Intelligent monitoring method based on electrocardiogram

Abstract

The invention discloses an intelligent monitoring method based on an electrocardiogram, and relates to the technical field of medical monitoring, and the method comprises the detection steps: S1, a patient wears a patch type electrocardiogram monitoring module, obtains electrocardiogram data in different periods, and uploads the data to an analysis module; s2, an analysis module analyzes waveform forms of the electrocardiogram data in different periods, electrocardiogram waveforms in abnormal time periods are screened out, the abnormal electrocardiogram waveforms are integrated to construct an electrocardiogram library, heart disease type electrocardiogram waveforms similar to the waveform of a current patient are screened out based on a Euclidean distance calculation formula, and the electrocardiogram waveforms are stored in the electrocardiogram library; and heart diseases similar to those of the patient are preliminarily screened out. The heart disease type electrocardiogram waveform similar to the current patient waveform is screened based on the Euclidean distance calculation formula, intelligent disease preliminary screening to a certain degree is achieved in the mode, the diagnosis range is objectively and efficiently narrowed through quantitative distance calculation, and the diagnosis efficiency can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical monitoring, and in particular to an intelligent monitoring method based on electrocardiogram. Background Art

[0002] Electrocardiogram is a common and important means of detecting heart health. It is widely used in clinical diagnosis and daily health monitoring. By recording the electrical change curve generated by the heart in each cardiac cycle, it can reflect the depolarization and repolarization process of the atrium and ventricle, and provide a key basis for the diagnosis of various heart diseases. Most of the existing intelligent electrocardiogram monitoring methods focus on the simple extraction and recognition of electrocardiogram waveform features. The algorithm automatically detects the key waveforms of QRS complex, P wave, and T wave, and calculates the basic parameters of heart rate and interval, and then compares them with the preset normal range to determine whether there is an abnormality. However, this method is often only isolated. On-site analysis of a single waveform or parameter. Although the ECG can be recorded for a long time, the subsequent interpretation of a large amount of long-term data mainly relies on manual analysis by professional ECG doctors to judge the patient's condition. This is not only a huge workload, time-consuming and labor-intensive, but also due to the subjective differences in manual interpretation and the influence of visual fatigue factors, it may lead to uneven diagnostic accuracy. Especially when faced with massive and complex ECG data, it is difficult to ensure that all subtle abnormalities can be discovered comprehensively and timely, resulting in inaccurate judgments and delayed patient diagnosis. In this regard, we propose an intelligent monitoring method based on ECG. Summary of the invention

[0003] In order to solve the above technical problems, an intelligent monitoring method based on electrocardiogram is provided. This technical solution solves the above-mentioned problems of inaccurate judgment and delayed patient diagnosis.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is: an intelligent monitoring method based on electrocardiogram, the detection steps are: S1. The patient wears a patch-type ECG monitoring module to obtain ECG data at different times and upload the data to the analysis module; S2, the analysis module analyzes the waveform morphology of the ECG data of different periods, screens out the ECG waveforms of abnormal time periods, integrates the abnormal ECG waveforms to build an ECG library, and screens out the ECG waveforms of heart disease types similar to the current patient's waveform based on the Euclidean distance calculation formula, and preliminarily screens out heart diseases similar to the patient; S3, converting the ECG library into time series data, combining the ECG time series data into a vector form in sequence based on the time series analysis vector, converting the preliminarily screened similar heart disease ECG data into a heart disease vector, calculating the similarity between the disease vector and the ECG vector collected by the patient based on the cosine similarity, and specifically determining the type of heart disease from the similar heart diseases; S4. Based on the obtained type results, diagnosis suggestions are given based on big data, doctors provide auxiliary diagnosis suggestions, and give diagnosis plans.

[0005] Preferably, the ECG collection period in step S1 is 6 hours. The patient's ECG is collected once after every 1 hour to obtain the patient's ECG data. After uploading the data, preprocessing is performed. The preprocessing includes filtering and denoising, data standardization and feature extraction. The extracted features are the waveform features of the ECG.

[0006] Preferably, in step S2, the abnormal electrocardiogram waveform is screened by judging the overall waveform difference, and the overall difference is measured by calculating the mean difference MSE between the actual waveform Es(t) and the normal reference waveform Ez(t) within the unit time (t1, t2), and the calculation formula is: ; Among them, MES (t1, t2) is the calculated output value. By setting the normal waveform threshold, when the calculated output value is greater than the set threshold, it is judged that the ECG waveform in the time period (t1, t2) is abnormal, and the collected ECG data is screened.

