TWI810901B - Remote electrocardiogram monitoring device and electrocardiogram monitoring method - Google Patents

Abstract

A remote electrocardiogram monitoring device includes a heartbeat sensing circuit and a processor. The heartbeat sensing circuit obtains an electrocardiogram (ECG) signal. The processor is electrically connected to the heartbeat sensing circuit to obtain the ECG signal, identify multiple peaks in the ECG signal, segment the ECG signal into complete cycle segments based on the peaks, and convert the complete cycle segments into multiple cycle images. The cycle images input to a convolutional neural network to determine whether it is normal.

Description

Remote electrocardiogram monitoring device and electrocardiogram monitoring method

The present disclosure relates to a remote electrocardiogram monitoring device, which can automatically identify whether the electrocardiogram is abnormal.

In some conventional technologies, if you want to monitor the patient's ECG, you need to rely on the medical staff to stick a patch next to it, record the patient's ECG through related instruments, and then let the doctor judge whether the ECG is abnormal. If the patient's electrocardiogram can be monitored or recorded in a remote manner, and whether the electrocardiogram is abnormal can be automatically judged, an instant judgment result can be obtained, and manpower can also be saved. However, there is still room for improvement in how to process ECG signals to provide accurate judgments.

Embodiments of the present disclosure provide a remote ECG monitoring device, including a heartbeat sensing circuit and a processor. The heartbeat sensing circuit is used for obtaining an electrocardiogram signal. The processor is electrically connected to the heartbeat sensing circuit to obtain the electrocardiogram signal, identify multiple peaks in the electrocardiogram signal, cut the electrocardiogram signal into multiple complete cycle segments according to the peak values, and convert the complete cycle segments into multiple cycle images respectively , and input the loop image to the convolutional neural network to identify whether the loop image is normal.

In some embodiments, the operation of identifying the peak value in the ECG signal includes: for the amplitude of the ith sampling point in the ECG signal with time , calculate the angle of the i-th sampling point according to the following mathematical formula 1 , and calculate the angle change according to the following mathematical formula 2 . [mathematical formula 1] [mathematical formula 2]

Next, the backward flat distance is calculated according to the following mathematical formula 3 , and calculate the forward flat distance according to the following mathematical formula 4 . [mathematical formula 3] [mathematical formula 4]

Next, the backward amplitude distance is calculated according to the following mathematical formula 5 , and calculate the forward amplitude distance according to the following mathematical formula 6 . [mathematical formula 5] [mathematical formula 6]

Next, calculate the backward angle according to the following mathematical formula 7 , and calculate the forward angle according to the following mathematical formula 8 . [mathematical formula 7] [mathematical formula 8]

Next, calculate the curvature of the i-th sampling point according to the following mathematical formula 9 . [mathematical formula 9]

If the curvature of the i-th sampling point is greater than the first critical value, it is judged that the i-th sampling point is a peak value.

In some embodiments, the operation of cutting the electrocardiogram signal into a plurality of complete cycle segments according to the peak value includes: for each peak value, if the corresponding amplitude is smaller than a second critical value, removing the peak value; If the time distance between the two peaks is smaller than the third critical value, the peak with a lower amplitude among the two peaks is removed; and the ECG signal is cut into complete cycle segments according to the remaining peaks.

In some embodiments, the processor is also used to calculate the mean amplitude and the standard deviation of the amplitude of all the sampling points in the ECG signal, and subtract n times the standard deviation of the amplitude from the mean amplitude to obtain the second critical value.

In some embodiments, the remote ECG monitoring device further includes a storage for storing ECG signals.

From another point of view, the embodiments of the present disclosure provide an electrocardiogram monitoring method suitable for a processor. The electrocardiogram monitoring method includes: obtaining an electrocardiogram signal through a heartbeat sensing circuit; identifying multiple peaks in the electrocardiogram signal; The peak cuts the ECG signal into multiple complete loop segments; and converts the complete loop segments into multiple loop images respectively, and inputs the loop images to the convolutional neural network to identify whether the loop images are normal.

In some embodiments, the above step of identifying the peak value in the ECG signal is performed according to the above formula 1 to formula 9.

In some embodiments, the step of cutting the electrocardiogram signal into a plurality of complete cycle segments according to the peak value includes: for each peak value, if the corresponding amplitude is smaller than a second critical value, removing the peak value; If the time distance is smaller than the third critical value, the peak with the lower amplitude among the two peaks is removed; and the electrocardiogram signal is cut into complete cycle segments according to the remaining peaks.

