ECG Data Processing Method and Device

Abstract

The present disclosure relates to a method and apparatus for electrocardiogram data processing. The method includes: collecting electrocardiogram data of a target object; determining a first feature vector according to the electrocardiogram data of the target object, where the first feature vector includes a plurality of heart rate variability indexes determined from the electrocardiogram data of the target object; inputting the first feature vector into a trained prediction model to obtain a prediction result, where the prediction result indicates whether the target object is suitable for receiving a specific medical treatment. According to the electrocardiogram data processing method of the embodiments of the present disclosure, the prediction accuracy of the adaptability of the target object to a specific medical treatment can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing, and in particular, to a method and device for electrocardiogram data processing. Background Art

[0002] Epilepsy, as a disease, affects the lives of patients. Most patients can control epileptic seizures through the combined use of one or more drugs. However, there are still some patients who are not sensitive to drug treatment, and these patients are called drug-resistant epilepsy patients. Some specific medical treatments, such as vagus nerve stimulation (VNS) surgery, can effectively control epileptic seizures in drug-resistant epilepsy patients. However, when specific medical treatments are applied to different individuals, the treatment effects vary greatly, and the quality of treatment effects has a high degree of uncertainty. According to statistical analysis results, for patients who have received specific medical treatments, only about 5%-9% of the patients' epileptic seizures are completely controlled, about 10% of the patients' epileptic seizures show no improvement at all, and the epileptic seizures of the remaining patients show a reduction in seizure frequency to varying degrees. Generally speaking, after receiving specific medical treatments, about 50%-60% of drug-resistant epilepsy patients can achieve the effect that the seizure frequency is reduced to less than half of that before treatment.

[0003] Predicting whether a patient is suitable for receiving a specific medical treatment, for example, for the problems of high uncertainty and large individual differences in the treatment effect of treating drug-resistant epilepsy patients by means of vagus nerve stimulation, predicting whether a patient is suitable for receiving VNS surgery through preoperative evaluation means has become a research hotspot in the related field. Summary of the Invention

[0004] In view of this, the present disclosure provides a method and device for electrocardiogram data processing. According to the electrocardiogram data processing method of the embodiments of the present application, the prediction accuracy of the adaptability of the target object to a specific medical treatment can be improved.

[0005] According to one aspect of the present disclosure, there is provided a method for electrocardiogram data processing, including: collecting electrocardiogram data of a target object; determining a first feature vector according to the electrocardiogram data of the target object, where the first feature vector includes a plurality of heart rate variability indexes determined from the electrocardiogram data of the target object; inputting the first feature vector into a trained prediction model to obtain a prediction result, where the prediction result indicates whether the target object is suitable for receiving a specific medical treatment.

[0006] In a possible implementation, the electrocardiogram data of the target object includes electrocardiogram data in the waking state and electrocardiogram data in the sleeping state. Determining a first feature vector according to the electrocardiogram data of the target object includes: determining the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state according to the electrocardiogram data of the target object; obtaining the first feature vector according to the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state.

[0007] In a possible implementation, obtaining the first feature vector according to the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state includes: determining a plurality of index types as the input of the trained prediction model; determining the plurality of heart rate variability indexes corresponding to the plurality of index types according to the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state; obtaining the first feature vector according to the plurality of heart rate variability indexes.

[0008] In a possible implementation, the prediction model is trained according to the electrocardiogram data of the sample object, and the method further includes: determining a plurality of heart rate variability indexes of the sample object according to the electrocardiogram data of the sample object; determining the plurality of heart rate variability indexes of the sample object whose effect meets the preset conditions as positive samples and the plurality of heart rate variability indexes of the sample object whose effect does not meet the preset conditions as negative samples according to the effect after the sample object receives the specific medical treatment; performing a significance analysis of the differences between the positive samples and the negative samples to obtain heart rate variability indexes with significant differences, where the difference between the heart rate variability indexes with significant differences of the positive samples and the negative samples is greater than the threshold; performing an importance ranking on the heart rate variability indexes with significant differences to obtain a second feature vector; training the prediction model according to the indexes in the plurality of heart rate variability indexes of the sample object that are of the same index type as the indexes in the second feature vector.

[0009] In a possible implementation, performing an importance ranking on the heart rate variability indexes with significant differences to obtain a second feature vector includes: obtaining a feature subset according to all the heart rate variability indexes with significant differences, and determining the prediction accuracy of the prediction model for the feature subset; determining whether the current feature subset is an empty set; when the current feature subset is an empty set, performing an importance ranking on the heart rate variability indexes in the feature subset with the highest prediction accuracy according to the importance scores of the heart rate variability indexes in the feature subset with the highest prediction accuracy to obtain a second feature vector.

[0010] In a possible implementation, importance ranking is performed on the heart rate variability indicators with significant differences to obtain a second feature vector, and it further includes: when the current feature subset is not an empty set, repeating the following operations: inputting the current feature subset into a preset random forest model to obtain the importance scores of each heart rate variability indicator in the current feature subset; eliminating at least one heart rate variability indicator with the lowest importance score in the current feature subset, and using the set of the remaining heart rate variability indicators in the current feature subset as a new current feature subset, determining the prediction accuracy of the prediction model for the current feature subset; and re-determining whether the current feature subset is an empty set.

[0011] In a possible implementation, the prediction model is trained based on the indicators in the multiple heart rate variability indicators of the sample object that have the same indicator type as those in the second feature vector, including: arranging and combining according to different indicators in the second feature vector to obtain multiple third feature vectors, where the combinations of indicators in different third feature vectors are different, or the combinations of indicators in different third feature vectors and the sorting of the indicators in the third feature vectors are different; calculating the prediction accuracies of the prediction model when using, as the input of the prediction model, the indicators in the multiple heart rate variability indicators of the sample object that have the same indicator type as those in the multiple third feature vectors respectively; and determining the input indicator type of the prediction model according to the third feature vector with the highest prediction accuracy to obtain the trained prediction model.

[0012] In a possible implementation, the electrocardiogram data of the sample object includes electrocardiogram data in the waking state and electrocardiogram data in the sleeping state. Determining the multiple heart rate variability indicators of the sample object according to the electrocardiogram data of the sample object includes: determining the RR interval sequence of the sample object in the waking state and the RR interval sequence of the sample object in the sleeping state according to the electrocardiogram data of the sample object; and determining the multiple heart rate variability indicators corresponding to a preset multiple indicator types according to the RR interval sequence of the sample object in the waking state and the RR interval sequence of the sample object in the sleeping state, where the multiple heart rate variability indicators of the sample object include one or more of at least one time-domain heart rate variability indicator, at least one frequency-domain heart rate variability indicator, and at least one non-linear heart rate variability indicator.

[0013] According to another aspect of the present disclosure, there is provided an electrocardiogram data processing device, including: an acquisition module for acquiring electrocardiogram data of a target object; a determination module for determining a first feature vector according to the electrocardiogram data of the target object, where the first feature vector includes a plurality of heart rate variability indexes determined from the electrocardiogram data of the target object; and a prediction module for inputting the first feature vector into a trained prediction model to obtain a prediction result, where the prediction result indicates whether the target object is suitable for receiving a specific medical treatment.

[0014] In a possible implementation manner, the electrocardiogram data of the target object includes electrocardiogram data in a waking state and electrocardiogram data in a sleeping state. Determining a first feature vector according to the electrocardiogram data of the target object includes: determining an RR interval sequence of the target object in the waking state and an RR interval sequence of the target object in the sleeping state according to the electrocardiogram data of the target object; and obtaining the first feature vector according to the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state.

[0015] In a possible implementation manner, obtaining the first feature vector according to the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state includes: determining a plurality of index types that are inputs of the trained prediction model; determining the plurality of heart rate variability indexes corresponding to the plurality of index types according to the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state; and obtaining the first feature vector according to the plurality of heart rate variability indexes.

[0016] In a possible implementation manner, the prediction model is trained according to the electrocardiogram data of a sample object, and the method further includes: determining a plurality of heart rate variability indexes of the sample object according to the electrocardiogram data of the sample object; determining the plurality of heart rate variability indexes of the sample object whose effect satisfies a preset condition as positive samples and the plurality of heart rate variability indexes of the sample object whose effect does not satisfy the preset condition as negative samples according to the effect after the sample object receives the specific medical treatment; performing a significance analysis of differences between the positive samples and the negative samples to obtain heart rate variability indexes with significant differences, where the difference between the heart rate variability indexes with significant differences between the positive samples and the negative samples is greater than a threshold; performing an importance ranking on the heart rate variability indexes with significant differences to obtain a second feature vector; and training the prediction model according to the indexes in the plurality of heart rate variability indexes of the sample object that are of the same index type as the indexes in the second feature vector.

[0017] In a possible implementation, importance ranking is performed on the heart rate variability indicators with significant differences to obtain a second feature vector, including: obtaining a feature subset based on all the heart rate variability indicators with significant differences, and determining the prediction accuracy of the prediction model for the feature subset; determining whether the current feature subset is an empty set; when the current feature subset is an empty set, performing importance ranking on the heart rate variability indicators in the feature subset with the highest prediction accuracy according to the importance scores of the heart rate variability indicators in the feature subset with the highest prediction accuracy, to obtain the second feature vector.

[0018] In a possible implementation, importance ranking is performed on the heart rate variability indicators with significant differences to obtain a second feature vector, and it further includes: when the current feature subset is not an empty set, repeating the following operations: inputting the current feature subset into a preset random forest model to obtain the importance scores of each heart rate variability indicator in the current feature subset; eliminating at least one heart rate variability indicator with the lowest importance score in the current feature subset, and using the set of the remaining heart rate variability indicators in the current feature subset as the new current feature subset, and determining the prediction accuracy of the prediction model for the current feature subset; re-determining whether the current feature subset is an empty set.

