Atrial fibrillation signal identification method and device, electronic equipment and computer readable medium

Abstract

The embodiment of the invention discloses an atrial fibrillation signal identification method and device, electronic equipment and a computer readable medium. A specific embodiment of the method comprises the following steps: acquiring an electrocardiogram signal through electrocardiogram acquisition equipment; performing noise reduction processing on the electrocardiogram signal to generate a denoised electrocardiogram signal; positioning an RR interval in the de-noised electrocardiogram signal to generate a to-be-marked RR interval sequence; identifying body position change interference and regular arrhythmia RR intervals of the RR interval sequence to be marked through XGboost and a random deep forest model, and generating an atrial fibrillation event group; performing morphological analysis on the de-noised electrocardiogram signals to generate electrocardiogram P-wave features; dynamically monitoring the electrocardiogram P wave features, and adjusting the atrial fibrillation starting time point and the atrial fibrillation ending time point of each atrial fibrillation event in the atrial fibrillation event group according to the dynamic monitoring result to obtain an optimized atrial fibrillation signal group. According to the embodiment, the accuracy of identifying the atrial fibrillation signal can be improved.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of signal recognition, and in particular to a method, device, electronic device, and computer-readable medium for recognizing atrial fibrillation signals. Background Art

[0002] Atrial fibrillation (AF) is one of the most common arrhythmias clinically, characterized by highly irregular atrial electrical activity, accompanied by atrial dysfunction and arrhythmias. AF not only affects the patient's quality of life, but also significantly increases the risk of stroke, heart failure and cardiovascular disease mortality. According to its duration, AF can be divided into paroxysmal AF (PAF), persistent AF (PAF) and permanent AF (PAF). Among them, paroxysmal AF (PAF) is characterized by intermittent attacks, which usually recover on its own within a few minutes to a few hours, posing a huge challenge to clinical diagnosis.

[0003] In the early stages of PAF, patients may have no obvious symptoms, or may only experience occasional non-specific symptoms such as palpitations, chest tightness, and dizziness. Due to the short duration of PAF, it is difficult to capture it through a conventional 12-lead electrocardiogram (ECG). As a tool that can continuously record ECG signals for 24 hours or even longer, dynamic electrocardiogram (Holter monitoring) has become an important means of diagnosing PAF. However, due to the randomness and brevity of PAF attacks, even with Holter monitoring, typical AF waveforms may not be recorded due to the absence of atrial fibrillation or extremely short duration of AF, resulting in missed diagnosis. In addition, factors such as patients' daily activities, changes in body position, and poor electrode contact may introduce noise, further affecting the accurate identification of PAF events.

[0004] Currently, the diagnosis of atrial fibrillation still mainly relies on manual ECG analysis, where doctors visually inspect the ECG waveform to identify AF events. Although this method has high reliability, it has the following major limitations: 1. Reliance on expert experience: Traditional manual interpretation methods are greatly affected by subjective factors. The doctor's experience and level of film review directly determine the accuracy of the diagnosis.

[0005] 2. High time cost: The recording time of a dynamic electrocardiogram is usually as long as 24 to 72 hours. Doctors need to analyze massive amounts of ECG data frame by frame, resulting in a huge workload for diagnosis.

[0006] 3. Difficulty in identifying short-term AF: Short-term AF events lasting less than 1 minute are more difficult to identify manually, and are prone to missed diagnosis or misdiagnosis due to heart rhythm variability or artifact interference.

[0007] The above information disclosed in this Background section is only for enhancement of understanding of the background of the inventive concept and therefore it may contain information that does not form the prior art that is already known in this country to a person of ordinary skill in the art. Summary of the invention

[0008] The content of this disclosure is used to introduce concepts in a brief form, which will be described in detail in the detailed implementation section below. The content of this disclosure is not intended to identify the key features or essential features of the technical solution claimed for protection, nor is it intended to limit the scope of the technical solution claimed for protection.

[0009] Some embodiments of the present disclosure provide atrial fibrillation signal recognition methods, devices, electronic devices, and computer-readable media to solve one or more of the technical problems mentioned in the above background technology section.

[0010] In a first aspect, some embodiments of the present disclosure provide a method for identifying an atrial fibrillation signal, the method comprising: acquiring an electrocardiogram signal through an electrocardiogram acquisition device; performing noise reduction processing on the electrocardiogram signal to generate a noise-reduced electrocardiogram signal, wherein the position of the R wave peak signal in the noise-reduced electrocardiogram signal is marked by R point detection; locating the RR interval in the noise-reduced electrocardiogram signal to generate a sequence of RR intervals to be marked, wherein each RR interval to be marked in the sequence of RR intervals to be marked corresponds to a group of atrial fibrillation start time points and atrial fibrillation end time points; identifying the RR intervals of the RR interval sequence to be marked with interference from body position changes and regular arrhythmia through XGboost and random deep forest models, and The RR interval sequences in these areas are excluded from the identification areas, and the purification process is completed. The purified RR interval sequences to be marked are subjected to atrial fibrillation event analysis to generate an atrial fibrillation event group, wherein the variability of each RR interval to be marked in the purified RR interval sequences to be marked is dynamically monitored through a sliding window; the above-mentioned denoised electrocardiogram signals are subjected to morphological analysis to generate electrocardiogram P wave features, wherein the above-mentioned electrocardiogram P wave features include the P wave starting point, the P wave peak and the P wave end point; the above-mentioned electrocardiogram P wave features are dynamically monitored, and according to the results of the dynamic monitoring, the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the above-mentioned atrial fibrillation event group are adjusted to obtain an optimized atrial fibrillation signal group.

