Cardiovascular disease analysis method and device, electronic equipment and storage medium

Abstract

The present application provides a kind of cardiovascular disease analysis method, device, electronic equipment and storage medium, the cardiovascular disease analysis method includes: in response to analysis request, the time series waveform of vital sign is acquired by near field acquisition;The time series waveform of vital sign is reconstructed by first neural network, and the electrocardiogram waveform is obtained;The time series waveform of vital sign is extracted, and the vital feature is obtained;The vital feature is analyzed by second neural network, and the analysis result of cardiovascular disease is obtained, and the analysis result is verified by electrocardiogram waveform.The present application realizes non-sensing measurement by using non-contact near field acquisition and analysis mode, is suitable for long-term monitoring, is suitable for family health care, is suitable for old-age care institutions, and the implementation mode is simple;Automatic monitoring and diagnosis are realized by the deep learning algorithm of artificial intelligence, and the work burden of medical staff is reduced.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, electronic device, and storage medium for analyzing cardiovascular diseases. Background Technology

[0002] With socio-economic development, changing lifestyles, increased life expectancy, and accelerated population aging and urbanization, the prevalence of cardiovascular health risk factors is evident, and the incidence and mortality rates of cardiovascular diseases continue to rise. Statistics show that over 17 million people die from cardiovascular diseases annually, and the mortality rate is extremely high, accounting for more than 40% of all disease-related deaths, exceeding that of cancer and other diseases, making it the undisputed leading cause of death. However, medical research has found that 80% of heart attacks and strokes are preventable. This means that with proper prevention, 80% of cardiovascular disease patients can avoid death; the prerequisite for prevention is regular checkups.

[0003] Several existing methods for diagnosing cardiovascular diseases include: (1) Electrocardiogram (ECG), which is the simplest and most commonly used method for diagnosing coronary heart disease; (2) Echocardiography, which examines the heart using the Doppler principle; (3) CT scan, which detects the degree of vascular calcification; (4) Radionuclide imaging, which takes pictures of the heart; (5) Myocardial enzyme testing; and (6) Angiography, which is the gold standard for diagnosing coronary heart disease. These methods are characterized by their complexity, requiring a professional physician to perform, making them unsuitable for home use and long-term monitoring. Summary of the Invention

[0004] The main objective of this invention is to provide a cardiovascular disease analysis method, device, electronic device, and storage medium, which improves the efficiency of cardiovascular disease analysis and lowers the barrier to entry for using cardiovascular disease analysis.

[0005] One aspect of the present invention provides a method for analyzing cardiovascular diseases, comprising:

[0006] In response to analysis requests, time-series waveforms of vital signs are acquired using near-field acquisition;

[0007] The time-series waveforms of the vital signs are reconstructed using a first neural network to obtain the electrocardiogram waveform;

[0008] Feature extraction is performed on the time-series waveforms of the vital signs to obtain the vital characteristics;

[0009] The biological characteristics are analyzed through a second neural network to obtain the analysis results of cardiovascular diseases, and the analysis results are verified by the electrocardiogram waveform.

[0010] According to some embodiments of the present invention, the acquisition of time-series waveforms of vital signs via near-field acquisition in response to an analysis request includes:

[0011] The time-series waveforms of the vital signs are acquired in a near-field manner using millimeter-wave radar. These waveforms are used to characterize multiple cardiac cycle waveforms at different locations within the region where the human heart is located. The cardiac cycle waveforms are used to characterize the mechanical vibrations generated by myocardial contraction and relaxation that are sensed by the millimeter-wave radar.

[0012] According to some embodiments of the present invention, the step of reconstructing the time-series waveforms of the vital signs through a first neural network to obtain an electrocardiogram waveform includes:

[0013] The first neural network is configured as a CNN-LSTM network, which is a joint network of a CNN deep convolutional neural network and an LSTM long short-term memory neural network. The CNN-LSTM network sequentially includes a 1D convolutional layer, a ReLU activation layer, a Max pooling layer, a dropout layer, a 1D convolutional layer, a ReLU activation layer, a Max pooling layer, a dropout layer, an LSTM layer, a dropout layer, an LSTM layer, a dropout layer, and a Dense layer. The electrocardiogram waveform reconstructed by the first neural network is the electrical activity of the same heart, which is synchronously acquired with the time-series waveform of the vital signs. The electrocardiogram waveform is acquired by an electrocardiograph.

