Electrocardiogram abnormal data analysis method and system based on artificial intelligence

Abstract

The invention provides an electrocardiogram abnormal data analysis method and system based on artificial intelligence, and relates to the field of medical data intelligent analysis. Analyzing the frequency components by adopting a data processing and image analysis algorithm based on artificial intelligence so as to extract waveform semantic features of different frequency components of the electrocardiogram, and performing information aggregation based on electrocardiogram kernel semantics on the waveform semantic features of the frequency components; according to the method, the electrocardiogram data is subjected to signal reduction, optimized electrocardiogram data is output, automatic noise reduction and optimization processing of the electrocardiogram data can be achieved, electrocardiogram anomaly analysis and detection are further conducted on the patient object based on the optimized electrocardiogram data, and the electrocardiogram anomaly detection effect is further improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent analysis of medical data, and more specifically, to an abnormal electrocardiogram data analysis method and system based on artificial intelligence. Background Art

[0002] Electrocardiogram (ECG) is an important medical test method used to record the electrical activity of the heart and is one of the important tools for diagnosing cardiovascular diseases. However, in the actual ECG data collection process, due to the influence of various internal and external factors, such as muscle movement, poor contact between electrodes and skin, environmental electromagnetic interference, etc., ECG signals are often mixed with noise. The presence of these noises will not only affect the doctor's visual interpretation of the ECG, but more importantly, they will interfere with the accuracy of the analysis algorithm and reduce the effect of ECG abnormality detection. Therefore, how to effectively recover high-quality ECG signals from noise has become a key issue in improving the accuracy of ECG analysis.

[0003] However, traditional ECG noise reduction methods usually use some basic digital filtering methods (such as low-pass filtering, high-pass filtering, etc.). Although these methods can reduce noise to a certain extent, they have limited noise reduction effects for complex and variable noise types, especially nonlinear noise, and are prone to losing useful ECG feature information, thus affecting the accuracy of subsequent ECG abnormality data analysis and detection.

[0004] Therefore, an optimized ECG abnormality data analysis scheme is needed to improve the accuracy of ECG abnormality data detection and analysis. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide an abnormal ECG data analysis method and system based on artificial intelligence.

[0006] According to one aspect of the present application, there is provided an abnormal ECG data analysis system based on artificial intelligence, which comprises: An electrocardiogram data collection module, used to obtain electrocardiogram data of a diagnosed patient object; An electrocardiogram data wavelet transform processing module, used for performing wavelet transform on the electrocardiogram data of the diagnosed patient object to obtain a plurality of frequency components; A frequency component feature extraction module, used for extracting features from the multiple frequency components to obtain multiple frequency component waveform semantic coding features; A frequency component feature dynamic aggregation analysis module is used to perform kernel feature dynamic aggregation analysis on the multiple frequency component waveform semantic coding features to obtain frequency multi-component kernel features, wherein the frequency component feature dynamic aggregation analysis module includes: a frequency component feature static and dynamic potential factor calculation unit, used to calculate the static potential factor and dynamic potential factor of each frequency component waveform semantic coding feature in the multiple frequency component waveform semantic coding features to obtain a set of frequency component waveform semantic static potential factors and a set of frequency component waveform semantic dynamic potential factors; a frequency component feature kernel semantic aggregation unit, used to perform kernel semantic aggregation on the multiple frequency component waveform semantic coding features based on the set of frequency component waveform semantic static potential factors and the set of frequency component waveform semantic dynamic potential factors to obtain the frequency multi-component kernel features; An ECG image semantic coding feature extraction module is used to restore the ECG signal based on the frequency multi-component kernel feature, and then extract ECG image features from the obtained optimized ECG data to obtain ECG image semantic coding features; The ECG abnormality analysis result generating module is used to determine whether the diagnosed patient object has ECG abnormality based on the semantic coding features of the ECG image.

[0007] In the above-mentioned artificial intelligence-based ECG abnormality data analysis system, the frequency component feature extraction module is used to: respectively pass each frequency component of the multiple frequency components through a component frequency domain waveform semantic feature extractor based on a 1D-CNN model to obtain multiple frequency component waveform semantic coding feature vectors as the multiple frequency component waveform semantic coding features.

[0008] In the above-mentioned artificial intelligence-based ECG data analysis system, the frequency component feature static and dynamic energy potential factor calculation unit includes: a frequency component waveform semantic initial aggregation representation vector generation subunit, which is used to calculate the positional mean vector of the multiple frequency component waveform semantic coding feature vectors to obtain the frequency component waveform semantic initial aggregation representation vector; a frequency component waveform semantic static energy potential factor calculation subunit, which is used to calculate the static energy potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors to obtain the set of frequency component waveform semantic static energy potential factors; a frequency component waveform semantic dynamic energy potential factor calculation subunit, which is used to calculate the dynamic energy potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors relative to the frequency component waveform semantic initial aggregation representation vector to obtain the set of frequency component waveform semantic dynamic energy potential factors.

[0009] In the above-mentioned artificial intelligence-based ECG data analysis system, the frequency component waveform semantic static potential factor calculation subunit is used to: calculate the mean and standard deviation of the frequency component waveform semantic coding feature vector to obtain the frequency component waveform semantic mean and the frequency component waveform semantic standard deviation; calculate the positional difference between the frequency component waveform semantic coding feature vector and the frequency component waveform semantic mean, and then divide it by the frequency component waveform semantic standard deviation by position point to obtain the frequency component waveform semantic distribution representation vector; calculate the fourth power of each position feature value in the frequency component waveform semantic distribution representation vector, and then calculate the mean of the obtained fourth power modulation vector to obtain the frequency component waveform semantic static potential factor.

[0010] In the above-mentioned artificial intelligence-based ECG data analysis system, the frequency component waveform semantic dynamic energy potential factor calculation subunit is used to: calculate the multiplication between the frequency component waveform semantic initial aggregation representation vector and the frequency component waveform semantic initial clustering weight matrix to obtain the frequency component waveform semantic initial aggregation representation modulation vector; calculate the multiplication between the frequency component waveform semantic encoding feature vector and the frequency component waveform semantic weight matrix to obtain the frequency component waveform semantic modulation vector; perform position point multiplication on the frequency component waveform semantic initial aggregation representation modulation vector and the frequency component waveform semantic modulation vector, and then perform position point multiplication on the frequency component waveform semantic bias vector to obtain the frequency component waveform semantic dynamic modulation representation vector; calculate the square of the Frobenius norm of the frequency component waveform semantic dynamic modulation representation vector to obtain the frequency component waveform semantic dynamic energy potential factor.

