Resting heart rate calculation method based on wearable device data

Abstract

The present application proposes a method for calculating resting heart rate based on wearable device data, which includes: a data preprocessing step: an acceleration resting judgment step: using the difference of three-axis acceleration signals to judge whether the electrocardiogram signal corresponding to the three-axis acceleration signal is in a resting state and determine the preliminary resting segment of the electrocardiogram signal, and screening out the resting segment based on the acceleration signal based on the length and interval of the preliminary resting segment; a heart rate recovery judgment step: obtaining the second-by-second heart rate trend from the electrocardiogram signal; calculating the inflection point of the heart rate trend decline and rise, which is the heart rate recovery point; a resting segment heart rate calculation step: the heart rate recovery point to the end point of the resting segment based on the acceleration signal is the resting segment of the acceleration combined heart rate; calculating the resting heart rate of the resting segment of the acceleration combined heart rate; a resting heart rate aggregation step: the resting heart rates of all the resting segments of the acceleration combined heart rate in a day are aggregated using the median to obtain the resting heart rate of the day.

Description

Technical Field

[0001] The present application relates to a method for calculating resting heart rate. Background Art

[0002] Clinical resting heart rate guidelines stipulate that after the subject has rested for 5 minutes, the heart rate should be counted by palpating the peripheral radial artery in a sitting position to obtain the resting heart rate, or the resting heart rate should be measured using a 12-lead electrocardiograph in a lying position.

[0003] One method for calculating resting heart rate is to determine the number of steps within a certain time window. Specifically, if there are less than 5 steps within a 10-minute time window, the heart rate recorded in the window is used as the resting heart rate.

[0004] To avoid confusion with sleep cycle HR, this method deletes data from the night time (10 pm to 6 am).

[0005] In this method, the timestamp information of heart rate measurement and step measurement does not match. Heart rate is recorded once per second, and step count is recorded once per minute. In order to obtain the closest heart rate measurement in time for each step measurement, this method only considers the heart rate measurement value within one minute after the step measurement. This problem of inconsistent time resolution will exclude some useful heart rate segments. For example, if you walk 10 steps in the first 10 seconds of a minute, and the next 50 seconds are in a resting state, due to the time resolution problem of step measurement, it is not possible to consider the inclusion of synchronized heart rate from the last 50 seconds. Instead, the ECG signals of the next minute are discarded and not included in the entire 10-minute window. In addition, the number of steps is generally judged by body movement greater than a certain threshold, and micro-movement in the absence of steps will also affect the heart rate; even if there are no steps for a 10-minute period, it does not mean that the entire period can be included as a resting heart rate segment, such as the 10 minutes of rest after strenuous exercise. This method cannot exclude this situation. Summary of the invention

[0006] The present invention aims to provide a resting heart rate calculation method based on wearable device data, which uses continuous electrocardiogram and three-axis acceleration signals collected by the wearable device to estimate the resting heart rate of the wearer in a daily natural state, without requiring the wearer to obtain the resting heart rate according to a standard procedure.

[0007] The resting heart rate calculation method based on wearable device data of the present application includes:

[0008] Data preprocessing step: preprocessing the ECG signal and acceleration signal obtained from the wearable device; wherein the acceleration signal is a three-axis acceleration signal;

[0009] Acceleration rest judgment step: for each point of the three-axis acceleration signal, calculate the absolute value of the difference sum of the three-axis acceleration signal. If the absolute value is less than the first threshold, the state of the point is recorded as 0, otherwise the state of the point is recorded as the absolute value of the difference sum of the three-axis acceleration signal of the point; then calculate the sum of the states of each point of the three-axis acceleration signal within each second, if the sum of the states is less than the second threshold, the states of the three-axis acceleration signal of the second are all recorded as 0, otherwise they are all recorded as 1; calculate the time interval of the segments of the three-axis acceleration signal whose states are all recorded as 0, and the states of the three-axis acceleration signal between the time periods whose time intervals are less than the third threshold are all recorded as 0, that is, based on the rest segment of the acceleration signal;

[0010] Heart rate recovery judgment steps: obtain the second-by-second heart rate trend from the ECG signal; calculate the inflection point of the heart rate trend in the time period corresponding to the resting segment based on the acceleration signal, and the inflection point is the heart rate recovery point;

[0011] The heart rate calculation steps for the resting segment are as follows: the period from the heart rate recovery point to the end point of the resting segment based on the acceleration signal is the resting segment of the acceleration combined with the heart rate; the resting heart rate of the resting segment of the acceleration combined with the heart rate is calculated;

[0012] Resting heart rate aggregation step: Aggregate the resting heart rate of all acceleration and heart rate resting segments in a day using the median to obtain the resting heart rate for the day.

