Remote sensing of vital signs by medical radar time-series signal using cardiac peak extraction and adaptive peak detection algorithm: Performance validation on healthy adults and application to neonatal monitoring at an NICU
Keisuke Edanami,Masaki Kurosawa,Hoang Thi Yen,Takeru Kanazawa,Yoshifusa Abe,Tetsuo Kirimoto,Yu Yao,Takemi Matsui,Guanghao Sun
DOI: https://doi.org/10.1016/j.cmpb.2022.107163
Abstract:Background and objective: Continuous monitoring of vital signs plays a pivotal role in neonatal intensive care units (NICUs). In this paper, we present a system for monitoring fully non-contact medical radar-based vital signs to measure the respiratory rate (RR), heart rate (HR), I:E ratio, and heart rate variability (HRV). In addition, we evaluated its performance in a physiological laboratory and examined its adaptability in an NICU. Methods: A non-contact medical radar-based vital sign monitoring system that includes 24 GHz radar installed in an incubator was developed. To enable reliable monitoring, an advanced signal processing algorithm (i.e., a nonlinear filter to separate respiration and heartbeat signals from the output of radar), template matching to extract cardiac peaks, and an adaptive peak detection algorithm to estimate cardiac peaks in time-series were proposed and implemented in the system. Nine healthy subjects comprising five males and four females (24 ± 5 years) participated in the laboratory test. To evaluate the adaptability of the system in an NICU setting, we tested it with three hospitalized infants, including two neonates. Results: The results indicate strong agreement in healthy subjects between the non-contact system and reference contact devices for RR, HR, and inter-beat interval (IBI) measurement, with correlation coefficients of 0.83, 0.96, and 0.94, respectively. As anticipated, the template matching and adaptive peak detection algorithms outperformed the conventional approach. These showed a more accurate IBI close to the reference Bland-Altman analysis (proposed: bias of -3 ms, and 95% limits of agreement ranging from -73 to 67 ms; conventional: bias of -11 ms, and 95% limits of agreement ranging from -229 to 207 ms). Moreover, in the NICU clinical setting, the IBI correlation coefficient and 95% limit of agreement in the conventional method are 0.31 and 91 ms. The corresponding values obtained using the proposed method are 0.93 and 21 ms. Conclusion: The proposed system introduces a novel approach for NICU monitoring using a non-contact medical radar sensor. The signal processing method combining cardiac peak extraction algorithm with the adaptive peak detection algorithm shows high adaptability in detecting IBI the time series in various application settings.