[0014] According to some embodiments of the present invention, the step of performing feature extraction on the time-series waveforms of the vital signs to obtain vital characteristics includes:

[0015] The time-series waveform of the vital signs is composed of multiple time-series waveforms. Continuous wavelet transform is performed on the multiple time-series waveforms in sequence to obtain multiple first time-frequency feature maps. The multiple first time-frequency feature maps are then cropped to obtain multiple second time-frequency feature maps. The cropping is achieved by extracting the required time-frequency regions from the multiple first time-frequency feature maps to obtain multiple second time-frequency feature maps. The multiple cropped second time-frequency feature maps are used as the vital signs.

[0016] According to some embodiments of the present invention, the calculation process of the continuous wavelet transform is as follows:

[0017] Based on the waveform of the time series of vital signs, a basic wavelet is selected. By changing the scale and position parameters of the basic wavelet, a cluster of wavelet functions is obtained. The wavelet function cluster and the time series waveform of vital signs are used to perform similarity operations to obtain continuous wavelet transform coefficients. The absolute value of the continuous wavelet transform coefficients is calculated to obtain the time-frequency feature map.

[0018] According to some embodiments of the present invention, the step of analyzing the vital characteristics through a second neural network to obtain the analysis results of cardiovascular diseases includes:

[0019] Multiple second time-frequency feature maps are sequentially input into a CNN backbone network and then passed through a Flatten layer to obtain multiple abstract representation vectors of the second time-frequency feature maps;

[0020] Perform a linear mapping process on the multiple abstract representation vectors to obtain multiple embedding vectors;

[0021] The multiple embedding vectors are input into the Transformer encoder model, and then classified by the MLP Head to obtain whether the patient has cardiovascular disease and the type of disease.

[0022] The CNN backbone network is a pre-trained deep convolutional neural network, which is a neural network trained on a large-scale dataset using supervised or self-supervised methods.

[0023] The abstract representation vector is a one-dimensional vector with m columns, the embedding vector is a one-dimensional vector with n columns, and the linear mapping maps the one-dimensional abstract representation vector with m columns to the one-dimensional embedding vector with n columns through linear operations, where m is greater than n, and the linear mapping is a dimension reduction operation;

[0024] The multiple embedded vectors input to the Transformer encoder model have no positional encoding beforehand. The Transformer encoder model is composed of multiple superimposed encoders, which include a multi-head self-attention layer and a feedforward layer.

[0025] According to some embodiments of the present invention, the MLP Head classification output includes at least one of noise, normal, atrial fibrillation, atrial flutter, second-degree atrioventricular block, ventricular bigeminy, third-degree atrioventricular block, ventricular voluntary rhythm, junctional escape rhythm, sinus rhythm, supraventricular tachycardia, trigeminy, and ventricular tachycardia.

[0026] This invention also discloses a cardiovascular disease analysis device, comprising: an acquisition module for acquiring time-series waveforms of vital signs via near-field acquisition according to an analysis request; an electrocardiogram (ECG) waveform reconstruction module for reconstructing the time-series waveforms of the vital signs using a first neural network to obtain ECG waveforms; a feature extraction module for extracting features from the time-series waveforms of the vital signs to obtain vital features; and a cardiovascular disease analysis module for analyzing the vital features using a second neural network to obtain cardiovascular disease analysis results, and verifying the analysis results using the ECG waveforms.

[0027] This invention also discloses an electronic device, including a millimeter-wave radar acquisition unit, a processor, and a memory; the millimeter-wave radar acquisition unit is used to acquire time-series waveforms of vital signs in the near field; the memory is used to store a program; the processor executes the program to implement any of the cardiovascular disease analysis methods described above.

[0028] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the methods described above.

[0029] This invention achieves non-contact measurement through non-contact near-field acquisition and analysis, making it suitable for long-term monitoring, home healthcare, and elderly care facilities, and is simple to implement. Through deep learning in artificial intelligence, it achieves automatic monitoring and diagnosis, reducing the workload of medical staff.

[0030] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0031] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0032] Figure 1 This is a flowchart illustrating the method of an embodiment of the present invention.

[0033] Figure 2 This is a schematic diagram of the internal structure of the method according to an embodiment of the present invention.

[0034] Figure 3 This is a schematic diagram of millimeter-wave near-field acquisition according to an embodiment of the present invention.

