Derivative Method to Detect Sleep and Awake States through Heart Rate Variability Analysis Using Machine Learning...

Area of Science:

• Cardiology
• Neurology
• Sleep Medicine

Background:

• Sleep disorders have significant short- and long-term health consequences, including attention deficits and cardiac issues.
• Polysomnography (PSG) is a common sleep assessment tool but is intrusive due to numerous cables and typically requires a clinical setting, potentially affecting accuracy.
• Assessing the central nervous system (CNS) via portable devices offers a simpler alternative for sleep disorder evaluation.

Purpose of the Study:

• To develop and validate a lightweight feature classification model for predicting awake and napping states.
• To investigate the efficacy of using RR intervals (RRI) and their second derivative for sleep state prediction.
• To determine if a short RRI time series window is sufficient for accurate sleep state classification.

Main Methods:

• Implemented a feature classification model utilizing the RR interval (RRI) and its second derivative.
• Trained and validated the model using heart rate variability (HRV) data from nine healthy young adults.
• Analyzed HRV data associated with light-on, light-off, sleep onset, and sleep offset events.

Main Results:

• A 30-minute RRI time series window was found to be sufficient for the model.
• The lightweight model accurately predicted whether subjects were awake or napping.
• The method demonstrates potential for non-intrusive sleep state assessment.

Conclusions:

• A simple, portable model using RRI can accurately predict awake and napping states.
• This approach offers a less intrusive alternative to traditional polysomnography (PSG).
• Further research could explore this method for broader sleep disorder diagnostics.