[0013] Preferably, in the data preprocessing step,

[0014] The EP Limited heart rate detection algorithm is used to detect the heartbeat position of the ECG signal to obtain the RR interval sequence of the ECG signal, and the heart rate sequence is obtained from the RR interval sequence; the EP Limited heart rate detection algorithm first filters the ECG signal, and then divides each signal peak into QRS wave or noise according to the detection rules; the three-axis acceleration signal is filtered.

[0015] Preferably, during the determination of the preliminary resting segment of the ECG signal:

[0016] For each point of the three-axis acceleration signal, the absolute value of the differential sum of the three-axis acceleration signals is calculated. If the absolute value is less than the first threshold, the state of the point is recorded as 0, otherwise the state of the point is recorded as the absolute value of the differential sum of the three-axis acceleration signals of the point; then the sum of the states of each point of the three-axis acceleration signal within each second is calculated. If the sum of the states is less than the second threshold, the states of the three-axis acceleration signals of that second are all recorded as 0, otherwise they are all recorded as 1; the time interval is calculated for the segments of the three-axis acceleration signal whose states are all recorded as 0, and the states of the three-axis acceleration signals between the time periods whose time intervals are less than the third threshold are all recorded as 0; the segment with a state of 0 is the initial resting segment.

[0017] Preferably, in the heart rate recovery determination step:

[0018] Outliers are removed from the heart rate sequence in the 2-minute window, missing values ​​are interpolated using the median in the window, and then the sliding average method is used to perform sliding average on the entire heart rate sequence to obtain the heart rate trend;

[0019] The heart rate is interpolated to synchronize with the three-axis acceleration signal, and the wave-by-wave heart rate is interpolated to the second-by-second heart rate using linear interpolation.

[0020] Preferably, in the step of calculating the heart rate in the resting segment:

[0021] For each resting heart rate segment, use the formula (60*f1) / rr to calculate the resting heart rate in turn;

[0022] Where: f1 is the sampling frequency of the ECG signal; rr is the RR interval of the ECG signal.

[0023] Preferably, for the above-mentioned situation where the resting segments are of different lengths, in order to make full use of the resting segment information, the resting segments are first divided, and those less than 30 seconds are directly calculated, and those longer than 30 seconds are first divided into multiple segments, and the resting heart rate of each segment is taken as the median of the segment.

[0024] The method of the present application starts from the original ECG and acceleration signals and combines heart rate and body movement data to overcome the problems of misjudgment of resting heart rate after exercise, heart rate fluctuations caused by slight exercise, and the inability to fully utilize the signal due to the time resolution of step judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A flowchart of a method for calculating resting heart rate based on wearable device data of the present application;

[0026] Figure 2 is the original signal graph of the subject;

[0027] Figure 3 This is the original signal detection diagram;

[0028] Figure 4 is a resting segment captured by the method of the present application;

[0029] Figure 5 The resting heart rate calculated by the method of the present application is compared with the clinical resting heart rate in sitting and lying positions;

[0030] Figure 6 This is the Bland-Altman plot of the clinical lying resting heart rate and the wearable device resting heart rate. DETAILED DESCRIPTION

[0031] Below, the resting heart rate calculation method based on wearable device data of the present application is described in detail with reference to the accompanying drawings.

[0032] Wearable device: This embodiment uses a medical-grade vest-type wearable accompanying monitoring device independently developed by the team, which can collect the subject's ECG, respiration, and three-axis acceleration information. This method uses ECG and three-axis acceleration data, where ECG data is single-lead ECG data with a sampling frequency of f1 = 200Hz, and three-axis acceleration data includes motion signals in three directions with a sampling frequency of f2 = 25Hz;

[0033] Data collection: This example collected data from 105 healthy subjects. The subjects wore the device to measure their resting heart rate in sitting and lying positions after a 5-minute rest in the standard process. The resting heart rate was measured by counting the heart rate through touching the peripheral radial artery. The subjects engaged in normal activities at other times, and the wearable data of the subjects was collected 24 hours a day.