[0007] Preferably, the Euclidean calculation formula in step S2 is: ; Where D j is the Euclidean distance between the current patient's ECG waveform and the ECG waveforms of j disease types, n is the total number of time points of the ECG waveform, that is, the length of the time series, p i is the amplitude of the current patient's electrocardiogram at the i-th time point, s ji is the amplitude of the ECG waveform of the jth disease type at the i-th time point, where i represents the index of the time point. The ECG waveform is a series of time series data. The formula calculates the square of the amplitude difference at each time point, sums the results of all time points, and finally takes the square root to get the Euclidean distance. The obtained value D j It represents the distance between the two. The smaller the distance, the more similar the waveforms are. It can be used to preliminarily determine that the patient has a disease with a similar ECG waveform.

[0008] Preferably, in step S3, the electrocardiogram is converted into time series data by setting the total duration of the electrocardiogram to A and the sampling frequency to b, then the total number of sampling points is: C=Ab, and E(t) represents the amplitude corresponding to the electrocardiogram signal collected at time t; Discretize the continuous time interval (0, A) according to the sampling frequency b to obtain discrete time points t n (n=0, 1, 2, ..., n-1), where t n=n / b, corresponding to the time corresponding to each sampling point, the corresponding time series data is expressed as X=(x0, x1, x2, ... x n-1 ), where x n is a discrete time point t n The ECG amplitude collected at .

[0009] Preferably, in step S3, the time range and corresponding sampling interval of the entire ECG time series data acquired are pre-defined, and the amplitude corresponding to the first time point becomes the first element of the vector, the amplitude corresponding to the second time point becomes the second element of the vector, and so on, until all the amplitude data corresponding to all time points in the entire time range are included to form a complete vector.

[0010] Preferably, in step S3, for each known heart disease, the disease-related ECG feature information is sequentially combined into a vector form, namely a disease vector. Each disease vector carries the key features of the corresponding disease in the ECG manifestation. The cosine similarity calculation method is used to measure the cosine value of the angle between two vectors to judge their similarity. The value range is between -1 and 1. When calculating the cosine similarity between the disease vector and the ECG vector collected by the patient, it is judged to be similar to the current disease vector.

[0011] Preferably, the cosine similarity calculation formula is: ; Where G represents the disease vector of the heart, H represents the ECG vector currently collected by the patient, is the angle between the two vectors. A specific similarity value is obtained by calculation. When this value is closer to 1, it means that the angle between the disease vector and the patient's ECG vector in the feature space is smaller, indicating that the similarity between them is higher, indicating that the ECG characteristics of the patient are more consistent with the typical ECG characteristics of the corresponding disease; conversely, the closer the value is to -1 or 0, the lower the similarity between the two, and the less consistent the disease is with the patient's actual situation, thereby specifically determining the type of heart disease the patient is currently suffering from.

[0012] Preferably, in step S4, through big data analysis, it is possible to understand the disease development trends, complications and effects of different treatments for patients of different age groups, genders and basic health conditions in the disease cases, and give suggestions to the current patients based on the acquired network information.

[0013] Preferably, in step S4, the doctor gives advice based on the patient's clinical symptoms, taking into account the patient's past medical history, family medical history, and current medications, and based on individual considerations and big data recommendations, tailors diagnostic recommendations for the patient that are tailored to their own situation.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention selects electrocardiogram waveforms of heart disease types similar to the current patient's waveform based on the Euclidean distance calculation formula. This method realizes a certain degree of intelligent preliminary disease screening. It no longer relies solely on manual experience to judge possible disease types, but through quantitative distance calculation, it objectively and efficiently narrows the diagnosis range, provides a reasonable reference direction for subsequent more accurate determination of disease types, reduces the blindness of screening, and helps to improve diagnostic efficiency.

[0015] The present invention uses cosine similarity to calculate the similarity between the disease vector and the ECG vector collected from the patient, and converts the preliminarily screened ECG data of similar heart diseases into heart disease vectors, so that the ECG features of different diseases can be quantitatively compared for similarity in the same vector space, making the distinction and judgment of diseases more scientific and accurate, and helping to more finely identify the type of disease that best suits the patient's actual situation from similar diseases.