[0035] Figure 4 This is a schematic diagram of the CNN-LSTM neural network for reconstructing an electrocardiogram according to an embodiment of the present invention.

[0036] Figure 5 This is a schematic diagram of life feature extraction according to an embodiment of the present invention.

[0037] Figure 6 This is a schematic diagram of the neural network for the cardiovascular disease analysis method according to an embodiment of the present invention.

[0038] Figure 7 This is a diagram of a device for treating cardiovascular diseases according to an embodiment of the present invention.

[0039] Figure 8 This is a block diagram of an electronic device for cardiovascular diseases according to an embodiment of the present invention. Detailed Implementation

[0040] The embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings. Throughout the description, the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions. In the following description, suffixes such as "module," "part," or "unit" used to denote elements are used only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, "module," "part," or "unit" can be used interchangeably. Terms such as "first," "second," etc., are used only to distinguish technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the sequential relationship of the indicated technical features. In the following description, the consecutive reference numerals for method steps are for ease of review and understanding. Adjusting the implementation order of steps, in conjunction with the overall technical solution of the present invention and the logical relationship between the various steps, will not affect the technical effect achieved by the technical solution of the present invention. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0041] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0042] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0043] Figure 1 This is a flowchart illustrating a method according to an embodiment of the present invention. It includes:

[0044] S100, in response to an analysis request, acquires time-series waveforms of vital signs using near-field acquisition;

[0045] S200 reconstructs the time-series waveforms of vital signs using the first neural network to obtain the electrocardiogram waveform;

[0046] S300 performs feature extraction on the time-series waveforms of vital signs to obtain the characteristics of the living organism;

[0047] S400 analyzes the characteristics of living organisms through a second neural network to obtain analysis results for cardiovascular diseases, and verifies the analysis results through electrocardiogram waveforms.

[0048] In some embodiments, millimeter-wave radar senses the mechanical vibrations of the human heart using millimeter waves.

[0049] In some embodiments, millimeter waves are emitted in a near-field acquisition method, reach a region where the human heart is located, and are reflected back. Multiple vital sign time-series waveforms are obtained by analyzing and calculating the echoes reflected from different locations within the region. These vital sign time-series waveforms are cardiac cycle waveforms. Cardiac cycle waves are mechanical vibrations generated by the contraction and relaxation of the myocardium. These multiple vital sign time-series waveforms carry information about multiple angles of the heart's mechanical vibrations.

[0050] In some embodiments, multiple vital sign time-series waveforms are acquired and input into a deep neural network algorithm to determine whether cardiovascular disease is present and its type. Deep neural networks are an effective method for analyzing complex data, but training deep neural network algorithms requires a large amount of data and has poor interpretability. Therefore, this invention proposes a CNN-LSTM deep neural network for reconstructing electrocardiograms (ECGs) using vital sign time-series waveforms monitored by millimeter-wave radar. The input to the CNN-LSTM deep neural network is the monitored vital sign time-series waveform, and the output is the ECG. The amount of information carried by the monitored vital sign time-series waveform is evaluated by reconstructing the ECG. Higher accuracy of the reconstruction method indicates a greater amount of information carried by the monitored vital sign time-series waveform. Complete reconstruction means that the amount of information carried by the vital sign time-series waveform is equivalent to that of the ECG. Since the ECG is a primary means of diagnosing cardiovascular diseases, this reconstruction method assesses the upper limit of cardiovascular disease diagnosis using monitored vital sign time-series waveforms and guides the optimization and improvement of deep neural network algorithms for cardiovascular disease analysis.

[0051] Figure 2This is a schematic diagram of the internal structure of the method according to an embodiment of the present invention. Vital sign time-series waveforms are acquired through near-field acquisition, wherein the vital sign time-series waveforms include multiple time-series waveforms [a1,a2,a3,…]. The multiple vital sign time-series waveforms [a1,a2,a3,…] are input into a CNN-LSTM network to obtain a reconstructed electrocardiogram waveform. Continuous wavelet transforms are sequentially performed on the multiple time-series waveforms [a1,a2,a3,…] to obtain multiple first time-frequency feature maps [b1,b2,b3,…]. The multiple first time-frequency feature maps [b1,b2,b3,…] are cropped to obtain multiple second time-frequency feature maps [c1,c2,c3,…]. The multiple second time-frequency feature maps [c1,c2,c3,…] are sequentially input into a CNN. The backbone network then obtains abstract representation vectors [d1, d2, d3, ...] of multiple time-frequency feature maps through the Flatten layer; these abstract representation vectors [d1, d2, d3, ...] are linearly mapped to obtain multiple embedding vectors [e1, e2, e3, ...]; these embedding vectors [e1, e2, e3, ...] are input into the Transformer encoder model, and then classified by the MLP Head to determine whether the patient has cardiovascular disease and the type of disease. In some embodiments, the internal structure of the CNN backbone network can be customized.