[0011] In the above-mentioned artificial intelligence-based ECG data analysis system, the frequency component feature kernel semantic aggregation unit includes: a frequency component waveform semantic superposition state potential factor calculation subunit, which is used to determine the superposition state potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors based on the frequency component waveform semantic static potential factor and the frequency component waveform semantic dynamic potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors to obtain a set of frequency component waveform semantic superposition state potential factors; a frequency component waveform semantic feature superposition potential weight coefficient calculation subunit, which is used to input the set of frequency component waveform semantic superposition state potential factors into the potential factor gated weighting network to obtain a set of frequency component waveform semantic feature superposition potential weight coefficients; a frequency multi-component kernel feature generation subunit, which is used to calculate the position-weighted sum of the multiple frequency component waveform semantic coding feature vectors based on the set of frequency component waveform semantic feature superposition potential weight coefficients to obtain a frequency multi-component kernel feature vector as the frequency multi-component kernel feature.

[0012] In the above-mentioned artificial intelligence-based ECG data analysis system, the frequency component waveform semantic superposition state energy potential factor calculation subunit is used to: calculate the natural exponential function value with the natural constant e as the base and the frequency component waveform semantic static energy potential factor of the frequency component waveform semantic encoding feature vector as the exponential power to obtain the frequency component waveform semantic static energy potential class support factor; calculate the natural exponential function value with the natural constant e as the base and the square of the frequency component waveform semantic dynamic energy potential factor of the frequency component waveform semantic encoding feature vector as the exponential power to obtain the frequency component waveform semantic dynamic energy potential class support factor; calculate the sum of the frequency component waveform semantic static energy potential class support factor and the frequency component waveform semantic dynamic energy potential class support factor to obtain the frequency component waveform semantic superposition state energy potential factor.

[0013] In the above-mentioned artificial intelligence-based ECG abnormality data analysis system, the ECG image semantic coding feature extraction module includes: an optimized ECG data decoding generation unit, which is used to input the frequency multi-component kernel feature vector into a decoder-based signal restorer to obtain the optimized ECG data; an optimized ECG data feature extraction unit, which is used to input the optimized ECG data into a trained convolutional neural network model-based ECG image feature extractor to obtain an ECG image semantic coding feature vector as the ECG image semantic coding feature.

[0014] In the above-mentioned artificial intelligence-based ECG abnormality data analysis system, the ECG abnormality analysis result generation module is used to: input the semantic encoding feature vector of the ECG image into a classifier-based intelligent analysis module to obtain an analysis result, and the analysis result is used to indicate whether the diagnosed patient object has ECG abnormality.

[0015] According to another aspect of the present application, there is provided a method for analyzing abnormal ECG data based on artificial intelligence, which comprises: obtaining electrocardiogram data of a diagnosed patient subject; Performing wavelet transformation on the electrocardiogram data of the diagnosed patient object to obtain a plurality of frequency components; Performing feature extraction on the multiple frequency components to obtain multiple frequency component waveform semantic coding features; Performing kernel feature dynamic aggregation analysis on the multiple frequency component waveform semantic coding features to obtain frequency multi-component kernel features, including: calculating the static potential factor and the dynamic potential factor of each frequency component waveform semantic coding feature in the multiple frequency component waveform semantic coding features to obtain a set of frequency component waveform semantic static potential factors and a set of frequency component waveform semantic dynamic potential factors; based on the set of frequency component waveform semantic static potential factors and the set of frequency component waveform semantic dynamic potential factors, performing kernel semantic aggregation on the multiple frequency component waveform semantic coding features to obtain the frequency multi-component kernel features; After restoring the electrocardiogram signal based on the frequency multi-component kernel feature, electrocardiogram image feature extraction is performed on the obtained optimized electrocardiogram data to obtain the electrocardiogram image semantic coding feature; Based on the semantic coding features of the electrocardiogram image, it is determined whether the diagnosed patient object has electrocardiogram abnormality.

[0016] This application has at least the following technical effects: Compared with the prior art, the artificial intelligence-based ECG data analysis method and system provided in the present application first utilizes wavelet transform to decompose the ECG data of the patient into multiple frequency components, and then uses artificial intelligence-based data processing and image analysis algorithms to analyze these frequency components, so as to extract the waveform semantic features of different frequency components of the ECG, and after information aggregation of these frequency component waveform semantic features based on the ECG kernel semantics, signal restoration is performed to output optimized ECG data, which can realize automatic noise reduction and optimization processing of ECG data, and further perform ECG abnormality analysis and detection on the patient based on the optimized ECG data, so as to further improve the effect of ECG abnormality detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 4 is a system block diagram of an abnormal ECG data analysis system based on artificial intelligence according to an embodiment of the present application.

[0018] Figure 2 Schematic diagram of data flow of an artificial intelligence-based ECG abnormality data analysis system according to an embodiment of the present application.

[0019] Figure 3It is a block diagram of a frequency component feature dynamic aggregation analysis module in an artificial intelligence-based ECG abnormality data analysis system according to an embodiment of the present application.

[0020] Figure 4 This is a block diagram of an ECG image semantic coding feature extraction module in an ECG abnormality data analysis system based on artificial intelligence according to an embodiment of the present application.

[0021] Figure 5 The present invention is a flowchart of an abnormal ECG data analysis method based on artificial intelligence according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0023] It should be noted that in this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where the device is located and with the authorization of the owner of the corresponding device.

[0024] As a vital medical detection method, the main function of the electrocardiogram is to record and reflect the electrophysiological activity of the heart. It is an indispensable auxiliary tool in the clinical diagnosis of cardiovascular diseases. However, in the actual electrocardiogram data collection process, the electrocardiogram signal is often interfered by a variety of internal and external factors, such as muscle activity, poor contact between electrodes and skin, and environmental electromagnetic interference, resulting in a variety of noise mixed in the signal. The presence of these noise components not only greatly hinders doctors' intuitive interpretation of the electrocardiogram, but more importantly, they also cause significant interference to the electrocardiogram analysis algorithm, thereby weakening the accuracy and reliability of ECG abnormality detection.

[0025] In order to improve the accuracy of ECG analysis, it is particularly important to recover clear ECG signals from noise. Although traditional ECG noise reduction methods, such as low-pass filtering and high-pass filtering, can reduce noise to a certain extent, the noise reduction effect of these traditional methods is not ideal for complex and variable noise types, especially nonlinear noise. While reducing noise, they are also prone to losing key feature information in the ECG, which affects the accuracy of subsequent ECG abnormality data analysis and detection.