[0019] In a possible implementation, the prediction model is trained based on the indicators with the same indicator types as those in the second feature vector among the multiple heart rate variability indicators of the sample object, including: arranging and combining according to different indicators in the second feature vector to obtain multiple third feature vectors, where the combinations of indicators in different third feature vectors are different, or the combinations of indicators in different third feature vectors and the sorting of the indicators in the third feature vectors are different; calculating the prediction accuracies of the prediction model when using, as the input of the prediction model, the indicators with the same indicator types as those in the multiple third feature vectors among the multiple heart rate variability indicators of the sample object respectively; determining the input indicator type of the prediction model according to the third feature vector with the highest prediction accuracy, to obtain the trained prediction model.

[0020] In a possible implementation, the electrocardiogram (ECG) data of the sample object includes the ECG data in the waking state and the ECG data in the sleeping state. Based on the ECG data of the sample object, a plurality of heart rate variability (HRV) indexes of the sample object are determined, including: based on the ECG data of the sample object, determining the RR interval sequence in the waking state and the RR interval sequence in the sleeping state of the sample object; based on the RR interval sequence in the waking state and the RR interval sequence in the sleeping state of the sample object, determining the plurality of HRV indexes corresponding to a plurality of preset index types, and the plurality of HRV indexes of the sample object include one or more of at least one time-domain HRV index, at least one frequency-domain HRV index, and at least one non-linear HRV index.

[0021] According to another aspect of the present disclosure, there is provided an ECG data processing device, including: a processor; a memory for storing processor-executable instructions; wherein, the processor is configured to execute the above method.

[0022] According to another aspect of the present disclosure, there is provided a non-volatile computer-readable storage medium, on which computer program instructions are stored, wherein, when the computer program instructions are executed by a processor, the above method is implemented.

[0023] According to the ECG data processing method of the embodiments of the present application, by collecting the ECG data of a target object and processing it to obtain a first feature vector, when the first feature vector is input into a prediction model, the prediction model can give a prediction result indicating whether the target object is suitable for receiving a specific medical treatment, so as to be able to predict in advance the adaptability of the target object to the specific medical treatment before receiving the specific medical treatment. And the first feature vector includes a plurality of HRV indexes, so the prediction result is related to all the plurality of HRV indexes, considering the comprehensive effect among different HRV indexes, and the accuracy of the prediction result can be improved.

[0024] According to the following detailed description of the exemplary embodiments with reference to the accompanying drawings, other features and aspects of the present disclosure will become clear. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 An exemplary application scenario of the ECG data processing method according to the embodiments of the present application is shown.

[0026] Figure 2 An exemplary schematic diagram of the ECG data processing method according to the embodiments of the present application is shown.

[0027] Figure 3 An exemplary method schematic diagram of obtaining HRV indexes according to the prior art based on ECG data is shown.

[0028] Figure 4 Shows an example of a Poincaré plot according to an embodiment of the present application.

[0029] Figure 5 Shows an example of an implementation manner for determining the importance ranking of heart rate variability indexes according to an embodiment of the present application.

[0030] Figure 6 Shows an example of an implementation manner for ranking heart rate variability indexes from high to low according to the scoring results in an embodiment of the present application.

[0031] Figure 7 Shows an example of obtaining a trained binary classification prediction model in an embodiment of the present application.

[0032] Figure 8 Shows an exemplary schematic diagram of an electrocardiogram data processing method according to an embodiment of the present application.

[0033] Figure 9 Shows an exemplary schematic diagram of a prediction model trained according to an embodiment of the present application.

[0034] Figure 10 Shows an exemplary schematic diagram of an electrocardiogram data processing device according to an embodiment of the present application.

[0035] Figure 11 Shows an exemplary block diagram of an electrocardiogram data processing device 800 according to an embodiment of the present application.

[0036] Figure 12 Shows an exemplary block diagram of an electrocardiogram data processing device 1900 according to an embodiment of the present application. Detailed implementation manners

[0037] The following will detail various exemplary embodiments, features, and aspects of the present disclosure with reference to the accompanying drawings. The same reference numerals in the drawings denote elements having the same or similar functions. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0038] The special word "exemplary" here means "serving as an example, embodiment, or illustration". Any embodiment described as "exemplary" here does not necessarily have to be construed as superior to or better than other embodiments.

[0039] In addition, for a better illustration of the present disclosure, numerous specific details are given in the following detailed implementation manners. Those skilled in the art should understand that the present disclosure can also be implemented without some specific details. In some instances, methods, means, elements, and circuits well-known to those skilled in the art are not described in detail so as to highlight the gist of the present disclosure.

[0040] Regarding the prediction of the adaptability of drug-refractory patients to specific medical treatments (such as vagus nerve stimulation surgery), there is currently no clear preoperative assessment method applied clinically. Based on data such as electrocardiography (ECG), electroencephalography (EEG), magnetic resonance imaging (MRI), patient demographics (gender, age, etc.), clinical history (course of disease), seizure characteristics (including seizure type, seizure frequency, lesion location, etc.), the research on the relevant factors of adaptability to specific medical treatments has also yielded inconsistent and even contradictory conclusions.

[0041] Currently, in predicting the adaptability of drug-resistant patients to specific medical treatments, a widely used method is to predict adaptability based on 24-hour ambulatory electrocardiogram signals. This method is simple to operate, relatively low-cost, and the acquisition device for ambulatory electrocardiogram signals is usually a wearable device, which makes the electrocardiogram signals less affected by the activities of patients and has strong data stability. Therefore, it has become a preferred adaptability prediction scheme for specific medical treatments. Specifically, this method involves wearing a portable ambulatory electrocardiogram information recording box on the sample subject before surgery, and collecting the preoperative 24-hour ambulatory electrocardiogram signals of the sample subject in a state of free movement, and recording the seizure frequency of the sample subject for a long period of time after undergoing a specific medical treatment (such as vagus nerve stimulation surgery). In this way, the preoperative 24-hour ambulatory electrocardiogram signals of multiple sample subjects are recorded and the postoperative seizure frequencies of multiple sample subjects are statistically analyzed. Multiple patients are classified into multiple groups according to the relationship between the seizure frequency and the preoperative seizure frequency. For example, sample subjects with a seizure frequency reduced to less than one-tenth of the preoperative level are grouped as one group, and sample subjects with a seizure frequency reduced to one-tenth to one-fifth of the preoperative level are grouped as one group, and so on. By analyzing the preoperative 24-hour ambulatory electrocardiogram signals of each group of sample subjects, the preoperative heart rate variability (HRV) index of each group of sample subjects is determined, and the heart rate variability index with statistically significant differences between different groups of sample subjects is found as the sensitive factor for predicting the adaptability of specific medical treatments. The relationship between each heart rate variability index with statistical differences and the postoperative seizure frequency is analyzed using statistical test methods (such as significance test methods), and the law between the index and the seizure frequency is summarized. On this basis, for a target subject who is going to receive this specific medical treatment (vagus nerve stimulation surgery) and needs to determine the adaptability to this specific medical treatment before surgery, the adaptability of the target subject to the specific medical treatment can be predicted by analyzing the preoperative 24-hour ambulatory electrocardiogram signals of the target subject, determining the heart rate variability index, and according to the value of the index and the law between the index and the seizure frequency summarized.

[0042] The existing method for adaptability prediction of specific medical treatments based on 24-hour ambulatory electrocardiogram signals has the following disadvantages: it ignores the combined effect among different heart rate variability indexes, resulting in a decrease in the accuracy of the prediction results; and the rules summarized by the statistical test method are uncertain. When using the summarized rules to repeatedly predict based on the same heart rate variability indexes, the prediction results obtained each time may also be different, making it difficult to convince people of the accuracy of the prediction results obtained from a single prediction. It may be necessary to perform multiple predictions to ensure the accuracy of the prediction results, leading to an increase in data processing costs. In addition, when the patient is in different states of sleep or wakefulness, there are significant differences in the values of the heart rate variability indexes. Therefore, the prediction results of the adaptability of specific medical treatments corresponding to the heart rate variability indexes in different states may be different. Assuming that the heart rate variability indexes in different states are not distinguished and directly using the heart rate variability indexes with unknown states for the adaptability prediction of specific medical treatments will also result in a decrease in the prediction accuracy.

[0043] In summary, the method of the prior art has the defect of low accuracy in clinical applications.

[0044] In view of this, the embodiments of the present application propose an electrocardiogram data processing method and device. According to the electrocardiogram data processing method of the embodiments of the present application, the accuracy of the prediction of the adaptability of specific medical treatments such as vagus nerve stimulation surgery can be improved.

[0045] Figure 1 An exemplary application scenario of the electrocardiogram data processing method according to the embodiments of the present application is shown. As Figure 1 shown, the electrocardiogram data processing method of the embodiments of the present application can be executed by a server or a terminal device. The server or the terminal device may include, but is not limited to, various electronic devices such as smart phones, personal computers, and tablet computers. A trained prediction model may be deployed on the server or the terminal device. The target object may wear or be equipped with a wearable device capable of collecting electrocardiogram data. The server or the terminal device can receive the electrocardiogram data from the wearable device through wired or wireless communication means. Among them, the terminal device may also be the wearable device itself. During any 24 hours before the target object undergoes a specific medical treatment, the wearable device continuously collects the electrocardiogram data of the target object and sends it to the server or the terminal device.

