PLI-Based Connectivity in Resting-EEG is a Robust and Generalizable Feature for Detecting MCI and AD: A Validation on...

Area of Science:

• Neuroscience
• Medical Technology
• Computational Biology

Background:

• Alzheimer's disease (AD) and mild cognitive impairment (MCI) pose significant public health challenges.
• Electroencephalography (EEG) combined with machine learning shows promise for AD/MCI diagnosis.
• Previous studies often lack generalizability due to single-center data collection.

Purpose of the Study:

• To reevaluate the effectiveness of EEG-based machine learning for AD/MCI detection using a large, multi-center dataset.
• To assess the performance of Phase Lag Index (PLI) as a key feature for AD/MCI classification.
• To compare PLI against other EEG features and classifiers for diagnostic accuracy.

Main Methods:

• Collected resting-state EEG data from 150 participants across six hospitals.
• Utilized Linear Discriminative Analysis (LDA) classifiers with PLI features.
• Compared PLI performance with other common EEG features and classifiers.
• Validated the model using leave-one-participant-out cross-validation (LOPO-CV) on a training set and an independent test set.

Main Results:

• Phase Lag Index (PLI) demonstrated superior performance compared to other EEG features.
• The LDA classifier with optimal PLI features achieved 82.50% LOPO-CV accuracy on the training set.
• The model maintained a 75.00% accuracy on the independent test set, indicating good generalizability.
• PLI-based functional connectivity emerged as a robust indicator for AD/MCI detection.

Conclusions:

• PLI-based functional connectivity is a reliable biomarker for detecting AD/MCI in real-world clinical settings.
• Multi-center validation enhances the generalizability of EEG-based machine learning models for neurological disorder diagnosis.
• This approach offers a promising, non-invasive method to aid in the early detection of AD/MCI.