MCI Detection From Odor-Evoked EEG Using a Multibranch Attention-Based Temporal-Spectral CNN.

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Dementia is a progressive neurodegenerative disorder.
• Mild Cognitive Impairment (MCI) often precedes dementia, characterized by memory loss and cognitive decline.
• Early MCI detection is critical for timely intervention and improved patient outcomes.

Purpose of the Study:

• To differentiate between individuals with MCI and healthy controls using EEG signals.
• To investigate the efficacy of odor-evoked brain potentials for MCI detection.
• To develop and evaluate an attention-based deep learning model for MCI classification.

Main Methods:

• Utilized publicly available multichannel EEG data.
• Extracted temporal-spectral features using wavelets, spectral grouping, and canonical correlation.
• Employed attention-based Convolutional Neural Network (CNN) models with individual feature branches, followed by a fully connected network for classification.

Main Results:

• The proposed method demonstrated superior performance in differentiating MCI from healthy subjects compared to other approaches.
• Ablation studies confirmed the significant contribution of individual feature sets and their synergistic effect when combined.
• The model successfully classified subjects based on odor-evoked brain potentials.

Conclusions:

• Odor-evoked brain potentials analyzed via EEG offer a promising biomarker for MCI detection.
• Attention-based CNN models provide an effective framework for classifying MCI using complex EEG features.
• This approach holds potential for early diagnosis and management of cognitive decline, aiding dementia prevention efforts.