[0011] In a second aspect, some embodiments of the present disclosure provide an atrial fibrillation signal recognition device, the device comprising: an acquisition unit, configured to acquire an electrocardiogram signal through an electrocardiogram acquisition device; a noise reduction processing unit, configured to perform noise reduction processing on the above electrocardiogram signal to generate a noise-reduced electrocardiogram signal, wherein the position of the R peak signal in the above noise-reduced electrocardiogram signal is marked by R point detection; a signal positioning unit, configured to locate the RR interval in the above noise-reduced electrocardiogram signal to generate a sequence of RR intervals to be marked, wherein each RR interval to be marked in the above sequence of RR intervals to be marked corresponds to a group of atrial fibrillation start time points and atrial fibrillation end time points; an atrial fibrillation event analysis unit, configured to perform body position change interference and regular arrhythmia RR on the sequence of RR intervals to be marked through XGboost and random deep forest models. The RR intervals in these areas are identified, and the RR interval sequences in these areas are excluded from the identified areas, the purification process is completed, and the RR interval sequences to be marked after the purification process are analyzed for atrial fibrillation events to generate an atrial fibrillation event group, wherein the variability of each RR interval to be marked in the RR interval sequences to be marked after the purification process is dynamically monitored through a sliding window; the morphological analysis unit is configured to perform morphological analysis on the above-mentioned denoised electrocardiogram signal to generate an electrocardiogram P wave feature, wherein the above-mentioned electrocardiogram P wave feature includes a P wave starting point, a P wave peak value and a P wave end point; the dynamic monitoring and optimization unit is configured to dynamically monitor the above-mentioned electrocardiogram P wave feature, and according to the result of the dynamic monitoring, adjust the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the above-mentioned atrial fibrillation event group to obtain an optimized atrial fibrillation signal group.

[0012] In a third aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner of the above-mentioned first aspect.

[0013] In a fourth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method described in any implementation manner of the above-mentioned first aspect is implemented.

[0014] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the atrial fibrillation signal recognition method of some embodiments of the present disclosure, the accuracy of atrial fibrillation event recognition can be improved. Specifically, the reason why missed diagnosis or misdiagnosis is easily caused by heart rhythm variability or artifact interference is that the traditional manual interpretation method is greatly affected by subjective factors, the doctor's experience and the level of film review directly determine the accuracy of the diagnosis, and for short-term AF events lasting less than 1 minute, manual recognition is more difficult. Based on this, the atrial fibrillation signal recognition method of some embodiments of the present disclosure, first, obtains an electrocardiogram signal through an electrocardiogram acquisition device. Noise is removed to ensure that the signal is clear. Secondly, the above electrocardiogram signal is subjected to noise reduction processing to generate a noise-reduced electrocardiogram signal. Among them, the R peak signal position in the above-mentioned noise-reduced electrocardiogram signal is marked by R point detection. The R point marking can provide basic data for subsequent RR interval calculation. Then, the RR interval in the above-mentioned noise-reduced electrocardiogram signal is located to generate a sequence of RR intervals to be marked. Among them, each RR interval to be marked in the above-mentioned RR interval sequence to be marked corresponds to a set of atrial fibrillation start time points and atrial fibrillation end time points. By preliminarily locating the RR interval, it can be used to determine the available RR interval and reduce data interference. After that, the RR interval sequence to be marked is identified by XGboost and random deep forest models for body position change interference and regular arrhythmia RR intervals, and the RR interval sequences in these areas are excluded from the identification area, and the purification process is completed, and the RR interval sequence to be marked after purification is analyzed for atrial fibrillation events to generate an atrial fibrillation event group, wherein the variability of each RR interval to be marked in the RR interval sequence to be marked after purification is dynamically monitored by a sliding window. Here, through dynamic monitoring, it can be used to preliminarily locate the occurrence interval of atrial fibrillation time. Then, the above-mentioned electrocardiogram signal after noise reduction is subjected to morphological analysis to generate electrocardiogram P wave features. Among them, the above-mentioned electrocardiogram P wave features include P wave starting point, P wave peak and P wave end point. Here, through morphological analysis, the P wave features can be further refined. Finally, the electrocardiogram P wave characteristics are dynamically monitored, and according to the results of the dynamic monitoring, the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the atrial fibrillation event group are adjusted to obtain an optimized atrial fibrillation signal group. Here, through more accurate P wave characteristics and dynamic monitoring, atrial fibrillation events can be more accurately located and the time points corresponding to atrial fibrillation events can be optimized. Thereby, the accuracy of atrial fibrillation signal recognition is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0016] Figure 1 is a flow chart of some embodiments of the atrial fibrillation signal identification method according to the present disclosure; Figure 2 is a schematic diagram marking the interference area; Figure 3 It is a schematic diagram marking the interval of atrial fibrillation events; Figure 4 It is a schematic diagram of optimizing the timing of atrial fibrillation; Figure 5 is a schematic diagram of the structure of some embodiments of the atrial fibrillation signal recognition device according to the present disclosure; Figure 6 It is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0017] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0018] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure can be combined with each other.

[0019] It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0020] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0021] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0022] At present, in terms of atrial fibrillation identification, AF identification is mainly based on the changes in the RR interval. Common methods include: 1. Threshold method: Set a fixed threshold for RR interval variability, such as when the root mean square difference (RMSSD) exceeds a certain set value, it is judged as AF. However, due to the significant differences in heart rate variability among different individuals, this method may lead to a high false positive rate (oversensitivity) or a high false negative rate (underidentification).

[0023] 2. Pattern matching algorithm: Use pre-set feature templates to match actual ECG data to improve the robustness of AF recognition. However, this method is easily affected by the diversity of arrhythmia types and has poor generalization ability.

[0024] 3. Artificial intelligence deep learning methods: such as convolutional neural networks, long short-term memory networks and their combination models.

[0025] However, the above method has the following disadvantages: Slow inference speed: Due to the huge amount of dynamic electrocardiogram signals, the use of complex deep neural networks (such as CNN-LSTM combined model) for full time series data analysis has a large amount of calculation and poor real-time performance, which makes it difficult to meet the needs of clinical immediate diagnosis. A 24-hour dynamic electrocardiogram takes several minutes or even dozens of minutes to analyze atrial fibrillation.

[0026] Strong data dependence: The training of deep learning models is highly dependent on large-scale labeled data, and the distribution differences between different data sets (such as device sampling rate, signal quality, patient characteristics, etc.) may affect the generalization ability of the model.

[0027] Insufficient ability to identify short-term AF: Due to the large individual differences in the performance of AF in the RR interval, traditional deep learning methods perform poorly in identifying short-burst AF and are easily affected by artifacts and heart rate variability.

[0028] In addition, although deep learning models can learn complex ECG signal features through massive data training, the diagnostic results are poorly interpretable due to the "black box" nature of the model, and it is difficult for doctors to understand the decision-making basis of the model, which may affect its promotion in medical applications. In addition, some deep learning methods have high computing resource requirements and are not suitable for resource-constrained wearable devices or remote medical monitoring systems, which further limits their application in actual clinical practice.

