阅读说明：本技术 一种动态心电信号心房颤动的检测方法 (Method for detecting atrial fibrillation of dynamic electrocardiosignals ) 是由 马亚全 王祥 于 2021-01-22 设计创作，主要内容包括：本发明涉及心房颤动检测技术领域,尤指一种动态心电信号心房颤动的检测方法。本发明动态心电信号心房颤动的检测方法中,将时间序列的RR间期,转化成Lorenz差值散点图,本检查方法绘制的Lorenz差值散点图加入了心拍数量因素,在心拍最多的地方显示红色,心拍最少的地方显示蓝色,其他的按数量从大到小由红色到蓝色渐变,这样做的好处是加入了数量信息,从而提高了房颤检测的准确性。(The invention relates to the technical field of atrial fibrillation detection, in particular to a method for detecting atrial fibrillation by using dynamic electrocardiosignals. In the method for detecting atrial fibrillation by using dynamic electrocardiosignals, RR intervals of a time sequence are converted into a Lorenz difference scatter diagram, the Lorenz difference scatter diagram drawn by the method is added with a heart beat number factor, red is displayed at the place with the most heart beats, blue is displayed at the place with the least heart beats, and other places are gradually changed from red to blue from large to small according to the number, so that the quantity information is added, and the accuracy of atrial fibrillation detection is improved.)

1. A detection method of atrial fibrillation of dynamic electrocardiosignals is characterized by comprising the following steps: the checking method comprises a training phase and a detection phase;

the training stage comprises the following steps:

a1, acquiring dynamic electrocardiogram data;

a2, labeling atrial fibrillation fragments and non-atrial fibrillation fragments to obtain a training data set;

a3, carrying out QRS wave detection on each data fragment of the data set, obtaining an RR period difference value of each heart beat, and drawing a Lorenz difference value scatter diagram of the fragments;

a4, inputting the Lorenz difference time scatter diagram of each segment of the data set into a neural network for training to obtain an atrial fibrillation classification model for atrial fibrillation detection;

the detection stage comprises the following steps:

b1, acquiring dynamic electrocardiogram data;

b2, carrying out QRS wave detection on the data, and simultaneously acquiring two values of difference between RR interval and RR interval of each heart beat;

b3, traversing the whole heartbeat sequence, and obtaining a deviation curve chart by calculating heartbeat deviation values in a period of time before and after the current heartbeat and obtaining the deviation value of each heartbeat;

b4, on the deviation curve, taking the segment with the heart beat deviation value larger than a certain threshold value as a candidate atrial fibrillation segment, and taking the segment smaller than the threshold value as a non-atrial fibrillation segment;

b5, drawing a Lorenz difference scatter diagram of the candidate atrial fibrillation fragments, and inputting the difference scatter diagram into the network model to obtain the final detection result of the fragments;

b6, traversing all the candidate atrial fibrillation fragments and outputting the result of the atrial fibrillation analysis of the whole data.

2. The method according to claim 1, wherein said step of detecting atrial fibrillation comprises the steps of: in step a1, acquiring dynamic electrocardiographic data by using the dynamic electrocardiographic acquisition box, and collecting dynamic electrocardiographic data of a known database.

3. The method according to claim 1, wherein said step of detecting atrial fibrillation comprises the steps of: in step a2, labeling atrial fibrillation segments and non-atrial fibrillation segments to obtain two types of electrocardiographic data, namely atrial fibrillation segments and non-atrial fibrillation segments, wherein the two types of electrocardiographic data are balanced by using SMOTE algorithm.

