Association Between Sleep Quality and Deep Learning-Based Sleep Onset Latency Distribution Using an...

Area of Science:

• Neuroscience
• Machine Learning
• Sleep Medicine

Background:

• Evaluating sleep quality typically requires extensive overnight monitoring, leading to inefficiencies in data management and analysis.
• Existing methods for sleep monitoring are often time-consuming and data-intensive.
• There is a need for more efficient methods to assess sleep patterns and quality.

Purpose of the Study:

• To develop a deep learning model for predicting sleep onset latency (SOL) distribution using a brief, early sleep electroencephalogram (EEG) recording.
• To explore the association between predicted SOL distribution and overall sleep quality (SQ).
• To determine if early EEG can effectively predict sleep features related to SQ.

Main Methods:

• A deep learning model was designed with signal decomposition/restoration and SOL distribution prediction structures.
• The model utilized a 30-second EEG segment from the early sleep cycle.
• The Sleep Heart Health Study public dataset was used for model training and evaluation.

Main Results:

• The model successfully estimated SOL distribution, categorizing it into four clusters.
• The model provides a temporal probability graph illustrating the process of falling asleep.
• A SOL under 10 minutes was found to correlate strongly with good SQ.
• SOL prediction from early EEG was more suitable than predicting total sleep time, sleep efficiency, or actual sleep time.

Conclusions:

• Deep learning enables the estimation of SOL distribution from brief, early EEG recordings.
• An SOL distribution within 10 minutes is a significant indicator of good SQ.
• This approach offers a more efficient method for sleep quality assessment.