Arrhythmia sudden death risk prediction method, device, storage medium and equipment

Abstract

This application discloses a method, device, storage medium, and equipment for predicting the risk of sudden cardiac death due to arrhythmia, relating to the field of biomedical technology. The method includes: acquiring multi-channel cardiac magnetic signals from a subject; performing fidelity enhancement processing on the multi-channel cardiac magnetic signals to extract core features characterizing the cardiac magnetic field; constructing a topological geometric model of the cardiac fiber plexus; extracting topological invariants of the cardiac electrical conduction system based on the core features and the topological geometric model; conducting non-equilibrium phase transition dynamics detection on the topological geometric model of the cardiac fiber plexus to obtain phase transition features reflecting the stability of the cardiac electrical conduction system; inputting the topological invariants and phase transition features into a preset topological neural network to complete a quantitative analysis of the degree of chaos in the cardiac electrical conduction system; and generating a risk score and early warning indicators for sudden cardiac death due to arrhythmia based on the quantitative analysis results. This application can achieve accurate early warning of the risk of sudden cardiac death due to arrhythmia.

Description

Technical Field

[0001] This application relates to the field of biomedical technology, and in particular to a method, device, storage medium and equipment for predicting the risk of sudden cardiac death due to arrhythmia. Background Technology

[0002] Sudden cardiac death due to arrhythmia is one of the most fatal emergencies in the field of cardiovascular disease. Its insidious onset and rapid progression have made it a major cause of death worldwide. With the development of medical diagnostic technology, the detection of cardiac magnetic biomarkers, due to its advantages such as being non-contact, having high spatiotemporal resolution, and reflecting raw information about cardiac electrical activity, has gradually become an important research direction for predicting the risk of sudden cardiac death due to arrhythmia.

[0003] In existing technologies, most methods for predicting the risk of sudden cardiac death due to arrhythmia are based on classical electromagnetic theory frameworks for analyzing cardiac magnetic signals. Generally, cardiac magnetic signals are collected using conventional magnetometers, and after simple filtering and noise reduction, feature parameters with limited dimensions such as magnetic field amplitude, frequency, and time-domain waveform are extracted. At the same time, a simplified representation of the heart structure is adopted using a Euclidean geometric model, assuming the heart to be a regular ellipsoid and ignoring its complex spiral muscle fiber structure and dynamic deformation characteristics. Finally, a risk assessment model is constructed by combining traditional machine learning algorithms to achieve a preliminary judgment of the risk of sudden cardiac death due to arrhythmia.

[0004] However, existing technologies cannot effectively capture the characteristics of the cardiac magnetic field and the topological nature of the electrical conduction system, and lack precise quantification of the non-equilibrium phase transition process and chaotic dynamics of the heart. On the one hand, conventional signal processing methods ignore key information such as quantum coherence and magnetic vortices contained in cardiac magnetic signals, making it difficult to extract features that reflect the deep-seated laws of cardiac electrical activity. On the other hand, Euclidean geometry modeling cannot characterize the topological invariants of the cardiac electrical conduction system and has poor adaptability to the dynamic deformation of the heart during contraction and relaxation. At the same time, traditional models do not analyze the stability of the cardiac system from the perspective of non-equilibrium thermodynamics and cannot quantify the degree of chaos in the system. Ultimately, this results in short early warning time, low individualization, and high false positive rate in risk prediction, making it difficult to meet the actual needs of early and accurate identification of high-risk patients in clinical practice. Summary of the Invention

[0005] In view of this, this application provides a method, device, storage medium and equipment for predicting the risk of sudden cardiac death due to arrhythmia, which can realize accurate early warning of the risk of sudden cardiac death due to arrhythmia.

[0006] According to a first aspect of this application, a method for predicting the risk of sudden cardiac death due to arrhythmia is provided, comprising:

[0007] Multi-channel magnetic field signals of the subject were collected, and the multi-channel magnetic field signals were subjected to fidelity enhancement processing to extract core features that can characterize the magnetic field properties of the heart.

[0008] A topological geometric model of the cardiac fiber tract is constructed, and topological invariants of the cardiac electrical conduction system are extracted based on the core features and the topological geometric model of the cardiac fiber tract.

[0009] Non-equilibrium phase transition dynamics were detected on the cardiac fiber plexus topological geometric model to obtain phase transition characteristics reflecting the stability of the cardiac electrical conduction system;

[0010] The topological invariants and phase transition features are input into a preset topological neural network to complete the quantitative analysis of the degree of chaos in the cardiac electrical conduction system.

