Real-time, automatic, open-source sleep stage classification system using single EEG for mice.

Area of Science:

• Neuroscience
• Computational Biology
• Sleep Science

Background:

• Accurate sleep stage classification is crucial for understanding sleep functions and disorders.
• Current methods often require complex multi-channel recordings and manual analysis, limiting scalability.
• Real-time, automated classification systems are needed for advanced sleep research and therapeutic interventions.

Purpose of the Study:

• To develop a novel, real-time sleep stage classification system using minimal data.
• To enable automatic, targeted interventions during specific sleep stages, particularly REM sleep.
• To create a scalable and adaptable system for broad research applications.

Main Methods:

• Utilized a convolutional neural network (CNN) with a single-channel electroencephalogram (EEG) from mice.
• Incorporated a long short-term memory (LSTM) recurrent neural network to process historical subject data.
• Employed universally available time-series features: raw signal, spectrum, and zeitgeber time.

Main Results:

• Achieved 90% overall accuracy and 81% multi-class Matthews Correlation Coefficient (MCC).
• Demonstrated high performance in classifying rapid eye movement (REM) sleep (91% sensitivity, 98% specificity).
• System (UTSN-L) eliminates the need for pre-calibration, electromyogram (EMG) recording, and manual scoring.

Conclusions:

• The developed system (UTSN-L) offers a scalable, accurate, and automated solution for real-time sleep stage classification.
• Enables precise, real-time interventions during REM sleep, facilitating research into its functions.
• Open-source availability and user-friendly interface promote widespread adoption for sleep research and therapeutic development.