A method and device for identifying ventricular fibrillation, an electronic device and a storage medium

Abstract

The application relates to a method and device for identifying ventricular fibrillation, an electronic device and a storage medium. The method comprises collecting and storing biological signals, differentially sorting the biological signals, constructing a quantile diagram according to a set step length based on the differentially sorted signals, selecting a predetermined number of feature points in the quantile diagram to calculate a target curvature of a connecting line of the feature points, and identifying whether ventricular fibrillation exists according to a comparison between the target curvature and a set threshold. The method can improve signal fidelity in the feature extraction process, reduce the misleading effect of local interference on identification, and improve the accuracy of ventricular fibrillation identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of signal control, and in particular to a method and device for identifying ventricular flutter and fibrillation, an electronic device and a storage medium. BACKGROUND

[0002] Ventricular flutter and fibrillation are serious ectopic heart rhythms, and the ventricles lose effective overall contraction ability and are replaced by fast and uncoordinated fibrillation of each myocardial part. Ventricular flutter is very short in duration and will turn into ventricular fibrillation within a few minutes, which is a fatal arrhythmia before death. Ventricular fibrillation refers to the sudden loss of overall coordination of the patient's myocardium, which is irregularly contracted and presents a clinical dangerous state of being unable to pump blood. Since both are malignant, extremely severe and fatal arrhythmias, the hemodynamic changes and clinical manifestations caused by them are almost the same, and they are both cardiac arrest in terms of blood discharge function. After occurrence, they almost all cause asphyxia syndrome and sudden death in a very short time, so rapid and effective cardiopulmonary resuscitation and early electric defibrillation and rhythm treatment are needed.

[0003] The equipment commonly used for electric defibrillation and rhythm treatment is AED (automatic external defibrillator) or WCD (wearable defibrillator). The AED or WCD connects the electrode to the human body, collects electrocardiogram data, automatically analyzes the heart rhythm, and decides whether to perform electric shock defibrillation according to whether it is a malignant arrhythmia such as ventricular fibrillation or ventricular flutter.

[0004] How to provide an accurate identification method for identifying ventricular fibrillation and flutter has become a hot spot in the field of electric defibrillation and rhythm treatment. SUMMARY

[0005] Embodiments of the present application provide a target signal control method and device, an electronic device and a storage medium.

[0006] In a first aspect, embodiments of the present application provide a method for identifying ventricular fibrillation and flutter, comprising:

[0007] Collecting and storing biological signals;

[0008] Differentially sorting the biological signals;

[0009] Constructing a quantile plot according to a set step length for the differentially sorted signals;

[0010] Selecting a predetermined number of feature points in the quantile plot to calculate a target curvature of a connecting line of the feature points;

[0011] Identifying whether there is ventricular flutter and fibrillation according to a comparison of the target curvature with a set threshold.

[0012] The selecting a predetermined number of feature points in the quantile plot to calculate a target curvature of a connecting line of the feature points comprises:

[0013] selecting three segmentation points in a predetermined area of the quantile graph;

[0014] obtaining the ordinate corresponding to the three segmentation points;

[0015] connecting the three ordinates, and calculating the cosine of the angle with the segmentation point in the middle as the vertex;

[0016] taking the obtained cosine of the angle as the target curvature.

[0017] Optionally, the quantile graph is a percentile graph.

[0018] Optionally, the range of the segmentation points is above 40% and below 99%.

[0019] As another implementation, the selecting a predetermined number of feature points in the quantile graph to calculate the target curvature of the connecting line of the feature points comprises:

[0020] selecting a first feature point and a second feature point in the quantile graph;

[0021] calculating the slope of the connecting line of the first feature point and the second feature point;

[0022] taking the slope as the target curvature.

[0023] Optionally, the quantile graph is a percentile graph.

[0024] Optionally, the range of the segmentation points of the first feature point is 40%-80%, and the range of the segmentation points of the second feature point is 90%-99%, and more preferably, the range of the segmentation points of the second feature point is 95%-99%.

[0025] As still another implementation, the selecting a predetermined number of feature points in the quantile graph to calculate the target curvature of the connecting line of the feature points comprises:

[0026] selecting a plurality of feature points in the quantile graph;

[0027] calculating Feature(i)=SECG(Ai1) / SECG(Ai0) for each of the plurality of feature points, where Ai0 is a high percentile point and Ai1 is a low percentile point;

[0028] inputting the calculated plurality of Feature(i) into a pre-trained learning model to learn and output the target curvature.