[0011] Based on the quantitative analysis results, risk scores and early warning indicators for sudden cardiac death due to arrhythmia were generated for the subjects.

[0012] According to a second aspect of this application, a device for predicting the risk of sudden cardiac death due to arrhythmia is provided, comprising:

[0013] The processing module is used to acquire multi-channel magnetic-cardiogram signals from the subject, perform fidelity enhancement processing on the multi-channel magnetic-cardiogram signals, and extract core features that can characterize the magnetic field properties of the heart.

[0014] The extraction module is used to construct a topological geometric model of the cardiac fiber tract and extract topological invariants of the cardiac electrical conduction system based on the core features and the topological geometric model of the cardiac fiber tract.

[0015] The detection module is used to perform non-equilibrium phase transition dynamics detection on the cardiac fiber bundle topological geometry model to obtain phase transition characteristics that reflect the stability of the cardiac electrical conduction system.

[0016] The input module is used to input the topological invariants and the phase transition features into a preset topological neural network to complete the quantitative analysis of the degree of chaos in the cardiac electrical conduction system.

[0017] The generation module is used to generate risk scores and early warning indicators for sudden cardiac death due to arrhythmia for the subjects based on the quantitative analysis results.

[0018] According to a third aspect of this application, a storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for predicting the risk of sudden cardiac death due to arrhythmia.

[0019] According to a fourth aspect of this application, an electronic device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor executes the program to implement the above-described method for predicting the risk of sudden cardiac death due to arrhythmia.

[0020] By employing the aforementioned technical solutions, this application provides a method, device, storage medium, and equipment for predicting the risk of sudden cardiac death due to arrhythmia. Through acquiring multi-channel cardiac magnetic signals and performing fidelity enhancement processing, it can accurately extract core features of the cardiac magnetic field, such as quantum coherence and magnetic vortices, overcoming the limitations of signal processing within the framework of traditional classical electromagnetic theory. By constructing a topological geometric model of the cardiac fiber plexus and extracting topological invariants, it can effectively characterize the topological essence of the cardiac electrical conduction system, solving the problem that Euclidean geometry cannot adapt to dynamic cardiac deformation. By conducting non-equilibrium phase transition dynamics detection on the cardiac fiber plexus topological geometric model and combining it with a topological neural network to quantify the degree of chaos in the cardiac electrical conduction system, it can achieve accurate quantification of the stability and chaotic characteristics of the cardiac system. The resulting individualized risk score and early warning index for sudden cardiac death due to arrhythmia can significantly extend the early warning time, improve the individualization of predictions, reduce the false positive rate, and meet the practical needs of early and accurate identification of high-risk patients in clinical practice.

[0021] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0022] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0023] Figure 1 A flowchart illustrating a method for predicting the risk of sudden cardiac death due to arrhythmia provided in an embodiment of this application is shown.

[0024] Figure 2 This illustration shows a schematic diagram of a network architecture for a topological neural network provided in an embodiment of this application;

[0025] Figure 3 A flowchart illustrating a method for predicting the risk of sudden cardiac death due to arrhythmia, according to another embodiment of this application, is shown.

[0026] Figure 4 This illustration shows a schematic diagram of the principle and process of predicting the risk of sudden cardiac death due to arrhythmia, provided in an embodiment of this application.

[0027] Figure 5 This illustration shows a schematic diagram of the structure of a device for predicting the risk of sudden cardiac death due to arrhythmia, provided in an embodiment of this application.

[0028] Figure 6 A schematic diagram of the structure of a device for predicting the risk of sudden cardiac death due to arrhythmia provided in another embodiment of this application is shown. Detailed Implementation

[0029] The present application will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the embodiments of the present application can be combined with each other.

[0030] Current technologies cannot effectively capture the characteristics of the cardiac magnetic field and the topological nature of the electrical conduction system, and lack precise quantification of the non-equilibrium phase transition processes and chaotic dynamics of the heart. On the one hand, conventional signal processing methods ignore key information such as quantum coherence and magnetic vortices contained in cardiac magnetic signals, making it difficult to extract features that reflect the deep-seated laws of cardiac electrical activity. On the other hand, Euclidean geometry modeling cannot characterize the topological invariants of the cardiac electrical conduction system and has poor adaptability to the dynamic deformation of the heart during contraction and relaxation. At the same time, traditional models do not analyze the stability of the cardiac system from the perspective of non-equilibrium thermodynamics and cannot quantify the degree of chaos in the system. Ultimately, this results in short early warning time, low individualization, and high false positive rate in risk prediction, making it difficult to meet the actual needs of early and accurate identification of high-risk patients in clinical practice.

