Sleep staging classification based on HRV: time-variant analysis.

Area of Science:

• Biomedical Engineering
• Sleep Medicine
• Signal Processing

Background:

• Assessing sleep quality typically requires polysomnography in sleep centers.
• There is a need for non-invasive, out-of-center methods to evaluate sleep macrostructure.
• Heart rate variability (HRV) contains information about autonomic nervous system activity during sleep.

Purpose of the Study:

• To develop and validate an algorithm for evaluating sleep macrostructure using ECG-derived heart rate fluctuations.
• To assess the feasibility of determining sleep quality outside of traditional sleep laboratory settings.
• To compare the performance of the developed algorithm against expert-scored sleep stages.

Main Methods:

• An algorithm combining a time-variant autoregressive model for feature extraction and a hidden Markov model (HMM) for classification was developed.
• Heart rate variability (HRV) features derived from the joint probability of the signal were used to train the HMM.
• The algorithm was tested on 17 full polysomnography recordings from healthy subjects.

Main Results:

• The automatic sleep stage classification achieved a total accuracy of 78.21% +/- 6.44% with two HRV features and 79.43% +/- 8.83% with three HRV features.
• Kappa index values of 0.41 +/- 0.1085 (two features) and 0.42 +/- 0.1493 (three features) were obtained.
• The algorithm demonstrated comparable performance to expert scoring for sleep macrostructure evaluation.

Conclusions:

• The developed algorithm effectively evaluates sleep macrostructure using ECG-derived HRV features.
• This approach shows potential for non-invasive, remote sleep quality assessment outside of sleep centers.
• Further validation in diverse populations and clinical settings is warranted.