Electrocardio data processing and coronary heart disease risk prediction method, device, system and medium

Abstract

A method, device, system and medium for electrocardio data processing and coronary heart disease risk prediction, motion electrocardio data output through at least one electrocardio lead is acquired; the motion electrocardio data is analyzed to obtain a heart rate sequence corresponding to each electrocardio lead; each QRS complex corresponding to each heartbeat is extracted from the motion electrocardio data; a plurality of heart rate intervals are divided according to the obtained heart rate sequence; each QRS complex is classified into a group corresponding to a heart rate interval according to the heart rate interval to which the heart rate corresponding to the QRS complex belongs; the average QRS complex corresponding to each heart rate interval is determined according to all QRS complexes in each group; the high-frequency energy of the average QRS complex corresponding to each heart rate interval is extracted; and an energy sequence is determined according to the high-frequency energy corresponding to each heart rate interval. The present scheme uses high-frequency components and dynamic load information which are more sensitive to ischemia, and can significantly improve the detection rate and discrimination ability for coronary heart disease, especially early and mild lesions.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical instruments, in particular to an electrocardiogram data processing method, a coronary heart disease risk prediction method, a processing device, an electrocardiogram data processing system and a computer readable storage medium. BACKGROUND

[0002] With the continuous improvement of people's living standards and the continuous increase of work pressure, heart disease is becoming more and more young and universal, and heart health problems are also attracting more and more attention and attention. Therefore, how to accurately identify the heart health status is a problem worth attention. Taking coronary heart disease as an example, early and accurate screening of coronary heart disease is crucial to reduce the risk of cardiovascular events. The current clinical diagnosis "gold standard" is coronary angiography (CAG), but this method belongs to invasive examination, which is high in cost and has certain risks. Therefore, exercise electrocardiogram (Exercise ECG) as a typical non-invasive and low-cost screening method is still widely used in clinical. The traditional exercise electrocardiogram examination method mainly relies on the low-frequency morphological changes such as horizontal type or downward sloping depression of ST segment. However, this method has the limitations of insufficient sensitivity and specificity, poor anti-interference ability, and dependence on subjective experience. SUMMARY

[0003] The purpose of the embodiments of the present application is to provide an electrocardiogram data processing method, a coronary heart disease risk prediction method, a processing device, an electrocardiogram data processing system and a computer readable storage medium.

[0004] In order to achieve the above-mentioned purpose, the first aspect of the present application provides an electrocardiogram data processing method, comprising:

[0005] Obtaining exercise electrocardiogram data output through at least one electrocardiogram lead;

[0006] Analyzing the exercise electrocardiogram data to obtain a heart rate sequence corresponding to each electrocardiogram lead;

[0007] Extracting a QRS complex corresponding to each heartbeat from the exercise electrocardiogram data;

[0008] Dividing a plurality of heart rate intervals according to the obtained heart rate sequence;

[0009] According to the heart rate interval to which the heart rate corresponding to each QRS complex belongs, the QRS complex is classified into a group corresponding to the corresponding heart rate interval;

[0010] Determining an average QRS complex corresponding to each heart rate interval according to all QRS complexes in each group;

[0011] Extracting the high-frequency energy of the average QRS complex corresponding to each heart rate interval;

[0012] Determine an energy sequence according to the high frequency energy corresponding to each heart rate interval.

[0013] In the embodiments of the present application, analyzing the exercise ECG data to obtain the heart rate sequence corresponding to each ECG lead comprises:

[0014] Detecting the R-wave peak position corresponding to each heartbeat in the exercise ECG data corresponding to each ECG lead;

[0015] Determining the time interval between the adjacent two R-waves according to the peak positions of the adjacent two R-waves, and determining the corresponding heart rate according to the time interval, to obtain the heart rate sequence corresponding to the exercise ECG data corresponding to each ECG lead.

[0016] In the embodiments of the present application, determining the time interval between the adjacent two R-waves according to the peak positions of the adjacent two R-waves, and determining the corresponding heart rate according to the time interval, to obtain the heart rate sequence corresponding to the exercise ECG data corresponding to each ECG lead, comprises:

[0017] Taking the corresponding heart rate as the heart rate corresponding to the QRS complex to which the former R-wave in the adjacent two R-waves belongs; and

[0018] Taking the average of the first preset number of heart rates at the end position in the exercise ECG data as the heart rate corresponding to the last QRS complex.

[0019] In the embodiments of the present application, determining the time interval between the adjacent two R-waves according to the peak positions of the adjacent two R-waves, and determining the corresponding heart rate according to the time interval, to obtain the heart rate sequence corresponding to the exercise ECG data corresponding to each ECG lead, comprises:

[0020] Taking the corresponding heart rate as the heart rate corresponding to the QRS complex to which the latter R-wave in the adjacent two R-waves belongs; and

[0021] Taking the average of the second preset number of heart rates at the start position in the exercise ECG data as the heart rate corresponding to the first QRS complex.

[0022] In the embodiments of the present application, the energy sequence comprises a plurality of high frequency energies, and each high frequency energy corresponds to a representative heart rate value of each heart rate interval.

