Method, system and storage medium for early screening of myocardial ischemia based on heart rate variability

Abstract

The application discloses a myocardial ischemia early screening method and system based on heart rate variability and a storage medium, and relates to the technical field of medical detection, and aims to provide a myocardial ischemia early screening method and system based on heart rate variability, which can be widely applied to the medical detection technical field.

Description

Technical Field

[0001] This invention relates to the field of medical testing technology, and in particular to a method, system, and storage medium for early screening of myocardial ischemia based on heart rate variability. Background Technology

[0002] Cardiovascular disease has long threatened human health and is the leading cause of death worldwide. Mental stress is a significant risk factor for the development and worsening of cardiovascular disease. Transient myocardial ischemia caused by psychological stress in daily life and work is clinically known as mental stress-induced myocardial ischemia (MSIMI). With increasing work and life pressures, sudden death due to excessive mental stress or fatigue has become increasingly frequent in recent years, representing a huge loss to families and society. MSIMI often presents clinically as transient chest pain, which is difficult for patients to notice. Timely warning and diagnosis at a hospital are crucial for improving patient prognosis. Because it does not involve organic changes in the heart, it is difficult to detect with routine examinations. Severe cases often present with chronic chest pain, yet remain difficult to diagnose. Currently, overall understanding of this disease is limited, and the clinical detection rate is low. However, this disease can induce acute myocardial ischemia under sudden mental stress. Sudden myocardial ischemia causes electrical activity disturbances in the heart, leading to malignant arrhythmias (mostly ventricular fibrillation). If the patient does not receive timely cardiopulmonary resuscitation or resuscitation fails, it can even lead to sudden death. Therefore, early screening for the disease is crucial. Excessive mental stress is a major contributing factor; however, there is currently a lack of early screening technologies for this disease.

[0003] Meanwhile, the incidence of MSIMI is as high as 50-70% in cardiovascular disease patients undergoing standardized psychological stress challenges. Cardiovascular disease patients with MSIMI cannot effectively prevent heart disease with drug treatment alone; their risk of adverse cardiac events increases, clinical prognosis worsens, and the risk of death increases exponentially. Therefore, routine screening for cardiovascular disease patients is also crucial. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a simple and efficient method, system and storage medium for early screening of myocardial ischemia based on heart rate variability.

[0005] In a first aspect, embodiments of the present invention provide a method for early screening of myocardial ischemia based on heart rate variability, comprising:

[0006] Acquire strong modal feature data and weak modal feature data; wherein, the strong modal feature data includes PET / CT data, and the weak modal data includes electrocardiogram data, psychological scale data, and basic information data;

[0007] Strong modal feature extraction is performed on the strong modal feature data to construct a strong modal pre-trained model;

[0008] The weak modal feature data is preprocessed and feature-filtered to construct a weak modal diagnostic model;

[0009] The strong modality pre-trained model is embedded into the weak modality diagnostic model, and the strong modality reuse assists the weak modality in feature mining.

[0010] By using a knockout strategy, the connections of the strong modality pre-trained model and the dependencies of the strong modality feature data are removed, and a myocardial ischemia early warning model is constructed.

[0011] Early screening and warning of myocardial ischemia are performed using the aforementioned myocardial ischemia early warning model.

[0012] Optionally, the method further includes a preliminary data preparation step, which includes:

[0013] Patients with symptoms of angina or myocardial ischemia within a predetermined age range were screened.

[0014] Target patients who meet the preset criteria are selected to construct a cohort of patients with stress-induced myocardial ischemia.

[0015] The preset conditions include non-obstructive coronary artery disease and a total cardiac score difference reaching a threshold; the mental stress-induced myocardial ischemia patient cohort is used for the collection of strong modality feature data and weak modality feature data.

[0016] Optionally, acquiring strong modal feature data and weak modal feature data includes:

[0017] Collect psychological scale data and basic information data;

[0018] And, collect electrocardiogram data corresponding to PET / CT data during the stress period.

