Self-supervised atrial fibrillation detection method based on cross-modal cross reconstruction and related equipment

Abstract

The invention provides a self-supervision atrial fibrillation detection method based on cross-modal cross reconstruction and related equipment. The method comprises the following steps: acquiring a to-be-detected electrocardiogram signal; performing time-frequency conversion on the electrocardiogram signal to be detected to obtain an electrocardiogram frequency domain signal; inputting the electrocardiogram signal to be detected and the electrocardiogram frequency domain signal into a preset self-supervised learning model to obtain a reconstructed electrocardiogram signal and a reconstructed electrocardiogram frequency domain signal; the self-supervised learning model comprises an encoder, a time domain decoder and a frequency domain decoder; the encoder is used for extracting time domain low-dimensional embedding features and frequency domain low-dimensional embedding features; the time domain decoder is used for generating a reconstructed electrocardiogram signal based on the frequency domain low-dimensional embedded feature; the frequency domain decoder is used for generating a reconstructed electrocardiogram frequency domain signal based on the time domain low-dimensional embedding feature; and calculating a reconstruction error between the reconstructed electrocardiogram frequency domain signal and the original electrocardiogram frequency domain signal, and if the reconstruction error is greater than a preset abnormal score, determining that the electrocardiogram signal to be detected is atrial fibrillation data.

Description

Technical Field

[0001] The present invention relates to the technical field of atrial fibrillation detection, and in particular to a self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction and related equipment. Background Art

[0002] Atrial fibrillation (AF), also known as AF, is a common arrhythmia. With the development of modern science and technology, AF detection has become an important topic in the field of medical health. At present, AF detection methods mainly include timing anomaly detection methods and deep learning-based methods. The former usually ignores the medical characteristics unique to AF electrocardiogram data, and its accuracy needs to be improved; the latter usually relies on a large number of labeled AF samples to train a deep learning model, and the labeling cost is relatively high. Summary of the invention

[0003] In order to reduce the labeling cost of atrial fibrillation samples and improve the average accuracy of atrial fibrillation detection, the present invention provides a self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction and related equipment.

[0004] In a first aspect, the present invention provides a self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction, comprising:

[0005] Acquire the electrocardiogram signal to be tested;

[0006] Performing time-frequency conversion on the electrocardiogram signal to be measured to obtain an electrocardiogram frequency domain signal;

[0007] The electrocardiogram signal to be measured and the electrocardiogram frequency domain signal are input into a preset self-supervised learning model to obtain a reconstructed electrocardiogram signal and a reconstructed electrocardiogram frequency domain signal; the self-supervised learning model is obtained by training based on a normal electrocardiogram signal; the self-supervised learning model includes an encoder, a time domain decoder and a frequency domain decoder; the encoder is used to extract time domain low-dimensional embedding features and frequency domain low-dimensional embedding features from the input electrocardiogram signal and the electrocardiogram frequency domain signal respectively; the time domain decoder is used to generate a reconstructed electrocardiogram signal based on the frequency domain low-dimensional embedding features; the frequency domain decoder is used to generate a reconstructed electrocardiogram frequency domain signal based on the time domain low-dimensional embedding features;

[0008] A reconstruction error between the reconstructed electrocardiogram frequency domain signal and the original electrocardiogram frequency domain signal is calculated. If the reconstruction error is greater than a preset abnormality score, the electrocardiogram signal to be measured is considered to be atrial fibrillation data.

[0009] Furthermore, the training process of the self-supervised learning model includes:

[0010] Construct a normal ECG signal dataset;

[0011] Inputting a normal electrocardiogram signal data set and its corresponding original electrocardiogram frequency domain signal data set into the self-supervised learning model to generate a reconstructed electrocardiogram signal data set and a reconstructed electrocardiogram frequency domain signal data set;

[0012] Minimize the joint loss to iteratively optimize the parameters of the self-supervised learning model and save the self-supervised learning model under the optimal parameters; wherein the joint loss is composed of the reconstruction loss between the reconstructed ECG signal data set and the original ECG signal data set and the reconstruction loss between the reconstructed ECG frequency domain signal data set and the original ECG frequency domain signal data set.