[0029] As still another implementation, the selecting a predetermined number of feature points in the quantile graph to calculate the target curvature of the connecting line of the feature points comprises:

[0030] The longitudinal coordinate values corresponding to all the segmentation points in the quantile graph are taken as deep learning input data, and are input into a deep learning model for learning, and a target curvature is output.

[0031] As a preferred mode, the differential sorting of the biological signal comprises:

[0032] Differential sorting of the biological signal by a predetermined step size;

[0033] Taking absolute values of the differential array data and recording to a differential buffer;

[0034] The data in the differential buffer are sorted in ascending order and recorded to a sorting buffer.

[0035] As a preferred mode, the differential sorting of the biological signal comprises:

[0036] Setting a quantile step size;

[0037] Taking the quantile sequence obtained according to the quantile step size as the horizontal coordinate;

[0038] Discretizing each array data after taking absolute values into a predetermined number of values, the predetermined number being the same as the number of segmentation points of the quantile sequence;

[0039] Corresponding the discrete values to the horizontal coordinates one by one and taking them as the quantile amplitude values of the horizontal coordinates to form the quantile graph.

[0040] In a second aspect, an embodiment of the present application provides a ventricular fibrillation and flutter recognition device, comprising:

[0041] A storage unit for collecting and storing biological signals;

[0042] A differential sorting unit for differentially sorting the biological signals;

[0043] A quantile graph construction unit for constructing a quantile graph according to a set step size from the differentially sorted signals;

[0044] A calculation unit for selecting a predetermined number of feature points in the quantile graph to calculate a target curvature of the connecting line of the feature points;

[0045] A recognition unit for recognizing whether there is ventricular fibrillation and flutter according to a comparison of the target curvature with a set threshold value.

[0046] In a third aspect, an embodiment of the present application provides an electronic device comprising a memory, a processor, a bus, and a computer program stored in the memory and executable on the processor, wherein the processor implements the ventricular fibrillation and flutter recognition method as described in the first aspect when executing the computer program.

[0047] In a fourth aspect, an embodiment of the present application provides a non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method for identifying ventricular fibrillation according to the first aspect.

[0048] According to the technical solution of the present application, in the technical solution of the present application, the biological signal is segmented into biological signal data segments, the data in each biological signal data segment is sorted by difference to construct a quantile graph, and the amplitudes of the biological signal in the quantile graph are sequentially sorted, and the arrangement order is usually from low-amplitude signal to high-amplitude signal. Since the energy feature of the large-amplitude body motion interference signal is higher than that of the biological signal, the large-amplitude body motion interference signal will be arranged to the rightmost side of the quantile graph according to the order. In the present application, the target curvature (curve bending rate) on the quantile graph is used as the identification feature based on the left side of the quantile graph, or an approximate feature reflecting the curve bending rate is extracted as the feature identification of ventricular fibrillation. Since the high-amplitude signal, i.e. the large-amplitude body motion interference signal, located at the rightmost side of the quantile graph is filtered out, the signal fidelity in the feature extraction process is improved, the local interference is reduced, and the accuracy of ventricular fibrillation identification is improved.

[0049] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present application, nor to limit the scope of the present application. Other features of the present application will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0050] The accompanying drawings are used to better understand the present application and do not limit the present application. Among them:

[0051] Figure 1 shows a flowchart of an embodiment of the ventricular fibrillation identification method of the present application;

[0052] Figure 2 is a quantile graph of an embodiment of the ventricular fibrillation identification method according to the present application;

[0053] Figure 3 is Figure 2 is a distribution diagram of the segmentation points of the quantile graph in the embodiment shown;

[0054] Figure 4 is a structure diagram of an embodiment of the ventricular fibrillation identification device of the present application;

[0055] Figure 5 shows an exemplary system architecture diagram applying the ventricular fibrillation identification method;

[0056] Figure 6Structure diagram of an embodiment of an electronic device of the present application. DETAILED DESCRIPTION

[0057] It should be noted that the following detailed description is merely exemplary in nature and is intended to provide further description of the application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0058] It is to be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, components, and / or groups thereof, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.

[0059] The embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0060] In order to make the technical solutions and advantages of the present application clearer, the following will further describe the present application in detail with reference to the drawings and specific embodiments.