[0031] To address the aforementioned technical problems, embodiments of the present invention provide a method for predicting the risk of sudden cardiac death due to arrhythmia, such as... Figure 1 As shown, the method includes:

[0032] Step 110: Collect multi-channel magnetic field signals from the subject, perform fidelity enhancement processing on the multi-channel magnetic field signals, and extract core features that can characterize the magnetic field properties of the heart.

[0033] Among them, the multi-channel magnetocardiogram (MCC) signal is a dynamic change signal of the magnetic field corresponding to the electrophysiological activity of the heart, which is synchronously acquired by a high-sensitivity multi-channel magnetic sensing device and covers the spatiotemporal distribution information of the magnetic field in both resting and excited states; the fidelity enhancement processing is a process that addresses environmental interference and quantum decoherence issues during the acquisition of MCC signals by correcting signal distortion and suppressing noise through a dedicated algorithm, maximizing the preservation of the original physiological and quantum characteristics of the signal; the characteristics of the cardiac magnetic field are the inherent topological structure, quantum coherence state and dynamic evolution law of the cardiac magnetic field, which are the essential attributes reflecting the electromechanical function and electrophysiological stability of the myocardium; the core features are key quantitative indicators that can accurately characterize the characteristics of the cardiac magnetic field, including topological charge, quantum coherence, decoherence time and other characteristic parameters that have both topological essence and quantum properties.

[0034] In this embodiment of the disclosure, a high-sensitivity multi-channel superconducting quantum interference array can be used to simultaneously acquire multi-channel magnetocardiogram signals from the subject in both resting and excited states under low-interference conditions. The decoherence channel is estimated using quantum process tomography, and a compensation strategy is designed to correct the quantum decoherence effect of the signal. Simultaneously, adaptive filtering and other methods are used to suppress environmental magnetic field noise and surface interference, achieving signal fidelity enhancement. The processed signal is converted into a differential form of the magnetic field for standardization. Integer magnetic topological charges are obtained through topological charge density calculation and integration. Simultaneously, the quantum fluctuation power spectrum is analyzed, order parameters and quantum coherence are calculated, and the decoherence time is estimated. Finally, core features that comprehensively characterize the topological structure and quantum coherence properties of the cardiac magnetic field are extracted.

[0035] This technical step can break through the limitations of signal analysis under the traditional classical electromagnetic theory framework. By accurately acquiring multi-channel cardiac magnetic signals containing quantum characteristics and implementing targeted fidelity enhancement processing, it effectively avoids the loss of effective signal information due to quantum decoherence. It successfully extracts core features that can reflect the deep quantum laws of cardiac electrical activity, filling the gap that traditional methods cannot capture the characteristics of cardiac magnetic fields. This provides basic data support with unique identification and physiological correlation for the subsequent construction of a high-precision arrhythmia and sudden death risk prediction model.

[0036] Step 120: Construct a topological geometric model of the cardiac fiber tract and extract topological invariants of the cardiac electrical conduction system based on the core features and the topological geometric model of the cardiac fiber tract.

[0037] Among them, the cardiac fiber plexus topological geometric model is a mathematical model of the cardiac electrical conduction system constructed based on fiber plexus theory and differential geometry. It uses the three-dimensional manifold of the heart as the base space and the U (1) phase space as the fiber. It can accurately characterize the structure of the cardiac spiral muscle fibers, dynamic deformation characteristics, and phase distribution law in the electrical conduction process. It can break through the simplified assumptions of the heart structure in the traditional Euclidean geometric model. The topological invariants of the cardiac electrical conduction system refer to the topological attribute quantification parameters that remain stable during the dynamic deformation process of the heart such as contraction and relaxation. Their values ​​are not affected by the continuous changes in the geometric shape of the heart. They can reflect the essential topological characteristics of the cardiac electrical conduction system. Typical parameters include the first Chern class value representing the total number of topological charges of the heart and the Pontryagin class value representing the degree of fiber plexus twisting.