4. The method according to claim 1, wherein said step of detecting atrial fibrillation comprises the steps of: in step a3, the method for obtaining the difference value of RR period is as follows: acquiring QRS wave position information of the electrocardio segment, acquiring RR interval of each heart beat according to the position information, wherein one QRS wave indicates that one heart beat exists, and subtracting the RR interval of the previous heart beat from the RR interval of the current heart beat to obtain the RR interval difference of the current heart beat;

in step a3, a Lorenz difference scattergram is plotted using the following method: and taking the RR interval difference value of the current heart beat as a horizontal axis, taking the RR interval difference value of the next heart beat as a vertical axis, obtaining a statistical point of the position, sequentially traversing all the heart beats, counting the heart beats at all the positions of the scatter diagram, and drawing the Lorenz scatter diagram according to the number.

5. The method according to claim 1, wherein said step of detecting atrial fibrillation comprises the steps of: in step a4, inputting all classified atrial fibrillation and non-atrial fibrillation Lorenz scatter diagrams into a neural network for learning, training to obtain an atrial fibrillation classification model, and outputting two types of results by the model: atrial fibrillation or non-atrial fibrillation.

6. The method according to claim 1, wherein said step of detecting atrial fibrillation comprises the steps of: in step B2, using QRS wave detection, the QRS wave position of the electrocardiographic data is obtained, and the RR interval of each beat is obtained from the position.

7. The method according to claim 1, wherein said step of detecting atrial fibrillation comprises the steps of: in step B3, the deviation value of each heartbeat is calculated by traversing the whole heartbeat sequence by taking the current heartbeat as the center, calculating the sum of the RR interval values of all heartbeats within 5 seconds before and after the current heartbeat and 10 seconds in total, and calculating the absolute value of the deviation between the RR interval values and the average value of the RR interval values in the current time period as the deviation value of the current heartbeat, so as to obtain a deviation curve graph.

8. The method according to claim 1, wherein said step of detecting atrial fibrillation comprises the steps of: in step B4, a threshold is set, and when the deviation of consecutive beats is greater than the threshold, the beat is considered as a potential atrial fibrillation segment, and a candidate atrial fibrillation segment is obtained.

9. The method according to claim 1, wherein said step of detecting atrial fibrillation comprises the steps of: in step B5, traversing each candidate atrial fibrillation segment one by one, using step A3 of the training phase to draw a Lorenz difference scatter diagram, and obtaining a final classification result of the segment through a trained model, wherein if the segment is an atrial fibrillation segment, the segment is retained, and otherwise, the segment is deleted.

10. The method according to claim 1, wherein said step of detecting atrial fibrillation comprises the steps of: in the Lorenz difference scatter diagram of the drawing fragment, red is displayed at the place with the most heartbeats, blue is displayed at the place with the least heartbeats, and other positions are gradually changed from red to blue according to the number from large to small, so that the information of the number of heartbeats is added.

Technical Field

The invention relates to the technical field of atrial fibrillation detection, in particular to a method for detecting atrial fibrillation by using dynamic electrocardiosignals.

Background

Atrial fibrillation (short for atrial fibrillation) is the most common tachyarrhythmia symptom, epidemiological research shows that the prevalence rate and the incidence rate of clinical atrial fibrillation in 1990-2010 are remarkably increased, the number of the clinical atrial fibrillation people in 2010 is estimated to be 3350 ten thousand, the number of the atrial fibrillation patients in China exceeds 1000 ten thousand, and the statistical data of the clinical atrial fibrillation related mortality rate is increased by 2 times within 10 years. Atrial fibrillation can form left atrial mural thrombus, and the resulting embolism is 90% cerebral artery embolism (ischemic stroke) and 10% peripheral artery embolism or mesenteric artery embolism and the like. According to the statistical data of 2017 of the global disease burden research, the disability caused by stroke accounts for 5.29% of the adjusted life-span years. Subclinical atrial fibrillation is generally regarded as atrial fibrillation with no or few typical clinical symptoms, wherein one part of atrial fibrillation is discovered by medical electrocardiography examination for physical examination or other reasons, the other part of atrial fibrillation is not discovered until after the occurrence of complications related to the atrial fibrillation, such as stroke, and even cannot be discovered after the occurrence of the stroke. Subclinical atrial fibrillation was first observed in patients with implanted pacemakers. It was found that only 17-21% of patients with pacemaker-implanted atrial fibrillation detected exhibited symptoms associated with the onset of atrial fibrillation. In patients with paroxysmal atrial fibrillation, only 10% of the episodes of atrial fibrillation cause symptoms, as assessed using 5-day dynamic electrocardiographic monitoring. Patients with asymptomatic episodes of atrial fibrillation are 12 times higher than the symptomatic number of episodes of atrial fibrillation, which makes detection of atrial fibrillation in such patients challenging. The number of patients with subclinical atrial fibrillation is multiple times that of patients with subclinical atrial fibrillation, and the harm of subclinical atrial fibrillation is the same as that of atrial fibrillation. The early detection of atrial fibrillation has great significance in preventing thromboembolism complications related to atrial fibrillation, particularly stroke.