[0013] Furthermore, it also includes a time domain predictor and a frequency domain predictor; the time domain predictor is used to predict the medical features of the electrocardiogram signal in the time domain based on the time domain low-dimensional embedding features, and the frequency domain predictor is used to predict the medical features of the electrocardiogram signal in the frequency domain based on the frequency domain low-dimensional embedding features;

[0014] Correspondingly, the training process of the self-supervised learning model also includes:

[0015] The time domain predictor and the frequency domain predictor are used as regularizers so as to introduce a regularization term in the joint loss.

[0016] Furthermore, the time-domain electrocardiogram signal medical characteristics include heart rate and atrial fibrillation index.

[0017] Furthermore, the medical characteristics of the ECG signal in the frequency domain include power in a high-frequency band and power in a low-frequency band.

[0018] In a second aspect, the present invention provides a self-supervised atrial fibrillation detection device based on cross-modal cross-reconstruction, comprising:

[0019] An acquisition module, used for acquiring an electrocardiogram signal to be tested;

[0020] A data preprocessing module, used for performing time-frequency conversion on the electrocardiogram signal to be measured to obtain an electrocardiogram frequency domain signal;

[0021] A cross-modal cross-reconstruction module, used for inputting the electrocardiogram signal to be measured and the electrocardiogram frequency domain signal into a preset self-supervised learning model to obtain a reconstructed electrocardiogram signal and a reconstructed electrocardiogram frequency domain signal; the self-supervised learning model is obtained by training based on a normal electrocardiogram signal; the self-supervised learning model includes an encoder, a time domain decoder and a frequency domain decoder; the encoder is used for extracting time domain low-dimensional embedding features and frequency domain low-dimensional embedding features from the input electrocardiogram signal and the electrocardiogram frequency domain signal respectively; the time domain decoder is used for generating a reconstructed electrocardiogram signal based on the frequency domain low-dimensional embedding features; the frequency domain decoder is used for generating a reconstructed electrocardiogram frequency domain signal based on the time domain low-dimensional embedding features;

[0022] The detection module is used to calculate the reconstruction error between the reconstructed electrocardiogram frequency domain signal and the original electrocardiogram frequency domain signal. If the reconstruction error is greater than a preset abnormality score, the electrocardiogram signal to be tested is considered to be atrial fibrillation data.

[0023] In a third aspect, the present invention provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first aspect when executing the program.

[0024] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in the first aspect.

[0025] The beneficial effects of the present invention are:

[0026] (1) The present invention models the atrial fibrillation detection task as an abnormality detection task and trains the self-supervised learning model based on normal electrocardiogram signal data. Therefore, it does not rely on a large number of labeled atrial fibrillation samples, thereby reducing the labeling cost. In addition, by simultaneously learning and reconstructing the global features of the electrocardiogram signal data from the time domain and frequency domain perspectives, the self-supervised learning model can learn more discriminative feature embeddings for atrial fibrillation detection, thereby improving the average accuracy of atrial fibrillation detection.

[0027] (2) When training the self-supervised learning model, the calculated time-frequency features are used as labels, and the time domain predictor and frequency domain predictor are introduced as regularizers, so that medical feature predictions are used to guide model learning and extract effective features from electrocardiogram data, so that the self-supervised learning model can learn more discriminative features for atrial fibrillation detection and further improve the accuracy of atrial fibrillation detection.

[0028] (3) Experimental results show that the AUC and AP of the proposed method are 6.29%, 9.22%, 3.57%, 9.37%, 11.94%, and 2.84% higher than those of the mainstream algorithms in CPSC2021, Icentia11k, and PTBXL, respectively, which proves the effectiveness of the proposed model in the task of atrial fibrillation detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A schematic flow chart of a self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction provided by an embodiment of the present invention;

[0030] Figure 2 A training process of a self-supervised learning model provided by an embodiment of the present invention;

[0031] Figure 3A framework diagram of a self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction provided by an embodiment of the present invention;

[0032] Figure 4 A schematic diagram of the structure of a self-supervised atrial fibrillation detection device based on cross-modal cross-reconstruction provided by an embodiment of the present invention;