[0052] Figure 3This is a schematic diagram of millimeter-wave near-field acquisition according to an embodiment of the present invention. Millimeter waves (mmWave) are a special type of radar technology that uses short-wavelength electromagnetic waves. Millimeter-wave radar can emit signals with wavelengths on the order of millimeters. This wavelength is considered short in the electromagnetic spectrum, which is one of the advantages of this technology. Indeed, the system components (such as antennas) required to process millimeter-wave signals are very small. Another advantage of short wavelengths is high accuracy. Millimeter-wave systems operating at frequencies of 76–81 GHz (corresponding to a wavelength of approximately 4 mm) can detect movements as small as a few tenths of a millimeter. The heartbeat is a periodic movement generated by the contraction and relaxation of the myocardium; there are fluctuations of 0.1 to 0.5 mm in the chest around the heart and 0.01 to 0.2 mm in the back around the heart. Clearly, millimeter-wave radar can detect these minute changes. Studies have found that the correlation between the timing waveforms of vital signs monitored by millimeter-wave radar and electrocardiograms is as high as 92%. Embodiments of this invention utilize millimeter-wave radar sensors to acquire vital sign time-series waveforms. These waveforms are then input into a deep neural network to determine whether the patient has cardiovascular disease and the type of disease. This invention also proposes a CNN-LSTM deep neural network for reconstructing electrocardiograms (ECGs) using vital sign time-series waveforms monitored by millimeter-wave radar. While an ECG records the electrical activity of the heart, vital sign time-series waveforms record its mechanical movement; both carry information about cardiac activity. ECGs are a primary means of diagnosing cardiovascular diseases, and the vital sign time-series waveforms acquired by millimeter-wave radar will become another method for diagnosing cardiovascular diseases. Compared to ECGs, the vital sign time-series waveforms acquired by millimeter-wave radar allow for non-contact measurement, making them suitable for long-term monitoring, home healthcare, and elderly care facilities. Furthermore, through deep learning in artificial intelligence, automated monitoring and diagnosis are achieved, reducing the workload of medical personnel.

[0053] Figure 4This is a schematic diagram of the CNN-LSTM neural network for reconstructing an electrocardiogram (ECG) according to an embodiment of the present invention. The CNN-LSTM network is a joint network of a CNN (Deep Convolutional Neural Network) and an LSTM (Long Short-Term Memory Neural Network). The CNN-LSTM network sequentially includes a 1D convolutional layer, a ReLU activation layer, a Max pooling layer, a dropout layer, another 1D convolutional layer, a ReLU activation layer, a Max pooling layer, a dropout layer, an LSTM layer, a dropout layer, an LSTM layer, a dropout layer, and a Dense layer. The ECG waveform reconstructed by the CNN-LSTM neural network is the electrical activity of the same heart, synchronously acquired with the time-series waveforms of vital signs. The ECG waveforms are acquired using an electrocardiograph (ECG). The training data of the CNN-LSTM network is organized as (train_x, train_y), where train_x represents multiple time-series waveforms of vital signs sampled by millimeter-wave radar, and train_y represents the ECG waveforms sampled by the ECG. The CNN-LSTM network relies on the CNN convolutional neural network to extract features and on the LSTM long short-term memory neural network to handle long-short-term dependencies. The internal structure of the CNN-LSTM network can be customized.

[0054] Figure 5 This is a schematic diagram of vital sign feature extraction according to an embodiment of the present invention. The time-series waveform of vital signs consists of multiple time-series waveforms. Continuous wavelet transform is performed sequentially on these multiple time-series waveforms to obtain multiple first time-frequency feature maps. These first time-frequency feature maps are then cropped to obtain multiple second time-frequency feature maps. The cropping step extracts the desired time-frequency region from the multiple first time-frequency feature maps to obtain multiple second time-frequency feature maps. Cropping is a filtering method, and the multiple second time-frequency feature maps represent vital sign features. Filtering algorithms in signal processing are processes that attenuate noise and enhance useful signals. Continuous wavelet transform converts the time-series waveform of vital signs to the time-frequency domain, and cropping the time-frequency feature maps extracts the time-frequency region containing the useful signal to achieve the filtering purpose.

