In a recent article published in the journal Sensors, researchers from Greece explored the feasibility of a wearable system for monitoring vital signs during sleep. The system utilizes five inertial measurement units (IMUs) positioned on the waist, arms, and legs.
Placement of the PDMonitor® sensors on the body. The sensors on the waist and the lateral compartments of the legs are attached with Velcro bands and StrapFrames, while wristbands are used to attach the sensors worn on the posterior compartment of the forearm. Image Credit: https://www.mdpi.com/1424-8220/24/13/4139
The aim is to provide a comprehensive understanding of sleep quality, particularly in individuals with sleep-disordered breathing (SDB). By incorporating IMUs in strategic locations, the system aims to enhance the accuracy and reliability of vital sign monitoring during sleep.
Background
Prior research has focused on analyzing sensor types and positioning to improve the precision of estimating cardiorespiratory parameters. Various methodologies have been developed to increase the signal-to-noise ratio (SNR) by selecting optimal axes, combining axes, or fusing estimated rates from different sensors. The use of multiple IMUs distributed across the body, as in this study, presents a novel approach to mitigate movement artifacts in ambulatory recordings and enhance data fusion techniques for vital sign estimation.
The Current Study
Twenty-three participants with varying severity of sleep-disordered breathing (SDB) were recruited for the study. Inclusion criteria encompassed individuals undergoing a sleep study to monitor vital signs, including respiratory rate (RR) and heart rate (HR), using polysomnography (PSG) as the reference standard.
The wearable system integrated five inertial measurement units (IMUs) strategically positioned on the waist, arms, and legs of the participants. Each IMU collected data on multiple axes, enabling comprehensive monitoring of body movements and vital signs during sleep.
Participants underwent overnight sleep monitoring using the wearable system. The IMUs recorded continuous data on accelerations and angular velocities, which were processed to estimate RR and HR. The dataset obtained from the IMUs was compared with the reference vital signs derived from PSG to evaluate the accuracy and reliability of the wearable system.
A novel weighted function was introduced to enhance the SNR for spectra with dominant rates, improving the quality of vital sign estimations. Signal processing techniques were applied to extract relevant features from the IMU data, allowing for the fusion of information across multiple axes and sensors.
The performance of the wearable system was assessed using various metrics, including Mean Absolute Error (MAE), time coverage (TC), Spearman’s correlation coefficient (ρ), and Bland–Altman analysis. MAE quantified the accuracy of RR and HR estimations, while T C indicated the proportion of time intervals with high-quality estimations. Correlation analysis evaluated the bivariate association between estimated and reference vital signs, providing insights into the system's performance.
Statistical tests, such as one-sample t-test and Kruskal–Wallis nonparametric analysis of variance, were conducted to assess the significance of bias and differences in performance metrics between different IMU combinations. Post hoc analyses, including Wilcoxon signed rank tests with Bonferroni correction, were employed to identify distinct distributions and evaluate the effectiveness of combining IMUs on the waist, arms, and legs.
Results and Discussion
The evaluation of the wearable system's performance revealed strong correlations between the estimated vital signs from the IMUs and the reference values obtained from PSG. The MAE values for respiratory and heart rate estimations indicated low errors, demonstrating the system's accuracy in monitoring vital signs during sleep. The TC metrics reflected the proportion of high-quality estimations over time intervals, with notable values for RR estimation and improvements observed for HR estimation when combining data from multiple IMUs.
Spearman’s correlation coefficient (ρ) was utilized to quantify the bivariate association between the estimated and reference vital signs, providing insights into the consistency and reliability of the wearable system. The Bland–Altman plots illustrated the agreement between the estimated and reference values, highlighting the system's ability to track vital signs effectively during sleep.
Combining data from IMUs positioned on the waist, arms, and legs enhanced the inter-participant time coverage of HR estimation, showcasing the benefits of data fusion techniques in improving monitoring capabilities. The results underscored the importance of strategic IMU placement and effective data fusion algorithms in optimizing vital sign monitoring during sleep.
Conclusion
In conclusion, the proposed inertial-based wearable system effectively monitors RR and HR during sleep, focusing on individuals with SDB. The novel methodology, incorporating multiple IMUs and data fusion techniques, enhances the accuracy and reliability of vital sign estimations.
The findings suggest that a single IMU sensor at the waist can provide robust vital sign monitoring, with the potential for further improvements by combining data from additional IMUs. This research opens avenues for exploring physiological variations in sleep patterns and circadian rhythms, ultimately contributing to improved health monitoring and sleep quality assessment without the need for PSG recordings.
Journal Reference
Kontaxis S., Kanellos F., et al. (2024). An Inertial-Based Wearable System for Monitoring Vital Signs during Sleep. Sensors, 24, 4139. DOI: 10.3390/s24134139,