Time-Frequency Analysis of Scalp EEG With Hilbert-Huang Transform and Deep Learning.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalography (EEG) is crucial in neuroscience and clinical applications.
• Conventional EEG analysis relies on fixed frequency bands, which may not capture individual neural variations.
• This limitation can lead to information loss and reduced classifier accuracy.

Purpose of the Study:

• To develop a systematic time-frequency analysis for scalp EEG signals.
• To introduce a data-driven method for computing subject-specific frequency bands using Hilbert-Huang Transform.
• To propose novel metrics for quantifying brainwave power and frequency.

Main Methods:

• Utilized Hilbert-Huang Transform for data-driven decomposition of EEG signals.
• Developed two novel metrics to quantify power and frequency of decomposed brainwave sub-signals.
• Validated the proposed metrics against conventional feature sets from wavelet and Hilbert-Huang Transform on two EEG datasets.

Main Results:

• The proposed metrics demonstrated superior discrimination compared to existing methods.
• Achieved high classification accuracies ranging from 94.93% to 99.84%.
• The novel metrics effectively quantify neural oscillations and brainwave characteristics.

Conclusions:

• The proposed subject-specific approach overcomes limitations of fixed frequency bands in EEG analysis.
• The novel metrics offer a more accurate and individualized quantification of neural oscillations.
• These metrics show significant potential for EEG-based classification and as biomarkers in neuroscience research.