[0033] Figure 5 A structural block diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution in the embodiment of the present invention will be clearly described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a part of the embodiment of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0035] like Figure 1 As shown, an embodiment of the present invention provides a self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction, comprising the following steps:

[0036] S101: Acquire an electrocardiogram signal to be tested;

[0037] Specifically, the electrocardiogram (ECG) signal can be collected based on traditional ECG signal acquisition equipment (such as placing electrodes on specific parts of the user's body to capture ECG signals), or collected based on portable and smart wearable devices (such as smart watches or smart phones that collect photoelectric volume pulse wave signals and then convert them into ECG signals), or ECG signals reconstructed based on contactless ECG acquisition technology (such as millimeter wave radar technology that reconstructs ECG signals by detecting tiny vibrations of the chest wall).

[0038] S102: performing time-frequency conversion on the electrocardiogram signal to be measured to obtain an electrocardiogram frequency domain signal;

[0039] Specifically, the time-frequency conversion methods include but are not limited to Fourier transform, short-time Fourier transform, fast Fourier transform and wavelet transform.

[0040] S103: inputting the electrocardiogram signal to be measured and the electrocardiogram frequency domain signal into a preset self-supervised learning model to obtain a reconstructed electrocardiogram signal and a reconstructed electrocardiogram frequency domain signal;

[0041] Specifically, the self-supervised learning model is trained based on a normal electrocardiogram signal; the self-supervised learning model includes an encoder, a time domain decoder and a frequency domain decoder; the encoder is used to extract time domain low-dimensional embedding features and frequency domain low-dimensional embedding features from the input electrocardiogram signal and the electrocardiogram frequency domain signal respectively; the time domain decoder is used to generate a reconstructed electrocardiogram signal based on the frequency domain low-dimensional embedding features; the frequency domain decoder is used to generate a reconstructed electrocardiogram frequency domain signal based on the time domain low-dimensional embedding features.

[0042] S104: Calculate a reconstruction error between the reconstructed electrocardiogram frequency domain signal and the original electrocardiogram frequency domain signal. If the reconstruction error is greater than a preset abnormality score, the electrocardiogram signal to be measured is considered to be atrial fibrillation data.

[0043] The self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction provided in an embodiment of the present invention models the atrial fibrillation detection task as an abnormality detection task, and trains the self-supervised learning model based on normal electrocardiogram signal data. Therefore, it does not rely on a large number of labeled atrial fibrillation samples, thereby reducing the labeling cost; and, by simultaneously learning and reconstructing the global features of the electrocardiogram signal data from the time domain and frequency domain perspectives, the self-supervised learning model can learn more discriminative feature embeddings for atrial fibrillation detection, thereby improving the average accuracy of atrial fibrillation detection.

[0044] In one embodiment, the present invention further provides a training process of a self-supervised learning model, comprising the following steps:

[0045] S201: construct a normal electrocardiogram signal data set;

[0046] S202: Inputting a normal electrocardiogram signal dataset and its corresponding original electrocardiogram frequency domain signal dataset into the self-supervised learning model to generate a reconstructed electrocardiogram signal dataset and a reconstructed electrocardiogram frequency domain signal dataset;

[0047] Specifically, the normal ECG signal dataset is converted into a corresponding ECG frequency domain signal dataset by performing time-frequency conversion. At the same time, in order to distinguish the ECG frequency domain signal dataset obtained by time-frequency conversion from the ECG frequency domain signal dataset generated by the self-supervised learning model, the former is called the original ECG frequency domain signal dataset, and the latter is called the reconstructed ECG frequency domain signal dataset.

[0048] S203: Minimize the joint loss to iteratively optimize the parameters of the self-supervised learning model and save the self-supervised learning model under the optimal parameters; wherein the joint loss is composed of the reconstruction loss between the reconstructed electrocardiogram signal data set and the original electrocardiogram signal data set and the reconstruction loss between the reconstructed electrocardiogram frequency domain signal data set and the original electrocardiogram frequency domain signal data set.

[0049] Specifically, the reconstruction loss may include but is not limited to mean square error loss, mean absolute error loss, or a mixed loss consisting of the above two loss functions.