[0046] According to the electrocardiogram data processing method of the embodiments of the present application, on the server or the terminal device, the electrocardiogram data of the target object undergoes a series of analysis and processing, and a feature vector for inputting into the prediction model can be obtained. The feature vector is input into the prediction model, and the prediction result of the target object can be obtained. According to this prediction result, it can be judged whether the target object is suitable for specific medical treatments such as vagus nerve stimulation surgery.

[0047] The following combinesFigures 2 - 6 Introduce the exemplary working mode of the electrocardiogram data processing method according to the embodiments of the present application. Figure 2 Show an exemplary schematic diagram of the electrocardiogram data processing method according to the embodiments of the present application. As Figure 2 shown, in a possible implementation, the method includes:

[0048] S11, Obtain the electrocardiogram data of the sample object and determine the heart rate variability index of the sample object.

[0049] Among them, the electrocardiogram data of the sample object can be two sets of electrocardiogram data corresponding to the awake state and the sleep state. For example, due to the physiological activities of the sample object, the 24-hour electrocardiogram data collected outside the body may be a mixture of the awake state and the sleep state, and the characteristics of the electrocardiogram data in the sleep state are different from those in the awake state. The 24-hour electrocardiogram data can be observed, the characteristics shown by the electrocardiogram data collected outside the body can be analyzed, the time period to which the electrocardiogram data that conforms to the characteristics of the electrocardiogram data in the sleep state is determined, and the continuous 4-hour electrocardiogram data in this time period is selected as the 4-hour electrocardiogram data in the sleep state. Similarly, the time period to which the electrocardiogram data that conforms to the characteristics of the electrocardiogram data in the awake state can be determined, and the continuous 4-hour electrocardiogram data in this time period is selected as the 4-hour electrocardiogram data in the awake state. The process of determining the electrocardiogram data in different states can be realized by manual screening.

[0050] Heart rate variability refers to the minute changes in instantaneous heart rate or instantaneous cardiac cycle, which is mainly affected by the autonomic nervous system (ANS) that controls the heart rhythm, where the autonomic nervous system includes the sympathetic nerve and the parasympathetic nerve. The heart rate variability index can, for example, include an index indicating the degree of heart rate variability statistically or calculated from the electrocardiogram data. Figure 3 Show an exemplary method schematic diagram of obtaining the heart rate variability index according to the electrocardiogram data in the prior art.

[0051] As Figure 3 shown, step S11 may include step S111 and step S112. After obtaining the electrocardiogram data of 4 hours in each of the awake and sleep states, the following steps S111 and step S112 can be executed to obtain the heart rate variability index.

[0052] S111, respectively obtain the RR interval sequences of 4 hours in each of the awake and sleep states from the electrocardiogram data of 4 hours in each of the awake and sleep states.

[0053] Among them, the RR interval is reflected on the electrocardiogram as the time (or distance, 1 mm on the electrocardiogram = 0.04 seconds of time) between adjacent heartbeats in the electrocardiogram. The time series composed of adjacent RR intervals is called the RR interval series. The electrocardiogram data can be digitized and denoised by using the methods of the prior art. For the processed electrocardiogram data, the acquisition time of the data showing the heartbeat waveform in the electrocardiogram is statistically counted, the time interval between adjacent heartbeats is determined, and the RR interval series is obtained based on multiple time intervals. For the electrocardiogram data of 4 hours each in the waking and sleeping states, the RR interval series corresponding to the waking state and the RR interval series corresponding to the sleeping state can be obtained respectively.

[0054] S112. Extract the heart rate variability characteristics from the RR interval series in the waking and sleeping states to obtain heart rate variability indexes.

[0055] For example, various methods such as the time-domain analysis method, frequency-domain analysis method, and non-linear analysis method of the prior art can be used to complete the extraction of heart rate variability characteristics. The following introduces the index types and specific determination methods determined by the time-domain analysis method, frequency-domain analysis method, and non-linear analysis method respectively in combination with Tables 1-3.

[0056] The time-domain analysis method includes statistical methods and geometric methods, which can relatively simply evaluate the change of heart rate. Among them, the statistical method determines the time of each RR interval by determining the time information of each heartbeat in the electrocardiogram data, and further calculates the average heart rate and the statistics related to the RR interval. The geometric method quantifies the distribution of the RR interval by transforming the RR interval series into a geometric figure and calculating the corresponding figure characteristics. The time-domain analysis method can quantitatively describe the regulatory effect of the autonomic nervous system on the heart rhythm. The following introduces the relevant information of the heart rate variability indexes determined according to the time-domain analysis method in combination with Table 1.

[0057] As shown in Table 1, according to the time-domain analysis method, multiple heart rate variability indexes can be determined, such as Mean RR (the mean of the RR interval series), RMSSD (the root mean square of the differences between adjacent RR intervals), SDNN (the standard deviation of the RR interval series), etc. Table 1 shows the names (such as Mean RR), units (such as ms), and corresponding definitions (such as the mean of the RR interval series) of various indexes determined by the time-domain analysis method.

[0058] Table 1

[0059]

[0060] Table 1 continued

[0061] MedianNN ms Median of the absolute differences between adjacent RR intervals MadNN ms Absolute median difference of RR intervals HCVNN —— Ratio of the absolute median difference of RR intervals to the median of the absolute differences between adjacent RR intervals IQRNN ms Interquartile range of the RR interval sequence pNN20 —— Proportion of the number of adjacent RR interval differences exceeding 20 ms in the total number of RR intervals

[0062] Among them, the numerical calculation method of the indicators in Table 1 can be calculated according to the defined instructions with reference to the corresponding methods in the prior art. For example, the indicator Mean RR is defined as the mean value of the RR interval sequence. In the prior art, the calculation method of the sequence mean value is the ratio of the sum of the elements in the sequence to the number of elements. The value of the indicator Mean RR can be determined according to the ratio of the sum of the RR intervals in the RR interval sequence to the number of RR intervals. Examples of the numerical calculation methods of other indicators in Table 1 will not be elaborated here.

[0063] The frequency domain analysis method includes power spectral density analysis (PSD). The result of the frequency domain analysis can indicate the distribution relationship of power with frequency. The following introduces the relevant information of the heart rate variability indicators determined according to the frequency domain analysis method in combination with Table 2.

[0064] As shown in Table 2, the frequency domain indicators determined according to the frequency domain analysis method may include ultra low frequency (ULF) power, very low frequency (VLF) power, low frequency (LF) power, high frequency (HF) power, the power ratio of low frequency to high frequency, and so on. Table 2 shows the names (such as ULF), units (such as ms 2 ) and corresponding definitions (such as frequency ≤ 0.0033 Hz) of multiple indicators determined by the frequency domain analysis method.

[0065] Table 2

[0066] Index Unit Definition ULF <![CDATA[ms 2 > Frequency ≤ 0.0033 Hz VLF <![CDATA[ms 2 > 0.0033 ≤ Frequency ≤ 0.04 Hz LF <![CDATA[ms 2 > 0.04 ≤ Frequency ≤ 0.15 Hz HF <![CDATA[ms 2 > 0.15 ≤ Frequency ≤ 0.4 Hz LF / HF —— ——

[0067] Among them, the typical heart rate variability power spectrum obtained by power spectral density analysis may include multiple separated spectral peaks. Different spectral peaks can be approximately regarded as being located in different frequency bands. The area under each frequency band can be used as the measured value of the power spectrum energy, that is, the value of the frequency domain indicator. The values of the ultra low frequency power ULF, very low frequency power VLF, low frequency power LF, and high frequency power HF can, for example, be respectively equal to the area of the frequency band corresponding to the corresponding definition. For example, the value of the ultra low frequency power ULF can be equal to the area of the frequency band less than 0.0033 Hz. The power ratio of low frequency to high frequency can, for example, be equal to the ratio of the area of the frequency band corresponding to the definition of the low frequency power LF to the area of the frequency band corresponding to the definition of the high frequency power HF.

[0068] Those skilled in the art should understand that the frequency domain analysis method is not limited to the power spectral density analysis method. The frequency domain analysis method may also include amplitude spectrum analysis, phase spectrum analysis, etc. As long as it can analyze the RR interval in the frequency domain and determine the heart rate variability indicators that can reflect the heart rate variability, the specific feature extraction method adopted by the frequency domain analysis method in this application is not limited.

[0069] The physiological activities of the heart determine the nonlinearity of the heart rate regulation mechanism. Therefore, in addition to time-domain and frequency-domain analyses, nonlinear analysis methods can be used to more comprehensively measure the characteristics of heart rate variability. The following introduces the relevant information on heart rate variability indexes determined according to nonlinear analysis methods in combination with Table 3.

[0070] In a possible implementation, the nonlinear indexes may include Poincaré Plot features and heart rate asymmetry (HRA). Among them, the Poincaré Plot features may include indexes SD1, SD2, CSI, CVI, GI, PI, SI, AI (see Table 3), and the heart rate asymmetry may include indexes SD1 d 、SD1 a 、C1 d 、C1 a 、SD2 d 、SD2 a 、C2 d 、C2 a (see Table 3). Table 3 shows the names (such as SD1), calculation methods (such as ) and corresponding definitions (such as describing the fluctuations of short-term RR intervals) of the indexes of the Poincaré Plot features and heart rate asymmetry indexes determined by the nonlinear analysis method.