[0026] It should be understood that wavelet transform is a mathematical tool used to decompose a signal into wavelet functions of different scales and positions. Compared with Fourier transform, wavelet transform can provide not only frequency information but also time information. Especially when processing non-stationary signals (such as electrocardiogram signals), it can provide multi-resolution analysis capabilities, which is very suitable for noise filtering. The noise in the electrocardiogram is usually nonlinear, such as electromyographic interference, baseline drift, etc. Traditional linear filters (such as low-pass filters) may not be able to effectively remove these nonlinear noises, while wavelet transform can process noise and useful signals separately at different scales through multi-resolution analysis.

[0027] Specifically, since the ECG signal contains multiple frequency components, components of different frequencies correspond to different physiological information. Noise often exists in a specific frequency range, while useful ECG signals are distributed in other frequency ranges. Wavelet transform can decompose the ECG signal into multiple frequency components through multi-resolution analysis, and then use artificial intelligence technology to analyze these frequency components to identify and remove noise in the ECG signal, while retaining useful ECG waveform information and improving the quality of ECG data, thereby providing a more reliable basis for subsequent feature extraction and anomaly detection.

[0028] Based on this, in the technical solution of the present application, an artificial intelligence-based ECG data analysis system is proposed, which can decompose the ECG data of the patient object into multiple frequency components by using wavelet transform, and then introduce artificial intelligence-based data processing and image analysis algorithms at the back end to analyze these frequency components, so as to extract the waveform semantic features of different frequency components of the ECG, and perform information aggregation based on the ECG kernel semantics on these frequency component waveform semantic features, and then perform signal restoration to output optimized ECG data, which can use wavelet transform and artificial intelligence technology to automatically perform noise reduction and optimization processing on the ECG data, and further perform ECG abnormality analysis and detection on the patient object based on the optimized ECG data, thereby automatically generating ECG abnormality analysis results in a more intelligent way, providing strong support for ECG abnormality detection and clinical diagnosis of patients.

[0029] Figure 1 4 is a system block diagram of an abnormal ECG data analysis system based on artificial intelligence according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of an abnormal ECG data analysis system based on artificial intelligence according to an embodiment of the present application. Figure 1 and Figure 2As shown, in the artificial intelligence-based ECG abnormality data analysis system 100, it includes: an ECG data collection module 110, which is used to obtain the ECG data of the diagnosed patient object; an ECG data wavelet transform processing module 120, which is used to perform wavelet transform on the ECG data of the diagnosed patient object to obtain multiple frequency components; a frequency component feature extraction module 130, which is used to extract features from the multiple frequency components to obtain multiple frequency component waveform semantic coding features; a frequency component feature dynamic aggregation analysis module 140, which is used to perform kernel feature dynamic aggregation analysis on the multiple frequency component waveform semantic coding features to obtain frequency multi-component kernel features; an ECG image semantic coding feature extraction module 150, which is used to restore the ECG signal based on the frequency multi-component kernel features, and then perform ECG image feature extraction on the obtained optimized ECG data to obtain ECG image semantic coding features; an ECG abnormality analysis result generation module 160, which is used to determine whether the diagnosed patient object has ECG abnormality based on the ECG image semantic coding features.

[0030] In an embodiment of the present application, the ECG data collection module 110 is used to obtain the ECG data of the diagnosed patient object. It should be understood that the original ECG data of the diagnosed patient object is the basis for subsequent analysis, and mainly includes ECG signal waveform information such as P wave, QRS complex and T wave. It records the electrical activity of the patient's heart and is an important basis for diagnosing whether the patient has ECG abnormalities. In this application, the original ECG data is subjected to noise reduction optimization processing, and feature analysis is performed based on the optimized ECG data to determine whether the patient object has ECG abnormalities, which helps to improve the accuracy and reliability of ECG abnormality detection.

[0031] In an embodiment of the present application, the electrocardiogram data wavelet transform processing module 120 is used to perform a wavelet transform on the electrocardiogram data of the diagnosed patient object to obtain multiple frequency components. It should be understood that the characteristics of the electrocardiogram signal vary with time and contain components of different frequencies. The wavelet transform can analyze the signal simultaneously in both the time and frequency domains, and is suitable for processing such non-stationary signals. In addition, the wavelet transform also provides the ability of multi-resolution analysis, which can analyze signals at different scales, which means that it is possible to effectively distinguish between high frequencies (such as noise) and low frequencies (such as heartbeat signals), which is conducive to better capturing the details and features of the electrocardiogram data to improve the accuracy of subsequent data analysis.

[0032] In an embodiment of the present application, the frequency component feature extraction module 130 is used to extract features from the multiple frequency components to obtain multiple frequency component waveform semantic coding features. Specifically, in an embodiment of the present application, the frequency component feature extraction module is used to: pass each frequency component of the multiple frequency components through a component frequency domain waveform semantic feature extractor based on a 1D-CNN model to obtain multiple frequency component waveform semantic coding feature vectors as the multiple frequency component waveform semantic coding features. It should be understood that the multiple frequency components are obtained by wavelet transform, wherein each frequency component contains different waveform semantic feature information in the electrocardiogram signal of the diagnosed patient object, and components of different frequencies correspond to different physiological information or noise interference characteristics. Therefore, in order to be able to identify and remove noise in the electrocardiogram signal while retaining useful electrocardiogram waveform information and improving the quality of electrocardiogram data, it is necessary to perform feature extraction and semantic importance analysis on these frequency components. Based on this, in the technical solution of the present application, each of the multiple frequency components is further subjected to feature mining in a component frequency domain waveform semantic feature extractor based on a 1D-CNN model to extract the waveform semantic feature information of each frequency component, thereby obtaining multiple frequency component waveform semantic coding feature vectors.

[0033] In the embodiment of the present application, the frequency component feature dynamic aggregation analysis module 140 is used to perform kernel feature dynamic aggregation analysis on the multiple frequency component waveform semantic coding features to obtain frequency multi-component kernel features. Specifically, Figure 3 FIG. 1 is a block diagram of a frequency component feature dynamic aggregation analysis module in an artificial intelligence-based ECG abnormality data analysis system according to an embodiment of the present application. Figure 3 As shown, the frequency component feature dynamic aggregation analysis module 140 includes: a frequency component feature static and dynamic potential factor calculation unit 141, which is used to calculate the static potential factor and dynamic potential factor of each frequency component waveform semantic coding feature in the multiple frequency component waveform semantic coding features to obtain a set of frequency component waveform semantic static potential factors and a set of frequency component waveform semantic dynamic potential factors; a frequency component feature kernel semantic aggregation unit 142, which is used to perform kernel semantic aggregation on the multiple frequency component waveform semantic coding features based on the set of frequency component waveform semantic static potential factors and the set of frequency component waveform semantic dynamic potential factors to obtain the frequency multi-component kernel feature.