[0029] Therefore, there is an urgent need for an efficient, accurate and real-time atrial fibrillation identification method in clinical practice to overcome the shortcomings of the existing technology. The present invention combines the absolute irregularity of the RR interval + the disappearance or variation of the P wave with the rule engine, statistical analysis and machine learning (RF (Random Forest) + XGBoost) to provide a high-precision, ultra-high-speed, real-time and low-computational cost AF identification solution.

[0030] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0031] Figure 1 The process 100 of some embodiments of the atrial fibrillation signal recognition method according to the present disclosure is shown. The atrial fibrillation signal recognition method comprises the following steps: Step 101: Acquire an electrocardiogram signal through an electrocardiogram acquisition device.

[0032] In some embodiments, the execution subject (eg, computing device) of the atrial fibrillation signal recognition method may obtain an electrocardiogram signal through an electrocardiogram acquisition device in a wired or wireless manner, wherein the electrocardiogram signal may be data collected by the electrocardiogram acquisition device for a single patient.

[0033] Step 102: Perform noise reduction processing on the electrocardiogram signal to generate a noise-reduced electrocardiogram signal.

[0034] In some embodiments, the execution subject may perform noise reduction processing on the electrocardiogram signal to generate a noise-reduced electrocardiogram signal, wherein the R peak signal position in the noise-reduced electrocardiogram signal may be marked by R point detection.

[0035] In some optional implementations of some embodiments, the execution subject performs noise reduction processing on the electrocardiogram signal to generate a noise-reduced electrocardiogram signal, including: In the first step, bandpass filtering is used to remove baseline drift and electromyographic interference from the above electrocardiogram signal to obtain a noise-reduced electrocardiogram signal.

[0036] In the second step, the R wave peak value of the above-mentioned denoised electrocardiogram signal is detected by adaptive threshold detection and second-order derivative method to mark the position of the R wave peak signal in the above-mentioned denoised electrocardiogram signal. Here, R point detection can be performed by adaptive threshold detection and second-order derivative method.

[0037] Step 103: locate the RR intervals in the noise-reduced electrocardiogram signal to generate a sequence of RR intervals to be marked.

[0038] In some embodiments, the execution subject may locate the RR interval in the noise-reduced electrocardiogram signal to generate a sequence of RR intervals to be marked. Each RR interval in the sequence of RR intervals to be marked may correspond to a set of atrial fibrillation start time points and atrial fibrillation end time points. Specifically, the RR interval refers to the interval between two adjacent QRS (QRS complex) waves in the electrocardiogram signal.

[0039] In some optional implementations of some embodiments, the execution subject locates the RR interval in the noise-reduced electrocardiogram signal to generate a sequence of RR intervals to be marked, including: The above-mentioned electrocardiogram signal after noise reduction is subjected to obvious abnormal points elimination to obtain the RR interval sequence to be marked. Among them, the abnormal RR variation threshold can be obtained based on the mean ± 3 times the standard deviation to eliminate obvious abnormal points. Here, the mean and standard deviation can be the mean and standard deviation of each RR interval. The abnormal RR variation threshold can be a range value.

[0040] Step 104, the marked RR interval sequence is used to identify the interference of body position changes and regular arrhythmia RR intervals through XGboost and random deep forest models, and the RR interval sequences in these areas are excluded from the identification area to complete the purification process, and the marked RR interval sequence is analyzed for atrial fibrillation events to generate an atrial fibrillation event group.

[0041] In some embodiments, the execution subject can identify the RR intervals of body position changes and regular arrhythmia through XGboost and random deep forest models, exclude the RR intervals of these areas from the identification area, complete the purification process, and perform atrial fibrillation event analysis on the RR interval sequence to be marked to generate an atrial fibrillation event group. The variability of the RR interval to be marked can be dynamically calculated through a preset sliding window (for example, the window size can be 15 heartbeats) to avoid the limitations brought by a fixed time window.

[0042] In some optional implementations of some embodiments, the execution subject uses XGboost and random deep forest models to identify the RR intervals of body position change interference and regular arrhythmia RR intervals for the RR interval sequence to be marked, excludes the RR interval sequences in these areas from the identification area, completes the purification process, and performs atrial fibrillation event analysis on the RR interval sequence to be marked after the purification process to generate an atrial fibrillation event group, including: In the first step, the above-mentioned RR interval sequence to be marked is input into the pre-trained random forest model to obtain the RR interval sequence after preliminary detection, wherein the above-mentioned random forest model is used to identify the change trend of the RR interval according to the smoothness of the RR interval sequence to be marked, so as to determine whether there is a sudden abnormal signal, so as to mark and eliminate the RR interval that interferes with the change of body position and regular arrhythmia. Here, the standard deviation corresponding to the electrocardiogram signal and the RR interval after removal (as the change trend of the RR interval) can be input into the pre-trained random forest model to obtain the RR interval sequence after preliminary detection.

[0043] Specifically, a pre-set regular arrhythmia detection method can be used to determine the changing trend of the RR interval to identify whether there is a sudden abnormality (such as premature beats).

[0044] In practice, the detection method of regular arrhythmia may include: 1. If the RR interval changes in a fixed multiple (such as bigeminy, tripeminy), it may be a premature beat and needs to be eliminated. 2. If the RR variation is highly regular and appears periodically, it is classified as a regular arrhythmia and is not used as an AF identification object. Here, the processing logic can be: If it meets the regular arrhythmia pattern (premature beat), the data is marked and eliminated and not included in subsequent analysis. If the RR change pattern has no obvious regularity, it enters XGBoost for fine classification.

[0045] As an example, see Figure 2 Based on the effective data purification of RR interval, regular RR interval changes caused by premature beats, bigeminy, tripeminy, etc., as well as RR identification error areas caused by electrode detachment and electromyographic interference are removed.

[0046] In addition, the above random forest model can be trained and generated by the following steps: A local atrial fibrillation case database is introduced, including ECG signals such as noise interference data, normal heart rhythm (including atrial fibrillation), regular arrhythmia (such as premature beats), etc., and has been annotated by ECG experts. Data preprocessing: Extract the RR interval in the ECG signal (calculate the time interval between adjacent R waves). Calculate features: RR interval mutation point (mean ± 3 times standard deviation). RR interval change trend (smoothness, variation pattern). Regular arrhythmia characteristics (fixed periodic changes such as bigeminy and tripeminy). The input of the model can include classification labels (including noise labels, normal signal labels, regular arrhythmia labels, etc.). After training is completed, preliminary classification is performed to eliminate noise and regular arrhythmia data. The rest is handed over to the XGBoost model for fine classification to identify RR interval variations related to atrial fibrillation.

