Primate brain pattern-based automated Alzheimer's disease detection model using EEG signals.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Alzheimer's disease (AD) is a progressive neurodegenerative disorder.
• Early detection of AD is crucial for effective management.
• Electroencephalography (EEG) shows potential for identifying early AD-related brain changes.

Purpose of the Study:

• To develop an automated model for early Alzheimer's disease detection using EEG signals.
• To leverage a novel directed graph approach inspired by primate brain patterns (PBP) for feature extraction.
• To evaluate the model's classification performance in distinguishing AD patients from healthy controls.

Main Methods:

• A novel directed graph based on primate brain connectome topology was used for local texture feature extraction from EEG signals.
• A tunable q-factor wavelet transform was combined with the PBP model to create a multilevel feature extractor.
• Iterative neighborhood component analysis selected discriminative features, which were fed into a weighted k-nearest neighbor (KNN) classifier.
• Leave-one subject-out (LOSO) and tenfold cross-validations were employed for performance evaluation.

Main Results:

• The PBP-based model, combined with wavelet transform, generated a rich set of 8,512 features per EEG signal.
• The model achieved 100% accuracy with tenfold cross-validation and 92.01% accuracy with LOSO cross-validation for AD vs. healthy classification.
• Both channel-wise and subject-level results demonstrated exemplary classification performance.

Conclusions:

• The developed multilevel PBP-based model effectively extracts discriminative features from EEG signals for Alzheimer's disease detection.
• This approach shows significant promise for the development of advanced, brain-inspired diagnostic tools.
• The findings pave the way for future research on connectome-inspired models for neurodegenerative disease diagnosis.