A Systematic Review of Detecting Sleep Apnea Using Deep Learning

Area of Science:

• Biomedical Engineering
• Computer Science
• Sleep Medicine

Background:

• Sleep apnea is a prevalent sleep disorder with diagnostic limitations.
• Polysomnography (PSG) is the gold standard but is costly and inaccessible.
• Automatic scoring methods using fewer sensors are being developed to overcome PSG challenges.

Purpose of the Study:

• To systematically review deep learning applications in sleep apnea detection over the past decade (2008-2018).
• To analyze implementation strategies, pre-processing, feature extraction, and network types for deep learning in sleep apnea research.
• To identify advantages, disadvantages, and challenges associated with deep learning approaches for sleep apnea diagnosis.

Main Methods:

• Systematic literature search conducted across five indexing services from 2008 to 2018.
• Inclusion and exclusion criteria applied to filter relevant studies.
• Analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.
• 255 papers were initially identified, with 21 selected for detailed review.

Main Results:

• Deep learning techniques are increasingly utilized in sleep apnea research due to improved performance and computational power.
• Various deep network architectures, pre-processing methods, and feature extraction techniques have been employed.
• Signals, sensors, databases, and implementation challenges specific to deep learning in sleep apnea were examined.

Conclusions:

• Deep learning offers a promising avenue for developing more accessible and automated sleep apnea detection systems.
• Further research is needed to optimize deep learning models and address implementation challenges for clinical practice.
• This review provides insights into the current landscape and future directions of deep learning in sleep apnea diagnosis.