[0071] Table 3

[0072]

[0073] Continued Table 3

[0074]

[0075] Figure 4 shows an example of a Poincaré Plot according to an embodiment of the present application. As Figure 4 shown, taking any RR interval (such as RRn) as the value of the abscissa X and the immediately following RR interval (such as RRn+1) as the value of the ordinate Y, the line X = Y can be, for example, the "identification line" in Table 3. The Poincaré Plot includes several scatter points (the number may be equal to n in Table 3). The position of each point is determined by two adjacent RR intervals. According to the position of the scatter points in the plot, they can be divided into scatter points above the identification line (the number may be equal to l in Table 3), scatter points below the identification line (the number may be equal to b in Table 3), and scatter points on the identification line (the number may be equal to n on ) in Table 3. According to the different states of the scatter points, they can be divided into accelerating scatter points (the number may be equal to n a ) and decelerating scatter points (the number may be equal to n d) Among them, the RR interval corresponding to the abscissa of the accelerated scatter points is longer than the RR interval corresponding to the ordinate, and the degree of prolongation is greater than the time change threshold of normal sinus rhythm (the RR intervals before and after); the RR interval corresponding to the ordinate of the decelerated scatter points is longer than the RR interval corresponding to the abscissa, and the degree of prolongation is greater than the time change threshold of normal sinus rhythm (the RR intervals before and after). According to the relevant parameters of the scatter points in different positions (above the identification line, below the identification line, on the identification line) or different states (acceleration or deceleration), such as the distance from the scatter point to the identification line, the phase angle of the scatter point, the cumulative sector area corresponding to the scatter point, etc., various non-linear indexes in Table 3 can be calculated.

[0076] Among them, "short-term" in Table 3 can represent a continuous short period of time in 4 hours in the waking or sleeping state. Taking "short-term" as 5 minutes as an example, 4 hours in the waking or sleeping state can correspond to 48 "short-terms", and each "short-term" corresponds to a set of indexes related to the "short-term", such as SD1 in Table 3 above. d 、SD1 a 、C1 d 、C1 a etc. Executing step S112 once can determine 48 sets of indexes related to "short-term". "Long-term" can represent a continuous period of time longer than "short-term" in 4 hours in the waking or sleeping state. Taking "long-term" as 4 hours as an example, 4 hours in the waking or sleeping state can correspond to 1 "long-term", corresponding to a set of indexes related to the "long-term", such as SD2 in Table 3 above. d 、SD2 a 、C2 d 、C2 a etc. Executing step S112 once can determine 1 set of indexes related to "long-term".

[0077] Among them, some non-linear heart rate variability indexes can also be calculated through the heart rate variability indexes determined by the previous time domain analysis. For example, referring to Table 3, the non-linear heart rate variability index SD1 describing the fluctuation of short-term RR intervals can be calculated through the time domain heart rate variability index SDSD (the standard deviation of the difference between adjacent RR intervals). In addition, the non-linear heart rate variability index SD2 in Table 3 can also be calculated through the time domain heart rate variability index. The indexes CSI and CVI in Table 3 can be calculated according to the non-linear heart rate variability indexes SD1 and SD2, and thus can also be regarded as non-linear indexes calculated according to the time domain heart rate variability indexes.

[0078] In a possible implementation, the non-linear index may further include heart rate fragmentation (HRF). Among them, the heart rate refers to the number of heartbeats per minute in a normal person at rest. The heart rate acceleration is determined by the frequency at which the vagus nerve regulates the sinoatrial node. Heart rate fragmentation is manifested as: the heart rate acceleration changes frequently at a frequency higher than that at which the vagus nerve regulates the sinoatrial node. Heart rate fragmentation may include one or more of the following indicators: the percentage of inflection points (turning points) in the entire RR interval series (PIP), the inverse of the average length of the acceleration / deceleration segments (IALS), the percentage of short segments (PSS), and the percentage of NN intervals in alternation segments (PAS). Among them, the "short segment" in the percentage of short segments indicator can be defined as the heart rate acceleration and deceleration process including 3 or more RR intervals. The "alternation segment" in the percentage of NN intervals in alternation segments indicator can be defined as an RR interval series including at least 4 RR intervals, and in this RR interval series, the heart rate acceleration changes sign before and after each heartbeat.

[0079] In a possible implementation, the non-linear index may further include the complexity of the heart rate. The complexity of the heart rate can be analyzed by approximate entropy. Approximate entropy (ApEn) was proposed by Pincus in 1991 and is a complexity measure analysis method without coarse-graining that can measure the complexity and regularity of time series. Its physical meaning is the magnitude of the probability of generating new patterns in a time series when the dimension changes. The greater the probability of generating new patterns, the more complex the time series, and the corresponding approximate entropy is also greater. For example, see Figure 4, the position of each point is determined by two adjacent RR intervals. Based on m + 1 consecutive RR intervals, m points can be determined. Based on the m points, m - 1 broken line segments can be connected. According to the similarity degree of the shapes of the m - 1 broken line segments, a mutual approximation probability can be determined. Similarly, based on m + 2 consecutive RR intervals (including the m + 1 RR intervals described above), m + 1 points (including the m points described above) can be determined. Based on the m + 1 points, m broken line segments can be connected. According to the similarity degree of the shapes of the m broken line segments, a mutual approximation probability can be determined. The difference between the two mutual approximation probabilities can indicate the complexity of the heart rate.

[0080] Those skilled in the art should understand that the heart rate variability indexes that can be obtained through time domain analysis methods, frequency domain analysis methods, and non - linear analysis methods should be more than just the indexes in the above examples. This application does not limit the specific types and specific quantities of the heart rate variability indexes determined according to the electrocardiogram data of the sample object.

[0081] In a possible implementation manner, after determining the heart rate variability indexes of the sample object through multiple analysis methods, the following step S12 can be executed to sort the heart rate variability indexes.

[0082] S12. Determine the importance ranking of the heart rate variability indexes according to the heart rate variability indexes of the sample object and the effect after the sample object undergoes a specific medical treatment.

[0083] Figure 5 An example showing an implementation manner of determining the importance ranking of the heart rate variability indexes according to the embodiments of the present application is shown. Refer to Figure 5 , step S12 may, for example, include the following steps S121 - S123:

[0084] S121. Classify the heart rate variability indexes of the sample object into positive samples and negative samples according to the effect after the sample object undergoes a specific medical treatment.

[0085] As described above, the sample object has undergone a vagus nerve stimulation surgery. The effect after the sample object undergoes a specific medical treatment, such as after the vagus nerve stimulation surgery, has a certain association with the heart rate variability indexes of the sample object. It can be preset that the sample object with a reduced postoperative seizure frequency compared to before the surgery is an effective object. It can be preset that the sample object with no obvious change in the postoperative seizure frequency compared to before the surgery is an ineffective object. In this case, the positive samples may include the heart rate variability indexes corresponding to the effective objects, and the negative samples may include the heart rate variability indexes corresponding to the ineffective objects.

[0086] S122. Perform a significance analysis of the inter-group differences in the heart rate variability indices of positive and negative samples to determine the heart rate variability indices with statistically significant differences.

[0087] For example, a significant difference is a statistical term that can be used to evaluate data differences. If there are significant differences between certain types of data, for example, the differences between the data participating in the comparison are greater than or equal to a threshold, it can be considered that the data participating in the comparison do not come from the same population, but from two different populations that are different in one or more aspects. If there are no significant differences between certain types of data, for example, the differences between the data participating in the comparison are less than the threshold, it can be considered that the data participating in the comparison come from the same population. In the embodiments of the present application, the data participating in the comparison have been determined to come from two different populations with differences. In this case, if there are significant differences between certain types of data, this type of data may be associated with the differences between the two populations. If there are no significant differences between certain types of data, this type of data may have nothing to do with the differences between the two populations.

[0088] For example, the two populations may be positive and negative samples respectively. When the data type is the cardio-sympathetic nerve function index CSI, the difference between the populations may be, for example, the difference in the effects after the sample subjects receive a specific medical treatment. If it is analyzed and determined that the difference in the cardio-sympathetic nerve function index CSI between the positive and negative samples is large, for example, greater than a preset threshold, it can be considered that there are significant differences between the cardio-sympathetic nerve function indices CSI used for analysis, and the cardio-sympathetic nerve function index CSI is associated with the effects after the sample subjects receive a specific medical treatment. When the data type is the cardio-vagal nerve function index CVI, if it is analyzed and determined that the difference in the cardio-vagal nerve function index CVI between the positive and negative samples is small, for example, less than another preset threshold, it can be considered that there are no significant differences between the cardio-vagal nerve function indices CVI used for analysis, and the cardio-vagal nerve function index CVI has nothing to do with the effects after the sample subjects receive a specific medical treatment.

[0089] Step S122 can be implemented based on the Mann-Whitney U test method in the prior art. Those skilled in the art should understand that other methods can also be used to implement step S122, as long as it can be used to statistically analyze and determine which of the heart rate variability indices of positive and negative samples have statistically significant differences. The present application does not limit the specific determination method of the heart rate variability indices with statistically significant differences.

[0090] S123. Calculate the importance scores of the heart rate variability indices with statistically significant differences, and sort the heart rate variability indices from high to low according to the scoring results.

[0091] Figure 6 An example of the implementation manner of sorting the heart rate variability indexes from high to low according to the scoring results in the embodiments of the present application is shown. Refer to Figure 6 , step S123 may include, for example, steps S21-S25 below:

[0092] S21, input the initial feature subset into the random forest machine learning model to obtain the importance score of each index; determine the prediction accuracy corresponding to the initial feature subset through the cross-validation method.

[0093] Among them, the initial feature subset may include all heart rate variability indexes with statistically significant differences, and N represents the number of indexes in the set. For example, assume that the heart rate variability indexes with statistically significant differences determined in step S122 are GI, PI, SI, and AI respectively. In this case, the initial feature subset may be S0 = {GI, PI, SI, AI}, and N = 4.

