Method and system for atrial fibrillation detection

Abstract

The application belongs to the technical field of atrial fibrillation detection, and specifically discloses an atrial fibrillation detection method and system. i The method is based on an ECG signal preprocessed by a band-pass filter, locates the position of an R wave, and calculates an RR interval. i1 The latter half of each RR interval data S i2 is intercepted and evenly divided into two data S i1 and S i2 . According to the data S i and S a , a P wave balance feature is calculated. The data S i is right-aligned and left-padded, and point-by-point superimposed averaging is performed to obtain data S a . The superimposed average feature is calculated through the data S i and the data S a . The atrial fibrillation detection index is calculated by using the heart rate variability index, the P wave balance feature and the superimposed average feature. The atrial fibrillation detection index is compared with a preset value. When the atrial fibrillation detection index is greater than the preset value, it is determined that the subject has atrial fibrillation. Otherwise, the subject does not have atrial fibrillation. The technical scheme improves the accuracy of atrial fibrillation recognition under complex ECG rhythm type conditions by comprehensively using heart beat information and ECG morphological information.

Description

Technical Field

[0001] The present invention belongs to the technical field of atrial fibrillation detection and relates to an atrial fibrillation detection method and system. Background Art

[0002] Atrial fibrillation, one of the most common clinically diagnosed tachyarrhythmias, is a significant risk factor for all-cause mortality, cardiovascular mortality, and stroke. Therefore, early detection of atrial fibrillation has attracted considerable attention. With the advancement of wearable physiological monitoring technology, wearable ECG monitoring devices have gradually become part of people's daily lives, providing a technical platform for routine monitoring of atrial fibrillation.

[0003] Currently, daily monitoring of atrial fibrillation primarily relies on real-time analysis of photoplethysmography (PPG) signals acquired by smartwatches. This approach distinguishes atrial fibrillation from normal ECG rhythms by analyzing the absolute irregularity of the heartbeat intervals. However, this method is only effective in distinguishing atrial fibrillation from normal ECG rhythms and has limited ability to distinguish other complex ECG rhythms, such as sinus arrhythmia and premature beats. To address this issue, the currently used deep learning algorithms are too complex to be suitable for real-time monitoring in wearable applications.

[0004] The effective information that can be extracted from PPG signals is the RR interval. Each heartbeat produces a pulse wave that corresponds one-to-one with an ECG R wave. However, the RR interval can vary significantly with conditions such as sinus arrhythmia, premature beats, and conduction block, significantly reducing the accuracy of atrial fibrillation detection based solely on the RR interval. Unlike PPG signals, ECG signals contain rich morphological information in addition to heartbeat information. The morphological characteristics of the P and F waves are also important indicators for distinguishing atrial fibrillation. This plays a crucial role in improving the ability to distinguish atrial fibrillation from other complex ECG rhythms.

[0005] The existing technology uses algorithms that locate the PQRST complex wave group characteristics to detect atrial fibrillation. The effectiveness of locating the various characteristic ECG waveforms (such as the P wave and T wave) will directly affect the accuracy of atrial fibrillation detection. This method has two possible misjudgment nodes: first, the algorithm needs to locate the R wave, and then locate the P wave based on this. If the P wave and other characteristic waveforms are located incorrectly, it will lead to incorrect atrial fibrillation identification. Secondly, even if the P wave is located correctly, misclassification may occur during atrial fibrillation identification. The combination of these two factors affects the accuracy of the algorithm. In addition, the algorithm for locating each characteristic waveform from the ECG signal is relatively complex, which is not conducive to embedded system deployment. Existing deep learning algorithms far exceed traditional algorithms in both time and space complexity, limiting their application in the field of wearable vital signs monitoring developed based on embedded systems.

[0006] Therefore, in response to the demand for early screening of atrial fibrillation in daily health monitoring, the development of a low-complexity atrial fibrillation detection method suitable for wearable ECG monitoring devices is of great significance for the daily application of wearable atrial fibrillation detection in the future. Summary of the Invention

[0007] The purpose of the present invention is to provide an atrial fibrillation detection method and system to improve the accuracy of atrial fibrillation recognition under conditions of complex ECG rhythm types.

