Method for heart rhythm recognition, medical device and medical system

Abstract

The application provides a heart rhythm recognition method, a medical device and a medical system. The heart rhythm recognition method comprises: a medical device collecting electrocardio characteristic data to be recognized; the medical device obtaining a first heart rhythm recognition result based on the electrocardio characteristic data to be recognized and an inherent heart rhythm recognition algorithm; the medical device obtaining a second heart rhythm recognition result based on the electrocardio characteristic data to be recognized and a pre-stored expansion model library; and the medical device obtaining a final heart rhythm recognition result based on the first heart rhythm recognition result and the second heart rhythm recognition result. The application reserves an expansion model library interface in the medical device, and does not need to update the inherent heart rhythm recognition algorithm of the medical device, so as to flexibly cope with the algorithm upgrading demand faced by the medical device after being put on the market.

Description

Technical Field

[0001] This invention relates to the field of medical device technology, specifically to a heart rhythm recognition method, medical device, and medical system. Background Technology

[0002] Cardiac arrest is a life-threatening emergency that can occur in people of all ages. Resuscitation during cardiac arrest is a race against time, with the golden window of opportunity being only four minutes. For high-risk groups, especially patients who have experienced malignant arrhythmic events such as ventricular tachycardia or ventricular fibrillation, implantable cardioverter-defibrillators (ICDs) are one of the best preventative measures. ICDs with cardioverter-defibrillator functionality mainly include traditional ICDs, newer subcutaneous implantable cardioverter-defibrillators (SICDs), and cardiac resynchronization therapy (CRT-D) devices with pacing capabilities. These types of implantable cardioverter-defibrillators monitor the patient's heart rhythm 24 hours a day and can automatically defibrillate the patient immediately upon the occurrence of a malignant arrhythmic event to restore a normal heart rhythm, thus saving the patient's life.

[0003] A rhythmic, normal heartbeat allows blood to circulate efficiently throughout the body. When a malignant arrhythmia occurs, the heart loses its normal rhythm, and its pumping function weakens or even completely ceases. If not promptly converted, this can lead to myocardial ischemia, loss of the heart's pumping function, and ultimately, death. Therefore, rapidly and accurately identifying malignant arrhythmias and promptly administering electrical defibrillation is crucial for effectively saving lives and is also a primary indicator for evaluating the performance of various implantable cardioverter-defibrillators (ICDs).

[0004] Treatment presupposes identification, so the algorithm for identifying malignant arrhythmias is a key area of ​​competition among manufacturers of implantable cardioverter-defibrillators (ICDs). This algorithm must ensure both timeliness and accuracy. Accuracy can be further subdivided into the missed defibrillation rate (algorithm sensitivity) and the false defibrillation rate (algorithm specificity), ensuring that no malignant arrhythmic events are missed while also preventing misdiagnosis of non-malignant arrhythmic events.

[0005] The existing cardiac rhythm recognition algorithm mainly involves the following steps: the algorithm monitors the patient's electrocardiogram (ECG) signal and determines whether a malignant arrhythmia event has occurred. If not, monitoring continues; if a malignant arrhythmia event occurs, pre-charging of the high-voltage capacitor begins. After pre-charging, the algorithm again determines whether a malignant arrhythmia event exists. If the malignant arrhythmia event persists, the patient is given electrical defibrillation; if the malignant arrhythmia event terminates spontaneously, defibrillation is abandoned, and ECG signal monitoring continues. The core of the algorithm consists of two malignant arrhythmia event determination steps, which can be achieved through the cooperation of multiple determination mechanisms to make a final judgment.

[0006] Existing cardiac rhythm recognition algorithms collect and analyze patients' electrocardiogram (ECG) signals to determine the presence of malignant arrhythmic events and perform electrical defibrillation treatment based on the calculation results. The various feature values ​​extracted by these algorithms include, but are not limited to: heart rate, overall and local morphology of the ECG signal, QRS pulse width, PP interval, RR interval, and PR interval. Although current algorithms possess high sensitivity and specificity, they still exhibit certain false defibrillation and missed defibrillation rates.

[0007] One major reason is that algorithm development and optimization are based on existing datasets. The accuracy of an algorithm directly depends on the completeness and diversity of the dataset. However, the completeness and diversity of the dataset are relative. During the algorithm's development process, the test datasets are relatively limited. Compared to the real-world datasets used after the algorithm is officially deployed, the datasets used in the development phase are often insufficient. Therefore, when facing real-world datasets, unforeseen special circumstances will always arise. In these cases, the algorithm's performance will decrease. Thus, even if an algorithm has extremely high accuracy during testing, it will inevitably decrease in real-world scenarios, and in some cases, algorithmic vulnerabilities may appear. When the algorithm is exposed to these vulnerabilities, its false positive rate will increase significantly.

[0008] The conventional solution is to enhance the robustness of the algorithm through continuous iteration and upgrades. However, for implantable cardioverter-defibrillators (ICDs), due to various safety considerations, algorithm upgrades are often only applied to later releases. The current algorithm is embedded in the ICD at the factory; once implanted, the algorithm cannot be upgraded or optimized, and only a few parameters can be fine-tuned via a programmer. Because fine-tuning these parameters makes it difficult to optimize the algorithm as a whole, when a relatively high false positive or false negative rate is later discovered for certain ECG signals, the algorithm is often unable to effectively address the issue, leaving patients to bear the potential risk of misdiagnosis.

[0009] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of the present invention, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0010] To address the problems in the existing technology, the present invention aims to provide a heart rhythm recognition method, a cardiac defibrillation method, a medical device, and a medical system. By reserving an interface for expanding the model library in the medical device, it eliminates the need to update the inherent heart rhythm recognition algorithm of the medical device, thus flexibly responding to the algorithm upgrade needs faced after the medical device is launched on the market.

[0011] This invention provides a method for heart rhythm recognition, comprising:

[0012] Medical equipment collects electrocardiogram (ECG) feature data to be identified;

[0013] The medical device obtains a first heart rhythm recognition result based on the electrocardiogram feature data to be identified and the inherent heart rhythm recognition algorithm;

[0014] The medical device obtains a second heart rhythm recognition result based on the ECG feature data to be identified and a pre-stored expanded model library;

[0015] The medical device obtains the final heart rhythm recognition result based on the first heart rhythm recognition result and the second heart rhythm recognition result.