[0034] It should be understood that since different frequency components in the electrocardiogram contain different physiological information and noise feature information respectively, at the same time, there are complex interactive semantics and association relationships between the semantic features of different frequency component waveforms in the electrocardiogram signal. Not only that, in the process of electrocardiogram noise reduction optimization and anomaly detection, these semantic information about the different frequency component waveforms of the electrocardiogram have different importance and contribution to the subsequent feature aggregation process and image optimization and electrocardiogram anomaly detection tasks. However, traditional feature aggregation methods are usually difficult to directly measure the specific contribution of different frequency component waveform semantic features to subsequent tasks and electrocardiogram noise reduction optimization processes, and it is also difficult to capture the complex internal structure and nonlinear relationship between different frequency component waveform semantic features, and simple linear transformation is not enough to fully describe these frequency domain semantic association feature information. Based on this, in order to be able to extract a compact and information-rich frequency component semantic kernel feature aggregation representation that can more accurately reflect the essential characteristics of the electrocardiogram from the set of frequency component waveform semantic encoding features, so as to better perform subsequent electrocardiogram anomaly detection tasks, in the technical solution of the present application, the kernel feature dynamic aggregation analysis of the multiple frequency component waveform semantic encoding features is further performed to obtain frequency multi-component kernel features.

[0035] In particular, the process of dynamic aggregation analysis of kernel features can extract compact and information-rich kernel features that can more accurately reflect the essential characteristics of the data from the multiple frequency component waveform semantic encoding feature vectors. In the process of dynamic aggregation analysis of kernel features, the feature force field can be used as an abstract space in which each frequency component waveform semantic encoding feature vector has a certain energy or influence, which includes both static properties (i.e., the information carried by the frequency component waveform semantic encoding feature vector itself) and dynamic properties (i.e., the interaction between each frequency component waveform semantic encoding feature vector). Dynamic analysis refers to the process of modeling the interaction pattern between these frequency component waveform semantic features, while kernel feature aggregation refers to converting the original multiple frequency component waveform semantic encoding feature vectors into a highly generalized core representation through a series of computational operations.

[0036] Specifically, the data processing flow of the kernel feature dynamic aggregation analysis includes constructing a center point or typical example that reflects the overall characteristics of the entire frequency component waveform semantic encoding feature set by calculating the positional mean vector of the multiple frequency component waveform semantic encoding feature vectors. This process helps to reduce noise and smooth data features. Next, for each individual frequency component waveform semantic encoding feature vector, its static potential factor is calculated. This step aims to evaluate the inherent properties or importance measures of each frequency component waveform semantic encoding feature vector when it is independent of other vectors. The static potential factor is used to distinguish which frequency component waveform semantic encoding features themselves carry more ECG semantic essential feature information. Subsequently, considering the interaction between the individual frequency component waveform semantic encoding feature vector and the overall data set, it is necessary to calculate the dynamic potential factor of each frequency component waveform semantic encoding feature vector relative to the frequency component waveform semantic initial aggregation representation vector. The dynamic potential factor is used to help capture the relative relationship between the data, that is, the association and complex interaction pattern between the semantic features of each frequency component waveform and the initial aggregate representation. This interactive semantics contains hidden, nonlinear relationships that may be ignored in traditional feature extraction methods, which helps to more effectively perform electrocardiogram noise reduction and optimization processing.

[0037] After obtaining the static and dynamic types of potential factors corresponding to the semantic encoding feature vector of each frequency component waveform, the next step is to combine these two types of information to form a superposition potential factor. This comprehensive indicator reflects the compound influence of the semantic features of a single frequency component waveform in the entire system. Next, these superposition potential factors are input into the potential factor gated weighting network, which can automatically learn the optimal weight allocation rules based on the data. Specifically, in order to enhance the interpretability and transparency of the model, an attention mechanism is introduced at this stage to explicitly mark the frequency components that contain key ECG semantic feature information, and to identify and highlight the frequency component waveform semantic features or their combinations that have the greatest impact on ECG denoising optimization and subsequent ECG abnormality detection tasks.

[0038] Finally, based on the set of frequency component waveform semantic features superimposed with potential weight coefficients, the multiple frequency component waveform semantic encoding feature vectors are weighted and summed by position, and the final frequency multi-component kernel feature vector can be obtained as the frequency multi-component kernel feature. This is an effective means of dimensionality reduction, which removes redundancy while retaining the main frequency component waveform semantic features. In addition, such processing can better capture the subtle differences and potential connections between the frequency component waveform semantic features, which can not only identify and retain the most core ECG waveform semantic information of multiple frequency component waveform semantic encoding features in the ECG data, but also remove redundant information and noise influence, provide a basis for subsequent ECG data optimization and auxiliary detection of ECG abnormalities in patients, and enhance the generalization ability of the model.

[0039] Specifically, in an embodiment of the present application, the frequency component feature static and dynamic energy potential factor calculation unit includes: a frequency component waveform semantic initial aggregation representation vector generation subunit, which is used to calculate the positional mean vector of the multiple frequency component waveform semantic coding feature vectors to obtain the frequency component waveform semantic initial aggregation representation vector; a frequency component waveform semantic static energy potential factor calculation subunit, which is used to calculate the static energy potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors to obtain the set of frequency component waveform semantic static energy potential factors; a frequency component waveform semantic dynamic energy potential factor calculation subunit, which is used to calculate the dynamic energy potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors relative to the frequency component waveform semantic initial aggregation representation vector to obtain the set of frequency component waveform semantic dynamic energy potential factors.

[0040] More specifically, in an embodiment of the present application, the frequency component waveform semantic static potential factor calculation subunit is used to: calculate the mean and standard deviation of the frequency component waveform semantic coding feature vector to obtain the frequency component waveform semantic mean and the frequency component waveform semantic standard deviation; calculate the positional difference between the frequency component waveform semantic coding feature vector and the frequency component waveform semantic mean, and then divide it by the frequency component waveform semantic standard deviation to obtain the frequency component waveform semantic distribution representation vector; calculate the fourth power of each position feature value in the frequency component waveform semantic distribution representation vector, and then calculate the mean of the obtained fourth power modulation vector to obtain the frequency component waveform semantic static potential factor.