[0047] In the second step, the above-mentioned preliminary detection RR interval sequence is input into a pre-trained machine learning model to obtain a sequence of RR intervals to be marked. Among them, the machine learning model can be used to eliminate the preliminary detection RR intervals that represent electrode detachment or electromyographic interference, and the RR intervals to be marked represent RR intervals with high variability. Here, the above-mentioned machine learning model can be a pre-trained XGBoost model. The above-mentioned machine learning model can eliminate the preliminary detection RR intervals that represent electrode detachment or electromyographic interference through electrode detachment detection and electromyographic interference detection.

[0048] Specifically, electrode detachment detection can be: by abnormally amplifying the continuous RR interval variation value, the erroneous RR interval recording caused by electrode detachment can be identified. Myoelectric interference detection can be: if the RR interval variation is too large, but the RR interval variation pattern is highly random, it may be myoelectric interference. It can be judged in combination with the noise power ratio in the ECG signal.

[0049] For example, if the machine learning model determines that the RR interval is electrode detachment or electromyographic interference, the RR interval segment is removed and not included in AF identification. If the machine learning model determines that the RR interval is a true high-variability RR interval segment, the data is retained and the next step of sliding window calculation of RR variability is entered.

[0050] In addition, the misjudgment of the machine learning model can be corrected: the self-similarity of the RR interval variation of this segment is calculated. If the variation pattern is relatively chaotic, the data is retained and entered into the sliding window analysis. If the RR interval variation has a fixed pattern (such as isometric mutation), it is still treated as a regular arrhythmia and eliminated. In this way, it can be ensured that the AF segment will not be mistakenly deleted due to regular errors. Thus, the sensitivity of AF recognition is improved.

[0051] As an example, self-similarity can be calculated by the following steps: RR self-similarity is mainly used to distinguish regular arrhythmias (such as PVC, PAC) from irregular RR variations associated with AF. The core idea of ​​self-similarity is to analyze the pattern stability of the RR interval sequence. If a certain segment of RR variation shows a periodic or fixed pattern, it may be a regular arrhythmia rather than AF. Here, the adjacent RR correlation matrix of the RR interval sequence can be constructed. Let the RR interval sequence be: .

[0052] Where N represents the total number of RR intervals. Secondly, the Pearson correlation coefficient between adjacent RR intervals can be constructed as: .

[0053] Where r represents the Pearson correlation coefficient, which characterizes the correlation between the variations of adjacent RR intervals.

[0054] In practice, the RR variability of atrial fibrillation is extremely irregular, but arrhythmias (such as atrial premature beats and ventricular premature beats) and data noise (such as electrode detachment and electromyographic interference) can also cause RR variation, which can easily interfere with atrial fibrillation identification. Therefore, a RR interval purification technology based on RF+XGBoost combined classification is proposed to ensure the high quality of RR variation data. Then, by eliminating non-atrial fibrillation related interference, the purity of RR variability data is improved to ensure the high accuracy of subsequent atrial fibrillation identification. In addition, the traditional deep learning methods (such as CNN and LSTM) have a slow atrial fibrillation detection speed, and it may take several minutes to tens of minutes to process 24-hour ECG data, which is not suitable for practical applications. This method adopts a feature extraction + RF + XGBoost machine learning model based on a rule engine to greatly optimize the computing efficiency, and it only takes a few seconds to complete the atrial fibrillation detection of 24-hour ECG data. It greatly improves the computing speed, far exceeding the computing efficiency of deep learning methods, and is suitable for real-time monitoring and telemedicine. At the same time, it can also be deployed on ordinary PCs or even embedded devices, suitable for wearable ECG devices, Holter monitoring, hospital ECG equipment, etc.

[0055] The third step is to determine the standard deviation and coefficient of variation corresponding to each RR interval to be marked in the above-mentioned RR interval sequence to be marked. Among them, the standard deviation can be used to characterize the amplitude of RR variation in the atrial fibrillation interval. Here, determining the coefficient of variation can be used to characterize the irregularity of RR variation in the atrial fibrillation interval. The coefficient of variation can be generated by the following formula: .

[0056] Among them, CV represents the coefficient of variation, and a high CV value usually indicates the disorder of RR changes. SDNN represents the standard deviation, and the larger the value, the greater the volatility of the heart rhythm. The mean RR interval can be the mean RR interval calculated using a sliding window of 15 heart beats.

[0057] Step 4: For each RR interval to be marked in the above RR interval sequence to be marked and the corresponding coefficient of variation and standard deviation, the following steps are performed to generate an atrial fibrillation event group: 1. In response to the coefficient of variation being greater than a preset coefficient of variation interval and the standard deviation being greater than a preset standard deviation interval, the RR interval to be marked is marked as a high-variability RR interval. Wherein, if the coefficient of variation is greater than a preset coefficient of variation interval (e.g., [5%, 12%]), the RR interval to be marked is characterized as a high-variability RR interval.

[0058] 2. In response to the coefficient of variation being in the preset coefficient of variation interval and the standard deviation being in the preset standard deviation interval, marking the RR interval to be marked as a critical RR variation interval. Wherein, the coefficient of variation being in the preset coefficient of variation interval and the standard deviation being in the preset standard deviation interval (e.g., 50-100 milliseconds) indicates that the RR interval to be marked is in the critical variation zone.

[0059] 3. In response to the above-mentioned coefficient of variation being less than the above-mentioned preset coefficient of variation interval, and the standard deviation being less than the above-mentioned preset standard deviation interval, the above-mentioned RR interval to be marked is deleted. Among them, the coefficient of variation being less than the above-mentioned preset coefficient of variation interval, and the standard deviation being less than the above-mentioned preset standard deviation interval, can indicate that the heart rhythm is normal, and the possibility of atrial fibrillation of the RR interval to be marked can be directly excluded.