[0094] The random forest machine learning model and the cross-validation method can be implemented based on the prior art. Step S21 may obtain the importance score of index GI as 0.5, the importance score of index PI as 0.2, the importance score of index SI as 0.2, and the importance score of index AI as 0.1 through the random forest learning model of the prior art. And it can be determined, for example, through the cross-validation method of the prior art that the prediction accuracy corresponding to the initial feature subset S0 is 60%.

[0095] S22, remove several indexes with the lowest importance scores from the current feature subset, and obtain a new feature subset according to the remaining indexes.

[0096] Among them, if there are multiple indexes with the lowest importance scores, at least one of the multiple indexes can be removed. In this case, step S22 may remove the index AI (0.1) with the lowest importance score in the initial feature subset S0, and obtain a new feature subset S1 = {GI, PI, SI} according to the remaining indexes GI, PI, and SI, and N = 3.

[0097] S23, input the new feature subset into the random forest machine learning model to obtain the importance score of each index; determine the prediction accuracy corresponding to the new feature subset through the cross-validation method.

[0098] Among them, the random forest machine learning model and the cross-validation method used in step S23 can be the same as those used in step S21. For example, step S23 can obtain the importance score of 0.6 for the index GI, the importance score of 0.2 for the index PI, and the importance score of 0.2 for the index SI through the random forest learning model of the prior art. And for example, the prediction accuracy corresponding to the feature subset S1 can be determined to be 70% through the cross-validation method of the prior art.

[0099] S24. Determine whether the current feature subset is an empty set. When it is not an empty set, repeat steps S22 - S24. When it is an empty set, execute step S25.

[0100] For example, the current feature subset judged in step S24 is actually obtained by executing step S22. As described above, the current feature subset S1 = {GI, PI, SI}, N = 3. It can be judged that the current feature subset is not an empty set, and enter the first repeated execution stage of steps S22 - S24. In this stage, execute step S22, and the index PI (score 0.2) and the index SI (score 0.2) with the lowest importance score in the feature subset S1 can be removed. According to the remaining index GI, a new feature subset S2 = {GI} is obtained, and N = 1. Execute step S23. For example, the importance score of 1.0 for the index GI of the feature subset S2 can be obtained, and for example, the prediction accuracy corresponding to the feature subset S2 can be obtained as 65%. Execute step S24. It can be judged that the current feature subset S2 = {GI} is not an empty set, and enter the second repeated execution stage of steps S22 - S24. In this stage, execute step S22, and the index GI (score 1.0) with the lowest importance score in the feature subset S2 can be removed. The obtained feature subset S3 is an empty set. The importance score cannot be calculated when executing step S23. Execute step S24. It can be judged that the current feature subset S3 is an empty set. In this case, execute step S25 below.

[0101] For another example, after step S24 determines that the current feature subset S1 = {GI, PI, SI} is not an empty set, it enters the first repeated execution stage of steps S22 - S24. In this stage, when step S22 is executed, the index SI in the feature subset S1 with the lowest importance score (score 0.2) among the indices PI (score 0.2) and SI can be removed, and the index PI is retained. Based on the remaining indices GI and PI, a new feature subset S4 = {GI, PI} is obtained, and N = 2. When step S23 is executed, for example, the importance score of the index GI in the feature subset S4 can be obtained as 0.6, the importance score of the index PI can be obtained as 0.4, and for example, the prediction accuracy corresponding to the feature subset S4 can be obtained as 75%. When step S24 is executed, it can be determined that the current feature subset S4 = {GI, PI} is not an empty set, and it enters the second repeated execution stage of steps S22 - S24. In this stage, when step S22 is executed, the index PI (score 0.4) with the lowest importance score in the feature subset S4 can be removed, and based on the remaining index GI, a new feature subset S5 = {GI} is obtained, and N = 1. When step S23 is executed, for example, the importance score of the index GI in the feature subset S5 can be obtained as 1.0, and for example, the prediction accuracy corresponding to the feature subset S5 can be obtained as 65%. When step S24 is executed, it can be determined that the current feature subset S5 = {GI} is not an empty set, and it enters the third repeated execution stage of steps S22 - S24. In this stage, when step S22 is executed, the index GI (score 1.0) with the lowest importance score in the feature subset S5 can be removed, and the resulting feature subset S6 is an empty set. When step S23 is executed, the importance score cannot be calculated. When step S24 is executed, it can be determined that the current feature subset S6 is an empty set. In this case, step S25 below is executed.

[0102] Those skilled in the art should understand that as long as it can be satisfied that each time S22 is executed, at least one index with the lowest importance score is removed from the current feature subset, the present application does not limit the specific removal method of the index with the lowest importance score.

[0103] S25. Select the feature subset with the highest prediction accuracy among all the feature subsets determined in steps S22 - S24, and sort the indices in this feature subset in descending order according to the importance scores of the indices in this feature subset to obtain a sorting result.

[0104] For example, among all the feature subsets S1, S2, and S3 determined in steps S22 - S24, the feature subset with the highest prediction accuracy can be, for example, feature subset S1, and the corresponding prediction accuracy is 70%. Feature subset S1 includes metrics GI, PI, and SI, with the importance score of metric GI being 0.6, the importance score of metric PI being 0.2, and the importance score of metric SI being 0.2. In this case, sorting the metrics in feature subset S1 can obtain a sorting result of GI, PI, SI or GI, SI, PI. The embodiments of the present application do not limit the sorting order of metrics with the same importance score.

[0105] Again, for example, among all the feature subsets S1, S4, S5, and S6 determined in steps S22 - S24, the feature subset with the highest prediction accuracy can be, for example, feature subset S4, and the corresponding prediction accuracy is 75%. Feature subset S4 includes metrics GI and PI, with the importance score of metric GI being 0.6 and the importance score of metric PI being 0.4. In this case, sorting the metrics in feature subset S4 can obtain a sorting result of GI, PI.

[0106] Step S123 can be implemented based on the recursive feature elimination feature selection method of the prior art. Those skilled in the art should understand that other methods can also be used to implement step S123, as long as it can complete the sorting of heart rate variability metrics with statistically significant differences and the importance of any metric in the sorting result is not higher than the previous metric. The present application does not limit the specific sorting method of heart rate variability metrics with statistically significant differences.

[0107] In a possible implementation manner, after step S12 determines the importance sorting of heart rate variability metrics, based on the sorting result and the heart rate variability metrics of the sample object, step S13 below can be executed to train a prediction model, and this model can be used in step S14 to predict the target object to determine the prediction result of the target object.

[0108] S13, according to the sorting of the importance and the heart rate variability metrics of the sample object, using the random forest machine learning classification algorithm, obtain a binary classification prediction model.

[0109] For example, according to the importance sorting of heart rate variability metrics, starting from the heart rate variability metric ranked first, add the heart rate variability metrics ranked later one by one to obtain a vector, and this vector can be used as the input feature vector of the binary classification prediction model to be trained. The binary classification prediction model can be trained according to the input feature vector. Figure 7 Show an example of obtaining a trained binary classification prediction model in the embodiments of the present application.

[0110] For example, as Figure 7As shown, after receiving the input feature vector, the binary classification prediction model to be trained can perform permutations and combinations based on multiple features in the feature vector to obtain multiple processed feature vectors including different combinations of features. For example, the input feature vector can be {GI, PI, SI}. If there is no restriction on the arrangement order of the elements in the processed feature vector, there can be ones, which can be {GI}, {PI}, {SI}, {GI, PI}, {GI, SI}, {PI, SI}, {GI, PI, SI} respectively; if it is restricted that the elements in the processed feature vector are arranged in order, there can be ones, which can be {GI}, {PI}, {SI}, {GI, PI}, {PI, GI}, {GI, SI}, {SI, GI}, {PI, SI}, {SI, PI}, {GI, PI, SI}, {GI, SI, PI}, {SI, PI, GI}, {SI, GI, PI}, {PI, SI, GI}, {PI, GI, SI} respectively. Using the indicators in the multiple heart rate variability indicators of the sample object that are of the same indicator type as each processed feature vector as the input of the binary classification prediction model to be trained, the binary classification prediction model to be trained can obtain a prediction accuracy rate. According to the processed feature vector corresponding to the input of the prediction model when the prediction accuracy rate is the highest, the optimal feature vector can be obtained. The binary classification prediction model determined according to the optimal feature vector is used as the trained binary classification prediction model.

[0111] Among them, the prediction accuracy rate of the binary classification prediction model can be determined by comparing the prediction result obtained by classifying the heart rate variability indicators of the sample object as a vector input into the binary classification prediction model with the effect after the sample object undergoes a specific medical treatment (such as vagus nerve stimulation surgery). For example, the binary classification prediction model can obtain samples with a prediction result of responder and samples with a prediction result of non-responder. Among them, the prediction result of responder corresponds to a reduction in postoperative epilepsy seizures of ≥50% after vagus nerve stimulation surgery, and the prediction result of non-responder corresponds to a reduction in postoperative epilepsy seizures of <50% after vagus nerve stimulation surgery. The number of samples that meet the prediction result of responder in the positive samples and the number of samples that meet the prediction result of non-responder in the negative samples can be counted respectively. According to the sum of the two, the number of samples with accurate prediction results can be obtained. In this case, the ratio of the number of samples with accurate prediction results to the total number of samples (the sum of positive samples and negative samples) can be used as the prediction accuracy rate of the binary classification prediction model.

[0112] The above method for determining the prediction accuracy rate of the binary classification prediction model is only an example. The embodiments of the present application do not limit the method for determining the prediction accuracy rate of the binary classification prediction model.

[0113] S14. Obtain the prediction result of the target object according to the binary classification prediction model and the electrocardiogram data of the target object.