[0019] Optionally, the step of extracting strong modal features from the strong modal feature data and constructing a strong modal pre-trained model includes:

[0020] The electrocardiogram data is preprocessed and subjected to first feature screening to construct a first weak modality diagnostic model;

[0021] And, perform a second feature screening on the psychological scale data to construct a second weak modality diagnostic model;

[0022] And, perform third feature filtering on the basic information data to construct a third weak modality diagnostic model.

[0023] Optionally, the step of preprocessing the electrocardiogram data and performing a first feature screening to construct a first weak modality diagnostic model includes:

[0024] The electrocardiogram data is subjected to smoothing filtering, bandpass filtering, and segmentation to obtain the first preprocessed data;

[0025] HRV feature calculation, feature autocorrelation calculation and redundant feature removal are performed on the first preprocessed data to construct a first weak modality diagnostic model;

[0026] The HRV feature calculation involves time-domain features, frequency-domain features, and nonlinear features, and the first weak modality diagnostic model is a multiple linear regression model.

[0027] Optionally, in the step of extracting strong modal features from the strong modal feature data and constructing a strong modal pre-training model, the second feature selection and the third feature selection include feature extraction, calculation of correlation between different features and removal of redundant features, and both the second weak modal diagnostic model and the third weak modal diagnostic model are multiple linear regression models.

[0028] Optionally, it also includes a weight propagation step, which includes:

[0029] The weight parameters are updated through backpropagation using the loss function;

[0030] The expression for the loss function is:

[0031]

[0032] in, Indicates input sample x i The feature output representation, l(x) i ,z i ,y i ) represents the prediction loss function. express The mapping function to the label space, where h represents the normalization indicator operation, g(z) i ) represents the strong modal feature z i The mapping function to the marked space, L reg λ represents the regularization term, and λ is used to balance the loss function and the regularization term.

[0033] Secondly, embodiments of the present invention provide a myocardial ischemia early screening system based on heart rate variability, comprising:

[0034] The first module is used to acquire strong modal feature data and weak modal feature data; wherein, the strong modal feature data includes PET / CT data, and the weak modal data includes electrocardiogram data, psychological scale data, and basic information data;

[0035] The second module is used to extract strong modal features from the strong modal feature data and construct a strong modal pre-training model.

[0036] The third module is used to preprocess and filter the weak modal feature data to construct a weak modal diagnostic model.

[0037] The fourth module is used to embed the strong modality pre-trained model into the weak modality diagnostic model, and to assist the weak modality in feature mining through strong modality reuse;

[0038] The fifth module is used to construct a myocardial ischemia early warning model by removing the connections of the strong modality pre-trained model and the dependencies of the strong modality feature data through a knockout strategy.

[0039] The sixth module is used for early screening and warning of myocardial ischemia through the aforementioned myocardial ischemia early warning model.

[0040] Thirdly, embodiments of the present invention provide an electronic device, including a processor and a memory;

[0041] The memory is used to store programs;

[0042] The processor executes the program to implement the method described above.

[0043] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a program that is executed by a processor to implement the method described above.

[0044] 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 method.

[0045] The embodiments of the present invention first acquire strong modality feature data and weak modality feature data; wherein, the strong modality feature data includes PET / CT data, and the weak modality data includes electrocardiogram data, psychological scale data, and basic information data; then, strong modality features are extracted from the strong modality feature data to construct a strong modality pre-trained model; then, the weak modality feature data is preprocessed and feature-filtered to construct a weak modality diagnostic model; then, the strong modality pre-trained model is embedded into the weak modality diagnostic model, and feature mining is performed by reusing strong modalities to assist weak modalities; subsequently, a knockout strategy is used to remove the connections of the strong modality pre-trained model and the dependencies of the strong modality feature data to construct a myocardial ischemia early warning model; finally, the myocardial ischemia early warning model is used for early screening and early warning of myocardial ischemia. This invention utilizes patient myocardial perfusion PET / CT data as strong modal data to assist in the mining of weak modal data such as electrocardiogram features, psychological assessment scale features, and basic information features. Through subsequent knockout of strong modal features, it constructs a myocardial ischemia early warning model that relies solely on data that patients can obtain daily (electrocardiogram data, psychological assessment scale, and basic information). Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 A schematic diagram of the overall steps of an early screening method for myocardial ischemia based on heart rate variability provided in an embodiment of the present invention;