In the conventional common atrial fibrillation detection, the Lorenz difference scatter diagram is not displayed in a color mode, but only the position is displayed or not displayed, the number of heartbeats at the position cannot be distinguished, and the accuracy of the method also needs to be improved.

Disclosure of Invention

In order to solve the problems, the invention provides a method for detecting atrial fibrillation of dynamic electrocardiosignals, which improves the detection accuracy.

In order to achieve the purpose, the invention adopts the technical scheme that: a detection method of atrial fibrillation of dynamic electrocardiosignals comprises a training phase and a detection phase;

the training stage comprises the following steps:

a1, acquiring dynamic electrocardiogram data;

a2, labeling atrial fibrillation fragments and non-atrial fibrillation fragments to obtain a training data set;

a3, carrying out QRS wave detection on each data fragment of the data set, obtaining an RR period difference value of each heart beat, and drawing a Lorenz difference value time scatter diagram of the fragments;

a4, inputting the Lorenz difference time scatter diagram of each segment of the data set into a neural network for training to obtain an atrial fibrillation classification model for atrial fibrillation detection;

the detection stage comprises the following steps:

b1, acquiring dynamic electrocardiogram data;

b2, carrying out QRS wave detection on the data, and simultaneously acquiring two values of difference between RR interval and RR interval of each heart beat;

b3, traversing the whole heartbeat sequence, and obtaining a deviation curve chart by calculating heartbeat deviation values in a period of time before and after the current heartbeat and obtaining the deviation value of each heartbeat;

b4, on the deviation curve, taking the segment with the heart beat deviation value larger than a certain threshold value as a candidate atrial fibrillation segment, and taking the segment smaller than the threshold value as a non-atrial fibrillation segment;

b5, drawing a Lorenz difference scatter diagram of the candidate atrial fibrillation fragments, and inputting the difference scatter diagram into the network model to obtain the final detection result of the fragments;

b6, traversing all the candidate atrial fibrillation fragments and outputting the result of the atrial fibrillation analysis of the whole data.

Further, in step a1, acquiring dynamic electrocardiographic data by using the dynamic electrocardiographic acquisition box, and collecting dynamic electrocardiographic data of a known database.

Further, in step a2, labeling atrial fibrillation segments and non-atrial fibrillation segments to obtain two types of electrocardiographic data, i.e., atrial fibrillation segments and non-atrial fibrillation segments, wherein the two types of electrocardiographic data are equalized by using SMOTE algorithm.

Further, in step a3, the method for obtaining the RR period difference is as follows: acquiring QRS wave position information of the electrocardio segment, acquiring RR interval of each heart beat according to the position information, wherein one QRS wave indicates that one heart beat exists, and subtracting the RR interval of the previous heart beat from the RR interval of the current heart beat to obtain the RR interval difference of the current heart beat;

in step a3, a Lorenz difference scattergram is drawn by the following method: and taking the RR interval difference value of the current heart beat as a horizontal axis, taking the RR interval difference value of the next heart beat as a vertical axis, obtaining a statistical point of the position, sequentially traversing all the heart beats, counting the heart beats at all the positions of the scatter diagram, and drawing the Lorenz scatter diagram according to the number.

