Original Research By: Pini et al., 2022
Summarized By: Neurobit
Polysomnography (PSG) is currently the gold standard for assessing sleep, but it is expensive and requires expert interpretation. There is a need for non-invasive, inexpensive, and reliable sleep monitors that can be used in the home. Recent technology has led to the development of consumer sleep monitors that use activity and physiological signals such as EEG, heart rate (HR), breathing, and pulse oximetry to monitor sleep. HR and HR variability have been shown to be accurate indicators of sleep stages.
The current study done by Nicolò Pini and colleagues (2022) aims to validate a new deep learning algorithm: Neurobit-HRV. Neurobit-HRV is a sleep staging algorithm that uses HR data from single-channel electrocardiography (ECG) to classify sleep into two, three, or four levels. The use of the algorithm with a HR device could provide a cost-effective and non-invasive solution for sleep staging.
The algorithm was tested using two datasets: one open-source dataset of PSG recordings from Physionet CinC with 994 participants and one proprietary dataset (Z3Pulse) with 52 participants. The algorithm was able to classify sleep into two, three, or four levels (wake, non-REM, REM, and deep sleep). The performance of the algorithm was evaluated using several measures including accuracy, Cohen's kappa, sensitivity, and specificity. The highest accuracy was achieved by the two-level model in both datasets. The three-level model had the best value of Cohen's kappa in both datasets. The four-level model had the lowest sensitivity and the highest specificity for classifying deep sleep segments in the open-source dataset. Age had a significant effect on the performance of the models, but sex and a measure called AHI did not.
The results of this study suggest that the Neurobit-HRV algorithm is accurate and robust. Furthermore, the researchers propose that using it with a heart rate device can be a cost-effective, publicly available, and non-invasive solution for sleep staging in the research and clinical communities.
Pini, N., Ong, J. L., Yilmaz, G., Chee, N. I. Y. N., Siting, Z., Awasthi, A., Biju, S., Kishan, K., Patanaik, A., Fifer, W. P., & Lucchini, M. (2022). An automated heart rate-based algorithm for sleep stage classification: Validation using conventional PSG and innovative wearable ECG device. MedRxiv, 2021.12.21.21268117. https://doi.org/10.1101/2021.12.21.21268117