[0023] In the embodiments of the present application, the representative heart rate value comprises any one of the following:

[0024] The lower limit value of the heart rate interval, the upper limit value of the heart rate interval, and the median value of the heart rate interval.

[0025] In the embodiments of the present application, the preset time period covers the entire exercise load detection process.

[0026] The second aspect of the present application provides a coronary heart disease risk prediction method, comprising:

[0027] Based on the trained coronary heart disease risk prediction model, the coronary heart disease risk is predicted according to the energy sequences corresponding to all electrocardiogram leads obtained according to the electrocardiogram data processing method.

[0028] In the embodiments of the present application, the coronary heart disease risk prediction model uses an L1 regularization logistic regression model, and the prediction of the coronary heart disease risk using the energy sequences corresponding to all electrocardiogram leads includes:

[0029] The target features are determined according to the energy sequences, and the target features include at least one of statistical features and morphological features;

[0030] The target features are normalized;

[0031] The normalized target features are input into the coronary heart disease risk prediction model to predict the coronary heart disease risk.

[0032] In the embodiments of the present application, the coronary heart disease risk prediction model uses a neural network model, and the prediction of the coronary heart disease risk using the energy sequences corresponding to all electrocardiogram leads includes:

[0033] The energy sequences and the demographic features are input into the coronary heart disease risk prediction model to predict the coronary heart disease risk.

[0034] The third aspect of the present application provides a processing device, including a processor and a memory, the memory storing instructions, and the processor being configured to call and execute the instructions from the memory to implement the electrocardiogram data processing method described above.

[0035] The fourth aspect of the present application provides a processing device, including a processor and a memory, the memory storing instructions, and the processor being configured to call and execute the instructions from the memory to implement the coronary heart disease risk prediction method described above.

[0036] The fifth aspect of the present application provides an electrocardiogram data processing system, including:

[0037] An electrocardiogram signal acquisition device including at least one electrode forming at least one electrocardiogram lead for acquiring electrocardiogram signals of a subject; and

[0038] The processing device described above.

[0039] The sixth aspect of the present application provides a machine-readable storage medium, the machine-readable storage medium storing instructions for causing a machine to execute the electrocardiogram data processing method described above.

[0040] The seventh aspect of the present application provides a machine-readable storage medium, the machine-readable storage medium storing instructions for causing a machine to execute the coronary heart disease risk prediction method described above.

[0041] The scheme provided by the embodiments of the present application can significantly improve the detection rate (sensitivity) and identification ability (specificity) for coronary heart disease, especially early and mild lesions, by using the high-frequency component more sensitive to ischemia and dynamic load information. In addition, the two-dimensional dynamic trend graph with heart rate as the horizontal axis and high-frequency energy as the vertical axis is constructed by rearranging, segmenting and averaging the whole process of exercise electrocardio signals according to the heart rate interval (not time). The trend graph analysis focuses on the overall change pattern and is not sensitive to accidental errors of individual data points. The method directly relates the cardiac load (heart rate) to the myocardial electrophysiological response (high-frequency energy), highlighting the physiological significance. In each heart rate interval, time registration and superimposed averaging are performed on multiple QRS waves, which greatly suppresses random noise and motion artifacts and highlights the stable changes of the high-frequency component related to heart rate.

[0042] Other features and advantages of the embodiments of the present application will be described in detail in the following specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0043] The accompanying drawings are included to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the embodiments of the present application together with the following specific embodiments, but do not constitute a limitation of the embodiments of the present application. In the drawings:

[0044] Figure 1 An example flowchart of an electrocardio data processing method according to the embodiments of the present application is schematically shown.

[0045] Figure 2 An example block diagram of an electrocardio data processing system according to the embodiments of the present application is schematically shown. DETAILED DESCRIPTION

[0046] The specific embodiments of the embodiments of the present application are described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to illustrate and explain the embodiments of the present application, and do not limit the embodiments of the present application.

[0047] It should be noted that the acquisition, transmission, storage, use, processing and the like of data in the technical scheme of the present application comply with the relevant provisions of laws and regulations. In the embodiments of the present application, some industry existing schemes such as software, components, models and the like may be mentioned, which should be considered as exemplary, and the purpose is only to illustrate the feasibility of the implementation of the technical scheme of the present application, but does not mean that the applicant has or will necessarily use the scheme.

[0048] The coronary heart disease detection method provided in the embodiments of the present application can be applied to a terminal, a server, an interactive system including a terminal and a server, and is implemented through the interaction of the terminal and the server, which is not specifically limited herein. The terminal can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, electrocardiogram monitoring devices, and portable wearable devices, and the server can be implemented by an independent server or a server cluster composed of multiple servers.

[0049] In the embodiments of the present application, the "exercise electrocardiogram data" or equivalent terms thereof can refer to the electrocardiogram data collected by the subject during the exercise load test. The exercise load test is to increase the heart load through a certain amount of exercise to collect the electrocardiogram data of the subject. Based on the collected exercise electrocardiogram data, the electrocardiogram detection method for analyzing the heart health status of the subject is widely used in the detection of heart disease and cardiovascular disease. The exercise electrocardiogram data can include a plurality of QRS complexes reflecting the changes of the left and right ventricular depolarization potentials and times. Each QRS complex is a set of Q wave, R wave and S wave in an electrocardiogram. Based on the exercise electrocardiogram data, the corresponding heart rate sequence and the heart rate corresponding to each QRS complex in the exercise electrocardiogram data can be analyzed.

