Deep learning-based algorithm accurately classifies sleep stages in preadolescent children with sleep-disordered...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computational Biology

Background:

• Manual sleep scoring is inconsistent, impacting diagnostic accuracy.
• Automatic sleep staging algorithms for adults may not generalize to children due to distinct EEG patterns.
• The preadolescent sleep stage classification remains understudied.

Purpose of the Study:

• To develop a deep learning-based automatic sleep stage classifier for preadolescent children (10-13 years old).
• To evaluate the algorithm's performance against human scorers and assess its reliability.

Main Methods:

• A dataset of 115 preadolescent children (with and without sleep-disordered breathing symptoms) was used.
• A combined convolutional and long short-term memory neural network utilized EEG, EOG, and EMG signals.
• Performance was validated against manual scoring and inter-rater reliability metrics.

Main Results:

• The deep learning model achieved 84.1% overall accuracy (Cohen's kappa = 0.78).
• Performance was consistent across subgroups with and without sleep-disordered breathing symptoms.
• The algorithm's agreement with manual scorers was comparable to inter-rater reliability.

Conclusions:

• The developed algorithm provides accurate and reliable automatic sleep scoring for preadolescents.
• This deep learning approach offers a less labor-intensive alternative to manual scoring.
• The findings support the use of automated methods for objective sleep analysis in pediatric populations.