[0016] In some embodiments, based on the ECG feature data to be identified and a pre-stored expanded model library, a second heart rhythm identification result is obtained, including the following steps:

[0017] Determine whether at least one heart rhythm recognition rule exists in the expanded model library;

[0018] If it does not exist, the second heart rhythm recognition result will be set to unmatched;

[0019] If present, a second heart rhythm recognition result is obtained based on the heart rhythm recognition rules and the ECG feature data to be recognized. Each heart rhythm recognition rule includes ECG feature conditions and recognition judgment results.

[0020] In some embodiments, obtaining a second heart rhythm recognition result based on the heart rhythm recognition rules and the electrocardiogram feature data to be recognized includes the following steps:

[0021] Determine whether the ECG feature data to be identified meets the ECG feature conditions in each of the heart rhythm recognition rules;

[0022] If the ECG feature data to be identified meets the ECG feature conditions in a heart rhythm identification rule, then the identification judgment result of the matching heart rhythm identification rule is taken as the second heart rhythm identification result.

[0023] If the ECG feature data to be identified does not meet the ECG feature conditions in all heart rhythm recognition rules, then the second heart rhythm recognition result is set as unmatched.

[0024] In some embodiments, obtaining a final heart rhythm recognition result based on the first heart rhythm recognition result and the second heart rhythm recognition result includes the following steps:

[0025] Determine whether the second heart rhythm recognition result is a mismatch;

[0026] If so, the first heart rhythm recognition result shall be taken as the final heart rhythm recognition result;

[0027] If not, then the final heart rhythm recognition result is generated by combining the first heart rhythm recognition result and the second heart rhythm recognition result.

[0028] In some embodiments, the final heart rhythm recognition result is generated by combining the first heart rhythm recognition result and the second heart rhythm recognition result, including the following steps:

[0029] The first weight and the second weight of the first heart rhythm recognition result and the second heart rhythm recognition result are obtained respectively;

[0030] The first heart rhythm recognition result and the second heart rhythm recognition result are weighted and averaged to obtain the final heart rhythm recognition result.

[0031] In some embodiments, obtaining the first weight and the second weight of the first heart rhythm recognition result and the second heart rhythm recognition result respectively includes the following steps:

[0032] Obtain the weight of the conforming heart rhythm recognition rule, and use it as the second weight of the second heart rhythm recognition result;

[0033] The first weight of the first heart rhythm recognition result is determined based on the second weight, so that the sum of the first weight and the second weight value is the preset weight sum.

[0034] In some embodiments, the following steps are also included:

[0035] The medical device receives expanded model library data from external devices and stores it locally as a pre-stored expanded model library.

[0036] In some embodiments, the following steps are also included:

[0037] The medical device sends electrocardiogram (ECG) monitoring records to an external device, and the ECG monitoring records include ECG characteristic data and device identification results.

[0038] In some embodiments, the following steps are also included:

[0039] Multiple ECG event records corresponding to multiple medical devices are collected, and each ECG event record includes ECG feature data, device identification results, and doctor identification results;

[0040] Records of ECG events where the device identification results differ from the doctor's identification results are extracted as records of misjudged ECG events.

[0041] Based on the misjudged ECG event records, heart rhythm recognition rules are generated as data to expand the model library. In each heart rhythm recognition rule, the recognition judgment result corresponds to the doctor's recognition result.

[0042] In some embodiments, after generating heart rhythm recognition rules based on the misjudged ECG event records, the following steps are further included:

[0043] Based on the analysis of the misjudged ECG event records, the correlation strength between each heart rhythm recognition rule and the misjudgment result is analyzed;

[0044] The weight of each heart rhythm recognition rule is set based on the correlation strength.

[0045] This invention also provides a method for cardiac defibrillation, comprising the following steps:

[0046] The medical device collects electrocardiogram (ECG) feature data to be identified, and obtains a first heart rhythm identification result based on the ECG feature data to be identified and the inherent heart rhythm identification algorithm.

[0047] The medical device determines whether the first heart rhythm recognition result is a defibrillable recognition result;

[0048] If the first heart rhythm recognition result is a defibrillation-ready result, the medical device precharges the defibrillation actuator.

[0049] After pre-charging is completed, the medical device executes the heart rhythm recognition method as described in claim 1 to obtain the final heart rhythm recognition result;

[0050] The medical device determines whether the final heart rhythm recognition result is a defibrillable recognition result;

[0051] If the final heart rhythm identification result is a defibrillation-ready result, the medical device controls the defibrillation actuator to perform defibrillation.

[0052] This invention also provides a medical device, comprising:

[0053] processor;

[0054] The memory stores the inherent heart rhythm recognition algorithm, the expanded model library, and the executable instructions of the processor.

[0055] The processor is configured to perform steps of the heart rhythm recognition method or the cardiac defibrillation method by executing the executable instructions.

[0056] This invention also provides a medical system, comprising:

[0057] The programmable controller is used to send expanded model library data to medical devices; and

[0058] The medical device is used to receive expanded model library data and store it locally as a pre-stored expanded model library.

[0059] In some embodiments, the device further includes an algorithm server, and the medical device is also used to send electrocardiogram (ECG) monitoring records to the programmable controller, the ECG monitoring records including ECG feature data and device identification results;

[0060] The programmable controller is also used to acquire the electrocardiogram monitoring record from the medical device and send it to the algorithm server;

[0061] The algorithm server is used to collect multiple ECG event records corresponding to multiple medical devices. Each ECG event record includes ECG feature data, device identification results, and doctor identification results. ECG event records where the device identification results differ from the doctor identification results are extracted as misjudged ECG event records. Based on the misjudged ECG event records, heart rhythm recognition rules are generated as data to expand the model library. In each heart rhythm recognition rule, the recognition judgment result corresponds to the doctor identification result.

[0062] The heart rhythm recognition method, cardiac defibrillation method, medical device, and medical system provided by this invention have the following advantages:

[0063] This invention, by reserving an interface for an expanded model library in medical devices, eliminates the need to update the inherent heart rhythm recognition algorithm of the medical devices. Furthermore, the expanded model library can strengthen the inherent heart rhythm recognition algorithm, flexibly addressing the algorithm upgrade needs faced after the medical devices are launched on the market. This improves the accuracy and safety of ECG monitoring, effectively avoids the additional risks brought about by modifying the inherent heart rhythm recognition algorithm architecture, and by supplementing the expanded model library, it can flexibly cope with various special situations, with a simple and fast upgrade method.

