EEG analysis and classification based on cardinal spline empirical mode decomposition and synchrony features

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Dementia diagnosis is crucial for early intervention to slow cognitive decline.
• Electroencephalography (EEG) offers a noninvasive, accessible, and cost-effective alternative to other dementia diagnostic methods.
• Existing signal processing techniques for EEG may have limitations in decomposing complex biosignals.

Purpose of the Study:

• To introduce a novel digital signal processing method, cardinal spline empirical mode decomposition (CS-EMD), for enhanced EEG analysis.
• To evaluate the efficacy of CS-EMD in classifying dementia using EEG data.
• To compare the performance of CS-EMD against classical empirical mode decomposition (EMD) and other existing methods.

Main Methods:

• Developed a novel CS-EMD algorithm for improved EEG signal decomposition into intrinsic mode functions (IMFs).
• Extracted longitudinal and transversal synchrony measures from CS-EMD derived IMFs.
• Utilized a support vector machine (SVM) classifier with synchrony measures as features for healthy and dementia classification.

Main Results:

• The CS-EMD method achieved high classification performance on a public EEG dataset.
• Optimal results were obtained using synchrony measures from five IMFs, yielding 90% accuracy, 96.67% specificity, 83.33% sensitivity, and 96.15% precision.
• CS-EMD outperformed the classical EMD method and other approaches in dementia classification.

Conclusions:

• The data-driven CS-EMD method demonstrates significant potential as an accurate and efficient dementia screening tool.
• CS-EMD's superior signal decomposition properties make it suitable for analyzing nonlinear and nonstationary biosignals beyond EEG.
• This novel approach could facilitate low-cost, widespread dementia screening.