Classifying Neurodegenerative Diseases from Selected Temporal EEG Electrodes: Towards Ear-EEG Applications

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Dementia affects over 55 million globally, necessitating early detection due to the lack of a cure.
• Machine learning (ML) applied to electroencephalography (EEG) shows potential for early dementia diagnosis.
• Accessible tools like ear-EEG are being explored for widespread brain health monitoring.

Purpose of the Study:

• To investigate the classification of Alzheimer's Disease (AD) and Fronto-temporal Dementia (FTD) using temporal EEG data.
• To assess the feasibility of a multiclass classification approach for differentiating between healthy individuals, AD, and FTD.
• To evaluate the potential of ear-EEG as an accessible tool for dementia detection.

Main Methods:

• Utilized EEG recordings from 88 participants (29 control, 36 AD, 23 FTD) at rest with eyes closed.
• Extracted features from temporal electrodes (T3, T4, T5, T6, F7, F8), including sub-band power and band power ratios.
• Employed binary and multiclass classification models to differentiate between cognitive states.

Main Results:

• Achieved 81.8% accuracy in binary classification of healthy vs. AD.
• Obtained 63.8% accuracy in multiclass classification of healthy, AD, and FTD.
• Identified challenges in multiclass classification for differentiating dementia subtypes using the selected features.

Conclusions:

• Temporal EEG data holds potential for developing dementia classification models, supporting the use of accessible tools like ear-EEG.
• Further advancements in multiclass classification models are needed to accurately differentiate between dementia types.
• Early detection through improved EEG-based tools can facilitate timely interventions and enhance patient outcomes.