[0064] This invention can be applied to implantable cardioverter-defibrillators (ICDs) to identify heart rhythms and further determine whether to perform defibrillation based on the heart rhythm identification results. However, this invention is not limited to this and can be applied to a wider range of algorithmic applications. The heart rhythm identification method of this invention can also be applied to other electrocardiogram (ECG) monitoring devices, such as external ECG monitoring devices and automated external defibrillators (AEDs). The cardiac defibrillation method is not limited to implantable defibrillators but can also be used with external defibrillators. Attached Figure Description

[0065] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0066] Figure 1 This is a schematic diagram of the structure of a heart rhythm recognition method according to an embodiment of the present invention;

[0067] Figure 2 This is a flowchart of an embodiment of the present invention for obtaining the second heart rhythm recognition result based on an expanded model library;

[0068] Figure 3 This is a flowchart of obtaining the final heart rhythm recognition result according to an embodiment of the present invention;

[0069] Figure 4 This is a schematic diagram of the structure of a medical device according to an embodiment of the present invention;

[0070] Figure 5 This is a schematic diagram of the structure of a medical system according to an embodiment of the present invention;

[0071] Figure 6 This is a flowchart of generating an expanded model library according to an embodiment of the present invention;

[0072] Figure 7 This is a flowchart of a cardiac defibrillation method according to an embodiment of the present invention. Detailed Implementation

[0073] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that the invention will be thorough and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The same reference numerals in the drawings denote the same or similar structures, and therefore repeated descriptions of them will be omitted. The words “or” and “or” in the specification may mean “and” or “or”. Although the terms “upper,” “lower,” “between,” etc., may be used in this specification to describe different exemplary features and elements of the invention, these terms are used herein only for convenience, such as the orientation according to the examples described in the accompanying drawings. Nothing in this specification should be construed as requiring a specific three-dimensional orientation of the structure to fall within the scope of the invention. Although “first” or “second,” etc., are used in this specification to denote certain features, they are only to indicate function and not to limit the number and importance of specific features.

[0074] This invention provides a heart rhythm recognition method for flexibly upgrading the internal heart rhythm recognition algorithm of medical devices, wherein the medical devices include at least a memory and a processor. Figure 1 As shown, the heart rhythm recognition method includes the following steps:

[0075] S100: Medical equipment collects electrocardiogram (ECG) feature data to be identified;

[0076] Specifically, the processor of the medical device collects real-time ECG feature data from the ECG data acquisition module as ECG feature data to be identified. The ECG data acquisition module may include, for example, ECG sensing electrodes, which are implanted or attached to the area around the heart via vascular implantation or subcutaneous implantation.

[0077] S200: The medical device obtains a first heart rhythm recognition result based on the electrocardiogram feature data to be identified and the inherent heart rhythm recognition algorithm;

[0078] Specifically, the inherent heart rhythm recognition algorithm is pre-stored in the memory. The processor of the medical device calls the inherent heart rhythm recognition algorithm to process the electrocardiogram feature data to be identified and obtain the first heart rhythm recognition result. The inherent heart rhythm recognition algorithm is the heart rhythm recognition algorithm that has been fixed in the medical device at the time of manufacture.

[0079] S300: The medical device obtains a second heart rhythm recognition result based on the ECG feature data to be identified and a pre-stored expanded model library;

[0080] Specifically, the expanded model library is pre-stored in the memory, and the processor of the medical device matches the ECG feature data to be identified with the expanded model library to obtain the second heart rhythm recognition result;

[0081] S400: The medical device obtains the final heart rhythm recognition result based on the first heart rhythm recognition result and the second heart rhythm recognition result.

[0082] Therefore, the medical device no longer relies solely on its own inherent heart rhythm recognition algorithm to obtain the first heart rhythm recognition result. At the same time, the medical device comprehensively considers the matching results in the expanded model library to strengthen the inherent heart rhythm recognition algorithm, flexibly responding to the algorithm upgrade needs faced after the medical device is launched on the market, thereby improving the accuracy and safety of ECG monitoring.

[0083] Therefore, by reserving an interface for an expanded model library in medical devices, this invention eliminates the need to update the inherent heart rhythm recognition algorithm of the medical devices. When it is necessary to strengthen the inherent heart rhythm recognition algorithm, heart rhythm recognition rules can be added to the expanded model library, effectively avoiding the additional risks brought about by modifying the inherent heart rhythm recognition algorithm architecture. By supplementing the expanded model library, various special situations can be flexibly addressed, and the upgrade method is simple and fast.

[0084] The heart rhythm recognition method of this invention can be applied to implantable cardioverter-defibrillators (ICDs) to identify heart rhythms and further determine whether to perform defibrillation based on the heart rhythm recognition results. However, this invention is not limited to this. The heart rhythm recognition method of this invention can also be applied to other ECG monitoring devices, such as ECG monitoring equipment, which only monitors and identifies heart rhythms and outputs the heart rhythm recognition results to external devices without performing defibrillation. For example, this heart rhythm recognition method can also be applied to automated external defibrillators (AEDs). Although AED upgrades can be performed externally, batch algorithm upgrades for a large number of commercially available AEDs are still very troublesome and involve additional risks from changing the architecture of the inherent heart rhythm recognition algorithm. Therefore, applying this heart rhythm recognition method to external monitoring devices or defibrillators can also strengthen the inherent heart rhythm recognition algorithm, flexibly addressing the algorithm upgrade needs faced after the medical devices are launched on the market, and improving the accuracy and safety of ECG monitoring.

[0085] In this embodiment, the ECG feature categories in the ECG feature data may include one or more of the following: heart rate, overall and local morphology of the ECG signal, QRS pulse width, PP interval, RR interval, and PR interval, and are not limited to the ECG feature categories listed here. The ECG feature categories used by the medical device to process the ECG feature data based on the inherent heart rhythm recognition algorithm and the expanded model library may be the same or different. For example, the inherent heart rhythm recognition algorithm may obtain a first heart rhythm recognition result based on a comprehensive judgment of the heart rate, overall and local morphology of the ECG signal, QRS pulse width, PP interval, RR interval, and PR interval, while the expanded model library may obtain a second heart rhythm recognition result based on the overall and local morphology analysis of the ECG signal. Furthermore, this invention is not limited to the cases listed here.

[0086] The inherent heart rhythm recognition algorithm is the algorithm that is already fixed inside the medical device at the factory. Subsequent adjustments, except for some parameters via a programmable controller, do not require further modification to the algorithm's architecture. This invention does not limit the specific algorithm type of the inherent heart rhythm recognition algorithm. Depending on the needs of different medical devices at the time of manufacture, various existing heart rhythm recognition algorithms can be selected. For example, a heart rhythm recognition algorithm based on electrocardiogram (ECG) signal morphology analysis, or a heart rhythm recognition algorithm based on time-domain and frequency-domain analysis of ECG signals, or inputting ECG feature data into a trained convolutional neural network or binary classification network for heart rhythm recognition, etc. The algorithm can also be a combination of the above algorithms, or a combination of the above algorithms with other algorithms.