[0016] The present invention gives diagnostic suggestions based on big data, making full use of the valuable experience and general rules contained in massive medical data. Big data can integrate the disease development trends, treatment effects, and complication information of many similar cases, thereby providing current patients with comprehensive and forward-looking diagnostic suggestions. Doctors can fully consider the patient's particularities based on the patient's specific clinical symptoms, past medical history, and individual physical signs, making the diagnostic suggestions more in line with the patient's actual situation, avoiding omissions that may occur due to mechanically applying general suggestions, and meeting the individual medical needs of patients to the greatest extent. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flow chart of the detection steps of the present invention. DETAILED DESCRIPTION

[0018] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0019] Reference Figure 1 As shown, an intelligent monitoring method based on electrocardiogram, the detection steps are: S1. The patient wears a patch-type ECG monitoring module to obtain ECG data at different times and upload the data to the analysis module; S2, the analysis module analyzes the waveform morphology of the ECG data of different periods, screens out the ECG waveforms of abnormal time periods, integrates the abnormal ECG waveforms to build an ECG library, and screens out the ECG waveforms of heart disease types similar to the current patient's waveform based on the Euclidean distance calculation formula, and preliminarily screens out heart diseases similar to the patient; S3, converting the ECG library into time series data, combining the ECG time series data into a vector form in sequence based on the time series analysis vector, converting the preliminarily screened similar heart disease ECG data into a heart disease vector, calculating the similarity between the disease vector and the ECG vector collected by the patient based on the cosine similarity, and specifically determining the type of heart disease from the similar heart diseases; S4. Based on the obtained type results, diagnosis suggestions are given based on big data, doctors provide auxiliary diagnosis suggestions, and give diagnosis plans.

[0020] By wearing a patch-type ECG monitoring module, the patient can achieve long-term and continuous monitoring of cardiac electrical activity during daily activities. Unlike traditional static ECG examinations that can only obtain data for a short period of time, this method can capture ECG data in different life scenarios and different physical conditions, greatly increasing the chance of discovering abnormal ECG conditions. Whether it is a short-term arrhythmia or changes in cardiac electrical activity induced by specific activities, they are more likely to be recorded, providing a rich data basis for subsequent comprehensive and accurate diagnosis; The Euclidean distance calculation formula is used to screen ECG waveforms of heart disease types similar to the current patient's waveform, realizing preliminary intelligent disease screening. It quickly narrows the range of possible diseases based on objective mathematical calculations. Compared with the traditional method of relying on doctors to manually check various diseases one by one, it greatly improves the efficiency of the early stage of diagnosis, allowing subsequent analysis to focus more on the types of diseases with strong correlation, reducing unnecessary workload and time costs; Based on the timing analysis vector, the ECG time series data are combined into a vector form, and the preliminarily screened similar heart disease ECG data are also converted into heart disease vectors. This process quantifies and integrates the ECG features. In the vector space, precise mathematical calculations can be used to measure the similarity between different ECG data, and the differences between the patient's ECG vector and different disease vectors can be compared more scientifically and objectively, thereby improving the ability to accurately determine the specific type of disease from similar heart diseases and avoiding the errors that may be caused by relying solely on subjective visual judgment of waveform similarity.

[0021] The ECG collection period in step S1 is 6 hours. After every 1 hour, the patient's ECG is collected once to obtain the patient's ECG data. After uploading the data, preprocessing is performed. The preprocessing includes filtering and denoising, data standardization and feature extraction. The extracted features are the waveform characteristics of the ECG.

[0022] The preprocessing step in this application is a prior art and will not be elaborated on here. Sampling at fixed time intervals helps to discover the potential patterns of changes in cardiac electrical activity over time, for example, whether there is a period of time in which the electrocardiogram waveform always shows similar subtle changes, or whether a certain abnormal waveform is more likely to appear after a specific time period. This regular data acquisition method makes it convenient for medical personnel to analyze the rhythmicity and trend characteristics of cardiac activity in a short period of time, which is of great significance for diagnosing some heart diseases or latent diseases related to the biological clock and daily activity rhythm, and can better understand the dynamic change pattern of cardiac function.