[0008] In order to achieve the above object, the basic scheme of the present invention is: a method for detecting atrial fibrillation, comprising the following steps:

[0009] Collect the ECG signal of the subject and perform band-pass filtering preprocessing;

[0010] Based on the ECG signal preprocessed by bandpass filtering, the position of the R wave is located and the RR interval is calculated;

[0011] Heart rate variability index was calculated using RR interval;

[0012] Cut off the second half of each RR interval data S i , and evenly divided into two segments S i1 With S i2 ;

[0013] According to the data S i1 With S i2 , calculate the P wave balance characteristics;

[0014] For each data S i Perform right alignment and left padding operations, and add and average point by point to obtain data S a ;

[0015] Through data S i and data S a , calculate the superposition average feature;

[0016] Atrial fibrillation detection indicators are calculated using heart rate variability indicators, P wave balance characteristics, and superposition average characteristics;

[0017] The atrial fibrillation detection index is compared with the preset value. When the atrial fibrillation detection index is greater than the preset value, it is judged as atrial fibrillation; otherwise, it is not atrial fibrillation.

[0018] The working principle and beneficial effects of this basic solution lie in the following: when extracting characteristic parameters, this solution only needs to locate the R wave to simultaneously obtain RR interval information and ECG signal morphology information, without the need to identify other characteristic waveforms such as P and T waves. The algorithm is simple and easy to implement, and it can also easily combine RR interval characteristics to construct a mathematical model, enabling atrial fibrillation monitoring and improving detection accuracy.

[0019] Further, the method for band-pass filtering and preprocessing the electrocardiosignal of the subject is:

[0020] The electrocardiosignal of the subject is collected by using a wearable electrocardio monitoring device;

[0021] The electrocardiosignal data x(n) with a time length of t seconds is synchronously intercepted in real time, and the data x(n) is band-pass filtered at 0.5-20 Hz to obtain a signal y(n), and n is the data sample index.

[0022] The electrocardiosignal is preprocessed, which is beneficial to subsequent data processing.

[0023] Further, the heart rate variability index is calculated as follows:

[0024]

[0025] Wherein, SDNN is the heart rate variability index, i is the RR interval number, n is the data sample index, RRI i is the RR interval, is the mean of all RR intervals.

[0026] The operation is simple and convenient to use.

[0027] Further, the data S i is intercepted, and the data S i1 and S i2 are obtained as follows:

[0028] The midpoint between the i th R wave position P R (i) and the i+1 th R wave position P R (i+1) in the electrocardiosignal y(n) is taken as the starting point, and the data sample 0.075s before the i+1 th R wave position P R (i+1) is taken as the end point, and the latter half of the data S i of each RR interval is intercepted.

[0029]

[0030] Wherein, f s is the signal sampling rate, and i is the RR interval number;

[0031] The intercepted data segment S i is equally divided into two parts S i1 and S i2 .

[0032] The operation is simple and convenient to use.

[0033] Further, the method for calculating the P wave balance feature is as follows:

[0034] The data segment Si Balance indicator BAL i :

[0035]

[0036] in, and Data segment S i1 With S i2 The mean of L is the data segment S i1 and S i2 The length of j is the index of the sampling point in the data segment, S i1 (j) is the data segment S i1 The jth sampling point in S i2 (j) is the data segment S i2 The jth sampling point in ;

[0037] Calculate the P wave balance characteristic BAL:

[0038]

[0039] Where N is the number of RR intervals in the ECG signal y(n), i is the RR interval sequence number, and n is the data sample index.

[0040] The calculation process is simple and easy to operate.