[0050] Figure 1 An example flowchart of the electrocardiogram data processing method according to the embodiments of the present application is schematically shown. Specifically, as shown in Figure 1 The electrocardiogram data processing method can include the following steps.

[0051] In step S101, exercise electrocardiogram data output by at least one electrocardiogram lead is acquired.

[0052] Specifically, during the electrocardiogram signal collection process, at least one electrode patch can be used for signal collection for the subject. For example, taking 10 electrode patches as an example, the electrode patches can be distributed on the chest and limbs of the human body to form 12 electrocardiogram leads (such as V1, V2, V3, V4, V5, V6, I, II, III, aVL, aVF and aVR), corresponding to output 12 groups of electrocardiogram data. It can be understood that 10 electrode patches are only used as an example and are not used to specifically limit the number of electrode patches. More or fewer electrode patches can be used according to actual needs.

[0053] In the embodiments of the present application, the subject can be subjected to a exercise load test using, for example, a standard Bruce or modified Bruce protocol. For the case of applying multiple electrocardiogram leads, multi-lead electrocardiogram data can be recorded synchronously to obtain exercise electrocardiogram data. In this case, the exercise electrocardiogram data can include exercise electrocardiogram data corresponding to each electrocardiogram lead.

[0054] In an example, in order to guarantee the integrity of high frequency components, the sampling rate of the electrocardiogram signal can be no less than 1000 Hz.

[0055] In the embodiments of the present application, the exercise load detection process can include multiple stages, for example, can include three stages of a resting stage, an exercise stage and a recovery stage in sequence, and the exercise electrocardiogram data includes electrocardiogram data of each stage. It can be understood that the division of stages is not limited to this, and can be divided according to actual conditions. In an example, the total duration of the three stages can be, for example, 12 to 15 minutes, wherein the duration of the resting stage can be, for example, 3 minutes, and the subject maintains a resting state in this stage. The duration of the exercise stage can be, for example, 6 to 9 minutes, and in this stage, the subject increases the heart load by exercise, so that the heart rate uniformly / steadily rises, and finally reaches the target heart rate. If the target heart rate is reached, the recovery stage is entered, and the duration of the recovery stage can be, for example, 3 minutes. The preset time period includes at least the resting stage, the exercise stage and the recovery stage. In a preferred example, if the duration of the exercise stage reaches 9 minutes, the subject still does not reach the target heart rate, and also enters the recovery stage.

[0056] In the embodiments of the present application, after obtaining the exercise electrocardiogram data, the obtained exercise electrocardiogram data can be preprocessed. In an example, the obtained exercise electrocardiogram data (original electrocardiogram data) can be band-stop filtered (for example, a band-stop filter of 0-0.5 Hz is selected) to remove the baseline drift and low frequency noise in the signal.

[0057] In the preferred embodiments of the present application, a QRS waveform template can be preset, the shape of each QRS complex can be compared with the template, QRS complexes with shapes significantly different from the template can be marked as abnormal QRS complexes, and the electrocardiogram data of the electrocardiogram cycle in which the abnormal QRS complex is located can be removed from the original electrocardiogram data to obtain the exercise electrocardiogram data to be processed. In an example, the reason for the occurrence of the abnormal QRS wave can be, for example, a wide and deformed ventricular premature beat.

[0058] In step S102, the exercise electrocardiogram data is analyzed to obtain a heart rate sequence corresponding to each electrocardiogram lead.

[0059] Specifically, step S102 can include:

[0060] In step S1021, the R-wave peak position corresponding to each heartbeat in the exercise electrocardiogram data corresponding to each electrocardiogram lead is detected. The detection method of the R-wave peak position can use the algorithm of the prior art, for example, the robust algorithm of Pan-Tompkins. Taking the improved Pan-Tompkins algorithm as an example, it can include the following steps:

[0061] The motion electrocardio data is band-pass filtered (for example, a band-pass filter of 5-15 Hz can be used) to avoid noise covering the edge features of QRS complex; the motion electrocardio data after band-pass filtering is subjected to differential operation to enhance the edge features of QRS complex; the differential electrocardio data obtained through differential operation is subjected to square operation to highlight the energy of QRS complex and suppress noise; the square electrocardio data obtained through square operation is subjected to sliding window integration operation (for example, the window length is 0.15 s, corresponding to 150 sampling points (sampling frequency 1000 Hz), to obtain integrated electrocardio data; the mean value of R wave peaks corresponding to the initial preset number (for example, 3-5) of heartbeats in the integrated electrocardio data is taken as the initial threshold value, and after each R wave is detected, its peak value in the integrated electrocardio data is recorded, and the threshold value for the next detection is dynamically adjusted according to the mean value of the N (for example, N=5) most recently successfully detected R wave peaks (for example, the threshold value for the next detection = 0.6 x the mean value of the previous N peak values), to realize adaptive detection of R wave and obtain a preliminary R wave peak position sequence; false detection and missed detection correction: the RR interval is calculated according to the positions of adjacent two R waves in the preliminary R wave peak position sequence, and then the heart rate is obtained, and the R wave is screened through the heart rate range (for example, 50-200 bpm) (the R wave beyond the range is regarded as false detection), and the abnormal interval between two consecutive R waves (for example, deviation > threshold value (for example, 30%) from the mean value of adjacent 3 intervals) is linearly interpolated to ensure the positioning accuracy of R wave (for example, the positioning accuracy is ≥99%).

