnostic accuracy and efficiency.
Purpose of the Study:
- To develop and evaluate a multi-objective optimization approach for selecting optimal EEG channels and features for MCI detection.
- To enhance the accuracy and efficiency of MCI diagnosis using machine learning techniques.
- To investigate the impact of feature and channel selection on classification performance.
Main Methods:
- EEG signals were decomposed into subbands using Variational Mode Decomposition (VMD) or Discrete Wavelet Transform (DWT).
- Features such as standard deviation, entropy, and fractal dimensions were extracted from subbands.
- The Non-dominated Sorting Genetic Algorithm II (NSGA-II) was employed for multi-objective optimization of channel and feature selection, aiming to minimize channel count and maximize classification accuracy.
Main Results:
- The NSGA-II algorithm significantly improved classification accuracy for MCI detection compared to using all 19 EEG channels.
- Selecting only five channels using NSGA-II increased accuracy to 91.56%, and using 8 features from 7 channels achieved 95.28% accuracy.
- Optimized selection of informative channels and features reduced the impact of noise, leading to substantial performance gains.
Conclusions:
- Multi-objective optimization using NSGA-II is effective for selecting informative EEG channels and features for accurate MCI detection.
- A reduced set of channels and features can yield higher diagnostic accuracy than using all available data, suggesting a more efficient approach.
- This method holds promise for the development of practical, clinically applicable tools for early MCI diagnosis.