[0041] More specifically, in an embodiment of the present application, the frequency component waveform semantic dynamic energy potential factor calculation subunit is used to: calculate the multiplication between the frequency component waveform semantic initial aggregation representation vector and the frequency component waveform semantic initial clustering weight matrix to obtain the frequency component waveform semantic initial aggregation representation modulation vector; calculate the multiplication between the frequency component waveform semantic coding feature vector and the frequency component waveform semantic weight matrix to obtain the frequency component waveform semantic modulation vector; perform position point multiplication on the frequency component waveform semantic initial aggregation representation modulation vector and the frequency component waveform semantic modulation vector and then perform position point multiplication on the frequency component waveform semantic bias vector to obtain the frequency component waveform semantic dynamic modulation representation vector; calculate the square of the Frobenius norm of the frequency component waveform semantic dynamic modulation representation vector to obtain the frequency component waveform semantic dynamic energy potential factor.

[0042] Specifically, in an embodiment of the present application, the frequency component feature kernel semantic aggregation unit includes: a frequency component waveform semantic superposition state potential factor calculation subunit, which is used to determine the superposition state potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors based on the frequency component waveform semantic static potential factor and the frequency component waveform semantic dynamic potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors to obtain a set of frequency component waveform semantic superposition state potential factors; a frequency component waveform semantic feature superposition potential weight coefficient calculation subunit, which is used to input the set of frequency component waveform semantic superposition state potential factors into the potential factor gated weighting network to obtain a set of frequency component waveform semantic feature superposition potential weight coefficients; a frequency multi-component kernel feature generation subunit, which is used to calculate the position-weighted sum of the multiple frequency component waveform semantic coding feature vectors based on the set of frequency component waveform semantic feature superposition potential weight coefficients to obtain a frequency multi-component kernel feature vector as the frequency multi-component kernel feature.

[0043] More specifically, in an embodiment of the present application, the frequency component waveform semantic superposition state potential factor calculation subunit is used to: calculate the natural exponential function value with the natural constant e as the base and the frequency component waveform semantic static potential factor of the frequency component waveform semantic encoding feature vector as the exponential power to obtain the frequency component waveform semantic static potential class support factor; calculate the natural exponential function value with the natural constant e as the base and the square of the frequency component waveform semantic dynamic potential factor of the frequency component waveform semantic encoding feature vector as the exponential power to obtain the frequency component waveform semantic dynamic potential class support factor; calculate the sum of the frequency component waveform semantic static potential class support factor and the frequency component waveform semantic dynamic potential class support factor to obtain the frequency component waveform semantic superposition state potential factor.

[0044] More specifically, in an embodiment of the present application, the frequency component waveform semantic feature superposition potential weight coefficient calculation subunit is used to: calculate the natural exponential function value with the natural constant e as the base and the negative number of the frequency component waveform semantic superposition state potential factor as the exponential power to obtain the frequency component waveform semantic superposition state potential class support factor; calculate the inverse of the sum of the frequency component waveform semantic superposition state potential class support factor and the constant one to obtain the frequency component waveform semantic feature superposition potential normalized representation coefficient; mask the frequency component waveform semantic feature superposition potential normalized representation coefficient to obtain the frequency component waveform semantic feature superposition potential weight coefficient; wherein, masking the frequency component waveform semantic feature superposition potential normalized representation coefficient to obtain the frequency component waveform semantic feature superposition potential weight coefficient includes: in response to the frequency component waveform semantic feature superposition potential normalized representation coefficient being greater than a predetermined threshold, taking the frequency component waveform semantic feature superposition potential normalized representation coefficient to its original value, otherwise setting it to 0 to obtain the frequency component waveform semantic feature superposition potential weight coefficient.

[0045] In the embodiment of the present application, specifically, the frequency component feature dynamic aggregation analysis module is used to: perform kernel feature dynamic aggregation analysis on the multiple frequency component waveform semantic coding features using the following feature dynamic aggregation formula to obtain frequency multi-component kernel features; wherein the feature dynamic aggregation formula is: in, semantically encoding feature vectors for the plurality of frequency component waveforms, are respectively the first, second, ..., first, second, third, fourth, fifth, fifth, sixth, sixth, fifth, sixth, seventh, fifth, seventh ... ,..., frequency component waveform semantic encoding feature vector, is the number of vectors in the semantic encoding feature vectors of the multiple frequency component waveforms, is the initial aggregation representation vector of the frequency component waveform semantics, It is The eigenvalues ​​of each position in the semantic encoding feature vector of the frequency component waveform are and They are The mean and standard deviation of yes The number of eigenvalues ​​in , yes The corresponding frequency component waveform semantic static energy potential factor, , and They are the frequency component waveform semantic initial clustering weight matrix, the frequency component waveform semantic feature weight matrix and the frequency component waveform semantic bias vector, It means point multiplication by position. represents the square of the Frobenius norm of a vector, yes The corresponding frequency component waveform semantic dynamic energy potential factor, yes The corresponding frequency component waveform semantic superposition state energy potential factor, yes The corresponding frequency component waveform semantic feature superposition energy potential normalized representation coefficient, Indicates masking processing. is the predetermined threshold, yes The corresponding frequency component waveform semantic feature superposition energy potential weight coefficient, is the frequency multi-component kernel eigenvector.

[0046] In the embodiment of the present application, the ECG image semantic coding feature extraction module 150 is used to restore the ECG signal based on the frequency multi-component kernel feature, and then extract ECG image features from the obtained optimized ECG data to obtain ECG image semantic coding features. Specifically, Figure 4 FIG. 1 is a block diagram of a semantic coding feature extraction module for ECG images in an ECG abnormality data analysis system based on artificial intelligence according to an embodiment of the present application. Figure 4 As shown, the ECG image semantic coding feature extraction module 150 includes: an optimized ECG data decoding generation unit 151, used for inputting the frequency multi-component kernel feature vector into a decoder-based signal restorer to obtain the optimized ECG data; an optimized ECG data feature extraction unit 152, used for inputting the optimized ECG data into a trained ECG image feature extractor based on a convolutional neural network model to obtain an ECG image semantic coding feature vector as the ECG image semantic coding feature.