[0062] In step S1022, the time interval (RR interval) between the adjacent two R waves is determined according to the peak positions of the adjacent two R waves, and the corresponding heart rate is determined according to the time interval, to obtain the heart rate sequence corresponding to the motion electrocardio data of each electrocardiogram lead. The formula for calculating the heart rate can be heart rate (bpm) = 60 / (RR interval (s)).

[0063] In step S103, the QRS complex corresponding to each heartbeat is extracted from the motion electrocardio data.

[0064] The existing technology disclosed for extracting QRS complex from electrocardio signal / data can be used to extract QRS complex from motion electrocardio data, for example, each QRS complex can be directly located and extracted from the motion electrocardio data, or the corresponding QRS complex can be determined and extracted based on the position of each R wave peak, which will not be described here. Each QRS complex extracted covers the Q wave, R wave and S wave of a single heartbeat completely, and does not contain P wave and T wave, to avoid interference of P wave and T wave.

[0065] In step S104, a plurality of heart rate intervals are divided according to the obtained heart rate sequence.

[0066] Specifically, the range of all heart rates (e.g., from minimum to maximum) occurring in the entire exercise load detection process (i.e., from rest to exercise to recovery period) can be divided into a preset number of heart rate intervals, that is, the heart rate sequence corresponding to the exercise electrocardiogram data is divided into a preset number of heart rate intervals. The preset number can be customized, for example, 40 to 70, and preferably can be 50. In an example, the preset number of heart rate intervals can be continuous. In another example, the preset number of heart rate intervals can be discontinuous, or partially continuous and partially discontinuous.

[0067] In step S105, according to the heart rate interval to which the heart rate corresponding to each QRS complex belongs, the QRS complex is classified into the group corresponding to the heart rate interval.

[0068] When the corresponding heart rate sequence is obtained based on the exercise electrocardiogram data corresponding to each electrocardiogram lead, the heart rate corresponding to each heartbeat can be obtained, that is, the heart rate corresponding to each QRS complex in the exercise electrocardiogram data can be obtained, and then each QRS complex can be classified into the group corresponding to the heart rate interval according to the heart rate interval to which the heart rate corresponding to each QRS complex belongs. In this way, all QRS complexes whose heart rates belong to the same heart rate interval are classified into the same group for subsequent analysis. In an example, assuming that there is a heart rate interval 80-85 bpm, and there are multiple heart rates of 80 bpm in the heart rate sequence, that is, the heart rates corresponding to multiple QRS complexes in the exercise electrocardiogram data are all 80 bpm, then the multiple QRS complexes corresponding to the heart rate of 80 bpm are all classified into the group corresponding to the heart rate interval 80-85 bpm.

[0069] In the embodiments of the present application, after the corresponding heart rate is determined based on the positions of the adjacent two R-wave peaks in the exercise electrocardiogram data, the heart rate can be taken as the heart rate corresponding to the QRS complex to which the former R-wave of the adjacent two R-waves belongs, and the average of the preset number of heart rates at the end position in the exercise electrocardiogram data (i.e., the preset number of heart rates with the latest time sequence in the heart rate sequence) can be taken as the heart rate corresponding to the last QRS complex, which can also be taken as the heart rate corresponding to the QRS complex to which the latter R-wave of the adjacent two R-waves belongs, and the average of the preset number of heart rates at the start position in the exercise electrocardiogram data (i.e., the preset number of heart rates with the earliest time sequence in the heart rate sequence) can be taken as the heart rate corresponding to the first QRS complex. The preset number can be customized according to actual conditions, for example, it can be 1-5.

[0070] In step S106, the average QRS complex corresponding to each heart rate interval is determined according to all QRS complexes in each group.

[0071] In the preferred embodiment of the present application, the QRS complexes in each group after classification can be registered and averaged. For example, all QRS complexes in the same group can be time-aligned based on the R-wave peak position, and then the aligned waveforms can be point-by-point arithmetic averaged to obtain the average QRS complex corresponding to the group, which is taken as the average QRS complex corresponding to the corresponding heart rate interval. This step can greatly reduce random noise. The core logic of QRS complex registration can use the dynamic time warping (DTW) algorithm to time-align all QRS complexes in the same group, which can specifically include: selecting the QRS complex with the largest R-wave peak and the most regular waveform in the group as a reference template, calculating the distance matrix of other QRS complexes and the reference template, finding the optimal matching path through dynamic programming to realize time axis alignment, and performing linear interpolation on all aligned QRS complexes to unify the number of sampling points.

