Heartbeat Cycle Length Detection by a Ballistocardiographic Sensor in Atrial Fibrillation and Sinus Rhythm

Abstract

Background: Heart rate monitoring is especially interesting in patients with atrial fibrillation (AF) and is routinely performed by ECG. A ballistocardiography (BCG) foil is an unobtrusive sensor for mechanical vibrations. We tested the correlation of heartbeat cycle length detection by a novel algorithm for a BCG foil to an ECG in AF and sinus rhythm (SR).

Methods: In 22 patients we obtained BCG and synchronized ECG recordings before and after cardioversion and examined the correlation between heartbeat characteristics.

Results: We analyzed a total of 4317 heartbeats during AF and 2445 during SR with a correlation between ECG and BCG during AF of r = 0.70 (95% CI 0.68-0.71, P < 0.0001) and r = 0.75 (95% CI 0.73-0.77, P < 0.0001) during SR. By adding a quality index, artifacts could be reduced and the correlation increased for AF to 0.76 (95% CI 0.74-0.77, P < 0.0001, n = 3468) and for SR to 0.85 (95% CI 0.83-0.86, P < 0.0001, n = 2176).

Conclusion: Heartbeat cycle length measurement by our novel algorithm for BCG foil is feasible during SR and AF, offering new possibilities of unobtrusive heart rate monitoring. This trial is registered with IRB registration number EK205/11. This trial is registered with clinical trials registration number NCT01779674.

Figures

Figure 1

Heartbeat measurement by a BCG foil: (a) the BCG sensor foil is positioned under the chest of the patient in a supine position; (b) mechanical contraction of the heart induces impedance change on the BCG sensor foil; (c) a BCG (blue signal) related signal is calculated and synchronized to an ECG (black signal).

Figure 2

Signal processing of BCG data: (a) the raw signal includes in its highest deflections inhalation and exhalation; (b) after time-domain filtering the breathing component is removed and repeating oscillations as a surrogate for the heart contraction are visible; (c) the local interval estimator defines the cycle length by beat-to-beat analysis (Figure 3); (d) additional calculations for the integral of the BCG complex and the maximal amplitude deflections are carried out; (e) the BCG signal is synchronized to the ECG.

Figure 3

Estimating the heartbeat cycle length: (a) the window of interest analyzes the signal for repeated amplitude patterns and estimates the cycle length; (b) a sliding window of interest performs basic threshold measurements and identifies two consecutive heartbeats; (c) three local interval estimators analyze the signals and each estimates a cycle length: the match between the three estimators is the quality index; (d) the window of interest moves forward and the estimated cycle length and quality index are displayed. *BCG: ballistocardiogram; CL: cycle length; QI: quality index.

Figure 4

ECG (black signal) synchronized BCG (blue signal), heartbeat count, ECG cycle length, and corresponding estimated BCG cycle length and quality index are simultaneously displayed. Beat 2 is a premature ventricular contraction resulting in a minor accordance of ECG and BCG cycle length. Also heartbeat 3 is affected by premature ventricular contraction; the following heartbeats show near-perfect accordance to the ECG cycle length with a high corresponding quality index. *BCG: ballistocardiogram; QI: quality index.

Figure 5

Correlation of analyzed cycle length of ECG and BCG in different quality index steps. Left AF, right SR (Bland-Altman Plot: y-axis: mean of difference ECG−BCG and 95% limits of agreement ±1.96 *SD): (a) all analyzed AF data; (b) all analyzed SR data; (c) AF data filtered by quality index >0.25; (d) SR data filtered by quality index >0.25; (e) AF data filtered by quality index >0.4; (f) SR data filtered by quality index >0.4.

Figure 6

Examples of cycle length detection by synchronized ECG and BCG with corresponding quality index. ECG (black signal) and synchronized BCG (blue signal). (a) AF shows a good correlation of the ECG and synchronized BCG interval detection. BCG#2 indexing a change in heartbeat cycle length resulting in an inaccurate BCG cycle length detection with corresponding decreased quality index. (b) SR after cardioversion with a near-perfect ECG and BCG cycle length correlation resulting in a high quality index for each heartbeat above 0.4. (c) SR after cardioversion with a premature atrial contraction (BCG#3). The corresponding quality index indicates a poor BCG quality for the premature contraction (BCG#3) and the following beat (BCG#4) due to a change in the deflection pattern and a good quality of BCG signal in normal SR. (d) SR after cardioversion with a good BCG cycle length detection interrupted by a high-energy artifact, most likely a moving artifact with a BCG interval marked as corrupt (BCG#3). The consecutive beats are all detected with an improving ECG and BCG cycle length correlation and an increasing quality index. Although the BCG pattern changed after the moving artifact no training phase was necessary for cycle length detection. *BCG: ballistocardiogram; QI: quality index.