[0087] The expanded model library is mainly aimed at situations where the inherent heart rhythm recognition algorithm may make misjudgments. For example, the inherent heart rhythm recognition algorithm of medical devices may mistakenly identify a heart rhythm recognition result that does not require defibrillation as a heart rhythm recognition result that can be defibrillated, and guide the defibrillation agency to perform defibrillation, causing discomfort to the wearer. Alternatively, the inherent heart rhythm recognition algorithm of medical devices may also mistakenly identify a heart rhythm recognition result that requires defibrillation as a heart rhythm recognition result that does not require defibrillation, thus delaying the best treatment time.

[0088] Since the extended model library is an enhancement to the inherent heart rhythm recognition algorithm, it does not require the construction of a complete and complex algorithm architecture like the inherent heart rhythm recognition algorithm. In this embodiment, an interface for the extended model library is reserved. The extended model library can be empty at the factory, that is, it does not contain any heart rhythm recognition rules. If no rules needing enhancement are subsequently found, the extended model library can remain empty. When rules needing enhancement are found during subsequent use, the extended model library can include at least one heart rhythm recognition rule, each of which includes ECG characteristic conditions and a recognition judgment result. The recognition judgment result is the doctor's recognition result when the ECG characteristic conditions are met, which is generally a result different from the recognition result of the inherent heart rhythm recognition algorithm of the medical device. For example, when the electrocardiogram (ECG) characteristics meet a heart rhythm recognition rule, if the inherent heart rhythm recognition algorithm is used, the result will be a defibrillable recognition, which is highly likely to be a misjudgment. Therefore, the recognition decision in the heart rhythm recognition rule will be a non-defibrillable recognition. Conversely, if the inherent heart rhythm recognition algorithm is used, the result will also be a non-defibrillable recognition, which is highly likely to be a misjudgment. Therefore, the recognition decision in the heart rhythm recognition rule will be a defibrillable recognition. Thus, this expanded model library can correct errors in the inherent heart rhythm recognition algorithm.

[0089] The ECG characteristic conditions may include, for example, numerical range conditions for one or more ECG characteristics and / or relationships between the numerical ranges of multiple ECG characteristics. The numerical range conditions for an ECG characteristic may include a single ECG characteristic, such as a heart rate between (x1, x2) for a rhythm recognition rule, where the values ​​of x1 and x2 can be selected and set as needed. The numerical range conditions for an ECG characteristic may also include a combination of two or more ECG characteristic conditions, such as a heart rate between (x1, x2) and a QRS pulse width between (y1, y2) for a rhythm recognition rule, where the values ​​of y1 and y2 can also be selected and set as needed. Relationships between the numerical ranges of multiple ECG characteristics may include, for example, the PP interval being m1 times the RR interval, or the value of characteristic a being a combination of the values ​​of characteristic b and characteristic c, where the value of m1 can also be selected and set as needed. The numerical range conditions of ECG features and the relationships between the numerical ranges of multiple ECG features can also be combined. For example, the ECG feature conditions of a heart rhythm recognition rule include a heart rate between (x1, x2) and a PP interval that is m1 times the RR interval. This is merely an example of the ECG feature conditions for several heart rhythm recognition rules and is not intended to limit the scope of this invention. The extended model library can, for example, store each heart rhythm recognition rule in the form of a table, with each row representing one heart rhythm recognition rule, corresponding to a column of ECG feature conditions and a column of recognition judgment results.

[0090] When the inherent heart rhythm recognition algorithm is highly sophisticated and accurate, the number of heart rhythm recognition rules in the expanded model library will be relatively small, with only a few rules requiring reinforcement. Conversely, when the inherent heart rhythm recognition algorithm is outdated or faces various unforeseen circumstances, the expanded model library may contain a larger number of heart rhythm recognition rules. Furthermore, the expanded model library can be updated as needed, ensuring that the medical device's memory always stores the latest expanded model library.

[0091] like Figure 2 As shown, in this embodiment, step S300: obtaining the second heart rhythm recognition result based on the ECG feature data to be identified and the pre-stored expanded model library, includes the following steps:

[0092] S310: Determine whether there is at least one heart rhythm recognition rule in the expanded model library;

[0093] If it does not exist, continue with S320: set the second heart rhythm recognition result to unmatched;

[0094] If it exists, continue to obtain the second heart rhythm recognition result according to the heart rhythm recognition rules and the ECG feature data to be identified.

[0095] like Figure 2 As shown, in this embodiment, step S300: obtaining the second heart rhythm recognition result based on the ECG feature data to be identified and the pre-stored expanded model library, when the expanded model library contains heart rhythm recognition rules, the process of obtaining the second heart rhythm recognition result according to the heart rhythm recognition rules and the ECG feature data to be identified includes the following steps:

[0096] S330: Determine whether the ECG feature data to be identified meets the ECG feature conditions in each of the heart rhythm recognition rules;

[0097] For example, the ECG feature conditions of a heart rhythm recognition rule include a heart rate between (x1, x2) and a QRS pulse width between (y1, y2). Then, it is determined whether the heart rate and QRS pulse width in the ECG feature data to be recognized conform to the corresponding numerical range. If yes, the ECG feature data to be recognized conforms to the heart rhythm recognition rule; if no, the ECG feature data to be recognized does not conform to the heart rhythm recognition rule.

[0098] If the ECG feature data to be identified meets the ECG feature conditions in the heart rhythm recognition rule, it means that the recognition result of the current ECG feature data using the inherent heart rhythm recognition algorithm has a certain probability of being wrong. Then continue to S340: take the recognition judgment result of the heart rhythm recognition rule that meets the rule as the second heart rhythm recognition result.

[0099] For example, if the ECG feature data to be identified conforms to the heart rhythm identification rule that the heart rate is between (x1, x2) and the QRS pulse width is between (y1, y2), then the identification result of the heart rhythm identification rule is taken as the second heart rhythm identification result.

[0100] If the ECG feature data to be identified does not meet the ECG feature conditions in all heart rhythm recognition rules, then continue to S320: set the second heart rhythm recognition result as unmatched.