[0072] In step S107, the high-frequency energy of the average QRS complex corresponding to each heart rate interval is extracted.

[0073] Specifically, each average QRS complex can be filtered to retain high-frequency components. For example, a digital high-pass filter (such as a Butterworth 4th order) can be used to high-pass filter the average QRS complex. The cutoff frequency of the high-pass filter can be, for example, 150 Hz, so that the components of the average QRS complex of 150 Hz and above can be retained. Alternatively, a band-pass filter can be selected according to experimental optimization, such as a 150-250 Hz band-pass filter.

[0074] The filtered average QRS complex can be referred to as a high-frequency QRS complex. For each high-frequency QRS complex, the total energy E in a QRS time window can be calculated based on the voltage of each sampling point on the high-frequency QRS complex, which is referred to as high-frequency energy, as follows:

[0075]

[0076] is the number of sampling points in the QRS time window, i.e., the number of sampling points of a single QRS complex, and the QRS time window refers to the continuous time interval corresponding to the extracted single QRS complex, which is the signal range for calculating the corresponding energy. N = sampling rate x QRS time limit (for example, when the sampling rate is 1000 Hz and the QRS time limit is 100 ms, N = 1000 x 0.1 = 100 sampling points), is the voltage of the i-th sampling point in the QRS time window.

[0077] In step S108, the energy sequence is determined according to the high-frequency energy corresponding to each heart rate interval. ​​

[0078] Specifically, after determining the high-frequency energy corresponding to each heart rate zone, a representative heart rate value can be determined for each heart rate zone to represent that zone. For example, the upper limit, lower limit, or median of each heart rate zone can be determined as its representative heart rate value, and the high-frequency energy corresponding to each heart rate zone can be associated with the representative heart rate value. For example, a data point sequence (denoted as energy sequence) can be generated with the representative heart rate value of each heart rate zone as the x-axis (heart rate, HR, unit: bpm) and the associated high-frequency energy E as the y-axis (unit: μV²): [(HR1, E1), (HR2, E2), ..., (HR... M E M [ ], serving as discrete data for the "high-frequency QRS heart rate-energy trend graph". A high-frequency QRS heart rate-energy trend graph can be plotted based on the data points of this sequence. In a preferred embodiment, if there is a missing data point for a certain heart rate interval (i.e., the corresponding high-frequency energy has not been calculated, usually because the heart rate corresponding to all QRS complexes in the exercise ECG data does not fall within that interval), but the two adjacent heart rate intervals have data (i.e., the corresponding high-frequency energy has been calculated) (for example, data is available for the heart rate intervals 80-85 bpm and 90-95 bpm, but not for the 85-90 bpm interval), linear interpolation can be used to supplement the missing data points to ensure the continuity of the trend graph.

[0079] This application provides a method for predicting the risk of coronary heart disease, which may include the following steps.

[0080] In step S201, based on the trained coronary artery disease risk prediction model, the risk of coronary artery disease is predicted according to the energy sequence corresponding to all electrocardiogram leads. This energy sequence can be obtained by the electrocardiogram data processing method described in the above embodiments.

[0081] In one embodiment of this application, the coronary heart disease risk prediction model (hereinafter referred to as the prediction model) can adopt, for example, an L1-regularized logistic regression (L1-LR) model. The L1 regularization of this model can make the coefficients of some features zero (0), achieving automatic feature selection and resulting in a sparse and interpretable model. Alternatively, the classifier can be replaced with a support vector machine, random forest, etc. L1-LR is a preferred option due to its interpretability.

[0082] In this embodiment, if an L1-LR model is used, step S201 may include:

[0083] In step S2011, target features are determined based on the energy sequence, and the target features include at least one of statistical features and morphological features.

[0084] Specifically, in the optional embodiments of the present application, a sliding average filter can be applied to the discrete energy sequence for smoothing to remove small fluctuations of high frequency energy in the energy sequence and to preserve and highlight the main trend. The window size of the filter can be, for example, 3 or 5, or can be adaptively selected. For example, when the length L of the energy sequence is ≥ 10, the window size = 5, and when 5 ≤ L < 10, the window size = 3. It can be understood that the window size and the values of the length of the heart rate energy sequence are only exemplary, and the values can be selected according to actual applications.

[0085] The smoothing formula used in the smoothing process can be as follows:

[0086]

[0087] wherein, is the i-th high frequency energy in the smoothed energy sequence, j is the window size, is the i-th high frequency energy in the energy sequence before smoothing.

[0088] The target features can be extracted from the energy sequence (if the smoothing process is performed, from the smoothed energy sequence), and the target features include at least one of statistical features and morphological features, wherein:

[0089] (1) Statistical features, which can include, but are not limited to, at least one of the mean, standard deviation, minimum value, maximum value, range, median, etc. of the high frequency energy in the energy sequence.