[0114] Among them, the target object can be an object who has not yet undergone vagus nerve stimulation surgery. In step S14, according to the method of step S11 above, the heart rate variability index of the target object can be obtained from the electrocardiogram data of the target object first. Among them, the obtained index at least includes the index indicated by the optimal feature vector input to the trained binary classification prediction model determined in step S13. After obtaining the index, multiple indexes can be freely combined to obtain a feature vector. In this case, the feature vector is used as the input data of the trained binary classification prediction model, and the binary classification prediction model can output the prediction result of the target object. This prediction result can be used to judge the adaptability of the target object to specific medical treatments such as vagus nerve stimulation surgery.

[0115] Those skilled in the art should understand that in addition to the binary classification prediction model described above, step S13 can also train a multi-classification prediction model, and step S14 can use the multi-classification prediction model to process the feature vector and output the prediction result of the target object. The specific form of the prediction model used to obtain the prediction result of the target object in this application is not limited.

[0116] Figure 8 An exemplary schematic diagram showing the electrocardiogram data processing method according to an embodiment of the present application. As Figure 8 shown, an embodiment of the present application proposes an electrocardiogram data processing method, the method includes:

[0117] S31. Collect the electrocardiogram data of the target object. Among them, the target object can be an object for predicting the adaptability of specific medical treatments such as vagus nerve stimulation surgery. For the exemplary acquisition method of the electrocardiogram data of the target object, reference can be made to the relevant description of step S11 above.

[0118] S32. Determine a first feature vector according to the electrocardiogram data of the target object, where the first feature vector includes multiple heart rate variability indexes determined from the electrocardiogram data of the target object. The heart rate variability index can be the exemplary heart rate variability index in Tables 1-3 above, or other indexes indicating heart rate variability that can be obtained by processing electrocardiogram data according to the prior art. The first feature vector can be the feature vector in step S14 above, and the exemplary implementation manner of step S32 can refer to the relevant description of step S14 above.

[0119] S33. Input the first feature vector into the trained prediction model to obtain a prediction result, where the prediction result indicates whether the target object is suitable for receiving a specific medical treatment. The trained prediction model can be the binary classification prediction model or the multi-classification prediction model described above. The specific medical treatment can be medical treatment means such as the vagus nerve stimulation surgery described above. The exemplary implementation manner of step S33 can refer to the relevant description of step S14 above.

[0120] According to the electrocardiogram data processing method of the embodiments of the present application, by collecting the electrocardiogram data of the target object and processing it to obtain the first feature vector, when the first feature vector is input into the prediction model, the prediction model can give a prediction result indicating whether the target object is suitable for receiving a specific medical treatment, so that the adaptability of the target object to the specific medical treatment can be predicted in advance before receiving the specific medical treatment. And the first feature vector includes multiple heart rate variability indexes, so the prediction result is related to all the multiple heart rate variability indexes, considering the comprehensive effect among different heart rate variability indexes, which can improve the accuracy of the prediction result.

[0121] In a possible implementation manner, the electrocardiogram data of the target object includes the electrocardiogram data in the waking state and the electrocardiogram data in the sleeping state, and step S31 includes: determining the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state according to the electrocardiogram data of the target object; obtaining the first feature vector according to the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state.

[0122] Among them, the definition of the RR interval sequence and the exemplary determination manner of the RR interval sequence can refer to the relevant description of step S111 above. The exemplary determination manner of the first feature vector can refer to the relevant description of step S14 above.

[0123] Since the RR interval sequences in the waking state and the sleeping state are distinguished, the values of the heart rate variability indexes in the first feature vector are associated with the state of the target object. Using the heart rate variability indexes with known states to predict the adaptability of the target object to a specific medical treatment can further improve the accuracy of the prediction result.

[0124] In a possible implementation manner, obtaining the first feature vector according to the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state includes: determining multiple index types as the input of the trained prediction model; determining the multiple heart rate variability indexes corresponding to the multiple index types according to the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state; obtaining the first feature vector according to the multiple heart rate variability indexes.

[0125] Among them, for the exemplary determination method of multiple index types of the input of the trained prediction model, reference can be made to the exemplary acquisition method of the heart rate variability index of the target object in step S14 above.

[0126] In this way, the index types in the first feature vector conform to the multiple index types of the corresponding input of the trained prediction model, so that when the first feature vector is input into the prediction model for prediction, the accuracy of the prediction result can be guaranteed.

[0127] In a possible implementation manner, the prediction model is trained according to the electrocardiogram data of the sample object. Figure 9 The exemplary schematic diagram showing the prediction model trained according to the embodiment of the present application.

[0128] Such as Figure 9 shown, the method further includes:

[0129] S41. Determine multiple heart rate variability indexes of the sample object according to the electrocardiogram data of the sample object. The exemplary implementation manner thereof can refer to the relevant description of step S11 above.

[0130] S42. According to the effect after the sample object receives the specific medical treatment, determine the multiple heart rate variability indexes of the sample object whose effect meets the preset conditions as positive samples, and determine the multiple heart rate variability indexes of the sample object whose effect does not meet the preset conditions as negative samples. An example of the preset condition may be that the postoperative seizure frequency is reduced compared with that before the operation as described above. The exemplary implementation manner of step S42 can refer to the relevant description of step S121 above.

[0131] S43. Perform a significance analysis on the positive samples and negative samples to obtain heart rate variability indexes with significant differences, where the difference between the heart rate variability indexes with significant differences of the positive samples and negative samples is greater than the threshold. The exemplary implementation manner thereof can refer to the relevant description of step S122 above.

[0132] S44. Rank the importance of the heart rate variability indexes with significant differences to obtain a second feature vector. Among them, the second feature vector can be, for example, the ranking result in the relevant description of step S123 above. The exemplary implementation manner of step S44 can refer to the relevant description of step S123 above.

[0133] S45. Train the prediction model according to the indexes in the multiple heart rate variability indexes of the sample object that are of the same index type as those in the second feature vector. The exemplary implementation manner thereof can refer to the relevant description of step S124 above.

[0134] In this way, a trained prediction model can be obtained. Compared with the method in the prior art of first statistically analyzing the relationship between a single indicator and the prediction result and then summarizing the rules, the prediction model of the present application is trained based on multiple heart rate variability indicators. Therefore, the prediction model is obtained by considering the comprehensive effect of multiple indicators on the prediction result, and the prediction result obtained by using the prediction model has higher accuracy.

[0135] In a possible implementation, perform importance ranking on the heart rate variability indicators with significant differences to obtain a second feature vector, including:

[0136] According to all the heart rate variability indicators with significant differences, obtain a feature subset, and determine the prediction accuracy of the prediction model for the feature subset; determine whether the current feature subset is an empty set; when the current feature subset is an empty set, perform importance ranking on the heart rate variability indicators in the feature subset with the highest prediction accuracy according to the importance scores of the heart rate variability indicators in the feature subset with the highest prediction accuracy, to obtain a second feature vector.

[0137] Among them, the feature subset obtained according to all the heart rate variability indicators with significant differences can be, for example, the initial feature subset in step S21 above. The exemplary manner of obtaining the second feature vector can refer to the relevant descriptions of steps S21 and S25 above.

[0138] In this way, the index type of the second feature vector can correspond to the index type in the feature subset with the highest accuracy, making the prediction accuracy of the prediction model higher.

[0139] In a possible implementation, performing importance ranking on the heart rate variability indicators with significant differences to obtain a second feature vector further includes:

[0140] When the current feature subset is not an empty set, repeat the following operations:

[0141] Input the current feature subset into a preset random forest model to obtain the importance scores of each heart rate variability indicator in the current feature subset; eliminate at least one heart rate variability indicator with the lowest importance score in the current feature subset, and the set of the remaining heart rate variability indicators in the current feature subset is used as the new current feature subset, and determine the prediction accuracy of the prediction model for the current feature subset; re-determine whether the current feature subset is an empty set.

[0142] Among them, the current feature subset can be, for example, the initial feature subset or the new feature subset described above. When the current feature subset is not an empty set, the exemplary implementation of the electrocardiogram data processing method can refer to the relevant descriptions of steps S22 - S24 above.

[0143] By eliminating the index with the lowest importance score each time, the indices with higher importance scores can be retained in the processed feature subset, which can ensure the prediction accuracy of the feature subset while reducing the dimension of the feature subset.

[0144] In a possible implementation, training the prediction model based on the indices in the multiple heart rate variability indices of the sample object that are of the same index type as those in the second feature vector includes:

[0145] Arrange and combine multiple third feature vectors according to the different indices in the second feature vector, where the combinations of indices in different third feature vectors are different, or the combinations of indices in different third feature vectors and the sorting of the indices in the third feature vector are different; calculate the prediction accuracy of the prediction model when using, as the input of the prediction model, the indices in the multiple heart rate variability indices of the sample object that are of the same index type as those in the multiple third feature vectors respectively; determine the input index type of the prediction model according to the third feature vector with the highest prediction accuracy to obtain the trained prediction model.

[0146] Among them, the second feature vector may be the input feature vector in the relevant description of step S13 above, and the third feature vector may be the processed feature vector in the relevant description of step S13 above. The exemplary implementation of step S45 may refer to the relevant description of step S13 above.

[0147] Further screen the indices included in the second feature vector after importance ranking, and perform permutation and combination on the screening results to obtain a third feature vector with a higher accuracy rate, so that when the prediction model is obtained according to the third feature vector, the prediction accuracy of the prediction model is further improved. And there are multiple selection methods for permutation and combination, which improves the flexibility of obtaining the prediction model.