[0047] In an embodiment of the present application, the optimized electrocardiogram data decoding generation unit 151 is used to input the frequency multi-component kernel feature vector into a decoder-based signal restorer to obtain the optimized electrocardiogram data. Accordingly, after obtaining the semantic kernel feature recognition and aggregation of multiple frequency component waveforms of the electrocardiogram, these key frequency multi-component kernel features are used to perform signal restoration to obtain the optimized electrocardiogram data. Specifically, the frequency multi-component kernel feature vector is input into a decoder-based signal restorer to obtain the optimized electrocardiogram data, thereby providing an accurate basis for subsequent feature extraction and electrocardiogram abnormality detection. It should be known to those of ordinary skill in the art that the decoder, as a neural network structure, can convert compressed feature vectors back to the original data space after training. Through the processing of the decoder, higher quality electrocardiogram data can be obtained, which is conducive to improving the reliability of information analysis in subsequent electrocardiogram data.

[0048] In the case where the multiple frequency component waveform semantic coding feature vectors respectively represent the frequency domain waveform semantic coding features of each frequency component determined by wavelet transform of the original electrocardiogram data of the diagnosed patient object, when they are input into the kernel feature extraction network based on feature force field dynamic analysis, the frequency multi-component kernel feature vector will also have frequency domain correlation feature force field dynamic differences under local frequency domain distribution differences, thereby causing differences in the spatial structure of the kernel feature distribution based on force field dynamics. As a result, when the frequency multi-component kernel feature vector is decoded and regressed through a decoder-based signal restorer, it will affect the accuracy of the decoding result.

[0049] Preferably, inputting the frequency multi-component kernel feature vector into a decoder-based signal restorer to obtain optimized electrocardiogram data comprises: The absolute value sum and the square root of the square sum of all eigenvalues ​​of the frequency multi-component kernel eigenvector are calculated to obtain the first frequency multi-component kernel feature space structure value and the second frequency multi-component kernel feature space structure value, that is: in, represents the frequency multi-component kernel feature vector eigenvalues, represents the first frequency multi-component kernel feature space structure value, represents the second frequency multi-component kernel feature space structure value; Determine the number of eigenvalues ​​of all eigenvalues ​​of the frequency multicomponent kernel eigenvector , i.e., the length of the frequency multi-component kernel eigenvector; For each eigenvalue of the frequency multi-component kernel eigenvector, the first frequency multi-component kernel feature spatial structure value minus the product of the eigenvalue and the number of eigenvalues ​​is calculated to obtain the first frequency multi-component kernel feature long-range dependency value ,in, represents the first frequency multi-component kernel feature space structure value, represents the frequency multi-component kernel feature vector eigenvalues, the number of eigenvalues ​​representing all eigenvalues ​​of the frequency multicomponent kernel eigenvector, represents the long-range dependency value of the first frequency multi-component kernel feature; Calculate the second frequency multi-component kernel feature long-range dependency value obtained by multiplying the square root of the number of eigenvalues ​​by the product of the eigenvalues ​​minus the second frequency multi-component kernel feature spatial structure value ,in, represents the second frequency multi-component kernel feature space structure value, represents the frequency multi-component kernel feature vector eigenvalues, the number of eigenvalues ​​representing all eigenvalues ​​of the frequency multicomponent kernel eigenvector, represents the long-range dependency value of the second frequency multi-component kernel feature; The exponential value calculated by taking the first frequency multi-component kernel feature long-range dependency value as the exponent of the natural constant and the inverse of the second frequency multi-component kernel feature long-range dependency value are weighted summed to obtain the optimized eigenvalue corresponding to each eigenvalue, in, represents the long-range dependency value of the first frequency multi-component kernel feature, represents the long-range dependency value of the second frequency multi-component kernel feature, represents a natural constant, and represents the weighted hyperparameter, Indicates the optimized eigenvalue corresponding to each eigenvalue; Combining the optimized eigenvalues ​​into an optimized frequency multi-component kernel eigenvector; The optimized frequency multi-component kernel feature vector is input into a decoder-based signal restorer to obtain optimized electrocardiogram data.

[0050] That is, in view of the possible spatial structure loss of the feature set of the frequency multi-component kernel feature vector in the high-dimensional space, which causes the decoder weight to implicitly infer spatial structure information based on the features, resulting in inconsistent convergence, a long-distance feature dependency is established based on the spatial structure representation of the frequency multi-component kernel feature vector relative to the overall feature scale of the frequency multi-component kernel feature vector to establish the feature local connectivity of the frequency multi-component kernel feature vector, and the spatial ambiguity information of the object feature value is captured by the unstructured feature value point prediction of the frequency multi-component kernel feature vector, thereby enhancing the spatial inductive bias perception ability of the feature set of the frequency multi-component kernel feature vector and improving the convergence consistency of the decoder. Improve the accuracy of the optimized electrocardiogram data obtained by the frequency multi-component kernel feature vector through the decoder-based signal restorer.

[0051] In the embodiment of the present application, the optimized ECG data feature extraction unit 152 is used to input the optimized ECG data into the trained ECG image feature extractor based on the convolutional neural network model to obtain the ECG image semantic coding feature vector as the ECG image semantic coding feature. It should be understood that the optimized ECG data is input into the trained ECG image feature extractor based on the convolutional neural network model for feature mining to extract the semantic coding feature information in the optimized ECG data after denoising and kernel semantic feature enhancement aggregation expression, thereby obtaining the ECG image semantic coding feature vector.

[0052] In the embodiment of the present application, the ECG abnormality analysis result generation module 160 is used to determine whether the diagnosed patient object has ECG abnormality based on the semantic coding features of the ECG image. Specifically, in the embodiment of the present application, the ECG abnormality analysis result generation module is used to: input the ECG image semantic coding feature vector into the intelligent analysis module based on the classifier to obtain the analysis result, and the analysis result is used to indicate whether the diagnosed patient object has ECG abnormality. Specifically, the semantic coding features of the ECG image are used for classification processing, so as to perform intelligent analysis and detection on whether the diagnosed patient object has ECG abnormality. In this way, wavelet transform and artificial intelligence technology can be used to automatically perform noise reduction and optimization processing on ECG data to improve the quality of ECG data, and further based on the optimized ECG data, ECG abnormality analysis and detection are performed on the patient object, so as to automatically generate ECG abnormality analysis results in a more intelligent way, which provides strong support for ECG abnormality detection and clinical diagnosis of patients. In particular, in a specific embodiment of the present application, the ECG image semantic coding feature vector is input into an intelligent analysis module based on a classifier to obtain an analysis result, and the analysis result is used to indicate whether the diagnosed patient object has ECG abnormalities, including: using the fully connected layer of the classifier to fully connect the ECG image semantic coding feature vector to obtain the ECG image semantic fully connected coding feature vector; inputting the ECG image semantic fully connected coding feature vector into the Softmax classification function of the classifier to obtain the probability value of the ECG image semantic coding feature vector belonging to each classification label, and the classification labels include those for indicating that the diagnosed patient object has ECG abnormalities and those for indicating that the diagnosed patient object does not have ECG abnormalities; and determining the classification label corresponding to the largest of the probability values ​​as the analysis result.

