A Hidden Markov Model for Seismocardiography
Executive Summary
This study introduces a Hidden Markov Model (HMM) for processing seismocardiograms (SCGs) to estimate heart rate, heart rate variability (HRV), and cardiac time intervals. Using the Baum-Welch algorithm for parameter learning and the Viterbi algorithm for state estimation, the method achieves superior accuracy compared to envelope and spectral-based approaches, with mean absolute errors of 5 ms for cardiac time intervals. The approach is adaptable to low-cost inertial sensors and does not require additional modalities, making it suitable for at-home medical services.
Answer Machine Insights
Q: What is the main advantage of the proposed HMM method for SCG processing?
The HMM method explicitly models sensor noise and beat-to-beat variations, enabling accurate estimation of heart rate, HRV, and cardiac time intervals without requiring additional sensor modalities.
The proposed algorithm... explicitly models sensor noise and beat-to-beat variations (both in amplitude and temporal scaling) in the seismocardiogram morphology.
Q: How does the HMM method compare to envelope and spectral-based methods?
The HMM method outperforms envelope and spectral-based methods in heart rate and cardiac time interval estimation, achieving lower mean absolute errors.
The proposed method is shown to outperform the current state-of-the-art among envelope and spectral-based methods for heart rate estimation.
Key Results
Mean absolute error for cardiac time intervals: 5 ms (IVCT) and 9 ms (LVET).
Heart rate variability measures achieved normalized mean absolute errors below 1% for SDNN and RMSSD.
Clinical Snapshot
Evidence Rating
Relevance
high Priority