[0060] Optionally, performing an atrial fibrillation event analysis on the purified RR interval sequence to be marked to generate an atrial fibrillation event group also includes: For each highly variable RR interval, perform the following AF event generation steps: In the first step, in response to determining that the high variability RR interval is continuously greater than a preset number of heartbeats, the high variability RR interval is marked as an atrial fibrillation event. Wherein, if the high variability RR interval is continuously greater than a preset number of heartbeats (for example, 8 heartbeats), it is marked as an atrial fibrillation event.

[0061] In the second step, the atrial fibrillation events corresponding to the high-variability RR intervals adjacent to the above-mentioned high-variability RR intervals and having an interval less than a preset time length (e.g., 2 seconds) are merged into the same atrial fibrillation event. In response to the RR variation critical interval being between the intervals of the two atrial fibrillation events, the RR variation critical interval and the two atrial fibrillation events are marked as a continuous atrial fibrillation area.

[0062] As an example, refer to Figure 3 . The sliding window was used to calculate the RR variability on the purified effective RR interval data (excluding various interferences) and the approximate range of atrial fibrillation was calibrated.

[0063] In practice, combined with the analysis of event sequences, the determination of AF start and end points can be further optimized. Thus, the temporal resolution of AF identification can be improved. Here, after that, through the method based on fixed threshold, the RR variation pattern of different patients can be adaptively adjusted to avoid misjudgment caused by individual differences.

[0064] Step 105, performing morphological analysis on the denoised electrocardiogram signal to generate electrocardiogram P wave features.

[0065] In some embodiments, the execution subject may perform morphological analysis on the noise-reduced electrocardiogram signal to generate electrocardiogram P wave features, wherein the electrocardiogram P wave features include a P wave starting point, a P wave peak, and a P wave end point.

[0066] Here, the P wave amplitude can be: the voltage difference between the P wave peak and the isoelectric line, identifying the energy change of the P wave. P wave duration (Duration): Calculate the time interval between the start point and the end point of the P wave to determine the stability of the P wave. If the P wave duration is shortened or deformed, it indicates that atrial premature beats or atrial fibrillation have occurred. P wave morphology stability: Calculate the cosine similarity of the P wave morphology to analyze the consistency of the P wave morphology. If the P wave morphology suddenly changes or disappears in a continuous heartbeat, it is marked as the starting point or end point of the atrial fibrillation event. P wave position (Position): Calculate the position of the P wave relative to the R wave (PR interval) to ensure the normal existence of the P wave. If the P wave disappears in multiple heartbeats and the RR variability increases, it indicates that atrial fibrillation is occurring.

[0067] In some optional implementations of some embodiments, the execution subject performs morphological analysis on the noise-reduced electrocardiogram signal to generate electrocardiogram P wave features, including: In the first step, the denoised ECG signal is decomposed by multi-scale wavelet transform to obtain a decomposed ECG signal, wherein the decomposition is used to remove the baseline drift and myoelectric noise for a second time.

[0068] The second step is to enhance the P wave feature in the above decomposed electrocardiogram signal to obtain an enhanced electrocardiogram signal. The P wave feature in the above decomposed electrocardiogram signal can be enhanced by the db4 (Daubechies) transformation function in the wavelet transform to obtain an enhanced electrocardiogram signal. Thus, the P wave feature is more obvious in the signal.

[0069] The third step is to filter the enhanced electrocardiogram signal to obtain a filtered electrocardiogram signal, wherein the high-frequency component is used to filter noise and the low-frequency component is used to maintain the P wave morphology information.

[0070] The fourth step is to improve the signal-to-noise ratio of the filtered electrocardiogram signal to obtain an adjusted electrocardiogram signal. The filtered electrocardiogram signal can be smoothed by a Savitzky-Golay filter to improve the signal-to-noise ratio and maintain the integrity of the P wave waveform.

[0071] The fifth step is to extract the P wave feature of the adjusted ECG signal to obtain the P wave feature of the ECG. Among them, the voltage difference between the equipotential lines of the adjusted ECG signal is determined as the P wave peak. The P wave starting point and the P wave end point of the P wave feature in the adjusted ECG signal are determined, and the difference between the P wave starting point and the P wave end point is determined as the P wave duration. Determine the PR interval of the P wave feature relative to the R wave in the adjusted ECG signal. When the P waves of multiple heartbeats disappear in the adjusted ECG signal and the variability of the PR interval increases, it is marked as the starting point or end point of the atrial fibrillation event. Here, if the P wave duration is shortened or deformed, it may indicate the occurrence of atrial premature beats or atrial fibrillation. The PR interval refers to the time interval from the starting point of the P wave on the ECG to the starting point of the QRS complex.

[0072] In addition, the P wave morphology pre-similarity can be calculated to analyze the consistency of the P wave morphology. If the P wave morphology suddenly changes or disappears in a continuous heartbeat, it is marked as the start or end of an atrial fibrillation event.

[0073] Step 106, dynamically monitor the electrocardiogram P wave characteristics, and adjust the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the atrial fibrillation event group according to the result of the dynamic monitoring to obtain an optimized atrial fibrillation signal group.

[0074] In some embodiments, the above-mentioned execution subject can dynamically monitor the above-mentioned electrocardiogram P wave characteristics, and according to the results of the dynamic monitoring, adjust the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the above-mentioned atrial fibrillation event group to obtain an optimized atrial fibrillation signal group.

[0075] In some optional implementations of some embodiments, the execution subject dynamically monitors the electrocardiogram P wave characteristics, and adjusts the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the atrial fibrillation event group according to the result of the dynamic monitoring, to obtain an optimized atrial fibrillation signal group, including: In the first step, the P wave characteristics of the electrocardiogram are dynamically monitored through the second sliding window to obtain dynamic monitoring results, wherein the dynamic monitoring results may include: whether the P wave disappears in the window, the variability of the P wave waveform in the window, whether the abnormal P wave is accompanied by an irregular increase in the RR interval, and the abnormal coefficient of the RR interval after the P wave disappears.

[0076] In the second step, based on the above dynamic monitoring results, the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the atrial fibrillation event group are time-optimized to obtain an optimized atrial fibrillation signal group.

[0077] Specifically, time optimization can be performed in the following ways: If more than 80% of the P waves disappear within the sliding window, it can be preliminarily determined that an AF event may have occurred.