[0055] In some embodiments, the calculation process of continuous wavelet transform is as follows: A basic wavelet is selected based on the waveform characteristics of the time-series waveform of vital signs. A wavelet function family is obtained by continuously changing the scale and position parameters of the basic wavelet. The wavelet function family and the time-series waveform of vital signs are used to perform similarity operations to obtain continuous wavelet transform coefficients. The absolute values ​​of the continuous wavelet transform coefficients are calculated to obtain the time-frequency feature map. Wavelet transform is essentially a correlation operation between the original signal and the scaled wavelet function family. By adjusting the scale, wavelets with different time-frequency widths can be obtained to match different positions of the original signal, achieving localized signal analysis. The wavelet transform window is an adjustable time-frequency window; a short window is used at high frequencies, and a wide window is used at low frequencies. Unlike the short-time Fourier transform, wavelet transform can better resolve the contradiction between time and frequency resolution.

[0056] Figure 6 This is a schematic diagram of the neural network for the cardiovascular disease analysis method according to an embodiment of the present invention. Multiple second time-frequency feature maps are sequentially input into a CNN backbone network and then passed through a Flatten layer to obtain multiple abstract representation vectors of the second time-frequency feature maps. Linear mapping is performed on the multiple abstract representation vectors to obtain multiple embedding vectors. These multiple embedding vectors are then input into a Transformer encoder model, and finally classified by an MLP Head to determine whether the patient has cardiovascular disease and the type of disease.

[0057] In some embodiments, the CNN backbone network is a pre-trained deep convolutional neural network, which is a neural network trained on a large-scale dataset using supervised or self-supervised methods. Training deep neural networks requires a large amount of data, and the collection and labeling of data consumes a lot of resources. Pre-trained deep convolutional neural networks are a means to reduce development costs and are reusable.

[0058] In some embodiments, the abstract representation vector is a one-dimensional vector with m columns, and the embedding vector is a one-dimensional vector with n columns. The linear mapping maps the one-dimensional abstract representation vector with m columns to the one-dimensional embedding vector with n columns through linear operations, where m is greater than n. The linear mapping is a dimensionality reduction operation. There is a dimensionality matching problem between the output of the CNN backbone network and the input of the Transformer encoder model. The purpose of the linear mapping is to achieve dimensionality matching.

[0059] In some embodiments, the multiple embedding vectors are not encoded at position before being input into the Transformer encoder model. The Transformer encoder model is composed of multiple stacked encoders, which include multi-head self-attention layers and feedforward layers. The Transformer encoder model is based on the attention mechanism, which can solve the problem of long-distance dependencies, meaning that the attention mechanism has better memory capabilities. In addition, attention supports parallel computation. The multiple embedding vectors are abstract features of multiple vital sign time-series waveforms. There is no temporal dependency between the multiple vital sign time-series waveforms, so there is no position encoding before they are input into the Transformer encoder model. The Transformer encoder model relies on the attention mechanism to extract the common attributes among the multiple vital sign time-series waveforms, and then performs classification through an MLP Head.

[0060] In some embodiments, the MLP Head classification output includes at least one of the following: noise, normal, atrial fibrillation, atrial flutter, second-degree atrioventricular block, ventricular bigeminy, third-degree atrioventricular block, ventricular voluntary rhythm, junctional escape rhythm, sinus rhythm, supraventricular tachycardia, trigeminy, and ventricular tachycardia. The above cardiovascular diseases are merely examples; the analysis results obtained through the technical solution of the cardiovascular analysis method of this invention are all within the protection scope of this invention. The correlation rate between the vital signs time-series waveforms monitored by millimeter-wave radar and electrocardiograms is as high as 92%, and the types of cardiovascular diseases obtained through electrocardiograms can also be obtained through the analysis of vital signs time-series waveforms monitored by millimeter-wave radar.

[0061] Figure 7 This is a diagram of a cardiovascular disease device according to an embodiment of the present invention. The device includes an acquisition module 701, an electrocardiogram waveform reconstruction module 702, a feature extraction module 703, and a cardiovascular disease analysis module 704.

