CN118161135A - Respiration rate estimation method based on non-contact electrocardiogram and heart rate variability - Google Patents

Abstract

The invention belongs to the technical field of health detection, and particularly relates to a respiration rate estimation method based on non-contact electrocardiogram and heart rate variability. The method comprises the steps of filtering and denoising non-contact electrocardiosignals; determining the position of R waves in the electrocardiograph by using two independent R wave detectors on the preprocessed electrocardiograph signals; calculating heart rate using the R-wave position; carrying out non-uniform cubic spline interpolation on the heart rate sequence to obtain an EDR waveform of the HRV; a respiration rate is estimated for the respiration signal. Compared with other respiration rate measuring methods, the method can be carried out on the basis of a convenient non-contact electrocardiogram, and the respiration rate is estimated by using the electrocardiogram, so that additional respiration measuring equipment is not needed, and the calculation complexity and the space complexity are low, thereby being beneficial to real-time estimation; the intelligent medical device can be used for constructing a non-inductive physiological monitoring environment and realizing the popularization of household application of portable intelligent medical treatment.

Description

A respiratory rate estimation method based on non-contact electrocardiogram and heart rate variability

Technical Field

The invention belongs to the technical field of health detection, and in particular relates to a respiratory rate estimation method.

Background technique

The electrocardiogram (ECG) is the most well-known tool in various biomedical applications, including measuring heart rate, detecting arrhythmias, diagnosing cardiac abnormalities, emotion recognition, biometrics, etc. ^[1] . With the continuous improvement of instrument performance and detection technology, more and more researchers have begun to study non-contact ECG signal detection methods. Breakthroughs in microelectronics technology and integrated circuit technology have promoted the development of portable, wearable, and non-contact ECG monitoring devices. Compared with ECG signals collected using the wet electrode method, non-contact ECGs usually use the principle of capacitive coupling to collect human surface potentials. This method has enhanced sensitivity ^[2] , but is also more sensitive to interference from environmental noise, the presence of motion artifacts, and even small movements of the electrodes during breathing. Although non-contact ECGs can clearly display characteristic waveforms such as P, QRS, and T waves, the fluctuations caused by baseline drift will also affect the analysis.

Respiration is a sign of human life and an important part of daily comprehensive monitoring. Respiration signals are usually recorded using techniques such as spirometry, pneumographs or volume pulsography. These techniques require the use of equipment that may interfere with natural breathing and are difficult to use in some situations, such as dynamic monitoring, stress testing and sleep studies ^[3] . In the application of electrocardiogram, based on the fact that respiratory activity affects ECG signals ^[4] , we can non-invasively extract respiratory signals from ECG signals, and then approximately reflect the pattern of respiratory signals, thereby providing sufficient respiratory-related information. ECG signal-derived respiration extraction method is a promising non-invasive respiratory activity monitoring method because respiration and ECG signals can be monitored simultaneously, and respiration can be obtained from available ECG sources without any other signals ^[3] . Compared with gas flow meters, respiratory straps and other methods, non-contact ECG signal monitoring is a non-sensing monitoring environment that can reduce interference with patients' natural breathing and improve comfort.

Based on the influence of respiration on the recorded ECG signals, many signal processing techniques have been developed to extract respiratory information. ECG-derived waveforms similar to recorded respiration are called ECG-derived respiration (EDR) signals. In the study of EDR, it can be divided into two types based on the number of signal leads: multi-lead and single-lead. The method based on the oscillation pattern of the average electrical axis rotation angle of the heart caused by the respiratory cycle is a multi-lead algorithm that uses vector cardiogram (VCG) signals ^[5,6] to analyze the offset of the ECG vector to extract EDR. However, non-contact ECG signals are not "standard lead" signals and may not provide information related to the ECG vector. Moreover, in multi-lead ECG monitoring systems, the use of multi-lead ECG may sacrifice patient convenience to obtain a more adequate EDR. The EDR time series method based on ECG morphology changes is a type of single-lead algorithm that generates EDR by sampling respiratory-related features hidden in the ECG signal, such as R amplitude ^[7,8] , RS amplitude ^[9] , QRS area ^[8,10] and QRS slope ^[4] . A disadvantage of these methods is that aliasing may occur when the ratio of heart rate to respiratory rate is lower than 2.

