Seismocardiographic Signal Variability During Regular Breathing and Breath Hold in Healthy Adults

Abstract

Seismocardiographic signals (SCG) are known to correlate with mechanical cardiac activity and may be used for monitoring patients with cardiovascular disease. However, SCG variability is not well understood and may interfere with signal utility. In the current study, the SCG signals were acquired in 5 healthy subjects during regular breathing along with ECG and respiratory flow measurements. In addition, SCG waveforms were recorded during breath hold at end inspiration as well as end expiration. The SCG events were identified and segmented using ECG events. SCG waveforms during regular breathing were separated into two clusters using unsupervised machine learning. The variability was assessed for the clustered and un-clustered SCG by analyzing the Dynamic Time Warping (DTW) distances of SCG waveforms in the time domain. The inter-group variability between the normal breathing clusters and breath hold suggested that cluster 2 events were more similar to end expiration events while no clear trend was observed for cluster 1. The intra-group variability was reduced by approximately 19% for regular breathing clusters and 42% during breath hold compared to the unclustered SCG during normal breathing. The reduced variability during breath hold suggests the utility of SCG recording at breath hold since variability reduction can lead to more robust methods for longitudinal patient monitoring.