[0062] The acquisition module is used to acquire the time-series waveforms of vital signs through near-field acquisition according to the analysis request; the electrocardiogram waveform reconstruction module reconstructs the time-series waveforms of vital signs through a first neural network to obtain the electrocardiogram waveforms; the feature extraction module extracts features from the time-series waveforms of vital signs to obtain vital features; the cardiovascular disease analysis module analyzes the vital features through a second neural network to obtain the analysis results of cardiovascular diseases, and verifies the analysis results through the electrocardiogram waveforms.

[0063] Figure 8 This is a block diagram of an electronic device for cardiovascular disease according to an embodiment of the present invention. The electronic device includes a millimeter-wave radar acquisition unit, a processor, and a memory; the millimeter-wave radar acquisition unit is used to acquire time-series waveforms of vital signs in the near field; the memory stores a program; the processor executes the program to perform the aforementioned cardiovascular disease analysis method; the electronic device has the function of carrying and running a software system for the cardiovascular disease analysis method provided in the embodiment of the present invention.

[0064] This invention also provides a computer-readable storage medium storing a program that is executed by a processor to implement the cardiovascular disease analysis method described above.

[0065] In some alternative embodiments, the functions / operations mentioned in the block diagrams may not occur in the order shown in the operation diagrams. For example, depending on the functions / operations involved, two consecutively shown blocks may actually be executed substantially simultaneously, or the blocks may sometimes be executed in reverse order. Furthermore, the embodiments presented and described in the flowcharts of this invention are provided by way of example to provide a more comprehensive understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is altered and sub-operations described as part of a larger operation are executed independently.

[0066] This invention also discloses a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium and execute the computer instructions, causing the computer device to perform the aforementioned cardiovascular disease analysis method.

[0067] Furthermore, although the invention has been described in the context of functional modules, it should be understood that, unless otherwise stated, one or more of the described functions and / or features may be integrated into a single physical device and / or software module, or one or more functions and / or features may be implemented in a separate physical device or software module. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the invention. Rather, given the properties, functions, and internal relationships of the various functional modules in the apparatus disclosed herein, the actual implementation of the module will be understood within the scope of conventional skill of an engineer. Therefore, those skilled in the art can implement the invention as set forth in the claims using ordinary techniques without excessive experimentation. It is also understood that the specific concepts disclosed are merely illustrative and not intended to limit the scope of the invention, which is determined by the full scope of the appended claims and their equivalents.

[0068] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0069] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0070] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0071] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0072] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0073] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

[0074] The above is a detailed description of the preferred embodiments of the present invention, but the present invention is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A cardiovascular disease analysis method characterized by, The method comprises the following steps: in response to an analysis request, acquiring a time waveform of vital signs by near-field acquisition; specifically, emitting millimeter waves in a near-field acquisition mode to a region where the heart is located in the human body, and obtaining a plurality of time waveforms of vital signs by analyzing and calculating the echoes reflected from different positions in the region, wherein the time waveform of vital signs is a heart cycle waveform, and the plurality of time waveforms of vital signs carry information of multiple angles of mechanical vibration of the heart; reconstructing the time waveform of vital signs by a first neural network to obtain an electrocardiogram waveform; specifically, the first neural network is configured as a CNN-LSTM network, which is a joint network of a CNN deep convolutional neural network and an LSTM long short-term memory neural network, and comprises a 1D convolutional layer, a ReLU activation layer, a Max pooling layer, a dropout layer, a 1D convolutional layer, a ReLU activation layer, a Max pooling layer, a dropout layer, an LSTM layer, a dropout layer, an LSTM layer, a dropout layer, and a Dense layer in sequence, the electrocardiogram waveform reconstructed by the first neural network is the electrical activity of the same heart collected synchronously with the time waveform of vital signs, and the electrocardiogram waveform is collected by an electrocardiograph device; performing feature extraction on the time waveform of vital signs to obtain vital features; analyzing the vital features by a second neural network to obtain an analysis result of cardiovascular diseases, and verifying the analysis result by the electrocardiogram waveform; wherein the amount of information carried by the time waveform of vital signs is detected by reconstructing the electrocardiogram waveform.

2. The cardiovascular disease analysis method according to claim 1, characterized by, The feature extraction on the time waveform of vital signs to obtain vital features comprises: The time waveform of vital signs is composed of a plurality of time waveforms, and a plurality of first time-frequency feature maps are obtained by sequentially performing continuous wavelet transform on the plurality of time waveforms, a plurality of second time-frequency feature maps are obtained by cropping the plurality of first time-frequency feature maps, and the plurality of second time-frequency feature maps obtained by cropping are used as the vital features.