Summary of the invention

The purpose of the present invention is to provide a method for estimating respiratory rate based on a non-contact electrocardiogram and heart rate variability based on the principle of capacitive coupling, so as to provide an estimation of the user's respiratory rate without relying on the use of additional equipment while non-contact electrocardiogram signal monitoring is being performed, which is particularly important during sleep.

The respiratory rate estimation method based on non-contact electrocardiogram and heart rate variability provided by the present invention comprises the following specific steps:

Step 1: Filter and denoise the non-contact ECG signal;

Step 2: using two independent R wave detectors to determine the position of the R wave in the ECG beat on the preprocessed ECG signal;

Step 3: Calculate heart rate using the R wave position;

Step 4: Perform non-uniform cubic spline interpolation on the heart rate sequence to obtain the EDR waveform of HRV;

Step 5: Estimate the breathing rate from the breathing signal.

In step 1, a 0.5-30 Hz bandpass filter is used to process the ECG signal. At the same time, the wavelet threshold denoising method is used to remove the low-frequency information below 0.5 Hz from the signal. The threshold function uses the unbiased risk estimation criterion, the wavelet base uses ‘db6’, and the decomposition level L=9.

In step 2, two independent stable R wave detectors are used to determine the position of the R wave in the electrocardiogram. Specifically, let X = [x ₁ , x ₂ , ..., x _i ] be the sequence obtained by R wave detector 1, and let Y = [y ₁ , y ₂ , ..., y _j ] be the sequence obtained by R wave detector 2 (converted into seconds). If:

|x _i -y _j |≤20ms, (1)

The R wave obtained by this test is considered correct, and x _i /y _j is inserted into the R wave sequence.

In step 3, the instantaneous heart rate is obtained by differentiating the R wave sequence in step 2. The formula for calculating the instantaneous heart rate is:

Among them, R_R_Interval is the R-R interval,

For the heart rate sequence, if the next instantaneous heart rate is greater than 1.5 times the current heart rate, it is regarded as an abnormal point and removed from the sequence.

In step 4, in order to obtain the original breathing-like waveform, the obtained heart rate values are interpolated using the cubic spline interpolation method to obtain the EDR waveform of HRV. It is worth noting that the position information of the R wave needs to be used for non-uniform interpolation here to obtain the position corresponding to the EDR and the reference signal for calculating the performance.

In step 5, the respiratory rate is estimated using the extreme value counting method, specifically:

(1) The respiratory signal is subjected to a bandpass filter with a frequency band of 0.1 to 0.5 Hz, and the filter is a 10th-order Butterworth filter;

(2) Find the maximum and minimum values of the filter curve; and take the third quartile Q ₃ (75th percentile) of all maximum values to suppress the influence of ultra-large breathing, and then use 0.2×Q ₃ as the threshold level;

(3) A breathing cycle is considered to start and end at a maximum above a threshold level; if the signal contains exactly one minimum below 0 and no other extremes, then the portion of the signal between two such maxima above 0.2 × Q ₃ is interpreted as a valid breathing cycle;

(4) The average length of all detected respiratory cycles is interpreted as the inverse of the respiratory frequency.

The present invention also includes a respiratory rate estimation system based on non-contact electrocardiogram and heart rate variability based on the above respiratory rate estimation method. The system includes five modules, namely: an electrocardiogram signal filtering and denoising processing module, an R wave position determination module in an electrocardiogram beat, a heart rate calculation module, a non-uniform cubic spline interpolation module, and a respiratory rate estimation module. These five modules respectively perform the operation functions of the five steps in the respiratory rate estimation method.

Compared with the prior art, the present invention has the following beneficial effects:

Compared with other respiratory rate measurement methods, the method of the present invention can be performed on the basis of a convenient non-contact electrocardiogram, and the respiratory rate can be estimated by electrocardiogram without the need for additional respiratory measurement equipment, which is more conducive to building a non-sensing physiological monitoring environment, thereby realizing the popularization of portable smart medical home applications. In addition, the method has low computational complexity and spatial complexity, which is conducive to real-time estimation.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a flow chart of the present invention.

FIG2 is a block diagram showing the steps and results of the EDR extraction experiment.

FIG3 is a waveform diagram of the signal after filtering and denoising.

FIG4 is a scatter plot of the instantaneous heart rate of the subjects.

