Identifying Individuals With Mild Cognitive Impairment Using Working Memory-Induced Intra-Subject Variability of
Thanh-Tung Trinh1, Chia-Fen Tsai2,3, Yu-Tsung Hsiao4
1Neural Engineering and Smart Systems Laboratory, Graduate Institute of Manufacturing Technology, College of Mechanical and Electrical Engineering, National Taipei University of Technology (Taipei Tech), Taipei, Taiwan.
Frontiers in Computational Neuroscience
|August 23, 2021
Summary
Electroencephalogram (EEG) variability shows promise for early detection of mild cognitive impairment (MCI). This novel method, analyzing task-induced changes in EEG spectral power, achieved high accuracy in distinguishing MCI and Alzheimer
Area of Science:
- Neuroscience
- Biomedical Engineering
- Medical Diagnostics
Background:
- Mild cognitive impairment (MCI) significantly increases the risk of developing dementia, such as Alzheimer's disease (AD).
- Early detection of MCI is crucial but challenging, with existing methods often being costly or invasive.
- Electroencephalogram (EEG) offers a non-invasive and accessible alternative for neurological assessment, yet specific EEG features for MCI detection require further exploration.
Purpose of the Study:
- To develop and validate a novel EEG feature extraction framework for early detection of MCI.
- To investigate the utility of spectral-power-based task-induced intra-subject variability as a potential biomarker for MCI.
- To compare the classification performance of the proposed EEG feature against established methods.
Main Methods:
- A new framework was designed to extract task-induced intra-subject spectral power variability from resting-state EEGs.
- This variability was measured as between-run similarity before and after participants completed a cognitively demanding working memory task.
- A Support Vector Machine (SVM) classifier with leave-one-participant-out cross-validation (LOPO-CV) was employed for classification tasks.
Main Results:
- The proposed method identified higher between-run similarity in frontal and central scalp regions for healthy controls (HC) compared to individuals with MCI or AD.
- The spectral-power-based intra-subject variability achieved 80.39% accuracy in classifying MCI vs. HC and 78% for AD vs. HC using LOPO-CV.
- This novel feature outperformed traditional EEG measures like spectral powers, coherence, and complexity (Katz's method).
Conclusions:
- The spectral-power-based task-induced intra-subject EEG variability is a promising neurophysiological feature for the early detection of MCI.
- The developed framework provides a novel and effective approach for identifying individuals at risk of cognitive decline.
- This non-invasive EEG-based method holds potential for clinical application in diagnosing MCI and AD.
Keywords:
Alzheimer's diseasebetween-run similaritybrain-computer interfaceelectroencephalographyintra-subject variabilitymachine learningmild cognitive impairment