[0034] Data preprocessing: The EP Limited heart rate detection algorithm is used to detect the heartbeat position of the ECG signal to obtain the RR interval sequence of the ECG signal. The heart rate detection algorithm first filters the original heartbeat signal, and then divides each signal peak into QRS wave or noise according to the detection threshold, peak height, position, maximum derivative value and other detection rules; the three-axis acceleration signal is filtered; the sleep segment is removed according to the recorded sleep time to avoid confusion between the sleeping heart rate and the resting heart rate.

[0035] Acceleration rest judgment: calculate the absolute value of the sum of the three-axis signal differentials, i.e. |ΔX + ΔY + ΔZ|, set the three-axis differential sum less than the threshold Ath1=15 to be recorded as 0, then calculate the three-axis differential sum within 1s, i.e. sum each f2 = 25 points, record it as 0 if the three-axis differential sum is less than Ath2=20, and record the others as 1, calculate the time interval for all segments recorded as 0, and convert the states between intervals less than Tth1=3 to 0, 0 is the rest segment judged based on acceleration, 1 is the activity segment judged based on acceleration;

[0036] Heart rate recovery resting judgment: First, calculate the heart rate trend, remove outliers in the heart rate in the 2-minute window in turn, use the median in the window to interpolate the missing values, and then use the sliding average method to perform sliding average on the entire heart rate segment to obtain the heart rate trend; secondly, interpolate the heart rate to synchronize with the acceleration per second judgment resting segment, and use linear interpolation to interpolate the wave-by-wave heart rate to the second-by-second heart rate; finally, calculate the heart rate recovery point, and calculate the heart rate trend decline and rise inflection point, i.e., the heart rate recovery point, in the acceleration judgment resting segment corresponding to the heart rate segment at the same time, define the rising threshold Tth2=10, and select the heart rate recovery point to the end of the acceleration resting segment as the acceleration combined with the heart rate resting judgment;

[0037] Resting heart rate calculation: The resting heart rate segments judged in the whole heart rate signal are calculated using the formula (60*f1) / rr. In the case where the resting segments judged above are of different lengths, in order to make full use of the resting segment information, the resting segments are first divided. If the length is less than 30s, it is directly calculated. If the length is longer than 30s, it is first divided into multiple segments. In this way, the information of the long resting segments can be fully utilized. Instead of generating only one resting heart rate value for each segment, the resting heart rate of each segment is the median of the segment.

[0038] Resting heart rate aggregation: The resting heart rates obtained from all resting heart rate segments are aggregated again using the median median(hr1, hr2, …, hri) to obtain the resting heart rate for the day;

[0039] The specific results are as follows:

[0040] The original signal of a subject, such as Figure 2 As shown, the first line is the ECG original signal, and the second line is the three-axis acceleration original signal; Figure 3 Shown Figure 2 The first line of the signal is the ECG signal detection signal; Figure 4 Shown are the resting heart rate segments selected by the method of the present application. The first row shows the original ECG and detection status; the second row shows the heart rate trend per second; the third row shows the acceleration signal and resting segment judgment; the fourth row shows the resting heart rate trend change; and the fifth row shows the resting judgment of three-axis acceleration.

[0041] The method of this application calculates the wearable resting heart rate and the clinical sitting and lying resting heart rate distribution. Figure 5 , it can be seen that the median resting heart rate calculated by the wearable device is lower than that in sitting position but higher than that in lying position.

[0042] The method of the present application calculates the resting heart rate value and compares it with the clinical lying resting heart rate using the Blant-Altman plot. Figure 6 The Bland-Altman plot of resting heart rate calculated based on wearable device data and resting heart rate in lying position is shown, where the difference in resting heart rate mostly falls within the 95% confidence interval, with a proportion of 95.24%.

[0043] The Bland-Altman results showed that the difference between the resting heart rate calculated based on this method and the average resting heart rate in the clinical lying position was 2.25; as the heart rate increased or decreased, the difference between the two did not show a fixed trend, and the degree of variation did not increase or decrease with the change of heart rate. The two measurement methods had good consistency.

