A systematic review on automatic identification of insomnia

Area of Science:

• Sleep Medicine
• Artificial Intelligence
• Biomedical Engineering

Background:

• Insomnia is a common sleep disorder affecting sleep quantity and quality.
• Machine learning (ML) and deep learning (DL) offer advanced tools for automated sleep analysis.
• Physiological signals are key data sources for detecting sleep disorders like insomnia.

Purpose of the Study:

• To explore and categorize algorithms for automatic insomnia detection.
• To compare the performance of various ML and DL techniques in identifying insomnia.
• To identify research gaps and future opportunities in automated insomnia detection.

Main Methods:

• Systematic review following PRISMA guidelines (2015-2023).
• Analysis of over 30 publications focused on automated insomnia identification.
• Evaluation of ML/DL models trained on annotated physiological signals, identifying 15 distinct algorithms.

Main Results:

• Automated insomnia detection systems utilize diverse data sources, ML/DL networks, and training frameworks.
• Classification of studies based on ML/DL model perspectives, learning structures, and input data types.
• Identified key components of automated techniques: data input, objectives, AI models, training, and databases.

Conclusions:

• Current automated insomnia detection methods show promise but require enhanced accuracy and reliability.
• Significant research gaps exist, highlighting opportunities for future technological and algorithmic advancements.
• Further development in AI can lead to more effective and efficient insomnia identification.