[0038] In this embodiment of the disclosure, the base space of the fiber tract can be determined based on the three-dimensional anatomical structure of the heart. The topological geometric model of the cardiac fiber tract is constructed using the U (1) phase space as the fiber. The core features representing the magnetic field characteristics of the heart extracted in the previous stage are substituted into the model. The curvature tensor is obtained by performing external differential operation on the connection form of the model through calculation. Then, the first Chern class value is calculated based on the curvature tensor. At the same time, the Pontryagin class value representing the degree of twisting of the fiber tract is calculated. The two classes of values ​​are integrated as topological invariants that can reflect the essential topological properties of the cardiac electrical conduction system. The whole process can realize the accurate transformation from quantum cardiac magnetic features to the inherent topological properties of the cardiac electrical conduction system.

[0039] This technique overcomes the limitations of traditional Euclidean geometric models in simplifying the complex structure of the heart. By using a fiber bundle topological geometric model, it achieves precise characterization of the spiral muscle fibers and dynamic deformation of the heart. The extracted topological invariants remain stable during the dynamic changes of cardiac contraction and relaxation. This solves the key problem that traditional feature parameters are sensitive to changes in cardiac morphology. It can provide topological input parameters with essential discriminative characteristics for subsequent non-equilibrium phase transition dynamics detection and topological neural network quantitative analysis, laying the core theoretical and data foundation for accurately assessing the risk of sudden cardiac death due to arrhythmia.

[0040] Step 130: Conduct non-equilibrium phase transition dynamics detection on the cardiac fiber plexus topological geometric model to obtain phase transition characteristics that reflect the stability of the cardiac electrical conduction system.

[0041] Among them, non-equilibrium phase transition dynamics detection is a dynamic analysis method for cardiac electrical conduction systems far from thermodynamic equilibrium. It analyzes the phase transition law and dynamic behavior of the system near the critical point by calculating the entropy generation rate, verifying the fluctuation theorem, detecting critical slowing phenomena, and judging multistable characteristics. The stability of cardiac electrical conduction system refers to the ability of cardiac electrical conduction system to resist external disturbances and internal fluctuations, maintain normal electrical signal conduction order, and avoid arrhythmia. Its core is directly related to the phase transition state and the degree of chaos of the system. Phase transition characteristics are a set of quantitative parameters that can reflect the phase transition law and stability state of cardiac electrical conduction system, mainly including entropy generation rate, critical slowing parameter, potential well depth, and corresponding multistable characteristics.

[0042] In this embodiment of the disclosure, thermodynamic data of the cardiac electrical conduction system characterized by the constructed cardiac fiber tract topological geometric model can be extracted. Based on this data, a suitable stochastic thermodynamic analysis framework can be built to calculate the entropy generation rate of the system and verify the detailed fluctuation theorem. Then, dynamic analysis is carried out on the order parameters of the system to detect the variation law of the order parameters near the critical point. The relaxation time of the system is calculated and its correlation with the system state is fitted by dynamic scaling theory to identify the critical slowing parameter. Then, the effective potential energy function of the cardiac electrical conduction system is reconstructed by combining the entropy generation rate, the results of the dynamic analysis of the order parameters, the relaxation time, and the critical slowing parameter. Based on the effective potential energy function, the potential well depth is calculated and the multi-stable characteristics of the system are judged. Finally, the above key parameters are integrated to form a phase transition characteristic that can reflect the stability of the cardiac electrical conduction system.

[0043] This technology overcomes the theoretical limitations of traditional methods that treat the cardiac system as an equilibrium state. By conducting non-equilibrium phase transition dynamics detection based on the topological geometric model of the cardiac fiber bundles, it can accurately capture the phase transition laws and deep dynamic characteristics of the cardiac electrical conduction system under conditions far from equilibrium. The extracted phase transition features can comprehensively characterize the stability state of the system from both thermodynamic and dynamic dimensions. This solves the problem that traditional feature parameters are difficult to reflect the essence of the system's dynamic changes. It provides key dynamic support for subsequent quantification of the chaos level of the cardiac electrical conduction system through topological neural networks, generation of individualized arrhythmia and sudden death risk scores, and early warning indicators.

[0044] Step 140: Input the topological invariants and phase transition features into the preset topological neural network to complete the quantitative analysis of the degree of chaos in the cardiac electrical conduction system.