[0078] If the disappearance of the P wave is accompanied by highly irregular changes in the RR interval (ie, the abnormal coefficient is greater than the above-mentioned preset coefficient of variation interval), the confidence level of the AF event in this segment can be enhanced.

[0079] If the P wave waveform variability exceeds the threshold, it indicates that the P wave is abnormal. Here, the abnormality may include atrial premature beats or the starting point of atrial fibrillation (such as the starting point of AF).

[0080] If the P wave abnormality is accompanied by an irregular increase in the RR interval, it can be determined as the early stage of atrial fibrillation, and then the AF event confirmation process can be executed. The AF event confirmation process can be to analyze the P wave characteristics of the electrocardiogram for atrial fibrillation events again, or it can be sent to the manual end for final review.

[0081] Here, the start and end time points of atrial fibrillation events can also be optimized by following the steps below: Calculate the P wave morphology change trend of the 10 heartbeats before the AF event. If the P wave gradually weakens and the CV exceeds 10%, it is the early stage of atrial fibrillation and the starting point is adjusted forward. If the P wave suddenly disappears, the heartbeat is the starting time point of atrial fibrillation.

[0082] If the P wave gradually recovers within 5-10 heart beats after the diagnosis of atrial fibrillation, and the RR interval variability decreases, then the termination time of atrial fibrillation is determined. If the P wave recovers briefly but the RR interval variability is large, atrial fibrillation persists or is not completely terminated, and further determination is required as to whether it is paroxysmal atrial fibrillation.

[0083] As an example, see Figure 4 In the approximate atrial fibrillation area located by the RR interval, beat-to-beat P wave feature extraction and further accurate identification of the atrial fibrillation time point are performed.

[0084] In practice, traditional deep learning and other traditional methods are based on fixed time windows of 1 minute, 30 seconds or 10 seconds, which cannot accurately identify the specific heartbeats where atrial fibrillation occurs. This method uses sliding window RR variability analysis + P wave accurate detection to achieve single heartbeat level atrial fibrillation event recognition: 15 heartbeat sliding window calculates SDNN and CV, and screens high-variability RR areas. P wave morphology analysis ensures the accuracy of the starting and ending points of AF events and avoids errors caused by fixed window division. Therefore, atrial fibrillation detection of 24-hour dynamic ECG data can be completed within a few seconds, and it can also be accurate to the single heartbeat level. Break through the fixed time window method to improve the detection accuracy of paroxysmal atrial fibrillation and persistent atrial fibrillation events. Suitable for dynamic ECG monitoring, wearable devices and telemedicine applications.

[0085] The above-mentioned embodiments of the present disclosure have the following beneficial effects: through the atrial fibrillation signal recognition method of some embodiments of the present disclosure, the accuracy of atrial fibrillation event recognition can be improved. Specifically, the reason why missed diagnosis or misdiagnosis is easily caused by heart rhythm variability or artifact interference is that the traditional manual interpretation method is greatly affected by subjective factors, the doctor's experience and the level of film review directly determine the accuracy of the diagnosis, and for short-term AF events lasting less than 1 minute, manual recognition is more difficult. Based on this, the atrial fibrillation signal recognition method of some embodiments of the present disclosure, first, obtains an electrocardiogram signal through an electrocardiogram acquisition device. Noise is removed to ensure that the signal is clear. Secondly, the above electrocardiogram signal is subjected to noise reduction processing to generate a noise-reduced electrocardiogram signal. Among them, the R peak signal position in the above-mentioned noise-reduced electrocardiogram signal is marked by R point detection. The R point marking can provide basic data for subsequent RR interval calculation. Then, the RR interval in the above-mentioned noise-reduced electrocardiogram signal is located to generate a sequence of RR intervals to be marked. Among them, each RR interval to be marked in the above-mentioned RR interval sequence to be marked corresponds to a set of atrial fibrillation start time points and atrial fibrillation end time points. By preliminarily locating the RR interval, it can be used to determine the available RR interval and reduce data interference. After that, the RR interval sequence to be marked is identified by XGboost and random deep forest models for body position change interference and regular arrhythmia RR intervals, and the RR interval sequences in these areas are excluded from the identification area, and the purification process is completed, and the RR interval sequence to be marked after purification is analyzed for atrial fibrillation events to generate an atrial fibrillation event group, wherein the variability of each RR interval to be marked in the RR interval sequence to be marked after purification is dynamically monitored by a sliding window. Here, through dynamic monitoring, it can be used to preliminarily locate the occurrence interval of atrial fibrillation time. Then, the above-mentioned electrocardiogram signal after noise reduction is subjected to morphological analysis to generate electrocardiogram P wave features. Among them, the above-mentioned electrocardiogram P wave features include P wave starting point, P wave peak and P wave end point. Here, through morphological analysis, the P wave features can be further refined. Finally, the electrocardiogram P wave characteristics are dynamically monitored, and according to the results of the dynamic monitoring, the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the atrial fibrillation event group are adjusted to obtain an optimized atrial fibrillation signal group. Here, through more accurate P wave characteristics and dynamic monitoring, atrial fibrillation events can be more accurately located and the time points corresponding to atrial fibrillation events can be optimized. Thereby, the accuracy of atrial fibrillation signal recognition is improved.

[0086] Further references Figure 5 As an implementation of the methods shown in the above figures, the present disclosure provides some embodiments of an atrial fibrillation signal recognition device. These device embodiments are similar to Figure 1 Corresponding to the method embodiments shown, the atrial fibrillation signal recognition device can be specifically applied to various electronic devices.

