Deep Learning and Insomnia: Assisting Clinicians With Their Diagnosis

Area of Science:

• Neuroscience
• Medical Technology
• Artificial Intelligence

Background:

• Effective sleep analysis is crucial but hindered by a lack of automated tools and complex hardware.
• Disordered sleep patterns, such as insomnia, require accurate diagnostic methods.

Purpose of the Study:

• To develop and evaluate a deep learning model for automated sleep analysis.
• To differentiate between patients with insomnia and healthy controls using electroencephalogram (EEG) features.

Main Methods:

• Applied deep learning to 57 EEG features from one or two EEG channels.
• Investigated both stage-independent and stage-dependent (NREM + REM) classification approaches.
• Trained and tested the model on data from 41 controls and 42 primary insomnia patients.

Main Results:

• The NREM + REM based classifier achieved 92% discrimination accuracy with two EEG channels and 86% with one channel.
• Deep learning models demonstrated high accuracy in distinguishing between insomnia patients and controls.
• The stage-dependent approach showed promising results in sleep disorder diagnosis.

Conclusions:

• Deep learning provides an effective automated tool for sleep analysis.
• This technology can assist in the clinical diagnosis of sleep disorders like insomnia.
• The findings highlight the potential of AI in improving sleep medicine.