[0048] Figure 2 A schematic diagram of the MSIMI cohort inclusion and exclusion criteria and its grouping diagnostic criteria provided in this embodiment of the invention;

[0049] Figure 3 This is a schematic diagram of the MSIMI queue data acquisition process provided in an embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram illustrating the principle and flow of the early screening method for myocardial ischemia based on heart rate variability provided in an embodiment of the present invention.

[0051] Figure 5 A schematic diagram illustrating the principle of weight propagation provided for an embodiment of the present invention;

[0052] Figure 6 A schematic diagram illustrating the principle of knockout measurement provided in an embodiment of the present invention;

[0053] Figure 7A graph showing the performance comparison of different data models provided in the embodiments of the present invention. Detailed Implementation

[0054] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0055] To address the problems existing in the prior art, firstly, this invention provides a method for early screening of myocardial ischemia based on heart rate variability, such as... Figure 1 As shown, the method includes the following steps:

[0056] Acquire strong modal feature data and weak modal feature data; wherein, the strong modal feature data includes PET / CT data, and the weak modal data includes electrocardiogram data, psychological scale data, and basic information data;

[0057] Strong modal feature extraction is performed on the strong modal feature data to construct a strong modal pre-trained model;

[0058] The weak modal feature data is preprocessed and feature-filtered to construct a weak modal diagnostic model;

[0059] The strong modality pre-trained model is embedded into the weak modality diagnostic model, and the strong modality reuse assists the weak modality in feature mining.

[0060] By using a knockout strategy, the connections of the strong modality pre-trained model and the dependencies of the strong modality feature data are removed, and a myocardial ischemia early warning model is constructed.

[0061] Early screening and warning of myocardial ischemia are performed using the aforementioned myocardial ischemia early warning model.

[0062] Optionally, the method further includes a preliminary data preparation step, which includes:

[0063] Patients with symptoms of angina or myocardial ischemia within a predetermined age range were screened.

[0064] Target patients who meet the preset criteria are selected to construct a cohort of patients with stress-induced myocardial ischemia.

[0065] The preset conditions include non-obstructive coronary artery disease and a total cardiac score difference reaching a threshold; the mental stress-induced myocardial ischemia patient cohort is used for the collection of strong modality feature data and weak modality feature data.

[0066] Optionally, acquiring strong modal feature data and weak modal feature data includes:

[0067] Collect psychological scale data and basic information data;

[0068] And, collect electrocardiogram data corresponding to PET / CT data during the stress period.

[0069] Optionally, the step of extracting strong modal features from the strong modal feature data and constructing a strong modal pre-trained model includes:

[0070] The electrocardiogram data is preprocessed and subjected to first feature screening to construct a first weak modality diagnostic model;

[0071] And, perform a second feature screening on the psychological scale data to construct a second weak modality diagnostic model;

[0072] And, perform third feature filtering on the basic information data to construct a third weak modality diagnostic model.

[0073] Optionally, the step of preprocessing the electrocardiogram data and performing a first feature screening to construct a first weak modality diagnostic model includes:

[0074] The electrocardiogram data is subjected to smoothing filtering, bandpass filtering, and segmentation to obtain the first preprocessed data;

[0075] HRV feature calculation, feature autocorrelation calculation and redundant feature removal are performed on the first preprocessed data to construct a first weak modality diagnostic model;

[0076] The HRV feature calculation involves time-domain features, frequency-domain features, and nonlinear features, and the first weak modality diagnostic model is a multiple linear regression model.

