Automatic Identification of Systolic Time Intervals in Seismocardiogram
Executive Summary
This study presents an automated approach for identifying systolic time intervals (STI) in seismocardiogram (SCG) signals using a sliding template methodology. The method leverages ensemble averaging to suppress noise and improve peak detection accuracy, validated across supine and seated postures with over 5600 heartbeats. Results demonstrate robust performance in noisy conditions, with significant improvements over existing envelope-based methods, suggesting potential for wearable cardiac health monitoring systems.
Answer Machine Insights
Q: How does the proposed method compare to existing envelope-based approaches?
The proposed method demonstrated significantly smaller limits of agreement (LoA) in peak detection, indicating higher accuracy and robustness.
The LoA widths of the proposed approach with generalized-15 parameters in supine and seated trials are 16.1 ms and 42.9 ms respectively, which are significantly smaller than the existing approach.
Q: What is the impact of noise on the detection accuracy?
The proposed method maintained high detection accuracy under both pre-filtering and post-filtering Gaussian noise scenarios, outperforming the envelope-based method.
The proposed approach outperforms the envelope-based method, especially in the post-filtering noise scenarios as shown in Fig. 7(b) and (d).
Key Results
The proposed method achieved a Bland-Altman limit of agreement (LoA) of 16.1 ms for supine trials and 42.9 ms for seated trials, significantly outperforming the envelope-based method.
AC peak detection accuracy remained robust under Gaussian noise scenarios, with minimal degradation even at low signal-to-noise ratios (SNR).
Visual Evidence
Figure 2. Experimental Setup - postures and data acquisition. Accelerometer placed at lower sternum was considered for this study.
Clinical Snapshot
Evidence Rating
Relevance
high Priority