[0023] In step S2, the abnormal ECG waveform is screened by judging the overall waveform difference, and the overall difference is measured by calculating the mean error MSE between the actual waveform Es(t) and the normal reference waveform Ez(t) within the unit time (t1, t2). The calculation formula is: ; Among them, MES (t1, t2) is the calculated output value. By setting the normal waveform threshold, when the calculated output value is greater than the set threshold, it is judged that the ECG waveform in the time period (t1, t2) is abnormal, and the collected ECG data is screened.

[0024] This application measures the degree of difference by calculating the mean square error (MSE) between the actual waveform Es(t) and the normal reference waveform Ez(t) in unit time, and transforms the process of judging whether the electrocardiogram waveform is abnormal, which originally relied on manual subjective visual judgment, into a quantitative evaluation method based on mathematical formulas. This quantitative method avoids the differences caused by the inconsistency of personal experience, observation angles and subjective judgment standards of different doctors, making the judgment on whether the electrocardiogram waveform is abnormal more objective and unified. No matter which medical personnel performs the analysis, as long as the same normal waveform threshold is set, a relatively consistent judgment result can be obtained, thereby improving the scientificity and accuracy of the diagnostic process and reducing misdiagnosis or missed diagnosis caused by subjective factors. Each wave shape, amplitude, and duration of the normal waveform Ez(t) has a standard range, which refers to the medically recognized electrocardiogram standard, which is derived from the standard healthy population electrocardiogram parameters specified in the "Guidelines for Standardized Electrocardiogram Measurement" (2023 Edition) issued by the American Heart Association (AHA). The QRS complex duration is 80-120ms, and the amplitude is 0.5-2.0mV. When calculating the mean square error (MSE), the median amplitude of the standard waveform at each time point is taken as the reference value Ez(t). Specifically, by collecting 10,000 healthy adult electrocardiogram data, a standard waveform is generated after the median value is calculated and stored in the analysis module. When the actual waveform Es(t) is compared with Ez(t) point by point, the MSE calculation formula can be clearly implemented; The specific steps of MSE calculation are: Determine the time interval and clarify the unit time interval to be analyzed, that is, (t1, t2); Obtain waveform data, obtain the specific values ​​of the actual waveform Es(t) and the normal reference waveform Ez(t) within the time interval, and obtain the specific values ​​after median calculation; Calculate the square of the difference, and for each time point t in the time interval, calculate ; Calculate the average value, sum the squares of the differences and then take the average value to get the mean square error MSE (t1, t2); To judge the abnormality, the calculated MSE (t1, t2) is compared with the set normal waveform threshold. If the MSE (t1, t2) is greater than the threshold, it is judged that the ECG waveform in this time period is abnormal. Assuming the amplitude range is 10-20V and the duration range is 0.1-0.2 seconds, after median calculation, the amplitude of 15V and the duration of 0.15 seconds are selected as the reference signal. The test signal is normalized so that its amplitude range is 0-1, and the MSE between the test signal and the reference signal is calculated.

[0025] The Euclidean calculation formula in step S2 is: ; Where D j is the Euclidean distance between the current patient's ECG waveform and the ECG waveforms of j disease types, n is the total number of time points of the ECG waveform, that is, the length of the time series, p i is the amplitude of the current patient's electrocardiogram at the i-th time point, s ji is the amplitude of the ECG waveform of the jth disease type at the i-th time point, where i represents the index of the time point. The ECG waveform is a series of time series data. The formula calculates the square of the amplitude difference at each time point, sums the results of all time points, and finally takes the square root to get the Euclidean distance. The obtained value Dj It represents the distance between the two. The smaller the distance, the more similar the waveforms are. This can be used to preliminarily determine the disease that the patient currently has with a similar ECG waveform. The abnormal ECGs were taken from the standardized waveforms of typical cases in the European Society of Cardiology (ESC) ECG Diagnostic Standard Database. For premature ventricular contractions (PVC), the average amplitude sequence of the QRS wave group distortion segment of 100 confirmed patients was taken. For atrial fibrillation (AF), the characteristic amplitude of the irregular RR interval segment was taken. After discretization, the exact value was obtained for each time point to ensure the operability of the Euclidean distance calculation. The specific calculation steps are: For the range value of the abnormal electrocardiogram, its specific characteristic parameters are determined, and the range value is converted into an exact fixed value. The average value of the range value is calculated as the fixed value by using a statistical method. After the fixed value of the abnormal electrocardiogram is determined, it is substituted into the Euclid calculation formula together with other related electrocardiogram characteristic parameters. The data points are the values ​​of different electrocardiogram characteristic parameters. The Python code is expressed as: import numpy as np # Current patient's ECG amplitude sequence p = np.array([1.2, 1.5, 1.1, 1.3, 1.4]) # Ventricular premature beat disease type ECG amplitude sequence s1 = np.array([1.0, 1.3, 1.2, 1.4, 1.3]) # Atrial fibrillation disease type ECG amplitude sequence s2 = np.array([1.1, 1.4, 1.0, 1.2, 1.5]) # Calculate Euclidean distance D1 = np.sqrt(np.sum((p - s1) ** 2)) D2 = np.sqrt(np.sum((p - s2) ** 2)) print(f"The Euclidean distance between the current patient's ECG and the type of ventricular premature beat disease ECG D1:{D1:.2f}") print(f"The Euclidean distance between the current patient's ECG and the atrial fibrillation disease type ECG D2:{D2:.2f}") if D1 <D2: print("Preliminary judgment is that the patient's electrocardiogram is more similar to the characteristics of ventricular premature beat disease.") else: print("Preliminary judgment shows that the patient's electrocardiogram is closer to the characteristics of atrial fibrillation.")

