Wavelet-Based Biphase Analysis of Brain Rhythms in Automated Wake-Sleep Classification.

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Electroencephalography (EEG) signal analysis for sleep detection faces challenges due to its nonstationary nature.
• Traditional methods like coherence may be unsuitable for automatic sleep staging.
• Novel features are needed to accurately classify sleep-wake states from EEG.

Purpose of the Study:

• To propose wavelet-based bi-phase (Wbiph) as a new feature for sleep-wake classification using EEG.
• To evaluate the efficacy of Wbiph against coherence in distinguishing sleep and wake states.
• To assess the performance of a convolutional neural network (CNN) classifier using Wbiph for sleep-wake classification.

Main Methods:

• Combined wavelet transform and bispectrum to create the wavelet-based bi-phase (Wbiph) feature.
• Applied statistical analysis to compare Wbiph with coherence, focusing on gamma rhythm.
• Utilized a convolutional neural network (CNN) for sleep-wake classification with Wbiph features.

Main Results:

• Statistical analysis highlighted the importance of gamma rhythm in sleep detection.
• Wbiph demonstrated superior performance over coherence in wake-sleep classification.
• CNN classification achieved 97.17% accuracy (nonLOSO) and 95.48% (LOSO) using Wbiph, outperforming previous studies.

Conclusions:

• Wavelet-based bi-phase (Wbiph) is a potent novel feature for EEG-based sleep-wake classification.
• Wbiph offers advantages over coherence due to its incorporation of wavelet and bispectrum analysis.
• The high classification accuracy achieved with Wbiph and CNN suggests its potential for advanced automatic sleep monitoring systems.