A deep learning algorithm based on 1D CNN-LSTM for automatic sleep staging.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Sleep staging is crucial for sleep research.
• Traditional machine learning methods for sleep staging require extensive feature engineering.
• A novel approach is needed to simplify and improve automatic sleep staging.

Purpose of the Study:

• To propose a deep learning algorithm for automatic sleep staging that eliminates the need for manual feature extraction.
• To utilize a combination of one-dimensional convolutional neural network and long short-term memory for sleep staging.
• To validate the algorithm's performance using electroencephalogram signals.

Main Methods:

• The proposed algorithm employs a deep learning architecture combining 1D CNN and LSTM.
• Electroencephalogram (EEG) signals are processed using wavelet transform.
• The processed EEG signals are directly input into the deep learning model for end-to-end sleep staging into 5 stages (awake, N1-N3, REM).

Main Results:

• The algorithm achieved an accuracy of 93.47% using a single Fpz-Cz EEG channel.
• Utilizing both Fpz-Cz and other EEG signals, the highest accuracy reached 94.15%.
• The model demonstrated robust performance across different physiological signals.

Conclusions:

• The developed deep learning algorithm enables end-to-end automatic sleep staging.
• The method is suitable for various physiological signals, reducing the complexity of traditional approaches.
• This algorithm offers a more efficient and accurate solution for sleep research.