[0026] The Euclidean distance calculation formula of the present application provides a clear and precise quantitative means for measuring the similarity between the current patient's ECG waveform and the ECG waveforms of different disease types. By converting the difference in waveform amplitude at each time point into a specific numerical value, the traditional fuzzy judgment method based on experience and subjective visual comparison is abandoned, making the similarity assessment objective and measurable. Medical personnel can intuitively judge which type of disease has an ECG waveform closer to the patient based on the specific distance value, providing accurate and convincing data support for subsequent diagnostic analysis, helping to reduce the risk of misdiagnosis due to subjective judgment differences and improve the scientificity and accuracy of diagnosis.

[0027] In step S3, the electrocardiogram is converted into time series data by setting the total duration of the electrocardiogram to A and the sampling frequency to b. Then the total number of sampling points is: C=Ab, and E(t) is used to represent the amplitude corresponding to the electrocardiogram signal collected at time t; Discretize the continuous time interval (0, A) according to the sampling frequency b to obtain discrete time points t n (n=0, 1, 2, ..., n-1), where t n =n / b, corresponding to the time corresponding to each sampling point, the corresponding time series data is expressed as X=(x0, x1, x2, ... x n-1 ), where x n is a discrete time point t n The ECG amplitude collected at .

[0028] The present application organizes electrocardiogram data in the form of a time series, which helps to improve the efficiency and accuracy of subsequent calculations and processing. When a computer processes such structured data, it can operate on the data at each time point in sequence, thus avoiding the analysis difficulties and errors that may occur when processing complex waveforms.

[0029] In step S3, the time range and corresponding sampling interval of the entire ECG time series data are pre-defined for the acquired ECG time series data. In order, the amplitude corresponding to the first time point becomes the first element of the vector, and the amplitude corresponding to the second time point becomes the second element of the vector, and so on, until all the amplitude data corresponding to all time points in the entire time range are included to form a complete vector.

[0030] The present application integrates the ECG amplitude data scattered at different time points into a vector, achieving a high degree of data integration, and simplifying a large amount of discrete data that was originally complex and arranged in chronological order into an orderly overall structure. The similarities and differences of the ECG amplitudes at different time points can be seen at a glance, and the similarity of the overall waveform can be judged. Moreover, when the similarity measurement and disease type judgment are subsequently performed based on the cosine similarity method, the data in vector form can seamlessly connect to the corresponding calculation logic, more accurately determine the degree of proximity between the patient's ECG vector and various disease vectors, and improve the accuracy and efficiency of disease diagnosis.

[0031] In step S3, for various known heart diseases, the disease-related ECG feature information is combined in sequence into a vector form, namely the disease vector. Each disease vector carries the key features of the corresponding disease in the ECG manifestation. The cosine similarity calculation method is used to measure the cosine value of the angle between two vectors to judge their similarity. The value range is between -1 and 1. When calculating the cosine similarity between the disease vector and the ECG vector collected from the patient, it is judged to be similar to the current disease vector.

