SLA-MLP: Enhancing Sleep Stage Analysis from EEG Signals Using Multilayer Perceptron Networks

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Sleep stage analysis is crucial for diagnosing sleep disorders and assessing sleep quality.
• Traditional sleep classification methods face limitations in accuracy, scalability, and objectivity.
• Existing deep learning models struggle with overfitting, computational demands, and imbalanced datasets.

Purpose of the Study:

• To introduce the Sleep Stage Analysis with Multilayer Perceptron (SLA-MLP) model for enhanced sleep stage classification.
• To overcome the limitations of traditional and existing deep learning approaches in sleep analysis.

Main Methods:

• Utilized advanced deep learning techniques, including Temporal Convolutional Networks (TCNs) for feature extraction and a Multilayer Perceptron (MLP) for classification.
• Implemented robust preprocessing steps: signal cropping, spectrogram conversion, and normalization.
• Employed data balancing techniques with adjusted class weights to manage imbalanced datasets.

Main Results:

• The SLA-MLP model achieved high accuracy rates: 97.23% on S-DSI, 96.23% on S-DSII, and 97.23% on S-DSIII datasets.
• Demonstrated superior performance compared to traditional sleep classification methods.
• Effectively addressed challenges like overfitting and data imbalance.

Conclusions:

• SLA-MLP offers a significant advancement in sleep stage analysis, providing a more precise tool for clinical applications and sleep research.
• The model's integrated approach of advanced feature extraction, robust preprocessing, and adaptive data balancing ensures reliable sleep stage classification.
• Achieved high accuracy, indicating its potential for improving the diagnosis and management of sleep disorders.