[0041] Further, we get data S a The method is:

[0042] Take all the intercepted data segments S in the ECG signal y(n) i The number of sampling points in the longest section L m As a basis, all S i Align the right side of the data segment and fill in the left side, then all the data segments S i The corresponding sample points are superimposed and averaged to obtain data S a :

[0043]

[0044] Where N is the number of RR intervals in the data y(n), j is the index of the sampling point in the data segment, S i (j) is the intercepted i-th data segment S i The jth sampling point in , n is the data sample index.

[0045] Get data S a , which is convenient for subsequent use.

[0046] Furthermore, the method for calculating the superposition average feature is as follows:

[0047] Calculate data Sa Peak-to-peak PPV a :

[0048] PPV a =max(S a )-min(S a );

[0049] Calculate each data segment S i The peak-to-peak value of

[0050]

[0051] Among them, max(S a ) is the data S a The maximum value, min(S a ) is the data S a The minimum value of , N is the number of RR intervals in the ECG signal y(n), i is the RR interval sequence number, and n is the data sample index;

[0052] Calculate the superposition average characteristic PEAK:

[0053]

[0054] Obtain superimposed average features to facilitate subsequent atrial fibrillation detection.

[0055] Furthermore, the atrial fibrillation detection index A is calculated:

[0056] A=SDNN+2×BAL+2×PEAK

[0057] Among them, SDNN is the heart rate variability index, BAL is the P wave balance feature, and PEAK is the superposition average feature.

[0058] The operation structure is simple, and the atrial fibrillation detection index A is obtained for atrial fibrillation identification, and the operation is simple.

[0059] Furthermore, the preset value is 5.

[0060] Set appropriate preset values ​​to ensure detection accuracy.

[0061] The present invention also provides an atrial fibrillation detection system, a wearable ECG monitoring device and a detection module. The wearable ECG monitoring device is used to collect ECG signals of a person to be tested. The output end of the wearable ECG monitoring device is connected to the input end of the detection module. The detection module executes the method described in the present invention to perform atrial fibrillation detection.

[0062] The system is used to detect atrial fibrillation, which is easy to operate and improves the accuracy of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 1 is a flow chart of the atrial fibrillation detection method of the present invention;

[0064] Figure 2 is the data S under the non-atrial fibrillation condition of the atrial fibrillation detection method of the present invention i Schematic diagram of the interception;

[0065] Figure 3 is the data S under the condition of atrial fibrillation in the atrial fibrillation detection method of the present invention i Schematic diagram of the interception;

[0066] Figure 4 is the data S under the non-atrial fibrillation condition of the atrial fibrillation detection method of the present invention i Schematic diagram of right-aligned and left-filled operation;

[0067] Figure 5 This is the right-alignment and left-padding operation in the atrial fibrillation detection method of the present invention under the condition of atrial fibrillation. DETAILED DESCRIPTION

[0068] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0069] In the description of the present invention, it should be understood that the terms "longitudinal", "transverse", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.

[0070] In the description of the present invention, unless otherwise specified and limited, it should be noted that the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a mechanical connection or an electrical connection, or it can be the internal communication between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to the specific circumstances.

[0071] The present invention discloses an atrial fibrillation detection method, which is applicable to wearable electrocardiogram monitoring equipment, integrates heartbeat information and electrocardiogram morphology information, and can improve the accuracy of atrial fibrillation recognition under complex electrocardiogram rhythm conditions. Figure 1 As shown, the atrial fibrillation detection method includes the following steps:

[0072] The wearable ECG monitoring device is used to collect the ECG signals of the subjects and perform band-pass filtering preprocessing;

[0073] Based on the ECG signal preprocessed by bandpass filtering, the position of the R wave P is located R , and calculate the RR interval RRI i , i is the RR interval number;

[0074] Heart rate variability index was calculated using RR interval;

[0075] like Figure 2 and Figure 3 As shown, the second half of each RR interval is intercepted (that is, half of the data of the RR interval, the data between each R wave and the adjacent R wave is divided into two halves, and the second half of the data is taken; when taking the second half of the data, the end point of the data segment needs to be moved forward 0.075s to avoid the rising edge of the R wave) data S i , and evenly divided into two segments S i1 With S i2 .

