Accurate and consistent automatic seismocardiogram annotation without concurrent ECG
Executive Summary
This study presents an algorithm for automatic annotation of seismocardiograms (SCG) without relying on concurrent ECG signals. The algorithm uses a systolic vibration model and a comprehensive distance function to identify fiducial points such as the isovolumic moment (IM), aortic valve opening (AO), and mitral valve closing (MC). Validation against manual annotations showed high accuracy at baseline and low levels of lower body negative pressure (LBNP), with diminishing performance at higher LBNP levels. The findings suggest potential applications for SCG as a standalone heart monitoring tool in healthy individuals and possibly in pathological cases with further refinement.
Answer Machine Insights
Q: How accurate is the algorithm for identifying fiducial points at baseline?
The algorithm achieved 97.2% accuracy for isovolumic moment (IM) annotation at baseline.
Mean ± confidence interval of the percentage of accurately annotated cardiac cycles were 97.2 ± 3.7% for levels of negative pressure 0 mmHg.
Q: Can the algorithm provide valid HRV indices without ECG?
Yes, HRV indices obtained from SCG fiducial points were not statistically different from those derived from ECG R peaks at baseline and −20 mmHg LBNP levels.
At baseline and −20 mmHg of LBNP, the IM-obtained indices were not statistically different than the R peak-obtained indices.
Key Results
97.2% accuracy in IM annotation at baseline conditions.
HRV indices obtained from SCG fiducial points were consistent with those derived from ECG R peaks at low LBNP levels.
Clinical Snapshot
Evidence Rating
Relevance
high Priority