Quantifying and Reducing Motion Artifacts in Wearable Seismocardiogram Measurements During Walking to Assess Left Ventricular Health
Executive Summary
This study introduces a framework for reducing motion artifacts in wearable seismocardiogram (SCG) measurements during walking using empirical mode decomposition (EMD). The methodology enables accurate estimation of pre-ejection period (PEP) from SCG signals, validated against impedance cardiogram (ICG) data, with significant improvements in signal-to-noise ratio (SNR). Clinical implications include enabling hemodynamic assessments during activity, such as the 6-minute walk test, which could provide deeper insights into cardiovascular health beyond stationary measurements.
Answer Machine Insights
Q: How does walking speed affect SCG signal quality?
Walking at higher speeds increases motion artifacts, requiring more heartbeats for accurate PEP estimation.
Walking at higher speeds requires a greater number of heartbeats. Hence, it can be concluded that some measure of intensity of the activity (speed of walking in this instance) can be incorporated into algorithms for better estimation of parameters.
Q: What is the minimum ensemble size for accurate PEP estimation during walking?
On average, 11 heartbeats are required for normal pace walking, increasing with walking speed.
The table shows that, if an RMSE of 3 ms can be tolerated in PEP estimation, then on average 11 heartbeats are required while the person is walking at normal pace.
Key Results
EMD-based de-noising reduced warping distance significantly, improving SNR during walking (e.g., dα: M=7.74, SD=5.54 vs. dβ: M=0.84, SD=1.11; p≪0.001).
Normalized PEP values from SCG correlated strongly with ICG during normal walking (r=0.86, p<0.01).
Visual Evidence
Fig. 3. (a) De-noising of the D-V heartbeat during walking using EMD algorithm. Four to five imfs (Im, where m = 1, 2, ...,) were generated after the application of EMD. The first imf (I1) was chosen as the de-noised D-V heartbeat for feature extraction. (b) Noisy heartbeats for walking at different speeds. (c) The first imf obtained after application of EMD to the noisy heartbeats in (b). The basic shape and characteristics of the SCG waveform are reconstructed accurately.
Clinical Snapshot
Evidence Rating
Relevance
high Priority