3. The cardiovascular disease analysis method according to claim 2, characterized by, The calculation process of the continuous wavelet transform comprises: selecting a basic wavelet according to the waveform of the time waveform of vital signs, obtaining a wavelet function cluster by changing the scale parameter and the position parameter of the basic wavelet, performing similarity operation on the wavelet function cluster and the time waveform of vital signs to obtain a continuous wavelet transform coefficient, and calculating the absolute value of the continuous wavelet transform coefficient to obtain a time-frequency feature map.

4. The cardiovascular disease analysis method according to claim 2, characterized by, The analysis of the vital features by the second neural network to obtain the analysis result of cardiovascular diseases comprises: inputting the plurality of second time-frequency feature maps into a CNN backbone network in sequence, and then obtaining a plurality of abstract representation vectors of the plurality of second time-frequency feature maps through a Flatten layer; performing linear mapping processing on the plurality of abstract representation vectors to obtain a plurality of embedding vectors; The plurality of embedding vectors are input into a Transformer encoder model, and then classified by an MLP Head to obtain whether the subject has a cardiovascular disease and the type of the disease; The CNN backbone network is a pre-trained deep convolutional neural network, which is a neural network trained on a large-scale dataset in a supervised or self-supervised manner; The abstract representation vector is a one-dimensional m-column vector, the embedding vector is a one-dimensional n-column vector, and the linear mapping maps the one-dimensional m-column abstract representation vector to the one-dimensional n-column embedding vector through linear operation, wherein m is greater than n, and the linear mapping is a dimension reduction operation; The plurality of embedding vectors are input into a Transformer encoder model, and then classified by an MLP Head to obtain whether the subject has a cardiovascular disease and the type of the disease; 5. The cardiovascular disease analysis method according to claim 4, characterized by, The MLP Head classification output includes at least one of noise, normal, atrial fibrillation, atrial flutter, second-degree atrioventricular block, ventricular bigeminy, third-degree atrioventricular block, ventricular automatic rhythm, junctional escape rhythm, sinus rhythm, supraventricular tachycardia, trigeminy, and ventricular tachycardia.

6. A cardiovascular disease analysis device, characterized by, Comprise: The acquisition module is configured to acquire time series waveforms of vital signs by near-field acquisition according to an analysis request; specifically, millimeter waves are emitted by near-field acquisition to a region where a heart is located, and the reflected waves from different positions in the region are analyzed and calculated to obtain a plurality of time series waveforms of vital signs, wherein the time series waveforms of vital signs are heart cycle waveforms, and the plurality of time series waveforms of vital signs carry information of multiple angles of mechanical vibration of the heart; The ECG waveform reconstruction module reconstructs the time series waveforms of vital signs by a first neural network to obtain an ECG waveform; specifically, the first neural network is configured as a CNN-LSTM network, which is a joint network of a CNN deep convolutional neural network and an LSTM long short-term memory neural network, and sequentially includes a 1D convolution layer, a ReLU activation layer, a Max pooling layer, a dropout layer, a 1D convolution layer, a ReLU activation layer, a Max pooling layer, a dropout layer, an LSTM layer, a dropout layer, an LSTM layer, a dropout layer, and a Dense layer; the ECG waveform reconstructed by the first neural network is the electrical activity of the same heart collected synchronously with the time series waveforms of vital signs, and the ECG waveform is collected by an electrocardiograph device; The feature extraction module extracts features from the time series waveforms of vital signs to obtain vital features; The cardiovascular disease analysis module analyzes the vital sign through a second neural network to obtain an analysis result of the cardiovascular disease, and verifies the analysis result through the electrocardiogram waveform; wherein information carried by the vital sign time sequence waveform is detected through reconstruction of the electrocardiogram waveform.

7. An electronic device, comprising: The system comprises a millimeter wave radar collector, a processor and a memory; The millimeter wave radar collector is configured to acquire the time sequence waveform of the vital sign in near field collection; The memory is configured to store a program; The processor executes the program to implement the cardiovascular disease analysis method according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The storage medium stores a program, and the program is executed by the processor to implement the cardiovascular disease analysis method according to any one of claims 1-5.

Images