[0087] like Figure 5 As shown, an atrial fibrillation signal recognition device 500 of some embodiments includes: an acquisition unit 501, a noise reduction processing unit 502, a signal positioning unit 503, an atrial fibrillation event analysis unit 504, a morphological analysis unit 505 and a dynamic monitoring and optimization unit 506. Among them, the acquisition unit 501 is configured to acquire an electrocardiogram signal through an electrocardiogram acquisition device; the noise reduction processing unit 502 is configured to perform noise reduction processing on the above electrocardiogram signal to generate a noise-reduced electrocardiogram signal, wherein the R peak signal position in the above noise-reduced electrocardiogram signal is marked by R point detection; the signal positioning unit 503 is configured to locate the RR interval in the above noise-reduced electrocardiogram signal to generate a sequence of RR intervals to be marked, wherein each RR interval to be marked in the above sequence of RR intervals to be marked corresponds to a group of atrial fibrillation start time points and atrial fibrillation end time points; the atrial fibrillation event analysis unit 504 is configured to identify the interference of body position changes and regular arrhythmia RR intervals in the sequence of RR intervals to be marked through XGboost and random deep forest models, and identify these areas The RR interval sequence to be marked is excluded from the identification area, the purification process is completed, and the atrial fibrillation event analysis is performed on the purified RR interval sequence to be marked to generate an atrial fibrillation event group, wherein the variability of each RR interval to be marked in the purified RR interval sequence to be marked is dynamically monitored through a sliding window; the morphological analysis unit 505 is configured to perform morphological analysis on the above-mentioned denoised electrocardiogram signal to generate an electrocardiogram P wave feature, wherein the above-mentioned electrocardiogram P wave feature includes a P wave starting point, a P wave peak value and a P wave end point; the dynamic monitoring and optimization unit 506 is configured to dynamically monitor the above-mentioned electrocardiogram P wave feature, and according to the result of the dynamic monitoring, adjust the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the above-mentioned atrial fibrillation event group to obtain an optimized atrial fibrillation signal group.

[0088] It can be understood that the units described in the atrial fibrillation signal identification device 500 are similar to those described in the reference Figure 1 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the atrial fibrillation signal identification device 500 and the units included therein, and will not be described in detail here.

[0089] Reference below Figure 6 , which shows a schematic structural diagram of an electronic device suitable for implementing some embodiments of the present disclosure. Figure 6 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure. Figure 6As shown, the computer device includes a processor, a memory and a network interface connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory. The non-volatile storage medium may store an operating system and a computer program. The computer program includes program instructions, which, when executed, enable the processor to execute any front-end page monitoring method. The processor is used to provide computing and control capabilities to support the operation of the entire computer device. The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium, which, when executed by the processor, enables the processor to execute any front-end page monitoring method. The network interface is used for network communications, such as sending assigned tasks, etc. Those skilled in the art will appreciate that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the scheme of the present disclosure, and does not constitute a limitation on the computer device to which the scheme of the present disclosure is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0090] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0091] In one embodiment, the processor is used to run a computer program stored in a memory to implement the following steps: acquiring an electrocardiogram signal through an electrocardiogram acquisition device; performing noise reduction processing on the electrocardiogram signal to generate a noise-reduced electrocardiogram signal, wherein the position of the R wave peak signal in the noise-reduced electrocardiogram signal is marked by R point detection; locating the RR interval in the noise-reduced electrocardiogram signal to generate a sequence of RR intervals to be marked, wherein each RR interval to be marked in the sequence of RR intervals to be marked corresponds to a group of atrial fibrillation start time points and atrial fibrillation end time points; and identifying the RR intervals to be marked by body position change interference and regular arrhythmia RR intervals through XGboost and random deep forest models. The RR interval sequences in these areas are identified and excluded from the identification area to complete the purification process, and the RR interval sequences to be marked after the purification process are subjected to atrial fibrillation event analysis to generate an atrial fibrillation event group, wherein the variability of each RR interval to be marked in the RR interval sequences to be marked after the purification process is dynamically monitored through a sliding window; a morphological analysis is performed on the above-mentioned denoised electrocardiogram signal to generate an electrocardiogram P wave feature, wherein the above-mentioned electrocardiogram P wave feature includes a P wave starting point, a P wave peak value and a P wave end point; the above-mentioned electrocardiogram P wave feature is dynamically monitored, and according to the result of the dynamic monitoring, the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the above-mentioned atrial fibrillation event group are adjusted to obtain an optimized atrial fibrillation signal group.

[0092] The embodiments of the present disclosure also provide a computer-readable storage medium, on which a computer program is stored. The computer program includes program instructions. The method implemented when the program instructions are executed can refer to the various embodiments of the atrial fibrillation signal recognition method of the present disclosure.

[0093] The computer-readable storage medium may be an internal storage unit of the computer device described in the above embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., provided on the computer device.

[0094] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0095] The above descriptions are only some preferred embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above inventive concept. For example, the above features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A method for identifying an atrial fibrillation signal, comprising: Obtaining electrocardiogram signals through an electrocardiogram acquisition device; Performing noise reduction processing on the electrocardiogram signal to generate a noise-reduced electrocardiogram signal, wherein the position of the R wave peak signal in the noise-reduced electrocardiogram signal is marked by R point detection; Locating the RR intervals in the noise-reduced electrocardiogram signal to generate a sequence of RR intervals to be marked, wherein each RR interval to be marked in the sequence of RR intervals to be marked corresponds to a set of atrial fibrillation start time points and atrial fibrillation end time points; The RR interval sequence to be marked is identified by XGboost and random deep forest model for the interference of body position change and regular arrhythmia RR interval, and the RR interval sequences in these areas are excluded from the identification area to complete the purification process, and the RR interval sequence to be marked after the purification process is analyzed for atrial fibrillation events to generate an atrial fibrillation event group, wherein the variability of each RR interval to be marked in the RR interval sequence to be marked after the purification process is dynamically monitored through a sliding window; Performing morphological analysis on the denoised electrocardiogram signal to generate electrocardiogram P wave features, wherein the electrocardiogram P wave features include a P wave starting point, a P wave peak, and a P wave end point; Dynamic monitoring is performed on the electrocardiogram P wave characteristics, and according to the results of the dynamic monitoring, the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the atrial fibrillation event group are adjusted to obtain an optimized atrial fibrillation signal group.

2. The method according to claim 1, characterized in that The performing noise reduction processing on the electrocardiogram signal to generate a noise-reduced electrocardiogram signal includes: Using bandpass filtering to remove baseline drift and electromyographic interference from the electrocardiogram signal to obtain a noise-reduced electrocardiogram signal; The R wave peak value of the denoised electrocardiogram signal is detected by adaptive threshold detection and second-order derivative method to mark the R wave peak signal position in the denoised electrocardiogram signal.

3. The method according to claim 1, characterized in that The locating the RR intervals in the noise-reduced electrocardiogram signal to generate a sequence of RR intervals to be marked includes: Obvious abnormal points are removed from the noise-reduced electrocardiogram signal to obtain a sequence of RR intervals to be marked.

