Preclinical Evaluation of a Noncontact Simultaneous Monitoring Method for Respiration and Carotid Pulsation Using Impulse-Radio Ultra-Wideband Radar

Figures

• 

  Figure 1. Simplified description of the experimental setting. IR-UWB radar and ECG measurements were obtained simultaneously from patients in a supine position breathing normally after a 5-minute rest. All patients were fully clothed and were asked not to move the upper part of their bodies during the measurements. The radar transmitter was placed at a 50-cm distance from the point 2 cm below the right mandibular angle and at a 45-degree angle with the bed on all three body planes, to aim at the right carotid bulb (red dot). Single lead ECG (lead II) was measured with 3 electrodes placed on the wrists and the right ankle.

• 

  Figure 2. Data processing flow. The radar sensor accumulated the received raw signals and generated a vital sign matrix for the measured period. Background data, also called clutter, were subtracted from the raw signal so that the target with respiratory motion could be detected. After ranging the target, the vital signal was obtained from the radar signal, including the respiratory and carotid pulse signals. The respiratory rate was estimated from the fast Fourier transform of the vital signal, and only the carotid pulse signal remained after passing the signal through a high-pass filter. From the carotid pulse signal, the peaks were identified and interbeat intervals were measured from the peak intervals.

• 

  Figure 3. Examples of signals measured using IR-UWB radar. The radar signals mainly consisted of cyclic oscillations generated by respiratory activity. Each time the subject moved, a large fluctuation occurred in the signal (A). In the power density spectrum, the signals from respiratory activity occupied the most prominent frequency, approximately between 0.25 Hz and 0.5 Hz, and the signals from carotid pulses occupied the second most prominent frequency, at approximately 0.89 Hz, as its harmonic component (B). The power of the respiratory activity signal was approximately 6.3 dB larger than that of the carotid pulsation signal.

• 

  Figure 4. Representative radar signal waveforms of respiratory activity and the carotid pulses. After the removal of clutter, the signal waveform was composed of large cyclic waves from respiratory activity and small but recognizable peaks from carotid pulsation (A). After high- and low-pass filters were applied, and the waveform was smoothed over 3 data points, the peak points on the carotid radar pulse waveform appeared to be well-correlated with R waves on ECG in both patients with NSR (B) and those with PeAF (C). The red circles at each peak in the radar waveform indicate the systolic phase of the carotid pulsation corresponding to each ECG R wave.

• 

  Figure 5. Agreement between the RRs measured by manual counting and those measured using the radar. RRs measured using radar were highly correlated with RRs measured by manual counting, and the BA plot shows no significant difference between the two RR measurement methods. †ICCR, intraclass correlation coefficient; SD, standard deviation.

• 

  Figure 6. Agreement between the radar and ECG in the NSR group. The PR (A), average R-R interval (B) and the individual R-R intervals (C) were measured in the NSR group using IR-UWB radar and ECG. There was excellent agreement between the two methods for both the PR and average R-R interval. The individual R-R intervals measured using radar also agreed well with those measured using ECG, although the agreement level was slightly lower than that for the PR and average R-R interval. †ECG, electrocardiography; ICCR, intraclass correlation coefficient.

• 

  Figure 7. Agreement between radar and ECG in the PeAF group. The PR (A), average R-R interval (B) and individual R-R intervals (C) were measured in the PeAF group using IR-UWB radar and ECG. Both the PR and average R-R interval showed strong agreement between radar and ECG. The level of agreement for the individual R-R intervals between the two methods was significantly lower than that for the PR and average R-R intervals but was still sufficiently high. †ECG, electrocardiography; ICCR, intraclass correlation coefficient.

• 

  Figure 8. Differentiation between NSR and PeAF by the analysis of the frequency domain. The frequencies of the radar signal intensities were color-coded and plotted against time on 2-dimensional spectrograms (A). The brighter-colored line represents the frequency at the peak signal intensity. The Max∆frq appeared to be greater in the patients with PeAF (1.1 Hz) than in those with NSR (0.1 Hz). In the receiver operating characteristic curve analysis, the Max∆frq showed very high diagnostic performance for PeAF at a threshold of 0.57 Hz (B). NSR, normal sinus rhythm; PeAF, persistent atrial fibrillation; Max, maximal; Min, minimal; frq, frequency; Max∆frq, maximum frequency variation in the peak signal intensity.