[0050] In one embodiment, in order to further optimize the detection performance of the self-supervised learning model, a regularizer is introduced in the training process, and the regularizer includes a time domain predictor and a frequency domain predictor; the time domain predictor is used to predict the medical features of the ECG signal in the time domain based on the time domain low-dimensional embedding features, and the frequency domain predictor is used to predict the medical features of the ECG signal in the frequency domain based on the frequency domain low-dimensional embedding features.

[0051] Specifically, the time domain electrocardiogram signal medical features (hereinafter referred to as the first time domain features) can be calculated based on the normal electrocardiogram signal data, and the frequency domain electrocardiogram signal medical features (hereinafter referred to as the first frequency domain features) can be calculated based on the original electrocardiogram frequency domain signal data. The predicted time domain electrocardiogram signal medical features are recorded as the second time domain features, and the predicted frequency domain electrocardiogram signal medical features are recorded as the second frequency domain features. By calculating the distance between the first time domain feature and the second time domain feature, and calculating the distance between the first frequency domain feature and the second frequency domain feature, a regularization term (such as L1 regularization or L2 regularization) is generated based on the two distances, and finally by introducing the regularization term into the original joint loss, the self-supervised learning model can learn more discriminative feature embeddings for atrial fibrillation detection, thereby improving its detection performance.

[0052] In one embodiment, the time-domain electrocardiogram signal medical features include heart rate (HR) and atrial fibrillation index (AFI). Specifically, for the first time-domain feature, the calculation process of the atrial fibrillation index is as follows: first, the NN interval (NNI) is calculated according to the R peak of the electrocardiogram (ECG) signal, and then the atrial fibrillation index AFI is determined according to each NNI obtained, and the formula is as follows:

[0053]

[0054] Where n represents the number of NNI. R peak is an important characteristic peak in the QRS complex, which usually represents the potential change during ventricular depolarization. The QRS complex reflects the entire process of ventricular depolarization. R peak is the first positive peak in the QRS complex, marking the beginning of ventricular depolarization.

[0055] In one embodiment, the frequency domain ECG signal medical features include the power of the high frequency band and the power of the low frequency band. Specifically, for the first frequency domain feature, a fast Fourier transform (FFT) is first applied to the ECG signal data, a time series signal, to obtain the spectrum feature distribution and the frequency domain signal Xf ; Next, extract the power of the high frequency band (HF) and the low frequency band (LF). The formula for the above process is as follows:

[0056]

[0057] F(K)=|X(f)| 2 (3)

[0058]

[0059] HF=sum(F(K)[i]),i∈[0.15hz,0.4hz] (5)

[0060] LF=sum(F(K)[i]),i∈[0.04hz,0.15hz] (6)

[0061] Where X is the electrocardiogram signal, X f is the corresponding frequency domain signal, F(K) is the power spectral density in the frequency domain, which can be calculated using X(f) obtained from FFT. HF is the power of high frequency, ranging from 0.15 Hz to 0.4 Hz. LF is the power of low frequency, ranging from 0.04 Hz to 0.15 Hz.

[0062] In one embodiment, the electrocardiogram signal is represented as X={x 1 ,x 2 ,x 3 …x n}, and define the encoder as E n , define the time domain decoder as D t , the frequency domain decoder is defined as D f The original ECG signal and its corresponding frequency domain signal are used as input, and the time-frequency signal is reconstructed crosswise using an autoencoder. The encoder uses a full-scale convolutional neural network to dynamically adjust the convolution kernel during the feature representation process, while the decoder consists of a multi-layer transposed convolutional neural network.