[0090] Mean: the average value of the high frequency energy in the (smoothed) energy sequence, reflecting the overall energy level;

[0091] Standard deviation (Std): reflecting the dispersion degree of the energy sequence, the energy fluctuation of the patient with coronary heart disease is larger;

[0092] Minimum value (Min): the minimum energy value in the energy sequence;

[0093] Maximum value (Max): the maximum energy value in the energy sequence;

[0094] Range: the difference between the maximum energy value and the minimum energy value in the energy sequence, reflecting the energy change amplitude;

[0095] Median: the middle value of the high frequency energy in the energy sequence after sorting, which is less affected by extreme values.

[0096] ​​(2) Morphological features, which can include, but are not limited to, at least one of the following: area under curve (AUC) of high-frequency QRS heart rate-energy curve (plotted based on energy sequence, with horizontal axis representing heart rate and vertical axis representing high-frequency energy corresponding to heart rate), number of peak and valley, number of times of slope sign change, etc.

[0097] Area under curve (AUC): the area enclosed by the (smoothed) high-frequency QRS heart rate-energy curve and the horizontal axis is calculated by trapezoidal integration method, which reflects the overall high-frequency energy level, and the calculation formula can be:

[0098]

[0099] wherein, HR is the interval between adjacent two heart rate values in the energy sequence, which is consistent with the step length (i.e. length) of the heart rate interval, is the smoothed energy value at the j th position, is the smoothed energy value at the j+ th position.

[0100] Number of peak and valley: local extreme points of high-frequency energy in the high-frequency QRS heart rate-energy curve or energy sequence are detected, which reflects the fluctuation of high-frequency energy in the energy sequence with heart rate. Alternatively, a sliding window method (e.g. window size = 3) can be used to identify the peak (middle point energy greater than that of two side points) and valley (middle point energy less than that of two side points) of high-frequency energy in the energy sequence, and the total number of determined peak and valley is counted. If it is a patient with coronary heart disease, the number of peak and valley of the energy sequence (high-frequency QRS heart rate-energy curve) of the patient with coronary heart disease is more.

[0101] Number of times of slope sign change: which reflects the frequency of “rise-fall” conversion of high-frequency energy in the energy sequence. The calculation formula can be, for example:

[0102] The slope (see formula:) of adjacent two data points (two-dimensional data points composed of heart rate and high-frequency energy) in the (smoothed) energy sequence is calculated, and the number of times of positive and negative alternation of slope sign is counted, which can reflect the frequency of energy change.

[0103] In this embodiment, the training of the prediction model (L1-LR model) can include the following steps.

[0104] Dataset construction: Collect samples diagnosed by coronary angiography (CAG), which can be divided into positive and negative groups. For example, the positive group includes 150 patients with coronary heart disease whose coronary angiography shows coronary stenosis ≥ 50%; the negative group includes 150 healthy / non-coronary heart disease patients whose coronary angiography shows normal coronary arteries or coronary stenosis < 20%. The two groups of people are matched in terms of age, gender, body mass index, and other demographic characteristics. Data labeling is performed using the results of coronary angiography as the gold standard. The healthy control group is labeled as "0", and the coronary heart disease group is labeled as "1".

[0105] Feature normalization: Standardize the target features extracted from the energy sequence of all samples, such as Z-score standardization, to eliminate the influence of dimensional differences on model training.

[0106] Model training: Input the standardized target feature matrix and corresponding labels (0 = healthy, 1 = coronary heart disease) into the prediction model. For example, 5-fold cross-validation and network search optimization of regularization strength hyperparameters (C value) can be used on the training set. The data set can be divided into training set (210 cases) and test set (90 cases) with a ratio of 7:3.

[0107] Parameter optimization: For example, determine the optimal regularization strength λ by 5-fold cross-validation, and the value range of λ can be [0.01, 0.1], with the goal of maximizing the test set AUC (area under the curve).

[0108] Optimization algorithm: The stochastic gradient descent (SGD) algorithm can be used to optimize the model parameters, with a learning rate of 0.001, an iteration number of 1000, and a batch size of 32.

[0109] Performance evaluation: The performance indicators of the model can be evaluated on an independent test set, including, for example, accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve (AUC-ROC).

[0110] In step S2012, the target features are normalized, and the normalized target features are input into the coronary heart disease risk prediction model to predict the risk of coronary heart disease.

[0111] In an example, the output of the prediction model can be a risk probability value P between 0 and 1 (e.g., P = 0.85), representing the likelihood of the subject having coronary heart disease. Clinicians can combine this probability value with patient clinical symptoms and other test results to make a comprehensive judgment, providing support for early screening, risk stratification, or treatment decisions for coronary heart disease. Further stratification can also be performed based on the risk probability value, such as low risk (P < 0.3), medium risk (0.3 ≤ P < 0.5), and high risk (P ≥ 0.5), providing a basis for clinical treatment plan development.

[0112] In another embodiment of the present application, the coronary heart disease risk prediction model can employ, for example, a neural network model, in which case the input to the neural network model is the demographic features and the energy sequences corresponding to all the ECG leads. The training method of the neural network model is similar to the training method of the L1-LR model of the above-mentioned embodiment and is known to those skilled in the art, which will not be described here. Specifically, the demographic features can include, but are not limited to, one or more of, for example, age, gender, height, weight, smoking history, drinking history, etc.