[0061] Electrical cardioversion therapy needs to be based on the measurement of the distribution and complexity of physiological signals. Currently, sample entropy is applied in the evaluation of the complexity of physiological time series (EEG, sEMG, etc.) and the diagnosis of pathological states. Sample entropy (SampEn) measures the complexity of time series by measuring the probability of generating new patterns in the signal. The greater the probability of generating new patterns, the greater the complexity of the sequence. That is, the lower the value of sample entropy, the higher the self-similarity of the sequence. The greater the value of sample entropy, the more complex the sample sequence.

[0062] When processing biological signals such as heart rate or blood pressure, by calculating the sample entropy of each segment of the entire time period, a sample entropy trend sequence can be generated, which helps to observe the complexity changes of biological signals such as heart rate or blood pressure. For example, irregular high-frequency fluctuations may result in a larger sample entropy, while a large fluctuation in heart rate may cause the sample entropy to decrease.

[0063] Specifically, when sample entropy is used to process data to identify ventricular fibrillation, the usual practice is to divide the entire analysis data segment into multiple segments according to the time length, and to calculate certain energy features for the multiple segments. Due to the differences in the scales of energy features, many methods also standardize the energy features, such as selecting the maximum value in the multiple segments to standardize the data, or selecting the maximum value in the multiple segments as the basis for filtering other data. The standardized values will be used as hyperparameters for data learning.

[0064] When the patient has clinical symptoms such as ventricular fibrillation and ventricular flutter, it is possible that the patient will have convulsions and other body movements due to discomfort, which will cause the physiological signals collected to have occasional large body movement interference. When the signal is analyzed by using the sample entropy method, two super parameter settings are needed, one is m, and the other is r; m is a certain step super parameter used in calculation. r is a threshold value, generally selected as 0.1-0.25 times the standard deviation of the signal. When calculating the sample entropy, a certain transformation is performed based on the signal, and then the transformed value is compared with the threshold value, and then the next step is processed. When the energy characteristics of the large body movement interference signal are higher than those of the heart signal, the fluctuation characteristics of the large body movement interference signal will affect the standard deviation of the signal. Moreover, the calculation method of the standard deviation is characterized in that even a small value with a large amplitude will have a great influence on the result, and thus will greatly affect the threshold value r. Therefore, in the ventricular fibrillation and ventricular flutter recognition method using sample entropy, the standard deviation characteristics of the large body movement interference signal will interfere with the threshold value calculation, resulting in errors in subsequent calculation, and thus affecting the accurate recognition of ventricular fibrillation and ventricular flutter.

[0065] Figure 1 The flow of one embodiment of the ventricular fibrillation and ventricular flutter recognition method of the present application is shown. As shown in FIG. 1, the ventricular fibrillation and ventricular flutter recognition method of the present embodiment can include the following steps: Figure 1

[0066] Step 101, collect and store the biological signal.

[0067] Specifically, the biological signal includes but is not limited to an electrocardiogram signal, a blood pressure signal, etc. The biological signal is collected in real time and stored in the original data buffer OECG.

[0068] For real-time data collection, data segments can be collected. Specifically, when real-time data collection starts, a counter can be used for counting, and when the collected data reaches a predetermined amount, the collection of the data segment is stopped, the collected data segment is stored in the data buffer and is prepared for the next step of processing, and the collection unit for collecting the biological signal is re-counted to start the collection of the next data segment.

[0069] Further, the data segment stored in the original data buffer OECG can be pre-processed, and the pre-processing includes but is not limited to filtering. As one of the optional schemes, when the biological signal is an electrocardiogram signal, a Butterworth filter of 5-20 Hz can be selected for digital filtering of the data. After data pre-processing, the data can remove power frequency interference, electromyographic interference, baseline drift and other problems, so that the biological signal is more pure, and the subsequent analysis and comparison of the signal are facilitated. The pre-processed data segment can be stored in the pre-processing buffer MECG.

[0070] ​Step 102, differentially sorting the biosignal.

[0071] The biosignal data segment stored in the original data buffer OECG or the original data buffer OECG by step 101 is differentially sorted by a predetermined step length, and then the absolute value of the differentially sorted array data is taken and recorded to the differential buffer DECG. The data in the differential buffer DECG is sorted in ascending order and recorded to the arrangement buffer SECG.