[0063] Specifically, the electrocardiogram signal X and the corresponding frequency domain signal X f Input to encoder E n In the time domain, we get the low-dimensional embedded signal Z t And the frequency domain low-dimensional embedded signal Z f , and then use the frequency domain decoder D f Based on the low-dimensional embedding signal Z in the time domain t Frequency domain decoder reconstructs the ECG frequency domain signal X f ′, and use the time domain decoder D t Based on the low-dimensional embedding signal Z in the frequency domain fThe electrocardiogram signal X′ is reconstructed using the time domain decoder. The specific process formula is as follows:

[0064] Z t =E n (X) (7)

[0065] Z f =E n (X f ) (8)

[0066] X′ f =D f (Z t ) (9)

[0067] X′=D t (Z f ) (10)

[0068] In one embodiment, in the prediction module, the medical features of the ECG signal in the time domain and frequency domain are predicted by low-dimensional features, thereby enhancing the ECG signal representation capability of the model. The original ECG signal X is used to obtain the values ​​of AFI, HR in the time domain and HF and LF in the frequency domain, and the calculated AFI value is defined as N a The calculated HR value is defined as N hr , the high frequency characteristic HF value is defined as N hf , the low frequency feature LF value is defined as N lf Then, the original ECG signal X and the frequency domain signal are input into the encoder E n , to obtain the low-dimensional embedding Z in the time domain t and frequency domain low-dimensional embedding Z f Subsequently, a multilayer perceptron (MLP) prediction head is used to process Z t Generate the predicted values ​​of time domain feature values ​​AFI and HR, and process Z f Generate the predicted values ​​of the frequency domain eigenvalues ​​HF and LF. The above prediction process can be expressed by the following formula:

[0069] N a ′,N hr ′=P t (Z t ) (11)

[0070] N hf ′,N lf ′=P f (Z f ) (12)

[0071] Among them, N a ′、N hr ′、N hf ′ and Nlf ′ represents the prediction result of ECG medical characteristics, and P t , P f It is a time domain and frequency domain predictor that obtains the predicted values ​​of ECG features from low-dimensional feature embeddings.

[0072] In order to verify the effectiveness of the scheme of the present invention, the present invention also quantitatively evaluates the scheme of the present invention (model framework such as Figure 3 The anomaly detection performance of the proposed method is shown in Table 1. Some experimental details are given below. The results of all compared algorithms on each dataset are shown in Table 1. The effectiveness of the method of the present invention is illustrated by showing the results of the model on the time series anomaly detection task. On the CPSC2021 dataset, compared with other anomaly detection algorithms with fixed network parameters, the AUC and AP indicators of the present invention are the highest, exceeding the optimal comparison algorithm by 6.29% and 9.37%, respectively. On the Icentia11k dataset, when the network parameters are fixed, the AUC and AP indicators of the present invention are 9.22% and 11.94% higher than those of other anomaly detection algorithms, respectively. On the PTBXL dataset, when the network parameters are fixed, the AUC and AP indicators of the present invention are 3.57% and 2.28% higher than those of other anomaly detection algorithms, respectively.

[0073] Table 1 Algorithm performance comparison

[0074]

[0075] The experimental results demonstrate that the proposed scheme is superior to traditional anomaly detection algorithms, showing its practical application potential in clinical settings.

[0076] Based on the same inventive concept, Figure 4 As shown, an embodiment of the present invention further provides a self-supervised atrial fibrillation detection device based on cross-modal cross-reconstruction, comprising: an acquisition module, a data preprocessing module, a cross-modal cross-reconstruction module and a detection module.

[0077] The acquisition module is used to acquire the electrocardiogram signal to be tested; the data preprocessing module is used to perform time-frequency conversion on the electrocardiogram signal to be tested to obtain an electrocardiogram frequency domain signal; the cross-modal cross-reconstruction module is used to input the electrocardiogram signal to be tested and the electrocardiogram frequency domain signal into a preset self-supervised learning model to obtain a reconstructed electrocardiogram signal and a reconstructed electrocardiogram frequency domain signal; the self-supervised learning model is obtained by training based on a normal electrocardiogram signal; the self-supervised learning model includes an encoder, a time domain decoder and a frequency domain decoder; the encoder is used to extract time domain low-dimensional embedding features and frequency domain low-dimensional embedding features from the input electrocardiogram signal and the electrocardiogram frequency domain signal respectively; the time domain decoder is used to generate a reconstructed electrocardiogram signal based on the frequency domain low-dimensional embedding features; the frequency domain decoder is used to generate a reconstructed electrocardiogram frequency domain signal based on the time domain low-dimensional embedding features; the detection module is used to calculate the reconstruction error between the reconstructed electrocardiogram frequency domain signal and the original electrocardiogram frequency domain signal. If the reconstruction error is greater than the preset abnormality score, the electrocardiogram signal to be tested is considered to be atrial fibrillation data.