Further, in step a4, inputting all classified atrial fibrillation and non-atrial fibrillation Lorenz scatter diagrams into a neural network for learning, and training to obtain an atrial fibrillation classification model, which outputs two types of results: atrial fibrillation or non-atrial fibrillation.

Further, in step B2, using QRS wave detection, the QRS wave position of the electrocardiographic data is obtained, and the RR interval of each heartbeat is obtained from the position.

Further, in step B3, the deviation value of each heartbeat is calculated by traversing the whole heartbeat sequence by taking the current heartbeat as the center, 5 seconds before and after the current heartbeat, and 10 seconds in total, and calculating the sum of the absolute values of the deviations of the RR intervals and the average value of the RR intervals in the current time period as the deviation value of the current heartbeat, so as to obtain a deviation curve graph.

Further, in step B4, a certain threshold is set, and when the deviation of consecutive beats is greater than the threshold, the beat is considered as a potential atrial fibrillation segment, and a candidate atrial fibrillation segment is obtained.

Further, in step B5, traversing each candidate atrial fibrillation segment one by one, using the step A3 of the training phase to draw a Lorenz difference scatter diagram, and obtaining a final classification result of the segment through the trained model, wherein if the segment is an atrial fibrillation segment, the segment is retained, and otherwise, the segment is deleted.

In addition, in the Lorenz difference value scatter diagram of the drawn segment, red is displayed at the place with the largest heart beat, blue is displayed at the place with the smallest heart beat, and other positions are gradually changed from red to blue according to the number from large to small, so that the heart beat number information is added.

The invention has the beneficial effects that: in the method for detecting atrial fibrillation by using dynamic electrocardiosignals, RR intervals of a time sequence are converted into a Lorenz difference scatter diagram, the Lorenz difference scatter diagram drawn by the method is added with a heart beat number factor, red is displayed at the place with the most heart beats, blue is displayed at the place with the least heart beats, and other places are gradually changed from red to blue from large to small according to the number, so that the quantity information is added, and the accuracy of atrial fibrillation detection is improved.

Drawings

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

FIG. 2 is a graph of the difference of normal ECG data segments, with the horizontal line indicating the set threshold.

FIG. 3 is a graph of the difference of atrial fibrillation electrocardiograph data segments with horizontal lines indicating the set threshold values.

FIG. 4 is a graph of the difference between normal and atrial fibrillation cardiac data segments with the horizontal lines indicating the set thresholds.

Fig. 5 is a diagram of the QRS wave position and RR interval of the heart beat.

FIG. 6 is a Lorenz difference scatter plot of the non-atrial fibrillation fragments of FIG. 2.

FIG. 7 is a Lorenz difference scatter plot of the atrial fibrillation fragment of FIG. 3.

Detailed Description

The present invention will be described in further detail with reference to the following detailed description and accompanying drawings. The present application may be embodied in many different forms and is not limited to the embodiments described in the present embodiment. The following detailed description is provided to facilitate a more thorough understanding of the present disclosure.

Referring to fig. 1, the present invention relates to a method for detecting atrial fibrillation of a dynamic cardiac signal, wherein the method includes a training phase and a detection phase;

the training stage comprises the following steps:

a1, acquiring dynamic electrocardiogram data by using a dynamic electrocardiogram acquisition box, and collecting the dynamic electrocardiogram data of a known database;

a2, labeling atrial fibrillation fragments and non-atrial fibrillation fragments of the obtained data by professional technicians to obtain two types of electrocardiogram data of the atrial fibrillation fragments and the non-atrial fibrillation fragments; under normal conditions, the data volume of the non-atrial fibrillation is larger than the quantity of the atrial fibrillation, and the two types of data are balanced by adopting an SMOTE algorithm, so that the obtained model is more reliable;