[0045] Among them, the topological neural network is a network model built by integrating topological data analysis and deep learning technology. It can adapt to the topological structure of the cardiac electrical conduction system, process deep feature data such as topological invariants and phase transition characteristics, complete node feature updates through message passing mechanism, and output analysis results that satisfy the covariant characteristics of canonical transformation. The degree of chaos in the cardiac electrical conduction system is a quantitative indicator describing the nonlinear dynamic behavior of the cardiac electrical conduction system. It reflects the trend of the system changing from a normal and ordered electrical signal conduction state to a disordered arrhythmia state. It can be accurately characterized by key parameters such as the maximum Lyapunov exponent and correlation dimension.

[0046] In specific application scenarios, such as Figure 2As shown, the pre-defined topological neural network can include an input layer, a graph convolutional layer, a manifold convolutional layer, and a topological attention mechanism. The input layer, as the network's data entry point, is responsible for receiving persistent landscape features characterizing the topological properties of the cardiac electrical conduction system, phase transition features reflecting the system's non-equilibrium phase transition laws, and a cardiac conduction graph constructed with myocardial regions as nodes and the electrical conduction connection strength between regions as edges, providing the initial data foundation for subsequent feature processing layers. The graph convolutional layer consists of two layers: GCN Layer 1 and GCN Layer 2. GCN Layer 1 aggregates the association information of nodes in the cardiac conduction graph through a message passing mechanism, completing the initial update of node features. Layer 2 further optimizes the initially updated features, strengthening the correlation between features. The manifold convolutional layer consists of two layers, Manifold Convolution Layer 1 and Manifold Convolution Layer 2, using hot kernel convolution as the core processing method. This not only adapts to the manifold structure characteristics of the cardiac electrical conduction system but also ensures that the feature processing meets the covariance requirements of canonical transformation, achieving precise feature optimization. The topological attention mechanism dynamically adjusts the weight ratio of different features, making the network more focused on the feature information that is more critical to quantifying the degree of chaos in the cardiac electrical conduction system. The layers work together to integrate and optimize the input features, and finally output high-dimensional features that can be used for chaos analysis.

[0047] In this embodiment of the disclosure, topological invariants reflecting the inherent topological properties of the cardiac electrical conduction system can be transformed into persistent landscape features L, and phase transition features Φ representing the non-equilibrium phase transition law of the system, together with the input layer cardiac conduction graph constructed with myocardial regions as nodes and the electrical conduction connection strength between regions as edges, denoted by the symbol G, are input into a pre-constructed topological neural network. The execution order of the topological neural network is as follows: first, L, Φ, and G are sent from the input layer to the core layer, and the features are updated by GCNLayer1 through message passing to aggregate node information. Then, the features are further optimized by GCN Layer2. Subsequently, the features are processed by manifold convolution Layer1 and Layer2 in the form of hot kernel convolution to ensure the canonical covariance properties. At the same time, the feature weights are adjusted by relying on the topological attention mechanism to complete the topological integration, feature enhancement, and interactive update and optimization of node features of the two types of input features. Finally, based on the high-dimensional feature vector H obtained from the output layer, the nonlinear dynamic model of the system is reconstructed, the phase space is reconstructed in sequence, strange attractors are obtained, and then the key indicators such as the maximum Lyapunov exponent and the correlation dimension D2 marked in the figure are calculated, thereby completing the accurate quantitative analysis of the degree of chaos of the cardiac electrical conduction system.

[0048] This technical step organically integrates the static topological properties of topological invariants with the dynamic phase transition laws of phase transition features. By leveraging the topological adaptability and feature learning capabilities of topological neural networks, it can overcome the limitations of traditional machine learning that relies solely on superficial physiological features. It can effectively capture the deep dynamic characteristics of the cardiac electrical conduction system, achieve precise quantification of the system's degree of chaos, and provide core dynamic-level evidence for subsequent assessment and early warning of arrhythmia and sudden cardiac death risks.

[0049] Step 150: Based on the quantitative analysis results, generate risk scores and early warning indicators for sudden cardiac death due to arrhythmia for the examinee.

[0050] The quantitative analysis results are a set of quantitative parameters that characterize the nonlinear dynamic behavior of the cardiac electrical conduction system, obtained by analyzing the topological invariants and phase transition characteristics of the cardiac electrical conduction system through a topological neural network. Key indicators reflecting the degree of chaos in the system include the maximum Lyapunov exponent, correlation dimension, and fractal characteristics of strange attractors. The arrhythmia-induced sudden death risk score is an individualized quantitative score that reflects the risk level of sudden cardiac death in the subject, generated by integrating multi-dimensional features such as the quantitative parameters of the degree of chaos in the cardiac electrical conduction system, topological properties, and phase transition laws. The score is positively correlated with the risk level. Early warning indicators are core parameters extracted from multi-dimensional features that can characterize the abnormal risk state of the cardiac electrical conduction system in advance, serving as a key basis for risk warning. Typical indicators include chaotic edge distance and topological vulnerability index.

