Automated Heart Rate Detection in Seismocardiograms Using Electrocardiogram-Based Algorithms-A Feasibility Study

Abstract

In recent decades, much work has been implemented in heart rate (HR) analysis using electrocardiographic (ECG) signals. We propose that algorithms developed to calculate HR based on detected R-peaks using ECG can be applied to seismocardiographic (SCG) signals, as they utilize common knowledge regarding heart rhythm and its underlying physiology. We implemented the experimental framework with methods developed for ECG signal processing and peak detection to be applied and evaluated on SCGs. Furthermore, we assessed and chose the best from all combinations of 15 peak detection and 6 preprocessing methods from the literature on the CEBS dataset available on Physionet. We then collected experimental data in the lab experiment to measure the applicability of the best-selected technique to the real-world data; the abovementioned method showed high precision for signals recorded during sitting rest (HR difference between SCG and ECG: 0.12 ± 0.35 bpm) and a moderate precision for signals recorded with interfering physical activity-reading out a book loud (HR difference between SCG and ECG: 6.45 ± 3.01 bpm) when compared to the results derived from the state-of-the-art photoplethysmographic (PPG) methods described in the literature. The study shows that computationally simple preprocessing and peak detection techniques initially developed for ECG could be utilized as the basis for HR detection on SCG, although they can be further improved.

Keywords: accelerometers; electrocardiogram; heart rhythm; seismocardiography.

Figure 1

The experimental scheme of the measurement. Patch 1 (yellow) is for sternal accelerometers; patch 2 (green) is for apex cordis accelerometers. The other patches belong to the reference ECGs.

Figure 2

A block diagram of the measurement system, showing the primary hardware modules (such as sensor patches, ECG, and processing unit), along with the main firmware blocks within the preprocessing unit (PU).

Figure 3

The detection example for the hamilton2002 preprocessing and nabian2018 peak detection combination of methods. Gray curve—raw signal, orange curve—signal after being processed with the hamilton2002 pipeline, black curve in the bottom plot—ECG signal, red vertical lines—timestamps of R-peaks detected on ECG, blue vertical lines—J-peaks detected on SCG, the numbers next to blue vertical lines show the distance from the previous and to the next peak in ms. For the selected interval: precision = 100%, recall = 100%, F1-score = 1.00, HR: 61.9 bpm, n(j_peaks) = n(R_peaks) = 9. The accuracy calculation does not include peaks close to the plot’s border. Signal magnitudes are normalized for demonstrational purposes.

Figure 4

The detection results for the hamilton2002 preprocessing and nabian2018 peak detection combination on the experimental data. Gray curve—raw signal, orange curve—signal after being processed with hamilton2002 algorithm, black curve in the bottom plot—ECG signal, red vertical lines—timestamps of R-peaks detected on ECG, blue vertical lines—J-peaks detected on SCG, the numbers around blue vertical lines show the distance from the previous and to the next peak in ms. The patch with the best precision is shown with a blue frame. For the selected interval: precision = 100%, recall = 100%, F1-score = 1.00, HR: 81.2 bpm, n(j_peaks) = n(R_peaks) = 12. The accuracy calculation does not include peaks close to the plot’s border. Signal magnitudes are normalized for demonstration purposes.

Figure 5

Five-second samples from the middle of the recording for different subjects while resting and during reading interference. All measurements were acquired from Patch 1, axis Z. Gray curve—raw signal, orange curve—signal after being processed with hamilton2002 pipeline, red vertical lines—timestamps of the R-peaks detected on the ECG, blue vertical lines—the J-peaks detected on the SCG, the numbers next to blue vertical lines show the distance from the previous and to the next peak in ms.

Figure 6

Averaged plots were created based on detected anchor points (in this case, J-peaks detected on the signal from Patch 1, axis Z were taken as anchor points for all signals); subject 2, relaxed, t = 60 s. The prominent peaks are labeled with letters. The number next to the letter shows the distance between each peak and the anchor point. Gray lines show the individual trajectories of each heartbeat.