[0076] According to the data S i1 With S i2 , calculate the P wave balance characteristics; for non-atrial fibrillation data, due to the presence of P waves, S i The first half of S i1 The amplitude and energy are smaller than the second half S i2 ; For atrial fibrillation data, the P wave disappears and a chaotic F wave appears, S i The first half of S i1 The amplitude, energy and the second half S i2 There is no stable size comparison relationship between them. By calculating S i1 ,S i2 The variance ratio of is used to construct the morphological feature BAL;

[0077] like Figure 4 and Figure 5 As shown, for each data S i Perform right alignment and left padding operations, and average the data point by point to obtain data S a ; Each segment S i The data lengths may not be equal (this is determined by the instantaneous heart rate. The longer the RR interval, the longer the S i The data segment contains more sample points, otherwise the sample points are less), so before performing superposition averaging, all S iAfter right-aligning and padding on the left, all data segments are of the same length. Next, each sample is stacked and averaged. In non-AF signals, the P wave appears at a fixed phase. After stacking and averaging, the P wave's characteristics are enhanced, increasing the signal amplitude and energy. However, the F wave of AF is chaotic and lacks a strict phase relationship. Therefore, after stacking and averaging, the positive and negative signals in the signal samples cancel each other out, reducing the signal energy. This characteristic is used to construct the stacking and averaging metric, PEAK.

[0078] Through data S i and data S a , calculate the superposition average feature;

[0079] Atrial fibrillation detection indicators are calculated using heart rate variability indicators, P wave balance characteristics, and superposition average characteristics;

[0080] The atrial fibrillation detection index is compared with a preset value. If the atrial fibrillation detection index is greater than the preset value, it is determined to be atrial fibrillation; otherwise, it is not atrial fibrillation. Preferably, the preset value is 5.

[0081] In a preferred embodiment of the present invention, the method for performing bandpass filtering preprocessing on the electrocardiogram signal of the subject is as follows:

[0082] Using wearable ECG monitoring equipment to collect the ECG signals of the subject;

[0083] Real-time synchronous interception of ECG signal data x(n) with a duration of t (t=5-10) seconds, and performing a 0.5-20 Hz bandpass on the data x(n) to obtain signal y(n), where n is the data sample point index.

[0084] Preprocessing of ECG signals facilitates subsequent data processing.

[0085] In a preferred embodiment of the present invention, the method for calculating the RR interval is:

[0086] The ECG signal y(n) is subjected to a 5Hz high-pass filter to obtain the signal z(n), which is then subjected to a nonlinear transformation to enhance the R-wave feature and obtain the signal s(n):

[0087]

[0088] The threshold th is calculated as follows:

[0089]

[0090] Where j is the data sample index and m is the length of the signal y(n).

[0091] Length The moving window is used to binarize the signal s(n) with a moving step of 1 sample point;

[0092] Output 1 when the maximum value of data in the moving window is greater than or equal to the threshold value th, and output 0 when the data in the moving window is less than the threshold value th, to obtain a rectangular wave signal s b (n);

[0093] Finding the signal s b (n) All intervals with all values being 1, locating the maximum value position of the signal y(n) in each interval, to obtain the R wave position P R Where f s is the sampling rate;

[0094] Calculate the RR interval RRI by the R wave position i

[0095] RRI i =P R (i+1)-P R (i)

[0096] Where i is the RR interval number.

[0097] In a preferred scheme of the present application, the heart rate variability index is calculated as follows:

[0098]

[0099] Where SDNN is the heart rate variability index, i is the RR interval number, n is the data sample index, RRI i is the RR interval, is the mean value of all RR intervals.

[0100] In a preferred scheme of the present application, the data S i is intercepted, and the data S i1 is obtained by the method as follows: i2

[0101] Taking the midpoint between the i th R wave position P R (i) and the i+1 th R wave position P R (i+1) as the starting point, and taking the data sample 0.075s (the width of the electrocardiogram signal is 0.12s, and 0.075s is set to avoid the rising edge of the R wave when intercepting data) before the i+1 th R wave position P R (i+1) as the end point, the latter half data S i of each RR interval is intercepted.