[0101] like Figure 3 As shown, in this embodiment, step S400: obtaining the final heart rhythm recognition result based on the first heart rhythm recognition result and the second heart rhythm recognition result includes the following steps:

[0102] S410: Determine whether the second heart rhythm recognition result is a non-match;

[0103] If so, it means that the recognition result of the current ECG feature data using the inherent heart rhythm recognition algorithm should be accurate, then continue to S420: take the first heart rhythm recognition result as the final heart rhythm recognition result;

[0104] If not, it means that the recognition result using the inherent heart rhythm recognition algorithm for the current ECG feature data has a certain probability of being misjudged, and it is necessary to combine the second heart rhythm recognition result to modify the recognition result. Then continue to S430: combine the first heart rhythm recognition result and the second heart rhythm recognition result to generate the final heart rhythm recognition result.

[0105] In this embodiment, step S430, which combines the first heart rhythm recognition result and the second heart rhythm recognition result to generate a final heart rhythm recognition result, includes the following steps:

[0106] S431: Obtain the first weight and the second weight of the first heart rhythm recognition result and the second heart rhythm recognition result respectively;

[0107] S432: The first heart rhythm recognition result and the second heart rhythm recognition result are weighted and averaged to obtain the final heart rhythm recognition result, that is, the final heart rhythm recognition result is calculated according to the following formula:

[0108] a=(a1 k1+a2 k2) / (k1+k2)

[0109] Where a is the final heart rhythm recognition result, a1 is the first heart rhythm recognition result, k1 is the first weight, a2 is the second heart rhythm recognition result, and k2 is the second weight.

[0110] In this embodiment, the heart rhythm recognition result is a heart rhythm recognition score. For example, the preset range for the heart rhythm recognition score is 0-5. A heart rhythm recognition score of 0 indicates the lowest probability that the current heart rhythm is malignant, i.e., the current heart rhythm is determined to be undefibrillable. A heart rhythm recognition score of 5 indicates the highest probability that the current heart rhythm is malignant, i.e., the current heart rhythm is determined to be defibrillable. For other heart rhythm recognition scores between 0 and 5, the higher the score, the higher the probability that the current heart rhythm is malignant. A scoring threshold can be set. When the heart rhythm recognition score is greater than or equal to the scoring threshold, it is determined to be a defibrillable heart rhythm; when the heart rhythm recognition score is less than the scoring threshold, it is determined to be an undefibrillable heart rhythm. When the medical device is a cardiac defibrillator, the decision to perform defibrillation is based on the final heart rhythm recognition score.

[0111] The first heart rhythm recognition result is a score obtained using an inherent heart rhythm recognition algorithm, which can be any value from 0 to 5. The second heart rhythm recognition result is fixed at 0 or 5. That is, the recognition judgment result set in the heart rhythm recognition rules has only two possibilities: one is that it is determined to be an undefibrillable heart rhythm, which corresponds to a recognition judgment result of 0, and the other is that it is determined to be a defibrillable heart rhythm, which corresponds to a recognition judgment result of 5.

[0112] The values ​​of the first and second weights determine the degree to which the second heart rhythm recognition result influences the final heart rhythm recognition result. For example, when the first weight equals the second weight, the final heart rhythm recognition score is the sum of the first and second heart rhythm recognition results divided by 2. When the first weight is greater than the second weight, the final heart rhythm recognition score is more influenced by the first heart rhythm recognition result; when the first weight is less than the second weight, the final heart rhythm recognition score is more influenced by the second heart rhythm recognition result. When the second weight is 0, the final heart rhythm recognition result still depends entirely on the first heart rhythm recognition result of the inherent heart rhythm recognition algorithm. When the first weight is 0, the final heart rhythm recognition result depends entirely on the second heart rhythm recognition result of the later updated and corrected expanded model library.

[0113] In this embodiment, the magnitude of the second weight is preset in the expanded model library and specifically corresponds to each heart rhythm recognition rule. Step S431: Obtaining the first weight and the second weight of the first heart rhythm recognition result and the second heart rhythm recognition result respectively includes the following steps:

[0114] Obtain the weight of the conforming heart rhythm recognition rule, and use it as the second weight of the second heart rhythm recognition result;

[0115] The first weight of the first heart rhythm recognition result is determined based on the second weight, such that the sum of the first weight and the second weight is the preset weight sum. For example, the preset weight sum is 1, the first weight and the second weight are both values ​​between 0 and 1, and the sum of the first weight and the second weight is 1. When the second weight is set to 0, the first weight is 1; when the second weight is set to 50%, the first weight is 50%; when the second weight is set to 1, the first weight is 0.

[0116] As mentioned above, the magnitude of the second weight determines the influence of the second heart rhythm recognition result on the final heart rhythm recognition result. Therefore, the magnitude of the second weight is preferably related to the correlation strength between the heart rhythm recognition rule and the misjudgment result. The stronger the correlation between the heart rhythm recognition rule and the misjudgment result, the greater the possibility of misjudgment in the first heart rhythm recognition result obtained using the inherent heart rhythm recognition algorithm when the ECG feature data conforms to the heart rhythm recognition rule. Therefore, the second weight needs to be set larger. Conversely, the weaker the correlation between the heart rhythm recognition rule and the misjudgment result, the smaller the possibility of misjudgment in the first heart rhythm recognition result obtained using the inherent heart rhythm recognition algorithm when the ECG feature data conforms to the heart rhythm recognition rule. Therefore, the second weight can be set smaller. Specifically, the optional setting methods for the second weight will be further described below.

[0117] like Figure 4As shown, this embodiment of the invention also provides a medical device M100, including: a processor M110; and a memory M120, which stores the inherent heart rhythm recognition algorithm, the extended model library, and executable instructions of the processor. The processor is configured to execute the steps of the heart rhythm recognition method by executing the executable instructions. The processor M110 can be implemented using an MCU, and the memory M120 can be the MCU's own storage area, external storage devices, various ROM memories, various Flash memories, etc.

[0118] This invention pre-stores an expanded model library in the medical device M100, eliminating the need to update the inherent heart rhythm recognition algorithm stored in the M100's memory. Furthermore, the expanded model library can strengthen the inherent heart rhythm recognition algorithm, flexibly addressing the algorithm upgrade needs faced after the M100 is launched on the market. This improves the accuracy and safety of ECG monitoring, effectively avoiding the additional risks associated with modifying the inherent heart rhythm recognition algorithm architecture. By supplementing the expanded model library, various special situations can be flexibly addressed, and the upgrade method is simple and fast.

[0119] Furthermore, the medical device M100 may further include a communication module M130 for communicating with external devices. When the medical device M100 is first manufactured and released to the market, the memory M120 may only store the inherent heart rhythm recognition algorithm and the executable instructions of the processor. As the medical device M100 is widely used, some vulnerabilities in the previously inherent heart rhythm recognition algorithm may be discovered during use. This can lead to the creation of an expanded model library, which is then sent to the medical device M100. After receiving the expanded model library data through the communication module M130, the medical device M100 stores it in the memory M120 as a pre-stored expanded model library.