FIG5 shows a diagram of a non-contact ECG signal, a reference respiration signal, and an EDR signal in the left column and a diagram of a reference ECG signal, a reference respiration signal, and an EDR signal in the right column.

FIG6 is a graph showing the changes in breathing rate over time for subjects in different lying positions.

Detailed ways

The present invention is further described below by way of examples in conjunction with the accompanying drawings. The specific examples described are only used to explain the present invention and are not intended to limit the present invention.

Example:

The present invention takes a breathing induction experiment as an example. The experiment recruited volunteers with no history of heart disease and respiratory disease. During the experiment, the volunteers adjusted their breathing rhythm according to the video (and audio) prompts. The rhythm and time were free breathing for 6 minutes, fixed-frequency breathing for 5 times for a total of 2 minutes (10, 12, 15, 20, 30c/min), deep breathing for 1 minute (7c/min), and irregular breathing for 1.5 minutes, a total of about 11 minutes. The volunteers collected signals in four lying positions: supine, left side, right side, and prone. The collected signals included non-contact ECG signals and reference signals for evaluating the respiratory rate estimation results (reference ECG signals and respiratory signals collected by the wet electrode method). The data analysis steps and process are shown in the figure:

The process of the respiratory rate estimation method based on non-contact ECG signal of the present invention is shown in FIG1 , and the following processing is performed on the data of different lying postures respectively:

Step 1: Perform 0.5-30Hz bandpass filtering and wavelet threshold denoising on the non-contact ECG signal and the reference ECG signal; at the same time, perform 0.1-2Hz bandpass filtering on the reference respiration signal and perform 256-point (sampling rate 256Hz) median smoothing filtering. The results of the non-contact ECG signal, the reference respiration signal and the reference ECG signal after filtering and denoising are shown in Figure 3.

Step 2: Use two independent R wave detectors to determine the position of the R wave in the ECG beat for the preprocessed ECG signal. The Pan-Tompkins method and the two-average method are used here to determine the "correctly matched" R wave position sequence. The R wave position is marked as shown in Figure 3, with '*' and 'o' marked in the non-contact ECG signal and the reference ECG signal, respectively.

Step 3: Calculate the instantaneous heart rate using the R wave position. If the next instantaneous heart rate is greater than 1.5 times the current heart rate, it is considered an abnormal point and removed from the sequence. Figure 4 shows the instantaneous heart rate correlation diagram of the non-contact ECG signal and the reference ECG signal, as well as the scatter plot of the current heart rate and the next heart rate of the two signals.

Step 4: Perform cubic spline interpolation on the heart rate sequence. The interpolated EDR signal is shown in Figure 5.

Step 5: Estimate the respiratory rate from the respiratory signal. Before using the extreme value estimation method, the signal needs to be removed from the mean.

Figure 6 shows the results of the experimental monitoring of the subjects, and the changes in the calculated respiratory rate over time under different sleeping positions. The respiratory rate calculated by the EDR signal can basically fit the respiratory rate calculated by the reference respiratory signal. Compared with the free breathing state of about 6 minutes before, the respiratory pattern obtained by EDR is more stable in the induced respiratory rate adjustment stage, and the trend is basically consistent with the results calculated by the reference respiratory signal and the guided paradigm.

references:

[1] KAPLAN BERKAYA S U A K, SORA GUNAL E, ET AL. A survey on ECG analysis[J]. Biomedical Signal Processing and Control, 2018, 43: 19.

[2] L.LEICHT E S, C.KNACKSTEDT. Capacitive ECG monitoring in cardiacpatients during simulated driving[J].IEEE Trans Biomed Eng, 2019, 66(3):9.

[3] R.BAILóN L S, P.LAGUNA.ECG-derived Respiratory Frequency Estimation[J].2006.

[4]JESU′S LA′ZARO A A,DANIEL ROMERO,EDUARDO GIL,PABLO LAGUNA,ESTHERPUEYO,AND RAQUEL BAILO′N.Electrocardiogram Derived Respiratory Rate from QRSSlopes and R-Wave Angle[J].2014.

[5] S.LEANDERSON PL,L. Estimation of the respiratory frequencyusing spatial information in the VCG[J].Med Eng Phys 2003,25(6):7.

[6] R.BAILóN L S, P.LAGUNA. A robust method for ecg-based estimation of the respiratory frequency during stress testing[J].IEEE Trans Biomed Eng 2006, 53(7): 13.