In some embodiments, the electrocardiogram monitoring method further includes: calculating the mean amplitude and the standard deviation of the amplitude of all sampling points in the electrocardiogram signal, and subtracting n times the standard deviation of the amplitude from the mean amplitude to obtain the second critical value.

In some embodiments, the ECG monitoring method further includes: storing the ECG signal through a memory.

In order to make the above-mentioned features and advantages of the present invention more comprehensible, the following specific embodiments are described in detail together with the accompanying drawings.

The terms "first", "second" and the like used herein do not specifically refer to a sequence or sequence, but are only used to distinguish elements or operations described with the same technical terms.

FIG. 1 is a schematic block diagram of a remote ECG monitoring device. Please refer to FIG. 1 , the remote ECG monitoring device 100 is a portable device that can be carried by a patient. The remote ECG monitoring device 100 includes a processor 110 , a heartbeat sensing circuit 120 , a storage 130 and a communication module 140 . The processor 110 is electrically connected to the heartbeat sensing circuit 120 , the storage 130 and the communication module 140 . The processor 110 may be a central processing unit, a microprocessor, a microcontroller, a digital signal processor, an application-specific integrated circuit, and the like. The heartbeat sensing circuit 120 is used to obtain the patient's ECG signal. For example, the heartbeat sensing circuit 120 can be an AD8232 heartbeat sensing chip produced by an Arduino manufacturer. The storage 130 can be a flash memory, a floppy disk, a hard disk, etc., and is used for storing ECG signals. The communication module 140 may be a circuit providing communication functions such as near field communication, infrared communication, Bluetooth, and Wi-Fi. In this disclosure, the processor 110 will execute an electrocardiogram monitoring method, using a convolutional neural network to automatically determine whether the electrocardiogram signal is abnormal, and the relevant steps will be described below.

First, a segment representing a complete cycle of a heartbeat must be cut out, for which multiple peaks in the ECG signal must first be identified. FIG. 2 is a schematic diagram of plotting ECG signals and peak values according to an embodiment. Please refer to Figure 2. After sampling, the ECG signal is a discrete signal. to represent the amplitude of the i-th sampling point, and Indicates the time of the i-th sampling point, and its unit is, for example, milliseconds. Here we need to calculate the curvature of each sampling point. When the signal is smoother, the curvature becomes smaller. Conversely, when the signal turns larger, the curvature becomes larger. The sampling point with a larger curvature may be the peak value. . First calculate the angle of the i-th sampling point according to the following mathematical formula 1 , and calculate the angle variation of the i-th sampling point according to the following mathematical formula 2 . [mathematical formula 1] [mathematical formula 2]

Where k is a positive integer, such as 2. Next, calculate a backward flat distance according to the following formula 3 , and calculate a forward flat distance according to the following mathematical formula 4 . [mathematical formula 3] [mathematical formula 4]

in A real number set by the user. The above mathematical formula 3 is to calculate the maximum length k, so that the angle change within the length k from the i-th sampling point is are all within a certain range. Similarly, the above mathematical formula 4 needs to calculate the maximum length k, so that the angle change within the length k from the i-th sampling point is are all within a certain range. Next, the backward amplitude distance is calculated according to the following mathematical formula 5 , and calculate the forward amplitude distance according to the following mathematical formula 6 . [mathematical formula 5] [mathematical formula 6]

The above backward amplitude distance is to calculate the i-th sampling point and the future The Euler distance between sampling points, similarly the forward amplitude distance is to calculate the i-th sampling point and the forward The Euler distance between sampling points. That is to say, these Euler distances represent not only the difference in time, but also the difference in amplitude. Next, calculate the backward angle according to the following mathematical formula 7 , and calculate the forward angle according to the following mathematical formula 8 . [mathematical formula 7] [mathematical formula 8]

Finally, calculate the curvature of the i-th sampling point according to the following mathematical formula 9 , if the curvature If it is greater than the first critical value, it can be judged that the i-th sampling point is a peak value. [mathematical formula 9]