[0148] In a possible implementation, the electrocardiogram data of the sample object includes electrocardiogram data in the waking state and electrocardiogram data in the sleeping state. Determining the multiple heart rate variability indices of the sample object according to the electrocardiogram data of the sample object includes:

[0149] Determine the RR interval sequence of the sample object in the awake state and the RR interval sequence of the sample object in the sleep state according to the electrocardiogram data of the sample object; determine the multiple heart rate variability indexes corresponding to a plurality of preset index types according to the RR interval sequence of the sample object in the awake state and the RR interval sequence of the sample object in the sleep state, and the multiple heart rate variability indexes of the sample object include one or more of at least one time-domain heart rate variability index, at least one frequency-domain heart rate variability index, and at least one non-linear heart rate variability index.

[0150] In this way, the flexibility of the determination method of the heart rate variability index can be improved.

[0151] Figure 10 An exemplary schematic diagram of an electrocardiogram data processing device according to an embodiment of the present application is shown.

[0152] As Figure 10 shown, the present application provides an electrocardiogram data processing device, including:

[0153] An acquisition module 100, configured to acquire electrocardiogram data of a target object;

[0154] A determination module 200, configured to determine a first feature vector according to the electrocardiogram data of the target object, where the first feature vector includes multiple heart rate variability indexes determined from the electrocardiogram data of the target object;

[0155] A prediction module 300, configured to input the first feature vector into a trained prediction model to obtain a prediction result, where the prediction result indicates whether the target object is suitable for receiving a specific medical treatment.

[0156] In a possible implementation manner, the electrocardiogram data of the target object includes electrocardiogram data in the awake state and electrocardiogram data in the sleep state. Determining a first feature vector according to the electrocardiogram data of the target object includes: determining the RR interval sequence of the target object in the awake state and the RR interval sequence of the target object in the sleep state according to the electrocardiogram data of the target object; obtaining the first feature vector according to the RR interval sequence of the target object in the awake state and the RR interval sequence of the target object in the sleep state.

[0157] In a possible implementation manner, obtaining the first feature vector according to the RR interval sequence of the target object in the awake state and the RR interval sequence of the target object in the sleep state includes: determining a plurality of index types that are inputs of the trained prediction model; determining the multiple heart rate variability indexes corresponding to the multiple index types according to the RR interval sequence of the target object in the awake state and the RR interval sequence of the target object in the sleep state; obtaining the first feature vector according to the multiple heart rate variability indexes.

[0158] In a possible implementation, the prediction model is trained according to the electrocardiogram data of the sample object, and the method further includes: determining a plurality of heart rate variability indexes of the sample object according to the electrocardiogram data of the sample object; according to the effect after the sample object receives the specific medical treatment, determining the plurality of heart rate variability indexes of the sample object whose effect meets the preset conditions as positive samples, and determining the plurality of heart rate variability indexes of the sample object whose effect does not meet the preset conditions as negative samples; performing a significance analysis of the differences between the positive samples and the negative samples to obtain heart rate variability indexes with significant differences, where the difference between the heart rate variability indexes with significant differences between the positive samples and the negative samples is greater than a threshold; performing an importance ranking on the heart rate variability indexes with significant differences to obtain a second feature vector; and training the prediction model according to the indexes in the plurality of heart rate variability indexes of the sample object that are of the same index type as the indexes in the second feature vector.

[0159] In a possible implementation, performing an importance ranking on the heart rate variability indexes with significant differences to obtain a second feature vector includes: obtaining a feature subset according to all the heart rate variability indexes with significant differences, and determining the prediction accuracy of the prediction model for the feature subset; determining whether the current feature subset is an empty set; when the current feature subset is an empty set, performing an importance ranking on the heart rate variability indexes in the feature subset with the highest prediction accuracy according to the importance scores of the heart rate variability indexes in the feature subset with the highest prediction accuracy, to obtain a second feature vector.

[0160] In a possible implementation, performing an importance ranking on the heart rate variability indexes with significant differences to obtain a second feature vector further includes: when the current feature subset is not an empty set, repeating the following operations: inputting the current feature subset into a preset random forest model to obtain the importance scores of each heart rate variability index in the current feature subset; eliminating at least one heart rate variability index with the lowest importance score in the current feature subset, and using the set of the remaining heart rate variability indexes in the current feature subset as a new current feature subset, and determining the prediction accuracy of the prediction model for the current feature subset; and re-determining whether the current feature subset is an empty set.

[0161] In a possible implementation, according to the indicators among the multiple heart rate variability indicators of the sample object that are of the same indicator type as those in the second eigenvector, training the prediction model includes: arranging and combining according to different indicators in the second eigenvector to obtain multiple third eigenvectors, where the combinations of indicators in different third eigenvectors are different, or the combinations of indicators in different third eigenvectors and the sorting of the indicators in the third eigenvector are different; calculating the prediction accuracy of the prediction model when using, as the input of the prediction model, the indicators among the multiple heart rate variability indicators of the sample object that are of the same indicator type as those in the multiple third eigenvectors; and determining the input indicator type of the prediction model according to the third eigenvector with the highest prediction accuracy to obtain the trained prediction model.

[0162] In a possible implementation, the electrocardiogram data of the sample object includes electrocardiogram data in the waking state and electrocardiogram data in the sleeping state. Determining the multiple heart rate variability indicators of the sample object according to the electrocardiogram data of the sample object includes: determining the RR interval sequence of the sample object in the waking state and the RR interval sequence of the sample object in the sleeping state according to the electrocardiogram data of the sample object; and determining the multiple heart rate variability indicators corresponding to a preset multiple indicator types according to the RR interval sequence of the sample object in the waking state and the RR interval sequence of the sample object in the sleeping state, where the multiple heart rate variability indicators of the sample object include one or more of at least one time-domain heart rate variability indicator, at least one frequency-domain heart rate variability indicator, and at least one non-linear heart rate variability indicator.

[0163] In a possible implementation, the present application provides an electrocardiogram data processing device, including: a processor; a memory for storing instructions executable by the processor; where the processor is configured to: call the instructions stored in the memory to execute the above electrocardiogram data processing method.

[0164] In a possible implementation, the present application provides a non-volatile computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the above electrocardiogram data processing method is implemented.

[0165] Figure 11 FIG. shows an exemplary block diagram of an electrocardiogram data processing device 800 according to an embodiment of the present application. For example, the electrocardiogram data processing device 800 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0166] Refer to Figure 11, the electrocardiogram (ECG) data processing device 800 may include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input / output (I / O) interface 812, a sensor component 814, and a communication component 816.

[0167] The processing component 802 generally controls the overall operation of the ECG data processing device 800, such as operations associated with display, telephone calls, data communication, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above methods. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0168] The memory 804 is configured to store various types of data to support the operation of the ECG data processing device 800. Examples of such data include instructions for any application or method operating on the ECG data processing device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk, or an optical disk.

[0169] The power supply component 806 provides power to the various components of the ECG data processing device 800. The power supply component 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the ECG data processing device 800.

[0170] The multimedia component 808 includes a screen that provides an output interface between the ECG data processing device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors can not only sense the boundaries of touch or swipe actions but also detect the duration and pressure associated with the touch or swipe operation.

[0171] In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the electrocardiogram data processing device 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each of the front camera and the rear camera can be a fixed optical lens system or have a focal length and optical zoom capabilities.

[0172] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), which is configured to receive external audio signals when the electrocardiogram data processing device 800 is in an operation mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals can be further stored in the memory 804 or transmitted via the communication component 816. In some embodiments, the audio component 810 further includes a speaker for outputting audio signals.

[0173] The I / O interface 812 provides an interface between the processing component 802 and a peripheral interface module, and the peripheral interface module can be a keyboard, a click wheel, buttons, etc. These buttons can include but are not limited to: a home button, a volume button, a start button, and a lock button.

[0174] The sensor component 814 includes one or more sensors for providing a status assessment of various aspects of the device 800. For example, the sensor component 814 can detect the on / off state of the device 800, the relative positioning of components, such as the display and the keypad of the electrocardiogram data processing device 800. The sensor component 814 can also detect a change in the position of the electrocardiogram data processing device 800 or a component of the electrocardiogram data processing device 800, the presence or absence of user contact with the electrocardiogram data processing device 800, the orientation or acceleration / deceleration of the electrocardiogram data processing device 800, and the temperature change of the electrocardiogram data processing device 800. The sensor component 814 can include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor component 814 can also include a light sensor, such as a CMOS or a CCD image sensor, for use in imaging applications. In some embodiments, the sensor component 814 can further include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0175] The communication component 816 is configured to facilitate communication, either wired or wirelessly, between the electrocardiogram (ECG) data processing device 800 and other devices. The ECG data processing device 800 may access a wireless network based on a communication standard, such as WiFi, 2G, or 3G, or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 further includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0176] In an exemplary embodiment, the ECG data processing device 800 may be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components for performing the above method.

[0177] In an exemplary embodiment, a non-transitory computer-readable storage medium is also provided, such as a memory 804 including computer program instructions, which can be executed by a processor 820 of the ECG data processing device 800 to complete the above method.

[0178] Figure 12 FIG. shows an exemplary block diagram of an electrocardiogram (ECG) data processing device 1900 according to an embodiment of the present application. For example, the ECG data processing device 1900 may be provided as a server. Referring to Figure 12 , the ECG data processing device 1900 includes a processing component 1922, which further includes one or more processors, and memory resources represented by a memory 1932 for storing instructions executable by the processing component 1922, such as application programs. The application programs stored in the memory 1932 may include one or more modules each corresponding to a set of instructions. In addition, the processing component 1922 is configured to execute instructions to perform the above method.

