Difference between eyes-open and eyes-closed resting state quantitative electroencephalography (qEEG) for predicting...

Area of Science:

• Neuroscience
• Medical Imaging
• Artificial Intelligence

Background:

• Electroencephalography (EEG) records brain's electrical activity, offering insights into cerebral cortex function.
• Quantitative EEG (qEEG) analysis of EEG signals serves as a neurophysiological indicator for early dementia detection.
• Mild cognitive impairment (MCI) and Alzheimer's disease (AD) represent significant challenges in neurological research.

Purpose of the Study:

• To introduce a deep learning (DL) classification approach for detecting MCI and AD.
• To utilize the difference between eyes-open resting (EOR) and eyes-closed resting (ECR) qEEG time-frequency (TF) images as input features.
• To evaluate the efficacy of DL models in differentiating between normal controls (NC), MCI, and AD subjects.

Main Methods:

• A dataset of 16,910 qEEG TF images from 890 subjects (NC, MCI, AD) was analyzed.
• EEG signals were converted to qEEG TF images using Fast Fourier Transform (FFT) across five frequency sub-bands.
• A convolutional neural network (CNN) within a DL framework processed TF images and age data for classification.

Main Results:

• The DL model differentiating NC versus cognitive impairment (CI = MCI + AD) achieved an AUC of 0.95, accuracy of 0.93, sensitivity of 0.97, and specificity of 0.92.
• Performance for NC versus AD classification was an AUC of 0.88, accuracy of 0.88, sensitivity of 0.89, and specificity of 0.86.
• For NC versus MCI classification, the model achieved an AUC of 0.85, accuracy of 0.83, sensitivity of 0.90, and specificity of 0.81.

Conclusions:

• The difference between EOR and ECR qEEG states provides a practical method for detecting cognitive impairment using DL.
• The developed DL models can serve as a decision-support system for clinicians in the early diagnosis of cognitive impairment.
• This approach offers a potential supportive reference for dementia research and clinical practice.