4. The method according to claim 1, characterized in that: The RR interval sequence to be marked is identified by using XGboost and random deep forest model for body position change interference and regular arrhythmia RR interval, and the RR interval sequences in these areas are excluded from the identification area to complete the purification process, and the RR interval sequence to be marked after the purification process is analyzed for atrial fibrillation events to generate an atrial fibrillation event group, including: Inputting the RR interval sequence to be marked into a pre-trained random forest model to obtain a preliminary detected RR interval sequence, wherein the random forest model is used to identify the change trend of the RR interval according to the smoothness of the RR interval sequence to be marked, so as to determine whether there is a sudden abnormal signal, so as to mark and eliminate the RR intervals that interfere with the body position change and regular arrhythmia; Inputting the RR interval sequence after the preliminary detection into a pre-trained machine learning model to obtain a sequence of RR intervals to be marked, wherein the machine learning model is used to eliminate the RR intervals after the preliminary detection that represent electrode detachment or electromyographic interference, and the RR intervals to be marked represent RR intervals with high variability; Determine the standard deviation and coefficient of variation corresponding to each RR interval to be marked in the RR interval sequence to be marked, wherein the standard deviation is used to characterize the amplitude of RR variation in the atrial fibrillation interval, and the coefficient of variation is used to characterize the irregularity of RR variation in the atrial fibrillation interval; For each RR interval to be marked in the RR interval sequence to be marked and the corresponding coefficient of variation and standard deviation, the following steps are performed to generate an atrial fibrillation event group: In response to the coefficient of variation being greater than a preset coefficient of variation interval, and the standard deviation being greater than a preset standard deviation interval, marking the RR interval to be marked as a high-variability RR interval; In response to the coefficient of variation being within the preset coefficient of variation interval and the standard deviation being within the preset standard deviation interval, marking the RR interval to be marked as a critical RR variation interval; In response to the coefficient of variation being smaller than the preset coefficient of variation interval and the standard deviation being smaller than the preset standard deviation interval, the RR interval to be marked is deleted.

5. The method according to claim 4, characterized in that The step of performing atrial fibrillation event analysis on the purified RR interval sequence to be marked to generate an atrial fibrillation event group also includes: For each highly variable RR interval, perform the following AF event generation steps: In response to determining that the high-variability RR interval is continuously greater than a preset number of heartbeats, marking the high-variability RR interval as an atrial fibrillation event; Atrial fibrillation events corresponding to high-variability RR intervals adjacent to the high-variability RR interval and with an interval less than a preset time length are merged into the same atrial fibrillation event, wherein, in response to the RR variation critical interval being between the intervals of two atrial fibrillation events, the RR variation critical interval and the two atrial fibrillation events are marked as continuous atrial fibrillation areas.

6. The method according to claim 5, characterized in that The performing morphological analysis on the noise-reduced electrocardiogram signal to generate electrocardiogram P wave features includes: Decomposing the denoised electrocardiogram signal by multi-scale wavelet transform to obtain a decomposed electrocardiogram signal, wherein the decomposition is used to remove baseline drift and myoelectric noise for a second time; enhancing the P wave feature in the decomposed electrocardiogram signal to obtain an enhanced electrocardiogram signal; Filtering the enhanced electrocardiogram signal to obtain a filtered electrocardiogram signal, wherein the high-frequency component is used to filter noise and the low-frequency component is used to maintain P-wave morphology information; Improving the signal-to-noise ratio of the filtered electrocardiogram signal to obtain an adjusted electrocardiogram signal; Perform P wave feature extraction on the adjusted ECG signal to obtain the ECG P wave feature, wherein the voltage difference between the equipotential lines of the adjusted ECG signal is determined as the P wave peak, the P wave start point and the P wave end point of the P wave feature in the adjusted ECG signal are determined, and the difference between the P wave start point and the P wave end point is determined as the P wave duration, and the PR interval of the P wave feature in the adjusted ECG signal relative to the R wave is determined, and when the P waves of multiple heartbeats disappear and the variability of the PR interval increases in the adjusted ECG signal, it is marked as the start point or end point of an atrial fibrillation event.

7. The method according to claim 6, characterized in that The dynamic monitoring of the electrocardiogram P wave characteristics and adjusting the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the atrial fibrillation event group according to the result of the dynamic monitoring to obtain an optimized atrial fibrillation signal group include: Dynamically monitoring the electrocardiogram P wave characteristics through a second sliding window to obtain a dynamic monitoring result; According to the dynamic monitoring results, the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the atrial fibrillation event group are time-optimized to obtain an optimized atrial fibrillation signal group.

8. An atrial fibrillation signal recognition device, comprising: An acquisition unit, configured to acquire an electrocardiogram signal through an electrocardiogram acquisition device; a noise reduction processing unit configured to perform noise reduction processing on the electrocardiogram signal to generate a noise-reduced electrocardiogram signal, wherein the position of the R wave peak signal in the noise-reduced electrocardiogram signal is marked by R point detection; a signal positioning unit configured to locate the RR interval in the noise-reduced electrocardiogram signal to generate a sequence of RR intervals to be marked, wherein each RR interval to be marked in the sequence of RR intervals to be marked corresponds to a set of atrial fibrillation start time points and atrial fibrillation end time points; The atrial fibrillation event analysis unit is configured to identify the RR intervals of body position change interference and regular arrhythmia through XGboost and random deep forest models for the RR interval sequence to be marked, exclude the RR interval sequences in these areas from the identification area, complete the purification process, and perform atrial fibrillation event analysis on the RR interval sequence to be marked after the purification process to generate an atrial fibrillation event group, wherein the variability of each RR interval to be marked in the RR interval sequence to be marked after the purification process is dynamically monitored through a sliding window; a morphological analysis unit configured to perform morphological analysis on the noise-reduced electrocardiogram signal to generate an electrocardiogram P wave feature, wherein the electrocardiogram P wave feature includes a P wave starting point, a P wave peak, and a P wave end point; The dynamic monitoring and optimization unit is configured to dynamically monitor the electrocardiogram P wave characteristics, and adjust the atrial fibrillation start time point and the atrial fibrillation end time point of each atrial fibrillation event in the atrial fibrillation event group according to the results of the dynamic monitoring to obtain an optimized atrial fibrillation signal group.

9. An electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A computer readable medium having a computer program stored thereon, wherein: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images