[0120] In this embodiment, the heart rhythm recognition method further includes the following steps:

[0121] The medical device receives expanded model library data from external devices and stores it locally as a pre-stored expanded model library.

[0122] Specifically, after receiving the expanded model library data, the medical device determines whether the expanded model library data is already stored in its local memory. If not, it stores the expanded model library data in its local memory as a pre-stored expanded model library. If it is, it updates the expanded model library in its local memory based on the expanded model library data. When updating the expanded model library in its local memory, it can perform an expansion update each time, such as adding recently generated heart rhythm recognition rules to the existing expanded model library. Alternatively, it can perform a complete update of the expanded model library each time, that is, setting a version number for each updated expanded model library and replacing the old version number of the expanded model library with the new version number.

[0123] Furthermore, the medical device can also send the real-time collected electrocardiogram (ECG) monitoring records out to external devices for data collection and further big data analysis. In this embodiment, the heart rhythm recognition method further includes the following steps:

[0124] The medical device sends electrocardiogram (ECG) monitoring records to an external device. These ECG monitoring records include ECG feature data and device identification results, and may include raw ECG data collected by the medical device. The ECG feature categories in the ECG feature data may include one or more of the following, but are not limited to: heart rate, overall and local ECG signal morphology, QRS pulse width, PP interval, RR interval, and PR interval. The device identification result is the first heart rhythm identification result obtained after processing the ECG feature data using the medical device's inherent heart rhythm identification algorithm.

[0125] Specifically, the external devices may include a programmable controller and an algorithm server. When the medical device is an in-vivo medical device, the programmable controller communicates with the medical device wirelessly. When the medical device is an external medical device, the programmable controller can communicate with the medical device via wired or wireless communication. Wireless communication methods include, but are not limited to, Wi-Fi, Bluetooth, radio frequency, and transceiver coils. The programmable controller periodically acquires ECG monitoring records from the medical device and further sends these records, including the raw ECG data, to an algorithm server in the cloud. Based on the ECG monitoring records and the doctor's further assessment, the algorithm server in the cloud generates the latest expanded model library data and can then distribute the latest expanded model library data to each of the medical devices via the programmable controller. The programmable controller and the algorithm server in the cloud can also transmit data wirelessly, including but not limited to, Wi-Fi, Bluetooth, and 4G.

[0126] like Figure 5 As shown, embodiments of the present invention also provide a medical system, comprising:

[0127] The programmable controller M200 is used to send expanded model library data to the medical device M100 and to acquire electrocardiogram (ECG) monitoring records from the medical device M100. The ECG monitoring records include ECG feature data and device identification results, and may further include raw ECG data collected by the medical device M100.

[0128] The medical device M100 is used to send electrocardiogram monitoring records to the programmer M200, and to receive expanded model library data from the programmer M200, storing it locally as a pre-stored expanded model library; and

[0129] The algorithm server M300 is used to obtain the electrocardiogram (ECG) monitoring records of the medical device M100 from the programmable controller M200, and send the ECG monitoring records to the doctor's end. The doctor's end can make a judgment based on the original ECG data. The algorithm server M300 obtains the corresponding doctor identification result from the doctor's end, generates multiple ECG event records, and each ECG event record includes ECG feature data, device identification result, and doctor identification result. After generating expanded model library data based on the ECG event records, the data is distributed to each of the medical devices M100 through the programmable controller M200.

[0130] In the medical system of this invention, by pre-storing an expanded model library in the medical device M100, there is no need to update the inherent heart rhythm recognition algorithm stored in the memory of the medical device M100. Furthermore, the expanded model library can strengthen the inherent heart rhythm recognition algorithm, flexibly addressing the algorithm upgrade needs faced after the medical device M100 is launched on the market. This improves the accuracy and security of ECG monitoring, effectively avoiding the additional risks associated with modifying the inherent heart rhythm recognition algorithm architecture. By supplementing the expanded model library, various special situations can be flexibly addressed, and the upgrade method is simple and fast. The expanded model library can be fully mined by the algorithm server M300 collecting ECG monitoring records from multiple medical devices M100, allowing for timely and accurate supplementation of the inherent heart rhythm recognition algorithm.

[0131] Therefore, in the medical system of this invention, the cloud-based algorithm server M300 can feed back the collected electrocardiogram (ECG) monitoring records to the doctor. Specifically, it can obtain specific ECG events (such as defibrillation actions performed, or the discovery of high-risk factors requiring defibrillation) based on the ECG monitoring records, and then send the ECG monitoring recorder for a period of time before and after the specific ECG event to the doctor. The doctor can provide a heart rhythm recognition result based on the received ECG monitoring records. This doctor's recognition result generally falls into one of two categories: a non-defibrillable rhythm or a defibrillable rhythm. Then, the cloud server can compare the device recognition result of the medical device's inherent heart rhythm recognition algorithm with the doctor's recognition result, focusing on ECG event records where the device recognition result and the doctor's recognition result differ. Data from these ECG event records is then mined, and based on big data analysis methods, potential algorithm vulnerabilities are located to expand the model library.

[0132] In this embodiment, the core idea of ​​mining these ECG event records is as follows: after filtering out those ECG event records where the device recognition results and doctor recognition results differ, analyze the ECG feature data in these ECG event records, and use statistical principles to try to find their common characteristics in order to identify vulnerabilities in the inherent heart rhythm recognition algorithm. When it is found that the algorithm's misjudgment rate is abnormally high for ECG event records that meet certain ECG feature conditions, a heart rhythm recognition rule is formed based on these ECG feature conditions, and the recognition judgment result in the heart rhythm recognition rule is the corresponding doctor's recognition result, that is, the doctor's recognition result is taken as the standard.

[0133] The following is combined with Figure 6 This embodiment details the method for forming an expanded model library through data mining. In this embodiment, the heart rhythm recognition method further includes the following steps:

[0134] S810: The programmable controller collects the electrocardiogram (ECG) monitoring records of the medical device and sends them to the algorithm server. The programmable controller can collect ECG monitoring records periodically, or it can actively send ECG monitoring records for a period of time before and after the ECG event to the programmable controller when a specific ECG event occurs in the medical device. Alternatively, the medical device can store the ECG monitoring records for a period of time before and after the ECG event locally when a specific ECG event occurs, and the programmable controller can read them periodically or irregularly.