[0072] In a specific embodiment, the biosignal data segment in the original data buffer OECG or the original data buffer OECG is differentially sorted by a step length of m (including but not limited to 4 milliseconds). The collection of the biosignal can be considered as time series data, and the differential model is not specifically limited in the present application. Any model that can eliminate fluctuation data in the original data buffer OECG or the original data buffer OECG and make the data more stable can be used to implement the differential processing in the present application.

[0073] Step 502, constructing a quantile plot according to a set step length for the differentially sorted signal.

[0074] Specifically, in this step, the quantile step length can be set first. The quantile sequence obtained according to the quantile step length is taken as the horizontal coordinate; each array data after taking the absolute value is discretized into a predetermined number of values, and the predetermined number is the same as the number of the segmentation points of the quantile sequence; the discretized values are one-to-one corresponding to the horizontal coordinates and taken as the quantile amplitude of the horizontal coordinates to form the quantile plot.

[0075] Taking 1% as an example, the quantile step length is 1%, and the quantile sequence is 1%, 2%, 3%,..., 100%. The number of segmentation points of the quantile sequence is 100.

[0076] The quantile sequence obtained according to the quantile step length is taken as the horizontal coordinate Xqq=1%, 2%,..., 100%.

[0077] The data in the arrangement buffer SECG is discretized into 100 values, which are taken as the quantile amplitude corresponding to the horizontal coordinates and taken as the vertical coordinates Yqq=SECG(Xqq), Xqq as the horizontal coordinate, Yqq as the vertical coordinate, and the horizontal and vertical coordinates constitute the quantile plot, please refer to Figure 2 .

[0078] It should be noted that the quantile step length of 1% is only exemplary, such as 1 / 20, 1 / 50, 1 / 200, etc., which can be applied to the production of the quantile plot. In order to ensure the collection density of the quantile amplitude, the quantile step length is preferably less than 1 / 20.

[0079] Step 503, selecting a predetermined number of feature points in the quantile graph to calculate the target curvature of the connecting line of the feature points.

[0080] In one embodiment of obtaining the target curvature, specifically, three segmentation points are selected in a predetermined region of the quantile graph; the ordinate corresponding to the three segmentation points is obtained; the three ordinates are connected, and the angle cosine is calculated with the segmentation point in the middle position as the angle point; and the obtained angle cosine is taken as the target curvature.

[0081] For example, the target curvature is calculated in detail with the percent graph shown in FIG. 8. Figure 2

[0082] Three segmentation points corresponding to the horizontal coordinates of 80%, 60% and 40% in the quantile graph are selected, that is, the feature points A, B and C corresponding to the three ordinates, as shown in FIG. 8. Figure 3 The angle cosine value is calculated with 60% as the angle point, and is taken as the target curvature. Specifically, the following steps are taken:

[0083] A. Coordinate calculation: the quantile number Ax is substituted into the sorting buffer SECG(Ax) to obtain the ordinate, and thus the coordinates of the three feature points (X80, Y80), (X60, Y60) and (X40, Y40) are obtained.

[0084] B. Side length calculation:

[0085] a = ((X80-X60)^2+(Y80-Y60)^2)^.05;

[0086] b = ((X40-X60)^2+(Y40-Y60)^2)^.05;

[0087] c = ((X80-X40)^2+(Y80-Y40)^2)^.05.

[0088] C. Angle cosine calculation:

[0089] Cosa = (a^2+b^2-c^2) / (2*a*b).

[0090] The obtained angle cosine value is taken as the target curvature.

[0091] Step 105, identifying whether there is ventricular fibrillation or ventricular flutter according to the comparison of the target curvature with the set threshold value.

[0092] The angle Thr = 150 degrees of the angle corresponding to the angle point is taken as the set threshold value,

[0093] If Cosa>cos(2*pi*150 / 360) is established, it is identified as ventricular fibrillation or ventricular flutter, otherwise it is identified as non-ventricular fibrillation or ventricular flutter. ​

[0094] wherein pi is pi, pi = 3.1415926.

[0095] It should be noted that the selection of Thr = 150 degrees as the threshold is only exemplary, and Thr can also be selected as any angle value greater than 150 degrees and less than 180 degrees.