[0078] In one embodiment, a self-supervised atrial fibrillation detection device based on cross-modal cross-reconstruction provided by an embodiment of the present invention also includes a training module; the training module is used to construct a normal electrocardiogram signal data set; the normal electrocardiogram signal data set and its corresponding original electrocardiogram frequency domain signal data set are input into the self-supervised learning model to generate a reconstructed electrocardiogram signal data set and a reconstructed electrocardiogram frequency domain signal data set; the joint loss is minimized to iteratively optimize the parameters of the self-supervised learning model and save the self-supervised learning model under the optimal parameters; wherein the joint loss is composed of the reconstruction loss between the reconstructed electrocardiogram signal data set and the original electrocardiogram signal data set and the reconstruction loss between the reconstructed electrocardiogram frequency domain signal data set and the original electrocardiogram frequency domain signal data set.

[0079] In one embodiment, in the self-supervised atrial fibrillation detection device based on cross-modal cross-reconstruction provided by an embodiment of the present invention, the training module also includes a time domain predictor and a frequency domain predictor; the time domain predictor is used to predict the medical features of the ECG signal in the time domain based on the time domain low-dimensional embedding features, and the frequency domain predictor is used to predict the medical features of the ECG signal in the frequency domain based on the frequency domain low-dimensional embedding features; correspondingly, the training module is used to use the time domain predictor and the frequency domain predictor as regularizers so as to introduce a regularization term in the joint loss.

[0080] It should be noted that the self-supervised atrial fibrillation detection device based on cross-modal cross-reconstruction provided in each embodiment of the present invention is for the purpose of realizing the above method. Its specific functions can be referred to the above method embodiments, which will not be described in detail here.

[0081] Figure 5 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 5 As shown, the electronic device may include: a processor (processor) 501, a communication interface (Communications Interface) 502, a memory (memory) 503 and a communication bus 504, wherein the processor 501, the communication interface 502, and the memory 503 communicate with each other through the communication bus 504. The processor 501 can call the logic instructions in the memory 503 to execute the self-supervised atrial fibrillation detection method, which includes: obtaining an electrocardiogram signal to be tested; performing time-frequency conversion on the electrocardiogram signal to be tested to obtain an electrocardiogram frequency domain signal; inputting the electrocardiogram signal to be tested and the electrocardiogram frequency domain signal into a preset self-supervised learning model to obtain a reconstructed electrocardiogram signal and a reconstructed electrocardiogram frequency domain signal; the self-supervised learning model is obtained by training based on a normal electrocardiogram signal; the self-supervised learning model includes an encoder, a time domain decoder and a frequency domain decoder; the encoder is used to extract time domain low-dimensional embedding features and frequency domain low-dimensional embedding features from the input electrocardiogram signal and the electrocardiogram frequency domain signal respectively; the time domain decoder is used to generate a reconstructed electrocardiogram signal based on the frequency domain low-dimensional embedding features; the frequency domain decoder is used to generate a reconstructed electrocardiogram frequency domain signal based on the time domain low-dimensional embedding features; the reconstruction error between the reconstructed electrocardiogram frequency domain signal and the original electrocardiogram frequency domain signal is calculated, and if the reconstruction error is greater than a preset abnormality score, the electrocardiogram signal to be tested is considered to be atrial fibrillation data.

[0082] In addition, when the logic instructions in the above-mentioned memory 503 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0083] An embodiment of the present invention also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the self-supervised atrial fibrillation detection method provided by the above-mentioned method embodiments.

[0084] An embodiment of the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the self-supervised atrial fibrillation detection method provided by the above-mentioned method embodiments is implemented.