The embodiment of FIG. 2 marks the positions of a plurality of peaks (eg, peaks 210 - 225 , 230 ). Theoretically, a complete cycle segment has P, Q, R, S, T and other waves, among which the R wave has the highest amplitude and can be used to identify the complete cycle segment. However, the above-mentioned peak judgment method may also consider the T wave as the amplitude, so some peaks must be filtered out first. For example, peaks can be filtered by amplitude, and a peak can be removed if its amplitude is smaller than a second threshold. In some embodiments, the amplitude average and amplitude standard deviation of all sampling points in the electrocardiogram signal can be calculated first, and the amplitude average is subtracted by n times the amplitude standard deviation to obtain the second critical value, wherein n is, for example, 1, 2 or other suitable values. In addition, peaks can also be filtered based on the temporal distance between peaks. For example, if the time distance between two adjacent peaks is smaller than the third critical value, the peak with a lower amplitude among the two peaks may be removed. The above-mentioned third critical value can be determined according to the number of heartbeats. If the patient has 80 heartbeats per minute, the length of each heartbeat is about 12.5 milliseconds. The third critical value can be set to 10 milliseconds or any other suitable value. When two When the time distance between peaks is less than 10 milliseconds, one of them may be an R wave and the other may be a T wave or other spike, where the higher amplitude peak is preserved. After the above processing, the peak 230 can be filtered out, and the ECG signal can be cut into multiple complete cycle segments according to the remaining peak values. For example, the middle point (on the time axis) of two adjacent peak values can be regarded as a complete cycle segment. The start point (and also the end point of the previous full loop segment).

Next, the cut complete loop segments are converted into two-dimensional images (also called loop images). FIG. 3 shows several examples of loop images 301-304. Finally, doctors can mark whether these circulatory images are normal to obtain labels, and a convolutional neural network can be trained according to these labels and circulatory images. In this embodiment, the pre-trained VGG19 convolutional neural network is used as the basis for further training. The last layer of the convolutional neural network can be a normalization (softmax) function, the error can be root mean square error or information entropy, etc., and the method of updating weights can be gradient descent, but this disclosure is not limited thereto.

The trained convolutional neural network can be stored in the memory 130 . When the patient carries the remote ECG monitoring device, the ECG signal can be cut into complete cycle segments according to the above method to generate cycle images, and the processor 110 executes the convolutional neural network to identify whether these cycle images are normal. If it is judged to be abnormal, a warning message can be sent to a remote device (such as the equipment of medical personnel) through the communication module 140, or the warning message can be provided to the patient in any way such as sound, video, text, etc. This disclosure is not limited The program after it is judged as abnormal.

FIG. 4 is a flowchart illustrating a method for monitoring an ECG according to an embodiment. Please refer to FIG. 4 , in step 401 , the heartbeat sensing circuit is used to obtain the electrocardiogram signal. In step 402, a plurality of peaks in the ECG signal are identified. In step 403, the ECG signal is segmented into a plurality of complete cycle segments according to the peak value. In step 404, the complete loop segments are respectively converted into a plurality of loop images, and the loop images are input to the convolutional neural network to identify whether the loop images are normal. However, each step in FIG. 4 has been described in detail above, and will not be repeated here. It should be noted that each step in FIG. 4 can be implemented as a plurality of program codes or circuits, and the present invention is not limited thereto. In addition, the method in FIG. 4 can be used in combination with the above embodiments, or can be used alone. In other words, other steps may also be added between the steps in FIG. 4 .

Although the present invention has been disclosed above with the embodiments, it is not intended to limit the present invention. Anyone with ordinary knowledge in the technical field may make some changes and modifications without departing from the spirit and scope of the present invention. The scope of protection of the present invention should be defined by the scope of the appended patent application.

100:Remote ECG monitoring device 110: Processor 120: Heartbeat sensing circuit 130: storage 140: Communication module 210~225,230: peak value 301~304: Loop image 401~404: steps

FIG. 1 is a schematic block diagram of a remote ECG monitoring device. FIG. 2 is a schematic diagram of plotting ECG signals and peak values according to an embodiment. FIG. 3 is an example of a drawing loop image according to an embodiment. FIG. 4 is a flowchart illustrating a method for monitoring an ECG according to an embodiment.