[0096] In this embodiment, the selection range of the three segmentation points is within the range of 40% to 99%. The original biological signal has signals with different amplitudes and large fluctuation signals. By difference processing, a part of the noise signal can be filtered, and after the data is subjected to absolute value and sorting, the biological signals in each biological signal data segment are sequentially arranged. Referring to Figure 2 , the amplitudes of the biological signals increase from left to right, and the signal with the largest amplitude is located at the rightmost side of the quantile plot. When the biological signal is an electrocardiogram signal, and the patient has a larger body movement due to the occurrence of ventricular fibrillation, the amplitude of the electrocardiogram signal will be larger than the normal electrocardiogram amplitude, corresponding to Figure 2 , the amplitude of the signal is often located at the rightmost side of the quantile plot. According to the characteristics of the distribution graph of the experimental analysis of the interference, usually, such signals are located between 99% and 100%, and in this embodiment, the segmentation points in the range of 40% to 99% are selected. This value range can effectively exclude the signals of larger body movements, so as to make the identification of ventricular fibrillation and ventricular flutter more accurate.

[0097] It should be noted that the method for obtaining the target curvature in the present application is not limited to the method in step 104, and other various methods can also be used to obtain the target curvature.

[0098] In another embodiment of obtaining the target curvature, the method for obtaining the target curvature comprises:

[0099] selecting a first feature point and a second feature point in the quantile plot; calculating the slope of the line connecting the first feature point and the second feature point; and taking the slope as the target curvature.

[0100] In an implementable scheme, the quantile plot can be a percentile plot, which can be referred to in Figure 2 When the quantile plot is a percentile plot, the segmentation point of the first feature point is selected in the range of 40% to 80%, and the segmentation point of the second feature point is selected in the range of 90% to 99%, and more preferably, the segmentation point of the second feature point is selected in the range of 95% to 99%. In this embodiment, the slope is used to represent the change rate of the fluctuation size difference, and since the biological signals in each biological signal data segment are sequentially arranged, referring to Figure 2, the amplitude of the biosignal increases from left to right, and the signal with the largest amplitude is located at the rightmost side of the quantile plot. In the range of 40%-80%, it belongs to the stationary section of the signal, and in the range of 90%-99%, it belongs to the fluctuation section. According to the characteristics of the distribution diagram of the interference condition analyzed through the experiment, it is found that the signal generated when the patient has a larger body movement due to the occurrence of ventricular flutter and ventricular fibrillation is usually located between 99%-100%. Therefore, when the second feature point selects a value in the range of 95%-99%, the abnormal signal can be excluded. At the same time, the high-amplitude electrocardiogram signal is usually concentrated in the range of 95%-99%, so whether ventricular fibrillation or ventricular flutter occurs can be accurately identified.

[0101] Specifically, taking the quantile plot shown in Figure 2 as an example, two percentile points and the quantile values corresponding to the two percentile points are selected. Among them, the first percentile point A0 is the 95% percentile point, and A1 is the 65% percentile point. Substituting the formula Feature=SECG(A0) / SECG(A1) into the formula, the result Feature calculated is used as the target curvature.

[0102] The target curvature is compared with a set threshold value. When the calculation result of Feature is greater than the threshold value, it is identified as ventricular fibrillation or ventricular flutter, otherwise it is identified as non-.

[0103] In another embodiment of obtaining the target curvature, the method for obtaining the target curvature comprises:

[0104] Selecting a predetermined number of feature points in the quantile plot to calculate the target curvature of the connecting line of the feature points comprises:

[0105] Selecting a plurality of feature points in the quantile plot; substituting the plurality of feature points into

[0106] Feature(i)=SECG(Ai1) / SECG(Ai0) respectively; wherein Ai0 is a high percentile point, and Ai1 is a low percentile point; the plurality of Feature(i) calculated is brought into a pre-trained learning model for learning, and the target curvature is output.

[0107] In this embodiment, the selection principle of Ai0 and Ai1 is that the percentile of Ai0 is only greater than the percentile of Ai1.

[0108] In some implementations, the plurality of Feature(i) can be brought into a machine trained in advance, including but not limited to a model such as a support vector machine (SVM) and an adaboost model.

[0109] In another embodiment of obtaining the target curvature, the method for obtaining the target curvature comprises:

[0110] The longitudinal coordinate values corresponding to all the segmentation points in the quantile graph are taken as deep learning input data, and are input into a deep learning model for learning, and a target curvature is output.

[0111] Specifically, still taking 1% as an example, the quantile step is 1%, and the quantile sequence is 1%, 2%, 3%,..., 100%. The number of segmentation points in the quantile sequence is 100.