[0085] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction, characterized in that: include: Acquire the electrocardiogram signal to be tested; Performing time-frequency conversion on the electrocardiogram signal to be measured to obtain an electrocardiogram frequency domain signal; The electrocardiogram signal to be measured and the electrocardiogram frequency domain signal are input into a preset self-supervised learning model to obtain a reconstructed electrocardiogram signal and a reconstructed electrocardiogram frequency domain signal; the self-supervised learning model is obtained by training based on a normal electrocardiogram signal; the self-supervised learning model includes an encoder, a time domain decoder and a frequency domain decoder; the encoder is used to extract time domain low-dimensional embedding features and frequency domain low-dimensional embedding features from the input electrocardiogram signal and the electrocardiogram frequency domain signal respectively; The time domain decoder is used to generate a reconstructed electrocardiogram signal based on the frequency domain low-dimensional embedding features; the frequency domain decoder is used to generate a reconstructed electrocardiogram frequency domain signal based on the time domain low-dimensional embedding features; A reconstruction error between the reconstructed electrocardiogram frequency domain signal and the original electrocardiogram frequency domain signal is calculated. If the reconstruction error is greater than a preset abnormality score, the electrocardiogram signal to be measured is considered to be atrial fibrillation data.

2. The self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction according to claim 1, characterized in that: The training process of the self-supervised learning model includes: Construct a normal ECG signal dataset; Inputting a normal electrocardiogram signal data set and its corresponding original electrocardiogram frequency domain signal data set into the self-supervised learning model to generate a reconstructed electrocardiogram signal data set and a reconstructed electrocardiogram frequency domain signal data set; Minimize the joint loss to iteratively optimize the parameters of the self-supervised learning model and save the self-supervised learning model under the optimal parameters; wherein the joint loss is composed of the reconstruction loss between the reconstructed ECG signal data set and the original ECG signal data set and the reconstruction loss between the reconstructed ECG frequency domain signal data set and the original ECG frequency domain signal data set.

3. The self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction according to claim 2, characterized in that: It also includes a time domain predictor and a frequency domain predictor; the time domain predictor is used to predict the medical features of the electrocardiogram signal in the time domain based on the time domain low-dimensional embedding features, and the frequency domain predictor is used to predict the medical features of the electrocardiogram signal in the frequency domain based on the frequency domain low-dimensional embedding features; Correspondingly, the training process of the self-supervised learning model also includes: The time domain predictor and the frequency domain predictor are used as regularizers so as to introduce a regularization term in the joint loss.

4. The self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction according to claim 3, characterized in that: The time-domain electrocardiogram signal medical features include heart rate and atrial fibrillation index.

5. The self-supervised atrial fibrillation detection method based on cross-modal cross-reconstruction according to claim 3, characterized in that: The electrocardiogram signal medical characteristics in the frequency domain include power in a high frequency band and power in a low frequency band.

6. A self-supervised atrial fibrillation detection device based on cross-modal cross-reconstruction, characterized in that: include: An acquisition module, used for acquiring an electrocardiogram signal to be tested; A data preprocessing module, used for performing time-frequency conversion on the electrocardiogram signal to be measured to obtain an electrocardiogram frequency domain signal; A cross-modal cross-reconstruction module, used for inputting the electrocardiogram signal to be measured and the electrocardiogram frequency domain signal into a preset self-supervised learning model to obtain a reconstructed electrocardiogram signal and a reconstructed electrocardiogram frequency domain signal; the self-supervised learning model is obtained by training based on a normal electrocardiogram signal; the self-supervised learning model includes an encoder, a time domain decoder and a frequency domain decoder; the encoder is used for extracting time domain low-dimensional embedding features and frequency domain low-dimensional embedding features from the input electrocardiogram signal and the electrocardiogram frequency domain signal respectively; The time domain decoder is used to generate a reconstructed electrocardiogram signal based on the frequency domain low-dimensional embedding features; the frequency domain decoder is used to generate a reconstructed electrocardiogram frequency domain signal based on the time domain low-dimensional embedding features; The detection module is used to calculate the reconstruction error between the reconstructed electrocardiogram frequency domain signal and the original electrocardiogram frequency domain signal. If the reconstruction error is greater than a preset abnormality score, the electrocardiogram signal to be tested is considered to be atrial fibrillation data.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method according to any one of claims 1 to 5 is implemented.

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

Images