100:Remote ECG monitoring device

110: Processor

120: Heartbeat sensing circuit

130: storage

140: Communication module

Claims (8)

A remote electrocardiogram monitoring device, comprising: a heartbeat sensing circuit for obtaining an electrocardiogram signal; and a processor electrically connected to the heartbeat sensing circuit for obtaining the electrocardiogram signal and identifying multiple signals in the electrocardiogram signal According to these peaks, the ECG signal is cut into a plurality of complete cycle segments, and these complete cycle segments are respectively converted into a plurality of cycle images, and these cycle images are input into a convolutional neural network to identify the Whether these cyclic images are normal, wherein the operation of identifying the peaks in the ECG signal includes: For the amplitude x _i and time t _i of the ith sampling point in the ECG signal, calculate the ith sample according to the following mathematical formula 1 point angle θ _i,k , and calculate an angle variation d _i,k according to the following mathematical formula 2;

[Mathematical formula 2] d _i,k = θ _{i +1 ,k} - θ _{i -1 ,k} calculates a backward flat distance k _b according to the following mathematical formula 3, and calculates a forward flat distance k according to the following mathematical formula 4 _f ;

Calculate a backward amplitude distance L _b according to the following mathematical formula 5, and calculate a forward amplitude distance L _f according to the following mathematical formula 6;

Calculate a backward angle θ _b according to the following formula 7, and calculate a forward angle θ _f according to the following formula 8; and

Calculate the curvature C _i of the ith sampling point according to the following mathematical formula 9; and

If the curvature C _i of the i-th sampling point is greater than a first critical value, it is determined that the i-th sampling point is one of the peaks. The remote electrocardiogram monitoring device as described in claim 1, wherein the operation of cutting the electrocardiogram signal into a plurality of complete cycle segments according to the peak values includes: for each of the peak values, if the corresponding amplitude is smaller than a second critical value , remove the peak; if the time distance between two adjacent peaks among the peaks is less than a third critical value, remove the peak with a lower amplitude among the two peaks; and Cutting the ECG signal into the complete cycle segments according to the remaining peaks. The remote electrocardiogram monitoring device as described in claim 2, wherein the processor is also used to calculate the amplitude average and amplitude standard deviation of all sampling points in the electrocardiogram signal, and subtract n times the amplitude standard deviation from the amplitude average to obtain the second critical value. The remote electrocardiogram monitoring device as described in claim 1 further includes a memory for storing the electrocardiogram signal. An electrocardiogram monitoring method, suitable for a processor, the electrocardiogram monitoring method includes: using a heartbeat sensing circuit to obtain an electrocardiogram signal; identifying multiple peaks in the electrocardiogram signal; cutting the electrocardiogram signal into a plurality of complete cycle segments; and converting these complete cycle segments into a plurality of cycle images respectively, and inputting the cycle images into a convolutional neural network to identify whether the cycle images are normal, wherein the electrocardiogram signal is identified The steps of these peak values include: for the amplitude x _i and time t _i of the i-th sampling point in the ECG signal, calculate the angle θ i,k of the i-th sampling point according to the following mathematical formula 1, and calculate the angle θ _i,k of the i-th sampling point according to the following mathematical formula Formula 2 calculates an angle variation d _i,k ;

[Mathematical formula 2] d _i,k = θ _{i +1 ,k} - θ _{i -1 ,k} calculates a backward flat distance k _b according to the following mathematical formula 3, and calculates a forward flat distance k according to the following mathematical formula 4 _f ;

Calculate a backward amplitude distance L _b according to the following mathematical formula 5, and calculate a forward amplitude distance L _f according to the following mathematical formula 6;

Calculate a backward angle θ _b according to the following formula 7, and calculate a forward angle θ _f according to the following formula 8; and

Calculate the curvature C _i of the ith sampling point according to the following mathematical formula 9; and

If the curvature C _i of the i-th sampling point is greater than a first critical value, it is determined that the i-th sampling point is one of the peaks. The electrocardiogram monitoring method as described in claim item 5, wherein the step of cutting the electrocardiogram signal into a plurality of complete cycle segments according to the peaks includes: for each of the peaks, if the corresponding amplitude is less than a second critical value, remove the peak; if the time distance between two adjacent peaks among the peaks is less than a third critical value, removing the peak with a lower amplitude among the two peaks; and cutting the ECG signal according to the remaining peaks become these full loop segments. The electrocardiogram monitoring method as described in claim 6, further comprising: calculating the amplitude average and amplitude standard deviation of all sampling points in the electrocardiogram signal, and subtracting n times the amplitude standard deviation from the amplitude average to obtain the second critical value value. The electrocardiogram monitoring method described in claim 5 further includes: storing the electrocardiogram signal through a memory.