[0032] The numerical value obtained by the cosine similarity calculation in this application is between -1 and 1, which can quantitatively measure the similarity between the disease vector and the patient's ECG vector. Compared with the traditional method of relying on doctors' naked eye observation and experience to judge the similarity of ECG features, this quantitative result is more objective and accurate, which greatly reduces the diagnostic bias caused by subjective factors, and enables medical personnel to measure the degree of fit between the patient's ECG performance and various known heart diseases based on the exact numerical value, providing a reliable basis for subsequent accurate diagnosis.

[0033] The formula for calculating cosine similarity is: ; Where G represents the disease vector of the heart, H represents the ECG vector currently collected by the patient, is the angle between the two vectors. A specific similarity value is obtained by calculation. When this value is closer to 1, it means that the angle between the disease vector and the patient's ECG vector in the feature space is smaller, indicating that the similarity between them is higher, indicating that the ECG characteristics of the patient are more consistent with the typical ECG characteristics of the corresponding disease; conversely, the closer the value is to -1 or 0, the lower the similarity between the two, and the less consistent the disease is with the patient's actual situation, thereby specifically determining the type of heart disease the patient is currently suffering from.

[0034] The cosine similarity calculation formula in this application is based on rigorous mathematical principles. The similarity between the disease vector and the patient's ECG vector is represented by a specific numerical value, and its value range is clearly limited to between -1 and 1. This quantitative method completely eliminates the ambiguity of relying on subjective visual observation and empirical judgment. No matter which medical personnel performs the analysis, as long as they use the same data and the calculation formula, they can obtain a consistent similarity value, which provides an objective and unified measurement standard for judging the degree of fit between the patient's ECG and the ECG of various diseases, greatly improving the scientificity and accuracy of the diagnosis process, and effectively avoiding misdiagnosis due to differences in human judgment.

[0035] In step S4, through big data analysis, we can understand the development trend of the disease, complications and effects of different treatment methods for patients of different age groups, genders and basic health conditions in the disease cases. Based on the obtained online information, we give suggestions to the current patients. In step S4, the doctor gives suggestions based on the patient's clinical symptoms, taking into account the patient's previous medical history, family medical history and current medication factors. Based on individual considerations and on the basis of big data suggestions, we tailor diagnostic suggestions for patients that suit their own conditions.

[0036] Through big data analysis, this application can comprehensively consider the impact of multiple dimensions of factors such as different age groups, genders, and basic health conditions on the disease. Due to differences in physical functions and physiological characteristics, the development trajectories of the disease in different groups of patients are often different. Some heart diseases may progress relatively slowly in the elderly but have more complications, while they may progress rapidly in young people but respond more positively to treatment. Big data covers a large amount of case information of different types of patients, and can fully present these diverse manifestations, allowing medical personnel to stand from a macro perspective and clearly grasp the universal development laws of the disease in different populations, providing a comprehensive reference basis for the current patient's condition estimation.

[0037] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments, and the above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention, and these changes and improvements fall within the scope of the present invention claimed.

Claims

1. An intelligent monitoring method based on electrocardiogram, characterized in that: The detection steps are: S1. The patient wears a patch-type ECG monitoring module to obtain ECG data at different times and upload the data to the analysis module; S2, the analysis module analyzes the waveform morphology of the ECG data of different periods, screens out the ECG waveforms of abnormal time periods, integrates the abnormal ECG waveforms to build an ECG library, and screens out the ECG waveforms of heart disease types similar to the current patient's waveform based on the Euclidean distance calculation formula, and preliminarily screens out heart diseases similar to the patient; S3, converting the ECG library into time series data, combining the ECG time series data into a vector form in sequence based on the time series analysis vector, converting the preliminarily screened similar heart disease ECG data into a heart disease vector, calculating the similarity between the disease vector and the ECG vector collected by the patient based on the cosine similarity, and specifically determining the type of heart disease from the similar heart diseases; S4. Based on the obtained type results, diagnosis suggestions are given based on big data, doctors provide auxiliary diagnosis suggestions, and give diagnosis plans.

2. The electrocardiogram-based intelligent monitoring method according to claim 1, characterized in that: The ECG collection period in step S1 is 6 hours. After every 1 hour, the patient's ECG is collected once to obtain the patient's ECG data. After uploading the data, preprocessing is performed. The preprocessing includes filtering and denoising, data standardization and feature extraction. The extracted features are the waveform characteristics of the ECG.