[0051] In this embodiment of the disclosure, based on the quantitative analysis results of the degree of chaos in the cardiac electrical conduction system in the early stage, and combined with multi-dimensional core parameters such as topological invariants that can reflect the inherent topological properties of the system and phase transition characteristics that characterize the stability of the system, various features are organically integrated through a multi-modal feature fusion architecture. Then, an appropriate risk assessment model is constructed to perform in-depth analysis and calculation on the integrated high-dimensional features, so as to obtain an individualized quantitative score that can accurately reflect the degree of risk of sudden cardiac death due to arrhythmia in the examinee. At the same time, key parameters that can intuitively characterize the risk status of the system are extracted, and finally a set of subject arrhythmia sudden cardiac death risk scoring and early warning index system with both quantitative assessment and early warning functions is formed.

[0052] This technology overcomes the limitations of traditional arrhythmia and sudden cardiac death risk assessment, which relies on superficial physiological indicators. By organically integrating the quantification parameters of the chaotic degree of the cardiac electrical conduction system, topological invariants, phase transition characteristics, and other deep-seated essential features, it achieves a precise characterization of the subject's risk status. The generated individualized risk score can clearly distinguish different levels of risk groups, and the extracted early warning indicators can capture abnormal changes in the system in advance. This can effectively extend the risk warning time, reduce the false positive rate of traditional assessment methods, and provide a scientific and reliable quantitative basis for early clinical identification of high-risk patients and the development of targeted intervention plans.

[0053] It should be noted that the technical solution in this application is not directly applied to the clinical diagnosis or treatment of subjects. Its core is based on multi-channel cardiac magnetic signal acquisition, cardiac fiber tract topological geometry modeling, non-equilibrium phase transition dynamics detection, and topological neural network analysis to extract multi-dimensional biomarkers that can reflect the quantum characteristics, topological properties, and degree of chaos of the cardiac electrical conduction system, thereby generating risk scores and early warning indicators for arrhythmic sudden death in subjects. This technical solution is only used to provide data support for scientific research related to cardiac electrophysiology, or to provide reference information for risk stratification and health management of healthy people, and does not directly give diagnostic conclusions for specific diseases.

[0054] In summary, the arrhythmia-induced sudden cardiac death risk prediction method provided by this invention can accurately extract core features of the cardiac magnetic field, such as quantum coherence and magnetic vortices, by acquiring multi-channel cardiac magnetic signals and performing fidelity enhancement processing, thus overcoming the limitations of signal processing under the traditional classical electromagnetic theory framework. By constructing a cardiac fiber tract topological geometric model and extracting topological invariants, the topological essence of the cardiac electrical conduction system can be effectively characterized, solving the problem that the Euclidean geometric model cannot adapt to the dynamic deformation of the heart. By conducting non-equilibrium phase transition dynamics detection on the cardiac fiber tract topological geometric model and combining it with a topological neural network to quantify the degree of chaos in the cardiac electrical conduction system, the stability and chaotic characteristics of the cardiac system can be accurately quantified. The finally generated individualized arrhythmia-induced sudden cardiac death risk score and early warning index can significantly extend the early warning time of risk, improve the individualization of prediction, reduce the false positive rate, and meet the actual needs of early and accurate identification of high-risk patients in clinical practice.

[0055] Furthermore, as a refinement and extension of the specific implementation of the above embodiments, and to fully illustrate the implementation of this embodiment, this embodiment also provides another method for predicting the risk of sudden cardiac death due to arrhythmia, such as... Figure 3 As shown, the method includes:

[0056] Step 310: Collect multi-channel magnetic field signals from the subject, perform fidelity enhancement processing on the multi-channel magnetic field signals, and extract core features that can characterize the magnetic field properties of the heart.

[0057] For embodiments of this disclosure, step 310 may include the following steps:

[0058] Step 310-1: Acquire multi-channel magnetocardiogram signals of the subject in the resting state and the provocation test state based on the OPM-MCG equipment array.