[0112] The quantile sequence obtained according to the quantile step is taken as the abscissa Xqq=1%, 2%,..., 100%.

[0113] The data in the arrangement buffer SECG is discretized into 100 values, and the 100 values are taken as the ordinate Yqq=SECG(Xqq) corresponding to the abscissa Xqq and as the quantile amplitude of the abscissa, please refer to Figure 2 .

[0114] Yqq is taken as the original data, and is input into a trained CNN (convolutional neural network), DNN (deep neural network) or other deep learning model, and a calculation result of a target curvature is output.

[0115] According to the above technical solution, the biological signal is segmented into biological signal data segments, the data in each biological signal data segment is differentially sorted to construct a quantile graph, and the amplitudes of the biological signals in the quantile graph are sequentially sorted, and the arrangement order is usually from low-amplitude signals to high-amplitude signals. Since the energy characteristics of large-amplitude body motion interference signals are higher than those of biological signals, the large-amplitude body motion interference signals will be arranged to the rightmost side of the quantile graph. The target curvature (curve bending rate) on the quantile graph on the left side of the application is taken as a recognition feature, or an approximate feature reflecting the curve bending rate is extracted as a feature recognition of ventricular fibrillation and ventricular flutter. Since the high-amplitude signals, i.e. large-amplitude body motion interference signals, on the rightmost side of the quantile graph are filtered out, the recognition method in the application improves the signal fidelity in the feature extraction process, reduces the misleading of local interference to recognition, and improves the accuracy of ventricular flutter and ventricular fibrillation recognition.

[0116] In addition, since the ventricular fibrillation and ventricular flutter recognition method in the application adopts data processing on biological signal data segments, compared with sample entropy which needs to select multiple segments and select the maximum value in multiple cuttings for data screening, the method shown in the application has the advantages of saving operation time and improving recognition efficiency.

[0117] The CUDB (CU ventricular tachyarrhythmia database) is used as a database for evaluation. The sample entropy feature and the quantile graph feature in the application are used as algorithm inputs. The classification accuracy of the quantile graph feature is about 10% higher than that of the sample entropy feature.

[0118] For the calculation of the accuracy, a confusion matrix can be used. The confusion matrix is a visualization tool mainly used for comparing the classification results and the actual measured values, and the accuracy of the classification results can be displayed in a confusion matrix.

[0119] The structure of the confusion matrix is generally shown in Table 1.

[0120] Table 1:

[0121]

[0122] In the confusion matrix, four symbols TP (True Positives), FP (False Positives), TN (True Negatives), and FN (False Negatives) are included, and the corresponding meanings are as follows: the sample is positive, the prediction result is positive, the sample is negative, the prediction result is positive, the sample is negative, and the prediction result is negative. The sample is positive, and the prediction result is negative.

[0123] The accuracy is the most commonly used classification performance indicator. It can be used to represent the accuracy of the model, i.e., the number of correct identifications by the model / the total number of samples. Generally, the higher the accuracy of the model, the better the effect of the model. The formula for calculating the accuracy using the confusion matrix is:

[0124] accuracy = (TP + TN) / (TP + FN + FP + TN)

[0125] Taking the total number of selected samples 745 as an example, the sample entropy feature and the quantile graph feature described in the application are used as algorithm inputs for accuracy calculation. The corresponding confusion matrices are shown in Table 2 and Table 3. Table 2 is the confusion matrix of the quantile graph feature, and Table 3 is the confusion matrix of the sample entropy feature.

[0126]

[0127] Substituting the above accuracy calculation formula, the accuracy is 0.87 and 0.79, respectively, which is about 10% higher. Therefore, the recognition method using the quantile graph feature in the application has better recognition effect.

[0128] Further reference Figure 4As an implementation of the method shown in the above figures, the application provides an embodiment of a device for identifying ventricular fibrillation and flutter, which corresponds to the method embodiment shown in Figure 1 The device can be applied in various electronic devices.

[0129] As shown in Figure 4 The control device 400 of the target signal in the embodiment includes a storage unit 401, a differential sorting unit 402, a quantile graph construction unit 403, a calculation unit 404, and an identification unit 405.

[0130] The storage unit 401 is configured to collect and store biological signals.

[0131] The differential sorting unit 402 is configured to perform differential sorting on the biological signals.

