Modeling Consistent Dynamics of Cardiogenic Vibrations in Low-Dimensional Subspace

Area of Science:

• Biomedical Engineering
• Cardiovascular Physiology
• Signal Processing

Background:

• Seismocardiogram (SCG) signals reflect cardiovascular events but vary greatly between patients.
• Existing methods struggle with SCG's transient and patient-specific morphology.
• Population-level patterns in SCG signals offer potential for improved analysis.

Purpose of the Study:

• To develop a method for robust SCG signal processing that accounts for morphological variability.
• To harness population-level patterns in SCG signals for enhanced clinical relevance.
• To enable reliable SCG analysis in ambulatory and outpatient settings.

Main Methods:

• Introduced seismocardiogram generative factor encoding (SGFE) to model SCG waveforms.
• Modeled SCG as stochastic samples from a low-dimensional subspace defined by generative factors.
• Validated learned factors against known cardiovascular events (aortic opening/closing) during dynamic processes.

Main Results:

• Learned factors strongly correlated with aortic opening (AO) and closing (AC) during exercise-recovery.
• Consistent trajectories in the learned subspace were observed despite morphological differences.
• Algorithmic compensation for sensor misplacement was enabled, yielding high R² values for AO and AC mapping.

Conclusions:

• Consistent low-dimensional behavior of SCG signals confirms population-level patterns.
• SGFE demonstrates potential for time-domain-abstracted processing of SCG signals.
• This approach may improve the feasibility of SCG utilization in real-world clinical settings.