Sleep Apnea Event Detection from Sub-frame Based Feature Variation in EEG Signal Using Deep Convolutional Neural...

Area of Science:

• Neuroscience
• Medical Technology
• Signal Processing

Background:

• Sleep apnea is a prevalent respiratory sleep disorder affecting millions globally.
• Electroencephalogram (EEG) signals offer a promising, non-invasive method for monitoring neural activity related to sleep.
• Accurate automatic detection of sleep apnea is crucial for timely diagnosis and treatment.

Purpose of the Study:

• To propose an innovative method for automatic sleep apnea detection using EEG signals.
• To enhance the accuracy of sleep apnea detection by analyzing local temporal feature variations.
• To evaluate the performance of the proposed method on a large-scale dataset.

Main Methods:

• EEG data frames were segmented into smaller sub-frames to extract local temporal features.
• A fully convolutional neural network (FCNN) was employed to process individual sub-frames and extract local features.
• A dense classifier analyzed extracted local features for apnea/non-apnea classification of the entire frame.
• A novel post-processing technique was applied to further refine detection accuracy.

Main Results:

• The proposed method demonstrated significant improvements in accuracy for automatic sleep apnea detection.
• Optimization of EEG frame length and post-processing parameters led to enhanced detection conditions.
• Large-scale experimentation on diverse patient data validated the method's effectiveness across varying apnea-hypopnea indices.

Conclusions:

• The developed FCNN-based approach effectively detects sleep apnea by integrating local and global EEG feature analysis.
• The innovative post-processing technique significantly boosts the performance of the automated detection system.
• This method holds potential for improving the clinical diagnosis and management of sleep apnea.