Graph-BrainConvNet: A One-class GCN-based approach for MCI detection from source-level MEG.

Area of Science:

• Neuroscience
• Gerontology
• Artificial Intelligence

Background:

• Mild cognitive impairment (MCI) is a precursor to dementia, necessitating early detection for intervention.
• Current neuroimaging methods for MCI detection are often limited to resting states or complex tasks unsuitable for general geriatric populations.
• Identifying reliable MCI biomarkers is crucial for developing effective risk reduction strategies.

Purpose of the Study:

• To investigate neural correlates of attention in MCI patients during a novel task involving silent movie viewing with auditory distractions.
• To develop and validate a machine learning model for detecting MCI using functional brain connectivity and spectral-temporal features.
• To explore potential auditory gating deficits as a biomarker for early MCI diagnosis.

Main Methods:

• A feasibility experiment was conducted with MCI patients and control participants (CPs) watching a silent movie with random auditory distractors.
• A novel one-class graph convolutional neural network (GCN)-based model was employed to analyze source-level functional brain connectivity and spectral-temporal features.
• Classification accuracy was assessed for distinguishing between MCI patients and CPs based on gamma rhythm activity.

Main Results:

• The GCN model achieved a high classification accuracy of 92.63±0.96% in differentiating MCI patients from CPs using gamma rhythm.
• MCI patients exhibited significantly higher neural responses to auditory distractors compared to CPs, suggesting impaired attentional processing.
• These findings indicate potential auditory gating deficits in individuals with MCI.

Conclusions:

• The study successfully identified neural correlates of attention in MCI patients during a unique experimental paradigm.
• The proposed GCN model demonstrates high efficacy in detecting MCI biomarkers, particularly related to auditory processing.
• The findings suggest that auditory gating deficits are a potential indicator of cognitive decline in MCI, offering a new avenue for early diagnosis.