[7] S. SARKAR S B A S P. Extraction of respiration signal from ECG for respiratory rate estimation[J]. Michael Faraday IET International Summit 2015, 2015.

[8]M.SCHMIDT AS,J.MULLER,K.-J.BAR,G.ROSE.Ecg derived respiration: comparison of time-domain approaches and application to altered breathing patterns of patients with schizophrenia[J].Physiol Meas,2017,38(4).

[9]D.DOBREV I D.Two-electrode telemetric instrument for infant heartrate and apnea monitoring[J].Med Eng Phys,1999,20(10):6.

[10]P.H.CHARLTON T B,L.TARASSENKO,D.A.CLIFTON,R.BEALE,P.J.WATKINSON.An assessment of algorithms to estimate respiratory rate from the electrocardiogram and photoplethysmogram[J].Physiol Meas,2016,37(4).

Claims (7)

1. A respiration rate estimation method based on non-contact electrocardiogram and heart rate variability is characterized by comprising the following specific steps:

step 1: filtering and denoising the non-contact electrocardiosignal;

Step 2: determining the position of R waves in the electrocardiograph by using two independent R wave detectors on the preprocessed electrocardiograph signals;

Step 3: calculating heart rate using the R-wave position;

step 4: carrying out non-uniform cubic spline interpolation on the heart rate sequence to obtain an EDR waveform of the HRV;

step 5: a respiration rate is estimated for the respiration signal.

2. The method according to claim 1, wherein the filtering and denoising the non-contact electrocardiosignal in step 1 comprises processing the electrocardiosignal with a 0.5-30 Hz band-pass filter; meanwhile, a wavelet threshold denoising method is used for removing low-frequency information below 0.5Hz from the signal, a threshold function selects an unbiased risk estimation criterion, a wavelet base selects 'db6', and a decomposition level L=9.

3. The respiratory rate estimation method according to claim 2, wherein in step 2, the position of the R wave in the electrocardiograph is determined by using two independent R wave detectors with stable performance, specifically, let x= [ X ₁,x₂,...,x_i ] be the sequence obtained by the R wave detector 1, let y= [ Y ₁,y₂,...,y_j ] be the sequence obtained by the R wave detector 2, and converted into the unit of seconds, if:

|x_i-y_j|≤20ms， (1)

Then the detected R-wave is considered correct and x _i/y_j is inserted into the R-wave sequence.

4. The respiratory rate estimation method according to claim 3, wherein the calculating the heart rate using the R-wave position in step 3 is obtained by the R-wave sequence difference in step 2, and the instantaneous heart rate calculation formula is:

Wherein R_R_Interval is R-R Interval;

for a heart rate sequence, if the next instantaneous heart rate is greater than 1.5 times the current heart rate, it is considered an outlier and is removed from the sequence.

5. The respiratory rate estimation method according to claim 4, wherein the non-uniform interpolation is performed using the position information of the R wave in step 5 to obtain the positions of EDR corresponding to the reference signal for calculating the performance.

6. The method according to claim 5, wherein the respiratory rate estimation in step 5 uses extremum counting, specifically:

(1) The breathing signal is subjected to band-pass filtering with the frequency band of 0.1-0.5 Hz, and the filter is a Butterworth filter with the order of 10;

(2) Solving the maximum value and the minimum value of the filtering curve; and taking the 3 rd quartile Q ₃, the 75 th percentile, of all maxima to suppress the effects of excessive respiration, then taking 0.2×q ₃ as the threshold level;

(3) The respiratory cycle is considered to begin and end at a maximum above a threshold level; if the signal happens to contain a minimum below 0 and no other extremum, then a portion of the signal between two such maxima above 0.2 XQ ₃ is interpreted as an effective respiratory cycle;

(4) The average length of all detected respiratory cycles is interpreted as the inverse of the respiratory frequency.

7. Respiratory rate estimation system based on the respiratory rate estimation method according to one of the claims 1-6, characterized in that it comprises 5 modules, respectively: the device comprises an electrocardiosignal filtering and denoising processing module, an R wave position determining module in an electrocardio-beat, a heart rate calculating module, a non-uniform cubic spline interpolation module and a respiration rate estimating module; the 5 modules sequentially execute the operation functions of the steps 1-5 in the respiratory rate estimation method.