[0113] In this embodiment, step S201 can include inputting the energy sequences corresponding to all the ECG leads and the corresponding demographic features into the coronary heart disease risk prediction model to predict the coronary heart disease risk.

[0114] In an embodiment of the present application, a processor is provided, which is configured to perform the electrocardio data processing method described in each of the above-mentioned embodiments.

[0115] In an embodiment of the present application, a processor is provided, which is configured to perform the coronary heart disease risk prediction method described in each of the above-mentioned embodiments.

[0116] In an embodiment of the present application, a processing device is provided, which includes a processor and a memory, the memory storing instructions, the processor being configured to call and execute the instructions from the memory to implement the electrocardio data processing method described in each of the above-mentioned embodiments.

[0117] In an embodiment of the present application, a processing device is provided, which includes a processor and a memory, the memory storing instructions, the processor being configured to call and execute the instructions from the memory to implement the coronary heart disease risk prediction method described in each of the above-mentioned embodiments.

[0118] Figure 2 An example block diagram of an electrocardio data processing system according to an embodiment of the present application is schematically shown. As shown, in an embodiment of the present application, an electrocardio data processing system can include: Figure 2

[0119] An electrocardio signal acquisition device 100, which includes at least one electrode forming at least one ECG lead for acquiring electrocardio signals of a subject; and

[0120] The processing device 200 described above.

[0121] The electrocardio signal acquisition device 100 can further include a display configured to display various data and information, including but not limited to, for example, electrocardio data, energy sequences, demographic features, coronary heart disease risk prediction results, etc.

[0122] ​In the embodiments of the present application, a machine readable storage medium is provided, and the machine readable storage medium stores instructions for causing a machine to execute the electrocardio data processing method described in the embodiments.

[0123] In the embodiments of the present application, a machine readable storage medium is provided, and the machine readable storage medium stores instructions for causing a machine to execute the coronary heart disease risk prediction method described in the embodiments.

[0124] Examples of the processor can include, but are not limited to, a single-chip microcomputer, a microprocessor, a digital signal processor (DSP), a field programmable gate array (FPGA), an application-specific integrated circuit (ASIC), and the like.

[0125] The memory can include a non-persistent memory in a computer readable medium, a random access memory (RAM) and / or a non-volatile memory such as a read-only memory (ROM) or a flash memory, and the memory includes at least one memory chip.

[0126] The scheme provided in the embodiments of the present application can significantly improve the detection rate (sensitivity) and discrimination ability (specificity) for coronary heart disease, especially early and mild lesions, by using the high-frequency component more sensitive to ischemia and dynamic load information. In addition, it is creatively proposed to rearrange, segment and average the whole process of exercise electrocardio signals according to the heart rate interval (rather than time), so as to construct a two-dimensional dynamic trend graph with heart rate as the horizontal axis and high-frequency energy as the vertical axis. The trend graph analysis focuses on the overall change pattern and is not sensitive to accidental errors of a single data point. The method directly relates the cardiac load (heart rate) to the myocardial electrophysiological response (high-frequency energy), highlighting the physiological significance. In each heart rate interval, time registration and superposition averaging are performed on multiple QRS waves, which greatly suppresses random noise and motion artifacts and highlights the stable changes of the high-frequency component related to heart rate.

[0127] The scheme provided in the embodiments of the present application not only extracts the conventional statistical features (such as mean and standard deviation) of the energy sequence, but also focuses on extracting its morphological features (such as peak-valley number, slope change number and area under curve AUC), which can effectively depict the abnormal patterns of high-frequency energy “attenuation”, “platform” or “abnormal fluctuation” of coronary heart disease patients under high cardiac load.

[0128] The coronary heart disease risk prediction model can use L1 regularization logistic regression as a classifier. L1 regularization can automatically perform feature selection, generate a sparse model, and only retain a few features with the highest discriminability, so that the model decision process is clear (can be explained by feature coefficients), while preventing overfitting, enhancing the generalization ability, and helping doctors understand the diagnostic basis of the model and increasing the clinical trustworthiness.

[0129] Those skilled in the art will appreciate that embodiments of the application can be readily used as a method, apparatus, or computer program product. Accordingly, the application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0130] The application is described in relation to flow diagrams and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It is understood that each flow and / or block in the flow diagrams and / or block diagrams, and combinations of flows and / or blocks in the flow diagrams and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of an apparatus, such as a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus, to produce a machine, so that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flow diagrams and / or block diagrams block or blocks. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the function specified by the flow or flows and / or block or blocks.

[0131] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flow diagrams and / or block diagrams flow or flows and / or block or blocks. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the function specified by the flow or flows and / or block or blocks.

[0132] The computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow diagrams and / or block diagrams flow or flows and / or block or blocks. Figure 1 one or more flows and / or blocks Figure 1 means for carrying out the function specified by the flow or flows and / or block or blocks.

[0133] In a typical configuration, a computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0134] Memory can include non-persistent memory, Random Access Memory (RAM), and / or non-volatile memory, such as Read Only Memory (ROM) or flash memory, in a computer readable medium. Memory is an example of computer readable media.