[0044] In the method of the present application, the use of the original acceleration signal to judge body movement rather than the number of steps can eliminate the heart rate changes caused by some micro-movements; in the method of the present application, the method of combining the acceleration signal with the heart rate is used to judge the resting segment and thus calculate the resting heart rate. Compared with the common method based only on the number of steps, it can more accurately eliminate the situation of heart rate recovery after exercise, especially for people with high activity levels, training or fitness; finally, after experiments, the method of mid-range statistical aggregation is used to define the resting heart rate for a certain period of time. Compared with the instability of a single measurement, such as measuring after getting up, as the duration increases, using multiple unequal time window heart rate assessments can be closer to the human resting heart rate.

[0045] Unless otherwise defined, all technical and / or scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art to which the invention relates. The materials, methods and examples mentioned in this application are illustrative only and not limiting.

[0046] Although the present invention has been described in conjunction with specific embodiments, within the spirit of the invention of this application, those skilled in the art may make appropriate substitutions, modifications and changes, and such substitutions, modifications and changes still fall within the scope of protection of this application.

Claims

1. A method for calculating resting heart rate based on wearable device data, comprising: Data preprocessing step: preprocess the ECG signal and acceleration signal obtained from the wearable device; The acceleration signal is a three-axis acceleration signal; Acceleration rest judgment step: for each point of the three-axis acceleration signal, calculate the absolute value of the difference sum of the three-axis acceleration signal. If the absolute value is less than the first threshold, the state of the point is recorded as 0, otherwise the state of the point is recorded as the absolute value of the difference sum of the three-axis acceleration signal of the point; then calculate the sum of the states of each point of the three-axis acceleration signal within each second, if the sum of the states is less than the second threshold, the states of the three-axis acceleration signal of the second are all recorded as 0, otherwise they are all recorded as 1; calculate the time interval of the segments of the three-axis acceleration signal whose states are all recorded as 0, and the states of the three-axis acceleration signal between the time periods whose time intervals are less than the third threshold are all recorded as 0, that is, based on the rest segment of the acceleration signal; Heart rate recovery judgment steps: obtain the second-by-second heart rate trend from the ECG signal; calculate the inflection point of the heart rate trend in the time period corresponding to the resting segment based on the acceleration signal, and the inflection point is the heart rate recovery point; The heart rate calculation steps for the resting segment are as follows: the period from the heart rate recovery point to the end point of the resting segment based on the acceleration signal is the resting segment of the acceleration combined with the heart rate; the resting heart rate of the resting segment of the acceleration combined with the heart rate is calculated; Resting heart rate aggregation step: Aggregate the resting heart rate of all acceleration and heart rate resting segments in a day using the median to obtain the resting heart rate for the day.

2. The method for calculating resting heart rate based on wearable device data according to claim 1, characterized in that: In the data preprocessing step, The EP Limited heart rate detection algorithm is used to detect the heartbeat position of the ECG signal to obtain the RR interval sequence of the ECG signal, and the heart rate sequence is obtained from the RR interval sequence; the EP Limited heart rate detection algorithm first filters the ECG signal, and then divides each signal peak into QRS wave or noise according to the detection rules; the three-axis acceleration signal is filtered.

3. The method for calculating resting heart rate based on wearable device data according to claim 2, characterized in that: In the heart rate recovery judgment step: Outliers are removed from the heart rate sequence in the 2-minute window, missing values ​​are interpolated using the median in the window, and then the sliding average method is used to perform sliding average on the entire heart rate sequence to obtain the heart rate trend; The heart rate is interpolated to synchronize with the three-axis acceleration signal, and the wave-by-wave heart rate is interpolated to the second-by-second heart rate using linear interpolation.

4. The method for calculating resting heart rate based on wearable device data according to claim 1, characterized in that: In the resting heart rate calculation step: For each resting heart rate segment, use the formula (60*f1) / rr to calculate the resting heart rate in turn; Where: f1 is the sampling frequency of the ECG signal; rr is the RR interval of the ECG signal.

5. The method for calculating resting heart rate based on wearable device data according to claim 4, characterized in that: For the above-mentioned situation where the resting segments are of different lengths, in order to make full use of the resting segment information, the resting segments are first divided. Those less than 30 seconds are directly calculated, and those longer than 30 seconds are first divided into multiple segments. The resting heart rate of each segment is the median of the segment.

Images