[0059] Among them, the OPM-MCG device array is a high-sensitivity magnetic sensing device array built based on the superconducting Josephson effect, which has the ability to capture weak magnetic field signals at the femtotes (fT) level and can accurately detect quantum-scale changes contained in the heart's magnetic field; the resting state refers to the state in which the subject is quiet, relaxed and without additional physiological load, and the cardiac magnetic signals collected in this state can be used as the benchmark data of the basic physiological state of the heart; the provocation test state refers to the state in which the heart generates physiological load through specific means such as exercise load and drug stimulation, and the cardiac magnetic signals collected in this state can capture the dynamic changes of the heart's magnetic field under stress conditions.

[0060] In this embodiment of the disclosure, an OPM-MCG device array can be used as the core sensing device. First, the subject is placed in a quiet, relaxed resting state, and multi-channel magnetic field signals are collected in this state. Then, the subject is induced into a stimulation test state through specific means, and multi-channel magnetic field signals under cardiac stress conditions are collected. The signals collected in both states are uniformly converted into a differential form of the magnetic field. (In the formula, For the magnetic field in directional components, (For the differential element in the corresponding direction), while considering the quantum fluctuation characteristics of the cardiac magnetic field, it satisfies (Where ξ(t) represents thermodynamic fluctuations, D is the diffusion coefficient, and δ is the Dirac function), this method comprehensively captures the characteristics and spatial distribution information of the cardiac magnetic field, providing high-fidelity raw data for subsequent high-fidelity enhancement processing and core feature extraction. For multi-channel magnetocardiogram (MCG) signal acquisition, a 32-channel optically pumped magnetocardiogram (OPM-MCG) array can be used. The sensors are arranged in a 3×11 matrix (central region), covering the main area of ​​the anterior wall of the adult heart (approximately 20cm×15cm), with a spatial resolution of approximately 1.25cm×1.125cm. The sampling frequency can be set to 1000 Hz to satisfy the Nyquist sampling theorem and cover the main frequency band of the MCG signal (0.05–100Hz). The device sensitivity is better than 1 fT / √Hz. Acquisition protocols may include 5 minutes of resting state recording (excluding the first minute of stable period) and 3 minutes of synchronous recording under low-intensity Bruce protocol exercise load to elicit potential electrophysiological abnormalities.

[0061] This technology utilizes the ultra-high sensitivity of the OPM-MCG array to break through the detection limits of traditional magnetometers, capturing key information such as quantum coherence and magnetic vortices contained in the heart's magnetic field. The multi-channel spatial distribution design enables high-resolution acquisition of the heart's magnetic field, accurately characterizing the spatial heterogeneity of the magnetic field. Signal acquisition in both resting and stimulated states can comprehensively cover the magnetic field characteristics of the heart under basal and stress conditions, effectively preserving the deep quantum information that is easily lost in traditional acquisition methods, and solving the problems of single dimension and insufficient sensitivity in traditional cardiac magnetic signal acquisition.

[0062] Step 310-2: Estimate the decoherence channel characteristics of the multi-channel magnetocardiogram (MCC) signal using quantum process tomography (QT), design a compensation pulse sequence based on the decoherence channel characteristics, and use the compensation pulse sequence to enhance the multi-channel MCC signal with high fidelity.

[0063] Among them, quantum process tomography is an analytical technique that accurately reconstructs the quantum evolution process by measuring the input and output quantum states of a quantum system. It can be used to characterize the decoherence evolution of multi-channel magnetocardiogram (MCC) signals during the acquisition and transmission stages. Decoherence channel characteristics refer to the channel evolution law of quantum coherence attenuation caused by environmental magnetic field interference, equipment noise, and other factors during the acquisition and transmission of multi-channel MCC signals. It is characterized by the interaction relationship between channel operators and the quantum state density matrix. Compensation pulse sequence is a pulse signal sequence customized for decoherence channel characteristics. It can cancel the quantum coherence attenuation caused by environmental interference and restore and maintain the effective quantum information of multi-channel MCC signals. Fidelity enhancement is a signal processing method that maintains the integrity of signal characteristics and effective information by using compensation pulse sequence to address the quantum coherence attenuation problem of MCC signals.

