Unveiling the temporal and spectral relationships between seismocardiogram signals, systolic time intervals and thorax characteristics

Abstract

Investigating the systolic time intervals (STIs) can provide information regarding cardiovascular health and autonomic nervous system. Among these STIs, the pre-ejection period (PEP) and left ventricular ejection time (LVET) have been found to be closely related to the ventricular failure and many other indicators. Recently, it has been shown that the seismocardiogram (SCG) signal can be used to derive hemodynamic parameters; thus, can be leveraged in the design of cardiovascular monitoring systems. Accordingly, in this work, (i) SCG-based PEP and LVET regression models leveraging both temporal and spectral analysis have been developed, (ii) the most distinctive features in STI estimation have been determined, and (iii) the relationship between the thorax impedance values and SCG-derived features has been investigated to quantity inter-subject variability. Overall, PEP and LVET could be estimated with low RMSE (7.436 and 11.022 ms, respectively) using extreme gradient boosting trees. Moreover, it was deduced that the thorax impedance characteristics indeed have high correlation with the SCG features, specifically in the 40 - 100 Hz range. Overall, these findings can potentially allow more useful and informed SCG-based STI analysis by providing valuable insights regarding SCG and thorax characteristics. © 2023 IEEE.