Dynamic Mode Decomposition-Based Clustered Pattern Projection for Reliable Alzheimer's Disease Detection from EEG.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Detecting Alzheimer's disease (AD) from normal aging using eyes-open (EO) electroencephalography (EEG) is difficult due to nonstationarity and fragmented responses.
• Conventional spectral methods often struggle with the complexities of EO photostimulation EEG data.

Purpose of the Study:

• To evaluate the effectiveness of prototype-based representations derived from Dynamic Mode Decomposition (DMD) for improving AD detection from EO EEG.
• To introduce and validate a novel framework, DMD-based Clustered Pattern Projection (DMD-CPP), for enhanced AD diagnosis.

Main Methods:

• Developed the DMD-CPP framework, which clusters segment-wise DMD representations to learn class-specific medoid prototypes.
• Encoded each EEG epoch as cosine-similarity coordinates relative to learned prototypes.
• Utilized a linear Support Vector Machine (SVM) classifier trained on DMD-CPP features and validated using leave-one-subject-out cross-validation.

Main Results:

• The DMD-CPP model demonstrated competitive classification accuracy and improved margin-based reliability for AD detection from EO photostimulation EEG.
• Observed enhanced decision margins in AD versus healthy control classification, with lower confidence assigned to misclassified normal epochs.
• Showed improvements in tasks involving frontotemporal dementia detection, though less pronounced than for AD.

Conclusions:

• Clustering-based pattern projection stabilizes EEG dynamics, yielding interpretable and confidence-aware feature representations.
• DMD-CPP offers a promising approach for reliable AD detection from EO EEG, outperforming conventional spectral methods.
• The findings support DMD-CPP as a valuable tool for neurological disorder diagnosis using EEG.