[0132] The quantile graph construction unit 403 is configured to construct a quantile graph according to a set step length based on the signals after differential sorting.

[0133] The calculation unit 404 is configured to select a predetermined number of feature points in the quantile graph to calculate a target curvature of a connecting line of the feature points.

[0134] The identification unit 405 is configured to identify whether ventricular fibrillation or flutter exists according to a comparison between the target curvature and a set threshold.

[0135] Figure 5 An exemplary system architecture 500 to which the method for identifying ventricular fibrillation and flutter can be applied is shown.

[0136] As shown in Figure 1 The system architecture 500 can include a biological signal collection device 501 and a terminal device 502. The biological signal collection device 501 and the terminal device 502 are connected through a network or a cable medium to realize data transmission.

[0137] The biological signal collection device 501 is configured to collect biological signals of a target object at the same time. It can be understood that the biological signal collection device 501 is configured to collect different types of biological signals, for example, the biological signal collection device 501 is configured to collect electrocardiogram signals, and can also be configured to collect blood pressure signals or venous signals, etc.

[0138] The biological signal collection device 501 can send the collected signals to the terminal device 502 in real time. After receiving the signals, the terminal device 502 can process the signals to identify the target signal in the biological signals.

[0139] The terminal device 502 can be hardware or software. When the terminal device 502 is hardware, it can be various electronic devices including but not limited to a tablet computer, a laptop computer, a desktop computer, and the like. When the terminal device 502 is software, it can be installed in the above-mentioned electronic devices. It can be implemented as multiple software or software modules (for example, used to provide a distributed service) or as a single software or software module. No specific limitation is made herein.

[0140] In some other applications, the above-mentioned system architecture can further include a database 503. The database 503 can be used to store the signals collected by the biological signal collection device 501, so that the terminal device 502 can use the above-mentioned signals for learning at an appropriate time.

[0141] It should be noted that the signal processing method provided by the embodiments of the present application is generally executed by the terminal device 502. Correspondingly, the signal processing apparatus is generally arranged in the terminal device 502.

[0142] It should be understood that, Figure 1 The number of biological signal collection devices, terminal devices, and databases in the above-mentioned system architecture is only illustrative. According to the implementation needs, there can be any number of biological signal collection devices, terminal devices, and databases.

[0143] In addition, in the technical solutions of the present application, an electronic device is also provided.

[0144] Figure 6 A structural schematic diagram of an electronic device provided by an embodiment of the present application is shown.

[0145] As Figure 6 shown, the electronic device can include a processor 601, a memory 602, a bus 603, and a computer program stored in the memory 602 and executable on the processor 601, wherein the processor 601 and the memory 602 complete mutual communication through the bus 603. The processor 601 executes the computer program to implement the steps of the above-mentioned method, for example, including: collecting and storing biological signals; differentially sorting the biological signals; constructing a quantile plot according to a set step length for the differentially sorted signals; selecting a predetermined number of feature points in the quantile plot to calculate a target curvature of a connecting line of the feature points; and identifying whether ventricular fibrillation exists according to a comparison of the target curvature with a set threshold.

[0146] In addition, an embodiment of the present application further provides a non-transitory computer-readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement steps of the above method, for example, comprising: acquiring a first biological signal and a second biological signal of a target object in real time; determining whether a first preset condition is met according to the first biological signal; determining whether a second preset condition is met according to the second biological signal; in response to determining that the first preset condition and the second preset condition are both met, determining a predicted trigger time according to the first biological signal and / or the second biological signal; adjusting the predicted trigger time according to a target time offset obtained in advance to obtain a target trigger time; and performing trigger control on the target signal according to the target trigger time.

[0147] In summary, in the technical solution of the present application, the biological signal is divided into biological signal data segments, the data in each biological signal data segment is differentially sorted to construct a quantile graph, and the amplitude of the biological signal in the quantile graph is sequentially sorted, and the arrangement order is usually from low amplitude signal to high amplitude signal. Since the energy feature of the large amplitude body motion interference signal is higher than that of the biological signal, the large amplitude body motion interference signal will be arranged to the rightmost side of the quantile graph according to the order. The left side of the quantile graph in the present application uses the target curvature (curve bending rate) on the quantile graph as the recognition feature, or extracts an approximate feature that can reflect the curve bending rate as the feature recognition of ventricular fibrillation and ventricular flutter. Since the high amplitude signal, i.e. the large amplitude body motion interference signal, located at the rightmost side of the quantile graph is filtered out, the recognition method in the present application improves the signal fidelity in the feature extraction process, reduces the misleading of local interference to recognition, and improves the accuracy of ventricular flutter and ventricular fibrillation recognition.

