A deep learning algorithm for sleep stage scoring in mice based on a multimodal network with fine-tuning technique.

Area of Science:

• Neuroscience
• Computational Biology
• Animal Models

Background:

• Accurate sleep stage scoring is crucial for understanding sleep architecture in both preclinical and clinical research.
• Current methods for sleep stage classification in mice can be labor-intensive and require expert manual analysis.

Purpose of the Study:

• To develop an automated sleep stage classification system for mice using a novel deep neural network algorithm.
• To enhance the accuracy and efficiency of sleep stage scoring in animal models.

Main Methods:

• Extracted defining features from mouse electromyogram (EMG) and electroencephalogram (EEG) signals to develop wake-sleep and REM/NREM sleep models.
• Integrated these models using three distinct algorithms: rule-based integration, ensemble stacking, and multimodal with fine-tuning.
• Utilized a deep learning approach for automated classification.

Main Results:

• The multimodal deep learning algorithm with fine-tuning demonstrated high potential for improving sleep stage scoring accuracy in mice.
• This approach showed promise in advancing automated sleep analysis in animal experiments.

Conclusions:

• The developed deep learning algorithm, particularly the multimodal with fine-tuning approach, offers a promising tool for accurate sleep stage scoring in mice.
• This advancement can significantly promote and facilitate sleep research in animal models.