[0077] Optionally, in the step of extracting strong modal features from the strong modal feature data and constructing a strong modal pre-training model, the second feature selection and the third feature selection include feature extraction, calculation of correlation between different features and removal of redundant features, and both the second weak modal diagnostic model and the third weak modal diagnostic model are multiple linear regression models.

[0078] Optionally, it also includes a weight propagation step, which includes:

[0079] The weight parameters are updated through backpropagation using the loss function;

[0080] The expression for the loss function is:

[0081]

[0082] in, Indicates input sample x i The feature output representation, l(x) i ,z i ,yi ) represents the prediction loss function. express The mapping function to the label space, where h represents the normalization indicator operation, g(z) i ) represents the strong modal feature z i The mapping function to the marked space, L reg λ represents the regularization term, and λ is used to balance the loss function and the regularization term.

[0083] Secondly, embodiments of the present invention provide a myocardial ischemia early screening system based on heart rate variability, comprising:

[0084] The first module is used to acquire strong modal feature data and weak modal feature data; wherein, the strong modal feature data includes PET / CT data, and the weak modal data includes electrocardiogram data, psychological scale data, and basic information data;

[0085] The second module is used to extract strong modal features from the strong modal feature data and construct a strong modal pre-training model.

[0086] The third module is used to preprocess and filter the weak modal feature data to construct a weak modal diagnostic model.

[0087] The fourth module is used to embed the strong modality pre-trained model into the weak modality diagnostic model, and to assist the weak modality in feature mining through strong modality reuse;

[0088] The fifth module is used to construct a myocardial ischemia early warning model by removing the connections of the strong modality pre-trained model and the dependencies of the strong modality feature data through a knockout strategy.

[0089] The sixth module is used for early screening and warning of myocardial ischemia through the aforementioned myocardial ischemia early warning model.

[0090] The content of the method embodiments of the present invention is applicable to the system embodiments. The specific functions implemented in the system embodiments are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above methods.

[0091] Another aspect of the present invention provides an electronic device, including a processor and a memory;

[0092] The memory is used to store programs;

[0093] The processor executes the program to implement the method described above.

[0094] The content of the method embodiments of the present invention is applicable to the embodiments of the present electronic device. The specific functions implemented by the embodiments of the present electronic device are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above methods.

[0095] Another aspect of this invention provides a computer-readable storage medium storing a program that is executed by a processor to implement the methods described above.

[0096] The content of the method embodiments of the present invention is applicable to the computer-readable storage medium embodiments. The specific functions implemented by the computer-readable storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above methods.

[0097] 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 method.

[0098] The implementation principle of the present invention for early screening of myocardial ischemia based on heart rate variability is described in detail below:

[0099] This invention is developed based on MSIMI cohort data. It uses patient myocardial perfusion PET / CT data as strong modality data to assist weak modality data HRV features, psychological assessment scale features, and basic information features (including personal history, family history, etc.) mining. Through subsequent strong modality feature knockout, it constructs a myocardial ischemia early warning model that relies solely on data that patients can obtain daily (ECG data, psychological assessment scale, basic information).