[0135] Computer readable media includes permanent and non-permanent, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital versatile disc (DVD), or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer readable media does not include transitory media, such as modulated data signals and carrier waves.

[0136] It should also be noted that the terms "comprising", "containing", or any other variant thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article or apparatus that comprises a list of elements does not include only those elements in the list, but can also include other elements not expressly listed or inherent to such process, method, article or apparatus. Without further limitation, an element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0137] The above merely provides an example of the present application and is not intended to limit the present application. The present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the scope of claims of the present application.

Claims

1. An electrocardiogram (ECG) data processing device, characterized in that, Configured to implement an electrocardiogram (ECG) data processing method, the ECG data processing method comprising: Acquire exercise electrocardiogram (ECG) data output from at least one ECG lead; Analyze the exercise ECG data to obtain the heart rate sequence corresponding to each ECG lead; Extract the QRS complex corresponding to each heartbeat from the exercise electrocardiogram data; The obtained heart rate sequence is divided into multiple heart rate zones; Based on the heart rate zone to which each QRS complex belongs, classify the QRS complex into the group corresponding to the corresponding heart rate zone. The average QRS complex for each heart rate zone is determined based on all QRS complexes within each group. Extract the high-frequency energy of the average QRS complex corresponding to each heart rate zone, filter each average QRS complex to retain the high-frequency components, and the filtered average QRS complex is the high-frequency QRS complex. For each high-frequency QRS group, based on the voltage at each sampling point on the high-frequency QRS group, the total energy E within a QRS time window is calculated according to the following formula, and is denoted as high-frequency energy: in, N represents the number of sampling points within the QRS time window, i.e., the number of sampling points for a single QRS group. The QRS time window refers to the continuous time interval corresponding to a single extracted QRS group. N = sampling rate × QRS time limit. For the first time window of QRS The voltage at each sampling point; The energy sequence is determined based on the high-frequency energy corresponding to each heart rate zone; The analysis of the exercise electrocardiogram data to obtain the heart rate sequence corresponding to each electrocardiogram lead includes: The peak position of the R wave corresponding to each heartbeat in the exercise electrocardiogram data of each electrocardiogram lead was detected; The time interval between two adjacent R waves is determined by the peak position of the two adjacent R waves, and the corresponding heart rate is determined based on the time interval to obtain the heart rate sequence corresponding to the exercise ECG data of each ECG lead. The time interval between two adjacent R waves is determined by identifying their peak positions, and the corresponding heart rate is then determined based on this time interval to obtain the heart rate sequence corresponding to the exercise ECG data for each ECG lead, including: This corresponding heart rate is taken as the heart rate corresponding to the QRS complex to which the preceding R wave belongs in the two adjacent R waves; and The average of the first preset number of heart rates at the end position in the exercise electrocardiogram data is taken as the heart rate corresponding to the last QRS complex. or This corresponding heart rate is taken as the heart rate corresponding to the QRS complex of the second R wave in the two adjacent R waves; and The average of the second preset number of heart rates at the starting position in the exercise electrocardiogram data is taken as the heart rate corresponding to the first QRS complex.

2. The electrocardiogram data processing device according to claim 1, characterized in that, The energy sequence includes multiple high-frequency energies, each high-frequency energy corresponding to a representative heart rate value for each heart rate zone.

3. The electrocardiogram data processing device according to claim 2, characterized in that, The representative heart rate value includes any one of the following: The lower limit of the heart rate zone, the upper limit of the heart rate zone, and the median of the heart rate zone.

4. A device for predicting the risk of coronary heart disease, characterized in that, Configured to implement a coronary artery disease risk prediction method, the coronary artery disease risk prediction method including: Based on a trained coronary heart disease risk prediction model, the risk of coronary heart disease is predicted according to the energy sequences corresponding to all electrocardiogram leads obtained using the electrocardiogram data processing method according to any one of claims 1 to 3.

5. The coronary heart disease risk prediction device according to claim 4, characterized in that, The coronary artery disease risk prediction model uses an L1 regularized logistic regression model and uses the energy sequences corresponding to all electrocardiogram leads to predict the risk of coronary artery disease, including: The target features are determined based on the energy sequence, and the target features include at least one of statistical features and morphological features; Normalize the target features; The normalized target features are input into the coronary heart disease risk prediction model to predict the risk of coronary heart disease.

6. The coronary heart disease risk prediction device according to claim 4, characterized in that, The coronary artery disease risk prediction model uses a neural network model to predict coronary artery disease risk using energy sequences corresponding to all electrocardiogram leads, including: The energy sequence and demographic features are input into the coronary heart disease risk prediction model to predict the risk of coronary heart disease.

7. An electrocardiogram (ECG) data processing system, characterized in that, include: An electrocardiogram (ECG) signal acquisition device includes at least one electrode forming at least one ECG lead for acquiring the ECG signal of a subject. as well as The electrocardiogram data processing device according to any one of claims 1 to 3 or the coronary heart disease risk prediction device according to any one of claims 4 to 6.

8. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions for causing the machine to perform the electrocardiogram data processing method according to any one of claims 1 to 3 or the coronary artery disease risk prediction method according to any one of claims 4 to 6.

Images