Deep learning for sleep analysis on children with sleep-disordered breathing: Automatic detection of mouth breathing...

Area of Science:

• Pediatric Sleep Medicine
• Artificial Intelligence in Healthcare
• Biomedical Signal Processing

Background:

• Sleep-disordered breathing (SDB) in children, including habitual snoring and obstructive sleep apnea (OSA), often involves mouth breathing.
• Mouth breathing as an SDB indicator in children is frequently overlooked and inconsistently diagnosed.
• Accurate detection of mouth breathing is crucial for timely SDB diagnosis and management in pediatric populations.

Purpose of the Study:

• To develop and evaluate a deep learning algorithm for the automatic detection of mouth breathing events in children.
• To utilize polysomnography (PSG) data, specifically mouth and nasal pressure signals, for identifying mouth breathing.
• To assess the performance of a convolutional neural network (CNN) model in classifying mouth breathing events.

Main Methods:

• Polysomnography (PSG) recordings from 20 children (aged 10-13 years) were analyzed.
• Mouth and nasal pressure signals from PSG were extracted and processed.
• Convolutional neural networks (CNNs) were employed to identify and classify mouth breathing events.

Main Results:

• The deep learning model achieved 93.5% accuracy and 97.8% precision on validation data.
• The model demonstrated an 89% true positive rate and a 2% false positive rate.
• Performance decreased on a secondary dataset, suggesting the need for larger training data.

Conclusions:

• Deep neural networks show significant potential for analyzing and classifying biological signals in sleep studies.
• Machine learning offers a valuable tool for enhancing sleep analysis and SDB diagnosis in children.
• Further development with larger datasets is recommended to improve model generalizability.