[0100] Preliminary Data Preparation: This invention targets individuals experiencing chest pain due to mental and psychological stress, aiming to differentiate between chest pain and ischemia. Therefore, it is necessary to extract characteristics of this population for the construction and validation of an early warning model, leading to the creation of a cohort of patients with stress-induced myocardial ischemia (MSIMI). Inclusion criteria were patients aged 18-75 years with angina or symptoms related to myocardial ischemia. The primary exclusion criterion was the presence of coronary artery obstruction or other cardiovascular-related organic changes or serious diseases that prevented cooperation with data collection; detailed criteria are detailed in [link to relevant documentation]. Figure 2Patients diagnosed with chest pain due to non-obstructive coronary artery disease (ANOCA) completed basic information such as personal history, medical history, and family history. Following this, the patients underwent controlled stress testing (12 minutes of mental arithmetic, public speaking, and color card identification tasks), 13NH3 PET / CT cardiac perfusion examination, and electrocardiogram examination (data acquisition process as follows). Figure 2 A total cardiac difference score (SDS) greater than or equal to 3 is diagnosed as myocardial ischemia, i.e., MSIMI patients. Among MSIMI patients, a persistent decrease in myocardial blood flow (MBF) after mental stress has a poor prognosis. Therefore, further risk classification is performed for MSIMI patients.

[0101] It should be noted that queue data, such as Figure 3 As shown, feature analysis was performed using electrocardiogram data corresponding to PET / CT data during the stress period.

[0102] The steps of this invention for achieving early screening of myocardial ischemia based on heart rate variability are as follows:

[0103] (1) As Figure 4 As shown, preprocessing, feature selection, and model construction were performed on electrocardiogram data, psychological scale data, and basic information data, respectively. Simultaneously, strong modality reuse from PET / CT was used to assist in the mining of weak modality features, followed by a stepwise knockout approach to remove strong modality features. Finally, the three weak modality models were integrated through ensemble learning to complete the diagnostic algorithm.

[0104] (2) In practical applications, the present invention mainly collects three types of data: electrocardiogram data, psychological stress scale data, and basic information data.

[0105] ①Basic information: Basic information mainly includes a description of personal condition, such as whether there is chest pain, frequency of pain, personal history, family history, etc. After the basic information is filled in, it can be supplemented and changed after other diseases are diagnosed.

[0106] ② Psychological scale completion: The psychological scale adopts a staged assessment and can be re-completed when the individual feels the need for a change in status.

[0107] ③ ECG data is collected using a smart bracelet, and an abnormal ECG signal is monitored using a 5-minute sliding window mode in the algorithm application.

[0108] The Poincaré diagram analyzes dynamic HRV data by moving a 5-minute window forward every 5 seconds, enabling real-time monitoring and timely early warning.

[0109] The core technology of this invention is the use of strong modal features from PET / CT scans to assist in the mining of weak modal features. A detailed description follows:

[0110] By reusing a deep network framework with strong modality models, a pre-trained strong modality model for a relevant task is embedded into a weak modality data diagnostic model structure. The deep network connects the strong modality model and the labeled information in parallel. During training, the strong modality model and target lesion-specific information are simultaneously used as supervision to optimize network parameters, providing more supervision information for the target domain network training and mitigating the error accumulation problem caused by segmented training to some extent. A knockout technique is also proposed. Considering that only weak modality information from the target domain is used for prediction during the testing phase, the training phase needs to be synchronized with the testing phase. Therefore, the training phase needs to gradually reduce the dependence on the strong modality model, i.e., gradually weaken the connection between the strong modality model and the current deep network. Discriminant information can be gradually migrated from the strong modality model to the weak modality model, achieving consistency between the training and testing networks, meaning that only weak modality input is needed during the testing phase.

[0111] The principle of using heart rate variability and PET / CT as weak and strong modes is explained in detail below:

[0112] Weight Propagation: Considering that weak modal data only contains feature labels that have already been mapped, the weight propagation step focuses on using the strong modality model to assist training and reduce the prediction loss of the weak modality network in the target domain. This step is a general weight update method with two parallel structures. The feature output of the weak modality neural network consists of fully connected label information, denoted as W, and a parallel fully connected strong modality model, denoted as U. The label information of strong and weak modalities belongs to the same label space, so the model Vp constructed from the strong modality input sample space and labels represents the strong modality model, where Vp is obtained through pre-training. The overall loss function can be expressed as:

[0113]

[0114] in:

[0115]

[0116] Indicates input sample x i The feature output representation, l(x) i ,z i ,y i ) represents the prediction loss function. express The mapping function to the label space, where h represents the normalization indicator operation, g(z) i ) represents the strong modal feature z iThe mapping function to the label space, i.e., the strong modal model, L reg λ represents the regularization term, and λ is used to balance the loss function and the regularization term.