[0102]

[0103] Where f s is the signal sampling rate, and i is the RR interval number;

[0104] ​The intercepted data segment S i Divided equally into two parts, front and back S i1 With S i2 .

[0105] In a preferred embodiment of the present invention, the method for calculating the P wave balance characteristic is as follows:

[0106] Calculate data segment S i Balance indicator (characteristic) BAL i :

[0107]

[0108] in, and Data segment S i1 With S i2 The mean of L is the data segment S i1 and S i2 The length of j is the index of the sampling point in the data segment, S i1 (j) is the data segment S i1 The jth sampling point in S i2 (j) is the data segment S i2 The jth sampling point in ;

[0109] Calculate morphological indicators and P wave balance characteristics BAL:

[0110]

[0111] Where N is the number of RR intervals in the ECG signal y(n), i is the RR interval sequence number, and n is the data sample index.

[0112] In a preferred embodiment of the present invention, the data S is obtained a The method is:

[0113] Take all the intercepted data segments S in the ECG signal y(n) i The number of sampling points in the longest section L m As a basis, all S i Align the right side of the data segment and fill in the left side, then all the data segments S i The corresponding sample points are superimposed and averaged to obtain data S a :

[0114]

[0115] Where N is the number of RR intervals in the data y(n), j is the index of the sampling point in the data segment, S i (j) is the intercepted i-th data segment S i The jth sampling point in , n is the data sample index.

[0116] In a preferred embodiment of the present invention, the method for calculating the superposition average feature is as follows:

[0117] Calculate data S a Peak-to-peak PPV a :

[0118] PPV a =max(S a )-min(S a );

[0119] Calculate each data segment S i The peak-to-peak value of

[0120]

[0121] Among them, max(S a ) is the data S a The maximum value, min(S a ) is the data S a The minimum value of , N is the number of RR intervals in the ECG signal data y(n), i is the RR interval sequence number, and n is the data sample index;

[0122] Calculate morphological indicators and superimpose average feature PEAK:

[0123]

[0124] In a preferred embodiment of the present invention, the atrial fibrillation detection index A is calculated:

[0125] A=SDNN+2×BAL+2×PEAK

[0126] SDNN is the heart rate variability indicator, BAL is the P wave balance feature, and PEAK is the superposition average feature. If the atrial fibrillation detection index A is greater than 5, it is judged as atrial fibrillation; if the atrial fibrillation detection index A is less than 5, it is judged as non-atrial fibrillation.

[0127] The present invention also provides an atrial fibrillation detection system, comprising a wearable ECG monitoring device and a detection module. The wearable ECG monitoring device is configured to collect ECG signals from a subject, and the output of the wearable ECG monitoring device is electrically connected to the input of the detection module. The detection module performs the method of the present invention to detect atrial fibrillation.

[0128] When extracting characteristic parameters, this technical solution only needs to locate the R wave to obtain the RR interval information and the morphological information of the ECG signal at the same time, without the need to identify other characteristic waveforms such as P wave and T wave. The algorithm is simple and easy to implement. Then, a mathematical model is established through linear regression analysis to construct a simple and feasible atrial fibrillation recognition algorithm to improve the detection accuracy. It has a good application effect for distinguishing between various ECG types (such as atrioventricular block, sinus arrhythmia, etc.) and atrial fibrillation. Both morphological indicators of the present invention do not require the separate positioning of characteristic waveforms such as P wave and T wave. After the position of the R wave is determined, the calculation can be performed very simply.