[0135] S820: The algorithm server collects ECG monitoring records from multiple medical devices and obtains corresponding doctor identification results to obtain multiple ECG event records corresponding to multiple medical devices. Each ECG event record includes ECG feature data, device identification results, and doctor identification results.

[0136] S830: The algorithm server extracts ECG event records where the device identification result differs from the doctor's identification result, and marks them as misjudged ECG event records; while ECG event records where the device identification result is the same as the doctor's identification result can be marked as non-misjudged ECG event records.

[0137] S840: The algorithm server generates heart rhythm recognition rules based on the misjudged ECG event records as data to expand the model library. In each heart rhythm recognition rule, the recognition result corresponds to the doctor's recognition result. In step S840, if an algorithmic pattern requiring reinforcement can be summarized based on the misjudged ECG event records, at least one heart rhythm recognition rule is generated. If no algorithmic pattern requiring reinforcement can be summarized, or if there is no need to reinforce the current inherent heart rhythm recognition algorithm, then no heart rhythm recognition rule needs to be generated.

[0138] Specifically, in step S840, the algorithm server's analysis of the misjudged ECG event records involves searching for patterns in the ECG feature data within these records. This is done using statistical and data mining methods to identify the ECG feature conditions at which misjudgments are most likely to occur. The pattern finding can employ existing statistical analysis methods, such as factor analysis. When using factor analysis, each ECG feature is treated as a factor. For each factor, it is determined whether there exists a range within which the probability of misjudgment is particularly high. Specifically, the numerical range of each factor can be divided into multiple segments, and a histogram can be generated based on the number of misjudgments in each segment for analysis. For two or more factors, it is also determined whether two or more factors simultaneously meet specific conditions, leading to a high probability of misjudgment. Similarly, histograms or other statistical methods can be used to analyze whether there are specific relationships between factors that result in a particularly high probability of misjudgment. In other alternative implementations, the data mining and analysis can also be implemented using machine learning models and / or manual judgment. Therefore, the ECG event records can also be analyzed manually to obtain multiple heart rhythm recognition rules through comprehensive evaluation.

[0139] Furthermore, after identifying the corresponding ECG feature conditions, it is necessary to verify them using non-false-judgment ECG event records. This involves determining whether the patterns corresponding to these ECG feature conditions exist outside of the non-false-judgment ECG event records. If they do not exist, the corresponding ECG feature conditions and the doctor's identification result can be used as an ECG recognition rule. If they exist, the probability of the ECG feature condition appearing in false-judgment and non-false-judgment ECG event records can be used to determine whether it should be used as an ECG recognition rule. If, based on the ECG event records, the number of false-judgments for the ECG feature condition is significantly greater than the number of non-false-judgments, it can be used as an ECG recognition rule. If the number of false-judgments is significantly less than the number of non-false-judgments, it should not be used as an ECG recognition rule. If the number of false-judgments and non-false-judgments are roughly equal, the weight value of the rhythm recognition rule corresponding to the ECG feature condition can be set according to the specific circumstances.

[0140] In this embodiment, after step S840: generating heart rhythm recognition rules based on the misjudged ECG event record, the following steps are also included:

[0141] S850: The algorithm server analyzes the correlation strength between each heart rhythm recognition rule and the misjudgment result based on the misjudged ECG event record;

[0142] S860: The algorithm server sets the weight of each heart rhythm recognition rule based on the correlation strength. The correlation strength is positively correlated with the weight of the heart rhythm recognition rule. The higher the correlation strength, the greater the weight of the heart rhythm recognition rule. The lower the correlation strength, the smaller the weight of the heart rhythm recognition rule.

[0143] The strength of the association is primarily influenced by two factors. One factor is the probability of misjudgment when the ECG characteristic conditions of the heart rhythm recognition rule are met. This can be obtained by comparing the number of misjudgments when these conditions are met with the total number of ECG event records that meet the condition. The other factor is the severity, which can be determined manually (by medical equipment maintenance personnel and / or doctors), i.e., the potential harm caused by a misjudgment under these ECG characteristic conditions. The value of the association strength can be obtained by a weighted sum of the probability of misjudgment and the severity. A mapping table between the association strength value and the weights can be pre-defined. When the association strength value is determined, the corresponding weight value is retrieved. This weight value influences the degree to which the recognition judgment result in the heart rhythm recognition rule affects the final heart rhythm recognition result when the ECG characteristic conditions are met, and may directly affect the final decision on whether to perform defibrillation.

[0144] For example, as illustrated above, the preset heart rhythm recognition score range is 0-5. A score of 0 indicates the lowest probability that the current heart rhythm is malignant, meaning it is an undefibrillable rhythm. A score of 5 indicates the highest probability that the current heart rhythm is malignant, meaning it is a defibrillable rhythm. For other scores between 0 and 5, the higher the score, the greater the probability of a malignant rhythm. A preset heart rhythm recognition score of 4 or higher indicates a malignant rhythm that can be defibrillated. Suppose that in a given recognition, the first heart rhythm recognition result is 4.5, indicating a defibrillable rhythm, and the second heart rhythm recognition result is 0. If the weight of the second heart rhythm recognition result is 0, the final heart rhythm recognition result is 4.5, meaning it is unaffected by the second heart rhythm recognition result. If the weight of the second heart rhythm recognition result is 50%, and the weight of the first heart rhythm recognition result is also 50%, the final heart rhythm recognition result is (4.5 + 0). 50% = 2.25, which means it becomes an undefibrillable rhythm.

[0145] In this embodiment, the medical device can be a standalone cardiac monitoring device or a defibrillator that integrates cardiac monitoring and defibrillation functions. When the medical device is a defibrillator, it can determine whether to perform defibrillation based on the final heart rhythm recognition result obtained in step S400. Furthermore, the medical device can use steps S100-S200 and S100-S400 in the above-described heart rhythm recognition method for the two stages of determining whether malignant heart rhythms have occurred and whether malignant heart rhythms persist after pre-charging, respectively.

[0146] like Figure 7 As shown, this embodiment of the invention also provides a method for cardiac defibrillation, comprising the following steps:

[0147] S910: The medical device executes steps S100~S200 in the heart rhythm recognition method to obtain a first heart rhythm recognition result;

[0148] S920: The medical device determines whether the first heart rhythm recognition result is a defibrillable recognition result; for example, when the heart rhythm recognition result is a score, it determines whether the score meets the preset score range requirements corresponding to the defibrillable recognition result.