[0117] like Figure 5 As shown, weight propagation uses strong modality and label information as supervision information to update the network parameters of the weak modality model. U is the mapping from the weak modality output features to the strong modality feature space and participates in the network weight update. The weight propagation step calculates the partial derivatives of the network parameters, i.e., θ, W, and U, and updates the parameters through backpropagation.

[0118] like Figure 6 As shown, the knockout strategy involves gradually weakening the connection between the weak modality network feature outputs and the strong modality model during training, and completely disconnecting them from the strong modality model after training. U is the mapping matrix from the weak modality output features (heart rate variability features) to the strong modality feature (PET / CT) space, and it participates in network weight updates. The knockout step randomly deletes some strong modality feature nodes z in each training round. i，j1 , z i，j2 ... and the corresponding connections, i.e., UU in matrix U. i，j1 U i，j2 U i，j3 ... are constrained to 0 after the nodes are knocked out. The knocked-out connections will not be used again in subsequent training. After training, all strong modal feature input layers z are completely disconnected from the deep network, thus eliminating the need for a strong modal model Vp.

[0119] For the knockout strategy, taking HRV as an example, experimental verification was conducted, such as... Figure 7 As shown, the performance of different models is demonstrated. The model with PET / CT as a strong modality-assisted HRV feature extraction has improved performance compared to the model that only uses HRV features.

[0120] In summary, for weak modal data such as electrocardiograms, echocardiograms, mental health assessment scales, biochemical tests, patient medical history, and environmental data, knockout can gradually eliminate the correlation between the reused model and the current deep structure. While the method is similar to Dropout, the purpose and effect are completely different. It is worth noting that in the training phase, Dropout nodes are reset in the next iteration, and all nodes in the testing phase also exist in the neural network, which is inconsistent with the setting that there is no strong modal model in the testing phase. In contrast, with the knockout strategy, once a node association is broken by the knockout operation, the connection will never function again. Therefore, after training, the connection between the target domain network and the strong modal model is completely broken, while the discriminative ability of the strong modal model has been gradually transferred to the currently learned deep network during training. This invention is better suited for MSIMI early screening diagnostic models when strong modal data is missing.

[0121] 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.

[0122] 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.

[0123] 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.

[0124] 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.

[0125] 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.

[0126] 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.

[0127] 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.

[0128] 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.

[0129] 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 method for early screening of myocardial ischemia based on heart rate variability, characterized in that, include: Acquire strong modal feature data and weak modal feature data; wherein, the strong modal feature data includes PET / CT data, and the weak modal feature data includes electrocardiogram data, psychological scale data, and basic information data; Strong modal feature extraction is performed on the strong modal feature data to construct a strong modal pre-trained model; The weak modal feature data is preprocessed and feature-filtered to construct a weak modal diagnostic model; The strong modality pre-trained model is embedded into the weak modality diagnostic model, and the strong modality reuse assists the weak modality in feature mining. A myocardial ischemia early warning model is constructed by stripping the connections of the strong modality pre-trained model and the dependencies of the strong modality feature data through a knockout strategy. The knockout strategy involves gradually weakening the connections between the weak modality network feature outputs and the strong modality model during training, and completely disconnecting them after training. U is a mapping matrix from the weak modality output features to the strong modality feature space, participating in network weight updates. The knockout step randomly deletes some strong modality feature nodes and their corresponding connections in each training round; that is, the corresponding element values ​​in matrix U are restricted to 0 after the nodes are knocked out. The knocked-out connections will not be used in subsequent training. After training, all strong modality feature input layers are completely disconnected from the deep network, thus eliminating the need for a strong modality model. Early screening and early warning of myocardial ischemia are performed using the aforementioned myocardial ischemia early warning model; This also includes a weight propagation step, which includes: The weight parameters are updated through backpropagation using the loss function; The expression for the loss function is: in, , Indicates input sample The feature output representation, Represents the prediction loss function. express Mapping function to the marked space, Indicates the normalization instruction operation. Represents strong modal features Mapping function to the marked space, Represents the regularization term. Used to balance the loss function and the regularization term.

