Using electromechanical signals recorded from the body for respiratory phase detection and respiratory time estimation: A comparative study

Electrocardiogram derived respiratory (EDR) is a non-invasive technique to estimate respiratory signal. As an alternative, recent studies suggest using Seismocardiogram to estimate respiratory signal, so called accelerometer derived respiration (ADR). In this study we compared the performance of ADR and EDR in precise detection of respiration phases (inhale and exhale timing). We also compared time lag between breath cycles extracted by use of ADR and EDR, and ground truth (respiratory signal recorded using chest band strain gauge). For this comparative analysis, Seismocardiogram, Single lead electrocardiogram (Lead II), and respiratory signal (using a chest band strain gauge) were recorded from 19 healthy subjects. Principal component analysis (PCA) and envelope detection methods are used to compute EDR and ADR. Initial results show that ADR in z-direction (back to front) seems promising approach in addition to the EDR with accuracy of above 85% in identifying respiration phases (inhale and exhale). 87% of breath cycles extracted from ADR had acceptable time lag compared to ground truth (respiratory signal recorded using a chest band strain gauge). ADR was able to correctly classify heartbeats to inhale and exhale classes with classification accuracy of around 76%.