[0129] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0130] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. An atrial fibrillation detection system, characterized in that: The invention comprises a wearable ECG monitoring device and a detection module. The wearable ECG monitoring device is used to collect ECG signals of a subject. The output end of the wearable ECG monitoring device is connected to the input end of the detection module. The detection module performs atrial fibrillation detection. The specific atrial fibrillation detection method comprises the following steps: Collect the ECG signal of the subject and perform band-pass filtering preprocessing; Based on the ECG signal preprocessed by bandpass filtering, the position of the R wave is located and the RR interval is calculated; Heart rate variability index was calculated using RR interval; Cut off the second half of each RR interval data S i , and evenly divided into two segments S i1 With S i2 ; According to the data S i1 With S i2 , calculate the P wave balance characteristics; For each data S i Perform right alignment and left padding operations, and average the data point by point to obtain data S a ; Through data S i and data S a , calculate the superposition average feature; Atrial fibrillation detection indicators are calculated using heart rate variability indicators, P wave balance characteristics, and superposition average characteristics; Compare the atrial fibrillation detection index with the preset value. When the atrial fibrillation detection index is greater than the preset value, it is judged as atrial fibrillation; otherwise, it is not atrial fibrillation; The method for calculating the P wave balance characteristics is as follows: Calculate data segment S i Balance indicator BAL i : in, and Data segment S i1 With S i2 The mean of L is the data segment S i1 and S i2 The length of j is the index of the sampling point in the data segment, S i1 (j) is the data segment S i1 The jth sampling point in S i2 (j) is the data segment S i2 The jth sampling point in ; Calculate the P wave balance characteristic BAL: Where N is the number of RR intervals in the ECG signal y(n), and i is the RR interval sequence number.

2. The atrial fibrillation detection system according to claim 1, wherein: The method for bandpass filtering preprocessing of the subject's ECG signal is as follows: Using wearable ECG monitoring equipment to collect the ECG signals of the subject; A data x(n) of t seconds of ECG signal is captured synchronously in real time, and the data x(n) is band-passed at 0.5 to 20 Hz to obtain the signal y(n), where n is the data sample index.

3. The atrial fibrillation detection system according to claim 1, wherein: Calculate heart rate variability metrics: Among them, SDNN is the heart rate variability index, i is the RR interval number, n is the data sample index, RRI i is the RR interval, is the mean of all RR intervals.

4. The atrial fibrillation detection system according to claim 1, wherein: Intercept data S i , and get data S i1 With S i2 The method is as follows: The position P of the i-th R wave in the ECG signal y(n) R (i) and the i+1th R wave position P R (i+1) as the starting point, with the i+1th R wave position P R The data sampling point 0.075s before (i+1) is the end point, and the second half of each RR interval data S is intercepted. i : Among them, f s is the signal sampling rate, i is the RR interval number; The intercepted data segment S i Divided equally into two parts, front and back S i1 With S i2 .

5. The atrial fibrillation detection system according to claim 1, wherein: Get data S a The method is: Take all the intercepted data segments S in the ECG signal y(n) i The number of sampling points in the longest section L m As a basis, all S i Align the right side of the data segment and fill in the left side, then all the data segments S i The corresponding sample points are superimposed and averaged to obtain data S a : Where N is the number of RR intervals in the data y(n), j is the index of the sampling point in the data segment, S i (j) is the intercepted i-th data segment S i The jth sampling point in .

6. The atrial fibrillation detection system according to claim 1, wherein: The method for calculating the superposition average feature is as follows: Calculate data S a Peak-to-peak PPV a : PPV a =max(S a )-min(S a ); Calculate each data segment S i The peak-to-peak value of Among them, max(S a ) is the data S a The maximum value, min(S a ) is the data S a The minimum value of , N is the number of RR intervals in the ECG signal y(n), and i is the RR interval sequence number; Calculate the superposition average characteristic PEAK: 。 7. The atrial fibrillation detection system according to claim 1, wherein: Calculate the atrial fibrillation detection index A: A=SDNN+2×BAL+2×PEAK Among them, SDNN is the heart rate variability index, BAL is the P wave balance feature, and PEAK is the superposition average feature.

8. The atrial fibrillation detection system according to claim 1, wherein: The preset value is 5.

Images