2. The method for early screening of myocardial ischemia based on heart rate variability according to claim 1, characterized in that, This also includes the preliminary data preparation step, which includes: Patients with symptoms of angina or myocardial ischemia within a predetermined age range were screened. Target patients who meet the preset criteria are selected to construct a cohort of patients with stress-induced myocardial ischemia. The preset conditions include non-obstructive coronary artery disease and a total cardiac score difference reaching a threshold; the mental stress-induced myocardial ischemia patient cohort is used for the collection of strong modality feature data and weak modality feature data.

3. The method for early screening of myocardial ischemia based on heart rate variability according to claim 1, characterized in that, The acquisition of strong modality feature data and weak modality feature data includes: Collect psychological scale data and basic information data; And, collect electrocardiogram data corresponding to PET / CT data during the stress period.

4. The method for early screening of myocardial ischemia based on heart rate variability according to claim 1, characterized in that, The step of preprocessing and feature filtering the weak modal feature data to construct a weak modal diagnostic model includes: The electrocardiogram data is preprocessed and subjected to first feature screening to construct a first weak modality diagnostic model; And, perform a second feature screening on the psychological scale data to construct a second weak modality diagnostic model; And, perform third feature filtering on the basic information data to construct a third weak modality diagnostic model.

5. The method for early screening of myocardial ischemia based on heart rate variability according to claim 4, characterized in that, The step of preprocessing and first feature screening of the electrocardiogram data to construct a first weak modality diagnostic model includes: The electrocardiogram data is subjected to smoothing filtering, bandpass filtering, and segmentation to obtain the first preprocessed data; HRV feature calculation, feature autocorrelation calculation and redundant feature removal are performed on the first preprocessed data to construct a first weak modality diagnostic model; The HRV feature calculation involves time-domain features, frequency-domain features, and nonlinear features, and the first weak modality diagnostic model is a multiple linear regression model.

6. The method for early screening of myocardial ischemia based on heart rate variability according to claim 4, characterized in that, In the step of preprocessing and feature filtering the weak modal feature data to construct a weak modal diagnostic model, the second feature filtering and the third feature filtering include feature extraction, correlation calculation of different features and removal of redundant features. Both the second weak modal diagnostic model and the third weak modal diagnostic model are multiple linear regression models.

7. A myocardial ischemia early screening system based on heart rate variability, characterized in that, The system for implementing the method of claim 1 includes: The first module is used to acquire strong modal feature data and weak modal feature data; wherein, the strong modal feature data includes PET / CT data, and the weak modal feature data includes electrocardiogram data, psychological scale data, and basic information data; The second module is used to extract strong modal features from the strong modal feature data and construct a strong modal pre-training model. The third module is used to preprocess and filter the weak modal feature data to construct a weak modal diagnostic model. The fourth module is used to embed the strong modality pre-trained model into the weak modality diagnostic model, and to assist the weak modality in feature mining through strong modality reuse; The fifth module is used to construct a myocardial ischemia early warning model by removing the connections of the strong modality pre-trained model and the dependencies of the strong modality feature data through a knockout strategy. The sixth module is used for early screening and warning of myocardial ischemia through the aforementioned myocardial ischemia early warning model.

8. An electronic device, characterized in that, Including the processor and memory; The memory is used to store programs; The processor executes the program to implement the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The storage medium stores a program that is executed by a processor to implement the method as described in any one of claims 1 to 6.

Images