[0179] The electrocardiogram (ECG) data processing device 1900 may further include a power supply component 1926 configured to perform power management of the device 1900, a wired or wireless network interface 1950 configured to connect the ECG data processing device 1900 to a network, and an input / output (I / O) interface 1958. The ECG data processing device 1900 may operate based on an operating system stored in the memory 1932, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, or the like.

[0180] In an exemplary embodiment, there is also provided a non-transitory computer-readable storage medium, such as the memory 1932 including computer program instructions, and the computer program instructions can be executed by the processing component 1922 of the ECG data processing device 1900 to complete the above method.

[0181] The present disclosure may be a system, a method, and / or a computer program product. The computer program product may include a computer-readable storage medium having thereon computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0182] A computer-readable storage medium may be a tangible device that can retain and store instructions for use by an instruction execution device. The computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer-readable storage medium include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disc (DVD), a memory stick, a floppy disk, a mechanically encoded device, such as a punch card or raised structures in grooves storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage medium used herein is not construed as being a transitory signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., an optical pulse through an optical fiber cable), or an electrical signal transmitted through a wire.

[0183] The computer-readable program instructions described herein can be downloaded to various computing / processing devices from a computer-readable storage medium or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network may include copper transmission cables, optical fiber transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing / processing device.

[0184] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state-setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages such as Smalltalk, C++, etc., and conventional procedural programming languages such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider). In some embodiments, by using the state information of the computer-readable program instructions to customize an electronic circuit, such as a programmable logic circuit, a field-programmable gate array (FPGA), or a programmable logic array (PLA), the electronic circuit can execute the computer-readable program instructions to implement various aspects of the present disclosure.

[0185] Aspects of the present disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0186] These computer-readable program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus to produce a machine such that the instructions, when executed by the processor of the computer or other programmable data processing apparatus, create an apparatus that implements the functions / acts specified in one or more boxes of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer-readable medium storing the instructions comprises a manufacture including instructions that implement various aspects of the functions / acts specified in one or more boxes of the flowchart and / or block diagram.

[0187] The computer-readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other devices to produce a computer-implemented process such that the instructions executed on the computer, other programmable data processing apparatus, or other devices implement the functions / acts specified in one or more boxes of the flowchart and / or block diagram.

[0188] The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of code, or a portion of an instruction, which comprises one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two consecutive blocks may in fact be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block of the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented by special-purpose hardware-based systems that perform the specified functions or acts, or by combinations of special-purpose hardware and computer instructions.

[0189] The embodiments of the present disclosure have been described above. The above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations are apparent to those of ordinary skill in the art without departing from the scope of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the marketplace, or to enable other ordinary skilled persons in the art to understand the embodiments disclosed herein.

Claims

1. A method for processing electrocardiogram data, characterized in that The method includes: Determining a plurality of heart rate variability indexes of the sample object according to the electrocardiogram data of the sample object; According to the effect after the sample object receives a specific medical treatment, determining the plurality of heart rate variability indexes of the sample object whose effect meets the preset conditions as positive samples, and determining the plurality of heart rate variability indexes of the sample object whose effect does not meet the preset conditions as negative samples; Performing a significance analysis of differences between the positive samples and the negative samples to obtain heart rate variability indexes with significant differences, where the difference between the heart rate variability indexes with significant differences between the positive samples and the negative samples is greater than a threshold; Performing an importance ranking on the heart rate variability indexes with significant differences to obtain a second feature vector; Training a prediction model according to the indexes in the plurality of heart rate variability indexes of the sample object that are of the same index type as the indexes in the second feature vector; Collecting the electrocardiogram data of the target object; Determining a first feature vector according to the electrocardiogram data of the target object, where the first feature vector includes a plurality of heart rate variability indexes determined from the electrocardiogram data of the target object; Inputting the first feature vector into the trained prediction model to obtain a prediction result, where the prediction result indicates whether the target object is suitable for receiving a specific medical treatment, and the specific medical treatment includes a vagus nerve stimulation surgery; Wherein, performing an importance ranking on the heart rate variability indexes with significant differences to obtain a second feature vector, including: Obtaining a feature subset according to all the heart rate variability indexes with significant differences, and determining the prediction accuracy of the prediction model for the feature subset; Judging whether the current feature subset is an empty set; When the current feature subset is not an empty set, repeating the following operations: Inputting the current feature subset into a preset random forest model to obtain the importance score of each heart rate variability index in the current feature subset; Eliminating at least one heart rate variability index with the lowest importance score in the current feature subset, and using the set of the remaining heart rate variability indexes in the current feature subset as a new current feature subset, and determining the prediction accuracy of the prediction model for the current feature subset; Re-judging whether the current feature subset is an empty set; When the current feature subset is an empty set, performing an importance ranking on the heart rate variability indexes in the feature subset with the highest prediction accuracy according to the importance scores of the heart rate variability indexes in the feature subset with the highest prediction accuracy to obtain a second feature vector.

2. The method according to claim 1, wherein The electrocardiogram data of the target object includes electrocardiogram data in the waking state and electrocardiogram data in the sleeping state. Determining a first feature vector according to the electrocardiogram data of the target object, including: Determining the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state according to the electrocardiogram data of the target object; Obtaining the first feature vector according to the RR interval sequence of the target object in the waking state and the RR interval sequence of the target object in the sleeping state.

3. The method according to claim 2, wherein Obtaining the first feature vector based on the RR interval sequence in the awake state and the RR interval sequence in the sleep state of the target object, includes: Determining multiple index types that serve as the input to the trained prediction model; Determining the multiple heart rate variability indexes corresponding to the multiple index types according to the RR interval sequence in the awake state and the RR interval sequence in the sleep state of the target object; Obtaining the first feature vector according to the multiple heart rate variability indexes.

4. The method according to claim 1, characterized in that, Training the prediction model according to the indexes in the multiple heart rate variability indexes of the sample object that are of the same index type as those in the second feature vector, includes: Performing permutations and combinations according to the different indexes in the second feature vector to obtain multiple third feature vectors, where the combinations of indexes in different third feature vectors are different, or the combinations of indexes in different third feature vectors and the sorting of the indexes in the third feature vectors are different; Calculating the prediction accuracy of the prediction model when using, as the input to the prediction model, the indexes in the multiple heart rate variability indexes of the sample object that are of the same index type as those in the multiple third feature vectors; Determining the input index type of the prediction model according to the third feature vector with the highest prediction accuracy to obtain the trained prediction model.

5. The method according to claim 1, characterized in that, The electrocardiogram data of the sample object includes electrocardiogram data in the awake state and electrocardiogram data in the sleep state. Determining the multiple heart rate variability indexes of the sample object according to the electrocardiogram data of the sample object, includes: Determining the RR interval sequence in the awake state and the RR interval sequence in the sleep state of the sample object according to the electrocardiogram data of the sample object; Determining the multiple heart rate variability indexes corresponding to the preset multiple index types according to the RR interval sequence in the awake state and the RR interval sequence in the sleep state of the sample object, and the multiple heart rate variability indexes of the sample object include one or more of at least one time-domain heart rate variability index, at least one frequency-domain heart rate variability index, and at least one non-linear heart rate variability index.

6. An electrocardiogram data processing device, characterized in that, The device includes: An acquisition module, configured to acquire the electrocardiogram data of the target object; A determination module, configured to determine a first feature vector according to the electrocardiogram data of the target object, where the first feature vector includes multiple heart rate variability indexes determined from the electrocardiogram data of the target object; A prediction module, configured to input the first feature vector into the trained prediction model to obtain a prediction result, where the prediction result indicates whether the target object is suitable for receiving a specific medical treatment, and the specific medical treatment includes a vagus nerve stimulation surgery; The device is further configured to: Determine the multiple heart rate variability indexes of the sample object according to the electrocardiogram data of the sample object; According to the effect after the sample object receives the specific medical treatment, determining the multiple heart rate variability indexes of the sample object whose effect meets the preset conditions as positive samples, and determining the multiple heart rate variability indexes of the sample object whose effect does not meet the preset conditions as negative samples; Perform a significance analysis on the positive samples and the negative samples to obtain heart rate variability indicators with significant differences, where the difference between the heart rate variability indicators with significant differences in the positive samples and the negative samples is greater than a threshold value; Rank the importance of the heart rate variability indicators with significant differences to obtain a second eigenvector; Train a prediction model based on the indicators with the same type as those in the second eigenvector among the multiple heart rate variability indicators of the sample object; Among them, ranking the importance of the heart rate variability indicators with significant differences to obtain a second eigenvector includes: Obtain a feature subset based on all heart rate variability indicators with significant differences, and determine the prediction accuracy of the prediction model for the feature subset; Judge whether the current feature subset is an empty set; When the current feature subset is not an empty set, repeat the following operations: Input the current feature subset into a preset random forest model to obtain the importance score of each heart rate variability indicator in the current feature subset; Eliminate at least one heart rate variability indicator with the lowest importance score in the current feature subset, and use the set of the remaining heart rate variability indicators in the current feature subset as the new current feature subset, and determine the prediction accuracy of the prediction model for the current feature subset; Re-judge whether the current feature subset is an empty set; When the current feature subset is an empty set, rank the importance of the heart rate variability indicators in the feature subset with the highest prediction accuracy according to the importance scores of the heart rate variability indicators in the feature subset with the highest prediction accuracy to obtain a second eigenvector.

7. An electrocardiogram data processing device, characterized in that, Include: A processor; A memory for storing instructions executable by the processor; Among them, the processor is configured to: Call the instructions stored in the memory to execute the method according to any one of claims 1 to 5.

8. A non-volatile computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, the method according to any one of claims 1 to 5 is implemented.

Images