[0148] The above only describes preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for identifying ventricular fibrillation, comprising: collecting and storing a biological signal; differentially sorting the biological signal; constructing a quantile plot according to a set step length for the differentially sorted signal; selecting a predetermined number of feature points in the quantile plot to calculate a target curvature of a line connecting the feature points; identifying whether ventricular fibrillation exists according to a comparison between the target curvature and a set threshold value; wherein the differentially sorting the biological signal comprises: differentially sorting the biological signal according to a predetermined step length; taking absolute values of the differentially sorted array data and recording them to a differential buffer; sorting the data in the differential buffer from small to large and recording them to a sorting buffer SECG.

2. The method of claim 1, wherein, The selecting a predetermined number of feature points in the quantile plot to calculate a target curvature of a line connecting the feature points comprises: selecting three split points in a predetermined area of the quantile plot; obtaining longitudinal coordinates corresponding to the three split points; connecting the three longitudinal coordinates and calculating a cosine of an angle with the split point in the middle as a vertex; taking the obtained cosine of the angle as the target curvature.

3. The method of claim 2, wherein, The quantile plot is a percentile plot.

4. The method of claim 2, wherein, The split points are selected in a range of more than 40% and less than 99%.

5. The method of claim 1, wherein, The selecting a predetermined number of feature points in the quantile plot to calculate a target curvature of a line connecting the feature points comprises: selecting a first feature point and a second feature point in the quantile plot; calculating a slope of a line connecting the first feature point and the second feature point; taking the slope as the target curvature.

6. The method of claim 5, wherein, The quantile plot is a percentile plot.

7. The method of claim 5, wherein, The first feature point is selected in a range of 40%-80% and the second feature point is selected in a range of 90%-99%.

8. The method of claim 7, wherein, The second feature point is selected in a range of 95%-99%.

9. The method of claim 5, wherein, The selecting a predetermined number of feature points in the quantile plot to calculate a target curvature of a line connecting the feature points comprises: selecting a plurality of feature points in the quantile plot; calculating Feature(i)=SECG(Ai1) / SECG(Ai0) for each of the plurality of feature points, where Ai0 is a high percentile point and Ai1 is a low percentile point; inputting the calculated plurality of Feature(i) into a pre-trained learning model for learning and outputting a target curvature.

10. The method of claim 1, wherein, The selecting a predetermined number of feature points in the quantile plot to calculate a target curvature of a line connecting the feature points comprises: taking longitudinal coordinate values corresponding to all split points in the quantile plot as deep learning input data and inputting them into a deep learning model for learning and outputting a target curvature.

11. The method of claim 1, wherein, The constructing a quantile plot according to a set step length for the differentially sorted signal comprises: setting a quantile step length; taking a quantile sequence obtained according to the quantile step length as a horizontal coordinate; discretizing each array data after taking absolute values into a predetermined number of values, the predetermined number being the same as the number of split points of the quantile sequence; corresponding the discretized values to the horizontal coordinate one by one and taking them as quantile amplitudes of the horizontal coordinate to form the quantile plot. 12.A device for identifying ventricular fibrillation, comprising: a storage unit configured to collect and store a biological signal; a differential sorting unit configured to sort the biological signal in a differential manner; a quantile map construction unit configured to construct a quantile map of the sorted signal according to a set step size; a calculation unit configured to select a predetermined number of feature points in the quantile map, and calculate a target curvature of a line connecting the feature points; an identification unit configured to identify whether there is a ventricular fibrillation according to a comparison between the target curvature and a set threshold value; the differential sorting of the biological signal comprises: differentially sorting the biological signal according to a predetermined step size; taking an absolute value of the array data after the differential sorting, and recording the absolute value to a differential buffer; sorting the data in the differential buffer from small to large, and recording the sorted data to a sorting buffer SECG.

13. An electronic device comprising a memory, a processor, a bus, and a computer program stored on the memory and executable on the processor, wherein, The processor executes the computer program to implement the ventricular fibrillation identification method according to any one of claims 1 to 11.

14. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the ventricular fibrillation identification method according to any one of claims 1 to 11.

Images