[0149] If the first heart rhythm recognition result is a defibrillation-ready result, then S930: the medical device precharges the defibrillation actuator;

[0150] If the first heart rhythm identification result is a non-defibrillable identification result, then S940: wait for the next defibrillable heart rhythm to be detected;

[0151] After pre-charging is completed, S950: The medical device executes steps S100~S400 in the heart rhythm recognition method to obtain the final heart rhythm recognition result;

[0152] S970: The medical device determines whether the final heart rhythm recognition result is a defibrillable recognition result; similarly, for example, when the heart rhythm recognition result is a score, it determines whether the score meets the preset score range requirements corresponding to the defibrillable recognition result;

[0153] If the final heart rhythm identification result is a defibrillable identification result, then S970: the medical device controls the defibrillation actuator to perform defibrillation;

[0154] If the final heart rhythm identification result is a non-defibrillable identification result, then S980: do not perform defibrillation, internally discharge energy to the defibrillation actuator, and S940: wait for the detection of the next defibrillable heart rhythm.

[0155] By employing the cardiac defibrillation method of the present invention and the above-mentioned heart rhythm recognition method, more accurate heart rhythm recognition results can be obtained, which can effectively avoid misjudgment of electrocardiogram feature data, avoid erroneous defibrillation due to misjudgment when defibrillation is not required, and avoid missing electrocardiogram events that require defibrillation due to misjudgment, thereby improving the accuracy and safety of medical equipment use.

[0156] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for recognizing heart rhythm, characterized in that, include: Medical equipment collects electrocardiogram (ECG) feature data to be identified; The medical device obtains a first heart rhythm recognition result based on the electrocardiogram feature data to be identified and the inherent heart rhythm recognition algorithm; The medical device obtains a second heart rhythm recognition result based on the ECG feature data to be identified and a pre-stored expanded model library; The medical device obtains the final heart rhythm recognition result based on the first heart rhythm recognition result and the second heart rhythm recognition result; The method further includes the following steps: The medical device sends electrocardiogram (ECG) monitoring records to an external device, and the ECG monitoring records include ECG characteristic data and device identification results. Multiple ECG event records corresponding to multiple medical devices are collected, and each ECG event record includes ECG feature data, device identification results, and doctor identification results; Records of ECG events where the device identification results differ from the doctor's identification results are extracted as records of misjudged ECG events. Based on the misjudged ECG event records, heart rhythm recognition rules are generated as data to expand the model library. In each heart rhythm recognition rule, the recognition judgment result corresponds to the doctor's recognition result.

2. The heart rhythm recognition method according to claim 1, characterized in that, Based on the ECG feature data to be identified and the pre-stored expanded model library, the second heart rhythm identification result is obtained, including the following steps: Determine whether at least one heart rhythm recognition rule exists in the expanded model library; If it does not exist, the second heart rhythm recognition result will be set to unmatched; If present, a second heart rhythm recognition result is obtained based on the heart rhythm recognition rules and the ECG feature data to be recognized. Each heart rhythm recognition rule includes ECG feature conditions and recognition judgment results.

3. The heart rhythm recognition method according to claim 2, characterized in that, Based on the heart rhythm recognition rules and the ECG feature data to be recognized, a second heart rhythm recognition result is obtained, including the following steps: Determine whether the ECG feature data to be identified meets the ECG feature conditions in each of the heart rhythm recognition rules; If the ECG feature data to be identified meets the ECG feature conditions in a heart rhythm identification rule, then the identification judgment result of the matching heart rhythm identification rule is taken as the second heart rhythm identification result. If the ECG feature data to be identified does not meet the ECG feature conditions in all heart rhythm recognition rules, then the second heart rhythm recognition result is set as unmatched.

4. The heart rhythm recognition method according to claim 3, characterized in that, Based on the first heart rhythm recognition result and the second heart rhythm recognition result, the final heart rhythm recognition result is obtained, including the following steps: Determine whether the second heart rhythm recognition result is a mismatch; If so, the first heart rhythm recognition result shall be taken as the final heart rhythm recognition result; If not, then the final heart rhythm recognition result is generated by combining the first heart rhythm recognition result and the second heart rhythm recognition result.

5. The heart rhythm recognition method according to claim 4, characterized in that, By combining the first heart rhythm recognition result and the second heart rhythm recognition result, a final heart rhythm recognition result is generated, including the following steps: The first weight and the second weight of the first heart rhythm recognition result and the second heart rhythm recognition result are obtained respectively; The first heart rhythm recognition result and the second heart rhythm recognition result are weighted and averaged to obtain the final heart rhythm recognition result.

6. The heart rhythm recognition method according to claim 5, characterized in that, The step of obtaining the first weight and the second weight of the first heart rhythm recognition result and the second heart rhythm recognition result respectively includes the following steps: Obtain the weight of the conforming heart rhythm recognition rule, and use it as the second weight of the second heart rhythm recognition result; The first weight of the first heart rhythm recognition result is determined based on the second weight, so that the sum of the first weight and the second weight is the preset weight sum.

7. The heart rhythm recognition method according to claim 2, characterized in that, It also includes the following steps: The medical device receives expanded model library data from external devices and stores it locally as a pre-stored expanded model library.

8. The heart rhythm recognition method according to claim 1, characterized in that, After generating heart rhythm recognition rules based on the misjudged ECG event records, the following steps are also included: Based on the analysis of the misjudged ECG event records, the correlation strength between each heart rhythm recognition rule and the misjudgment result is analyzed; The weight of each heart rhythm recognition rule is set based on the correlation strength.

9. A medical device, characterized in that, include: processor; The memory stores the inherent heart rhythm recognition algorithm, the expanded model library, and the executable instructions of the processor. The processor is configured to perform the steps of the heart rhythm recognition method of claim 1 by executing the executable instructions.

10. A medical system, characterized in that, include: The programmable controller is used to send expanded model library data to medical devices; as well as The medical device of claim 9 is used to receive expanded model library data and store it locally as a pre-stored expanded model library.

11. The medical system according to claim 10, characterized in that, It also includes an algorithm server, and the medical device is also used to send electrocardiogram (ECG) monitoring records to the programmable controller, wherein the ECG monitoring records include ECG feature data and device identification results; The programmable controller is also used to acquire the electrocardiogram monitoring record from the medical device and send it to the algorithm server; The algorithm server is used to collect multiple ECG event records corresponding to multiple medical devices. Each ECG event record includes ECG feature data, device identification results, and doctor identification results. ECG event records where the device identification results differ from the doctor identification results are extracted as misjudged ECG event records. And based on the misjudged ECG event records, heart rhythm recognition rules are generated as data to expand the model library, wherein in each heart rhythm recognition rule, the recognition judgment result corresponds to the doctor's recognition result.

Images