[0053] To summarize, an artificial intelligence-based ECG abnormality data analysis system 100 based on an embodiment of the present application is explained, which first uses wavelet transform to decompose the ECG data of the patient subject into multiple frequency components, and then uses artificial intelligence-based data processing and image analysis algorithms to analyze these frequency components, so as to extract the waveform semantic features of different frequency components of the ECG, and after information aggregation of these frequency component waveform semantic features based on the ECG kernel semantics, signal restoration is performed to output optimized ECG data, which can realize automatic denoising and optimization processing of ECG data, and further perform ECG abnormality analysis and detection on the patient subject based on the optimized ECG data to further improve the effect of ECG abnormality detection.

[0054] As described above, the ECG abnormal data analysis system 100 based on artificial intelligence according to the embodiment of the present application can be implemented in various terminal devices, such as a server for ECG abnormal data analysis based on artificial intelligence. In one example, the ECG abnormal data analysis system 100 based on artificial intelligence according to the embodiment of the present application can be integrated into the terminal device as a software module and / or a hardware module. For example, the ECG abnormal data analysis system 100 based on artificial intelligence can be a software module in the operating system of the terminal device, or can be an application developed for the terminal device; of course, the ECG abnormal data analysis system 100 based on artificial intelligence can also be one of the many hardware modules of the terminal device.

[0055] Alternatively, in another example, the artificial intelligence-based ECG abnormality data analysis system 100 and the terminal device may also be separate devices, and the artificial intelligence-based ECG abnormality data analysis system 100 may be connected to the terminal device via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0056] Figure 5 FIG. 1 is a flow chart of an abnormal ECG data analysis method based on artificial intelligence according to an embodiment of the present application. Figure 5 As shown, in the abnormal ECG data analysis method based on artificial intelligence, it includes: S110, acquiring the ECG data of the diagnosed patient object; S120, performing wavelet transform on the ECG data of the diagnosed patient object to obtain multiple frequency components; S130, performing feature extraction on the multiple frequency components to obtain multiple frequency component waveform semantic coding features; S140, performing kernel feature dynamic aggregation analysis on the multiple frequency component waveform semantic coding features to obtain frequency multi-component kernel features; S150, performing ECG signal restoration based on the frequency multi-component kernel features, and performing ECG image feature extraction on the obtained optimized ECG data to obtain ECG image semantic coding features. S160, based on the semantic coding features of the ECG image, determine whether the diagnosed patient object has ECG abnormalities, wherein step S140 includes: calculating the static potential factor and the dynamic potential factor of each frequency component waveform semantic coding feature in the multiple frequency component waveform semantic coding features to obtain a set of frequency component waveform semantic static potential factors and a set of frequency component waveform semantic dynamic potential factors; based on the set of frequency component waveform semantic static potential factors and the set of frequency component waveform semantic dynamic potential factors, performing kernel semantic aggregation on the multiple frequency component waveform semantic coding features to obtain the frequency multi-component kernel feature.

[0057] The specific operations of each step in the above-mentioned artificial intelligence-based ECG abnormal data analysis method have been referenced above. Figures 1 to 4 The artificial intelligence-based ECG abnormality data analysis system has been introduced in detail, and therefore, its repeated description will be omitted.

[0058] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in this application (but not limited to) by each other to form a technical solution.

[0059] In addition, although each operation is described in a specific order, this should not be construed as requiring these operations to be performed in the specific order shown or to be performed in a sequential order. Under certain circumstances, multitasking and parallel processing may be advantageous. Similarly, although some specific implementation details are included in the above discussion, these should not be construed as limiting the scope of the application. Some features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination mode.

[0060] Although the subject matter has been described in language specific to structural features and / or method logic actions, it should be understood that the subject matter defined is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely exemplary forms of implementation. With respect to the apparatus in the above-mentioned embodiments, the specific manner in which each module performs the operation has been described in detail in the embodiments related to the method, and will not be described in detail here.

Claims

1. An artificial intelligence-based ECG abnormality data analysis system, characterized in that: include: An electrocardiogram data collection module, used to obtain electrocardiogram data of a diagnosed patient object; An electrocardiogram data wavelet transform processing module, used for performing wavelet transform on the electrocardiogram data of the diagnosed patient object to obtain a plurality of frequency components; A frequency component feature extraction module, used for extracting features from the multiple frequency components to obtain multiple frequency component waveform semantic coding features; A frequency component feature dynamic aggregation analysis module is used to perform kernel feature dynamic aggregation analysis on the multiple frequency component waveform semantic coding features to obtain frequency multi-component kernel features, wherein the frequency component feature dynamic aggregation analysis module includes: a frequency component feature static and dynamic potential factor calculation unit, used to calculate the static potential factor and dynamic potential factor of each frequency component waveform semantic coding feature in the multiple frequency component waveform semantic coding features to obtain a set of frequency component waveform semantic static potential factors and a set of frequency component waveform semantic dynamic potential factors; a frequency component feature kernel semantic aggregation unit, used to perform kernel semantic aggregation on the multiple frequency component waveform semantic coding features based on the set of frequency component waveform semantic static potential factors and the set of frequency component waveform semantic dynamic potential factors to obtain the frequency multi-component kernel features; An ECG image semantic coding feature extraction module is used to restore the ECG signal based on the frequency multi-component kernel feature, and then extract ECG image features from the obtained optimized ECG data to obtain ECG image semantic coding features; The ECG abnormality analysis result generating module is used to determine whether the diagnosed patient object has ECG abnormality based on the semantic coding features of the ECG image.

2. The artificial intelligence-based ECG abnormality data analysis system according to claim 1, characterized in that: The frequency component feature extraction module is used to: pass each frequency component of the multiple frequency components through a component frequency domain waveform semantic feature extractor based on a 1D-CNN model to obtain multiple frequency component waveform semantic coding feature vectors as the multiple frequency component waveform semantic coding features.