a3, analyzing each electrocardiogram data segment of the two types of data as follows: QRS wave detection, namely acquiring QRS wave position information of the electrocardio segment, acquiring an RR interval of each heart beat (one QRS wave indicates that one heart beat exists) through the position information, subtracting the RR interval of the previous heart beat from the RR interval of the current heart beat to obtain an RR interval difference of the current heart beat, and drawing by using the RR interval difference to obtain a Lorenz difference scatter diagram of the electrocardio segment;

a4, inputting all classified atrial fibrillation and non-atrial fibrillation Lorenz scatter diagrams into a neural network for learning, and training to obtain an atrial fibrillation classification model; the model will output two types of results: atrial fibrillation or non-atrial fibrillation.

The detection stage comprises the following steps:

b1, acquiring dynamic electrocardiogram data;

b2, using QRS wave detection to obtain QRS wave position of the electrocardio data, and obtaining RR interval of each heart beat according to the position;

b3, traversing the whole heartbeat sequence, and calculating a deviation value of each heartbeat by taking the current heartbeat as the center, calculating the sum of the absolute values of the deviations of the RR intervals and the average value of the RR intervals in the current time period as the deviation value of the current heartbeat, wherein the RR intervals of all heartbeats are 5 seconds before and after the current heartbeat and 10 seconds in total;

b4, setting a certain threshold, and when the deviation of a plurality of continuous heart beats is larger than the threshold, considering the heart beats as potential atrial fibrillation fragments to obtain candidate atrial fibrillation fragments;

b5, traversing each candidate atrial fibrillation fragment one by one, drawing to obtain a Lorenz difference scatter diagram by utilizing the step 3 of the training phase, obtaining a final classification result of the fragment through a trained model, if the fragment is an atrial fibrillation fragment, reserving the fragment, and if not, deleting the fragment;

b6, traversing all the candidate atrial fibrillation fragments and outputting the result of the atrial fibrillation analysis of the whole data.

Referring to fig. 2-7, the Lorenz difference scattergram is plotted by the following method: and taking the RR interval difference of the current heartbeat as a horizontal axis and the RR interval difference of the next heartbeat as a vertical axis to obtain a statistical point of the position. And traversing all heartbeats in sequence, and drawing the number of heartbeats at all positions of the scatter diagram according to the number to obtain the Lorenz scatter diagram. In addition, in the Lorenz difference scatter diagram of the drawing fragment, red is displayed at the position with the most heart beats, blue is displayed at the position with the least heart beats, and other positions are gradually changed from red to blue according to the quantity from large to small, so that the heart beat quantity information is added, and the atrial fibrillation detection accuracy is improved.

Compared with the prior art, the Lorenz difference scatter diagram is drawn as the core of the whole algorithm, the RR intervals of the time sequence are converted into the Lorenz difference scatter diagram, and the Lorenz difference scatter diagram is different from the common Lorenz difference scatter diagram in that: the ordinary Lorenz difference scatter diagram is not displayed in a color mode, but only displays the position with or without heart beat, and the number of heart beats at the position cannot be distinguished; the Lorenz difference scatter diagram drawn by the checking method is added with a heart beat number factor, red is displayed at the place with the most heart beats, blue is displayed at the place with the least heart beats, and other positions are gradually changed from red to blue according to the number from large to small, so that the quantity information is added, and the atrial fibrillation detection accuracy is improved.

It is further noted that, unless otherwise specifically stated or limited, the terms "QRS wave detection", "Lorenz difference scattergram", "RR interval", and the like are prior art, and those skilled in the art will understand the specific meaning of the above terms in the present invention according to specific situations.

The above embodiments are merely illustrative of the preferred embodiments of the present invention, and not restrictive, and various changes and modifications to the technical solutions of the present invention may be made by those skilled in the art without departing from the spirit of the present invention, and the technical solutions of the present invention are intended to fall within the scope of the present invention defined by the appended claims.