3. The electrocardiogram-based intelligent monitoring method according to claim 1, characterized in that: In step S2, the abnormal ECG waveform is screened by judging the overall waveform difference, and the overall difference is measured by calculating the mean error MSE between the actual waveform Es(t) and the normal reference waveform Ez(t) within the unit time (t1, t2). The calculation formula is: ； Among them, MES (t1, t2) is the calculated output value. By setting the normal waveform threshold, when the calculated output value is greater than the set threshold, it is judged that the ECG waveform in the time period (t1, t2) is abnormal, and the collected ECG data is screened.

4. The electrocardiogram-based intelligent monitoring method according to claim 1, characterized in that: The Euclidean calculation formula in step S2 is: ； Where Dj is the Euclidean distance between the current patient's ECG waveform and the ECG waveform of the jth disease type, n is the total number of time points of the ECG waveform, that is, the length of the time series, pi is the amplitude of the current patient's ECG at the i-th time point, sji is the amplitude of the ECG waveform of the j-th disease type at the i-th time point, i represents the index of the time point, and the ECG waveform is a series of time series data. The formula calculates the square of the amplitude difference at each time point, sums the results of all time points, and finally takes the square root to get the Euclidean distance. The obtained value Dj represents the distance between the two. The smaller the distance, the more similar the waveforms are, and the disease similar to the patient's current ECG waveform can be preliminarily determined.

5. The electrocardiogram-based intelligent monitoring method according to claim 1, characterized in that: In step S3, the electrocardiogram is converted into time series data by setting the total duration of the electrocardiogram to A and the sampling frequency to b. Then the total number of sampling points is: C=Ab, and E(t) is used to represent the amplitude corresponding to the electrocardiogram signal collected at time t; Discretize the continuous time interval (0, A) according to the sampling frequency b to obtain discrete time points t n (n=0, 1, 2, ..., n-1), where t n =n / b, corresponding to the time corresponding to each sampling point, the corresponding time series data is expressed as X=(x0, x1, x2, ... x n-1 ), where x n is a discrete time point t n The ECG amplitude collected at .

6. The electrocardiogram-based intelligent monitoring method according to claim 1, characterized in that: In step S3, the time range and corresponding sampling interval of the entire ECG time series data are pre-defined for the acquired ECG time series data. In order, the amplitude corresponding to the first time point becomes the first element of the vector, and the amplitude corresponding to the second time point becomes the second element of the vector, and so on, until all the amplitude data corresponding to all time points in the entire time range are included to form a complete vector.

7. The electrocardiogram-based intelligent monitoring method according to claim 1, characterized in that: In step S3, for various known heart diseases, the disease-related ECG feature information is combined in sequence into a vector form, namely the disease vector. Each disease vector carries the key features of the corresponding disease in the ECG manifestation. The cosine similarity calculation method is used to measure the cosine value of the angle between two vectors to judge their similarity. The value range is between -1 and 1. When calculating the cosine similarity between the disease vector and the ECG vector collected from the patient, it is judged to be similar to the current disease vector.

8. The electrocardiogram-based intelligent monitoring method according to claim 7, characterized in that: The formula for calculating cosine similarity is: ； Where G represents the disease vector of the heart, H represents the ECG vector currently collected by the patient, is the angle between the two vectors. A specific similarity value is obtained by calculation. When this value is closer to 1, it means that the angle between the disease vector and the patient's ECG vector in the feature space is smaller, indicating that the similarity between them is higher, indicating that the ECG characteristics of the patient are more consistent with the typical ECG characteristics of the corresponding disease; conversely, the closer the value is to -1 or 0, the lower the similarity between the two, and the less consistent the disease is with the patient's actual situation, thereby specifically determining the type of heart disease the patient is currently suffering from.

9. The electrocardiogram-based intelligent monitoring method according to claim 1, characterized in that: In step S4, through big data analysis, we can understand the development trends of the disease, complications and effects of different treatments for patients of different ages, genders and basic health conditions in the disease cases, and give suggestions for the current patients based on the obtained online information.

10. The electrocardiogram-based intelligent monitoring method according to claim 1, characterized in that: In step S4, the doctor gives advice based on the patient's clinical symptoms, taking into account the patient's medical history, family history, and current medications. Based on individual considerations and big data recommendations, the doctor provides the patient with customized diagnostic advice that suits their situation.

Images