3. The artificial intelligence-based ECG abnormality data analysis system according to claim 2, characterized in that: The frequency component characteristic static and dynamic energy potential factor calculation unit includes: A frequency component waveform semantic initial aggregate representation vector generating subunit, used for calculating the positional mean vector of the plurality of frequency component waveform semantic encoding feature vectors to obtain a frequency component waveform semantic initial aggregate representation vector; A frequency component waveform semantic static energy potential factor calculation subunit, used to calculate the static energy potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors to obtain a set of frequency component waveform semantic static energy potential factors; The frequency component waveform semantic dynamic potential factor calculation subunit is used to calculate the dynamic potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors relative to the frequency component waveform semantic initial aggregation representation vector to obtain a set of frequency component waveform semantic dynamic potential factors.

4. The artificial intelligence-based ECG abnormality data analysis system according to claim 3, characterized in that: The frequency component waveform semantic static energy potential factor calculation subunit is used to: Calculating the mean and standard deviation of the frequency component waveform semantic encoding feature vector to obtain the frequency component waveform semantic mean and the frequency component waveform semantic standard deviation; Calculate the position difference between the frequency component waveform semantic coding feature vector and the frequency component waveform semantic mean, and then divide it by the frequency component waveform semantic standard deviation to obtain the frequency component waveform semantic distribution representation vector; After calculating the fourth power of the characteristic values ​​of each position in the frequency component waveform semantic distribution representation vector, the mean value of the fourth power modulation vector is calculated to obtain the frequency component waveform semantic static energy potential factor.

5. The artificial intelligence-based ECG abnormality data analysis system according to claim 4, characterized in that: The frequency component waveform semantic dynamic energy potential factor calculation subunit is used to: Calculating the multiplication between the frequency component waveform semantic initial aggregation representation vector and the frequency component waveform semantic initial clustering weight matrix to obtain the frequency component waveform semantic initial aggregation representation modulation vector; Calculating the multiplication between the frequency component waveform semantic coding feature vector and the frequency component waveform semantic weight matrix to obtain a frequency component waveform semantic modulation vector; The frequency component waveform semantic initial aggregation representation modulation vector is multiplied by the frequency component waveform semantic modulation vector by the position point, and then multiplied by the frequency component waveform semantic bias vector by the position point to obtain the frequency component waveform semantic dynamic modulation representation vector; The square of the Frobenius norm of the frequency component waveform semantic dynamic modulation representation vector is calculated to obtain the frequency component waveform semantic dynamic energy potential factor.

6. The artificial intelligence-based ECG abnormality data analysis system according to claim 5, characterized in that: The frequency component feature kernel semantic aggregation unit includes: A frequency component waveform semantic superposition state potential factor calculation subunit is used to determine the superposition state potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors based on the frequency component waveform semantic static potential factor and the frequency component waveform semantic dynamic potential factor of each frequency component waveform semantic coding feature vector in the multiple frequency component waveform semantic coding feature vectors to obtain a set of frequency component waveform semantic superposition state potential factors; A frequency component waveform semantic feature superposition potential weight coefficient calculation subunit, used for inputting the set of frequency component waveform semantic superposition state potential factors into the potential factor gated weighting network to obtain a set of frequency component waveform semantic feature superposition potential weight coefficients; The frequency multi-component kernel feature generating subunit is used to calculate the position-weighted sum of the multiple frequency component waveform semantic encoding feature vectors based on the set of potential weight coefficients superimposed on the frequency component waveform semantic features to obtain the frequency multi-component kernel feature vector as the frequency multi-component kernel feature.

7. The artificial intelligence-based ECG abnormality data analysis system according to claim 6, characterized in that: The frequency component waveform semantic superposition state energy potential factor calculation subunit is used to: Calculate the natural exponential function value with the natural constant e as the base and the frequency component waveform semantic static potential factor of the frequency component waveform semantic encoding feature vector as the exponential power to obtain the frequency component waveform semantic static potential class support factor; Calculate the natural exponential function value with the natural constant e as the base and the square of the frequency component waveform semantic dynamic potential factor of the frequency component waveform semantic encoding feature vector as the exponential power to obtain the frequency component waveform semantic dynamic potential class support factor; The sum of the frequency component waveform semantic static potential class support factor and the frequency component waveform semantic dynamic potential class support factor is calculated to obtain the frequency component waveform semantic superposition state potential factor.

8. The artificial intelligence-based ECG abnormality data analysis system according to claim 7, characterized in that: The electrocardiogram image semantic coding feature extraction module comprises: An optimized electrocardiogram data decoding and generating unit, used for inputting the frequency multi-component kernel feature vector into a decoder-based signal restorer to obtain the optimized electrocardiogram data; The optimized electrocardiogram data feature extraction unit is used to input the optimized electrocardiogram data into a trained electrocardiogram image feature extractor based on a convolutional neural network model to obtain an electrocardiogram image semantic coding feature vector as the electrocardiogram image semantic coding feature.

9. The artificial intelligence-based ECG abnormality data analysis system according to claim 8, characterized in that: The ECG abnormality analysis result generating module is used to: input the ECG image semantic coding feature vector into a classifier-based intelligent analysis module to obtain an analysis result, and the analysis result is used to indicate whether the diagnosed patient object has an ECG abnormality.

10. An artificial intelligence-based ECG abnormality data analysis method, characterized in that: include: obtaining electrocardiogram data of a diagnosed patient subject; Performing wavelet transformation on the electrocardiogram data of the diagnosed patient object to obtain a plurality of frequency components; Performing feature extraction on the multiple frequency components to obtain multiple frequency component waveform semantic coding features; Performing kernel feature dynamic aggregation analysis on the multiple frequency component waveform semantic coding features to obtain frequency multi-component kernel features, including: calculating the static potential factor and the dynamic potential factor of each frequency component waveform semantic coding feature in the multiple frequency component waveform semantic coding features to obtain a set of frequency component waveform semantic static potential factors and a set of frequency component waveform semantic dynamic potential factors; based on the set of frequency component waveform semantic static potential factors and the set of frequency component waveform semantic dynamic potential factors, performing kernel semantic aggregation on the multiple frequency component waveform semantic coding features to obtain the frequency multi-component kernel features; After restoring the electrocardiogram signal based on the frequency multi-component kernel feature, electrocardiogram image feature extraction is performed on the obtained optimized electrocardiogram data to obtain the electrocardiogram image semantic coding feature; Based on the semantic coding features of the electrocardiogram image, it is determined whether the diagnosed patient object has electrocardiogram abnormality.

Images