[0064] Because multi-channel magnetocardiogram (MCC) signals are subject to environmental magnetic noise interference during acquisition, they are equivalent to a quantum decoherence channel. In this embodiment, the input and output quantum states of the multi-channel MCC signals can be measured using quantum process tomography (QT): six linearly independent calibration signals (simulating different "input states") are used for multiple measurements, and the supercomputing operator of the decoherence channel is estimated using linear reconstruction, with the target estimation error controlled within 5%. Based on the measurement and reconstruction results, the evolution of the decoherence channel is characterized as follows: (In the formula, To decoherent channel, For channel operation elements, (where ρ is the Kraus operator and ρ is the quantum density of states matrix). Based on this precisely characterized decoherent channel characteristics, an adaptive finite impulse response (FIR) digital filter is designed in the digital signal processing domain as a compensation pulse sequence. The coefficients of this filter are optimized so that it can compensate for channel distortion and suppress noise bands within the signal frequency band (0.05-100Hz). This compensation pulse sequence is applied to the original multi-channel magnetocardiogram (MCC) signal. Through the interaction between the pulse sequence and the signal quantum state, the quantum coherence loss caused by environmental interference and equipment noise is offset, thereby achieving fidelity enhancement of the multi-channel MCC signal and fully preserving the key information such as quantum coherence and magnetic vortex contained in the signal.

[0065] This technology overcomes the limitations of traditional magnetocardiogram (MCC) signal processing, which focuses solely on classical electromagnetic features and ignores quantum coherence information. By employing quantum process tomography, it achieves precise characterization of decoherence channel properties. The compensation pulse sequence designed based on these properties can specifically counteract coherence attenuation, effectively preserving deep features such as quantum tunneling and spin coherence in multi-channel MCC signals. This provides high-fidelity quantum-level data support for subsequent core steps such as magnetic topological charge calculation, cardiac fiber bundle topological modeling, and non-equilibrium phase transition detection, significantly improving the accuracy and early identification capabilities for predicting the risk of arrhythmias and sudden cardiac death.

[0066] Step 310-3: Convert the enhanced multi-channel magnetocardiogram signal into a differential form of the magnetic field to complete the signal standardization process.

[0067] Among them, the differential form of the magnetic field is a precise mathematical representation of the spatial distribution of the magnetic field using differential geometry. It can characterize the vector distribution characteristics of the magnetic field in three-dimensional space and provide a suitable mathematical carrier for subsequent topological characteristic analysis. The signal standardization processing is a unified processing operation carried out on the magnetic field signal converted into differential form. The purpose is to eliminate the dimensional differences and baseline drift between different acquisition channels and different subjects, suppress residual noise interference, and ensure the consistency and comparability of signal data.

[0068] In the embodiments of this disclosure, the multi-channel magnetocardiogram signal that has undergone fidelity enhancement processing can be converted into a differential form of the magnetic field using a differential geometry representation method. (In the formula, For the magnetic field in directional components, The spatial vector distribution characteristics of the magnetic field are accurately characterized by the differential element in the corresponding direction. Then, the converted differential form magnetic field signal is subjected to unified processing. By eliminating the dimensional differences between different acquisition channels, correcting signal baseline drift, and suppressing residual noise interference, the signal is standardized, providing basic data with unified format and complete information for subsequent core steps such as magnetic topological charge density calculation.

[0069] This technical step converts the enhanced cardiac magnetic signal into a differential form, enabling a precise mathematical characterization of the spatial distribution of the cardiac magnetic field. This breaks through the limitations of signal representation under traditional classical electromagnetic theory. Subsequent standardization processing can effectively eliminate interference factors and individual differences in the signal, ensuring data consistency and comparability. This provides a high-quality, unified data foundation for subsequent core analysis steps such as magnetic topological charge density calculation and cardiac fiber tract topological modeling.

[0070] Step 310-4: Calculate the magnetic topological charge density based on the standardized multi-channel cardiac magnetic signals, and perform an integral operation on the magnetic topological charge density to obtain the topological charge value characterizing the magnetic vortex structure of the heart, which serves as the core feature.

[0071] Among them, the magnetic topological charge density is the core quantitative parameter describing the distribution intensity of magnetic vortices in the heart's magnetic field. It can characterize the density characteristics of magnetic field singularities and is the basis for calculating the topological charge value. The topological charge value is an integer obtained by integrating the magnetic topological charge density in three dimensions. Its positive or negative sign represents the chirality of the magnetic vortex, and its magnitude represents the number of magnetic vortices. It is a key topological parameter reflecting the magnetic vortex structure of the heart. The magnetic vortex structure of the heart consists of Abelian gauge field singularities in the heart's electromagnetic field, corresponding to the topological protective current loop inside the heart, and is the topological origin of reentrant arrhythmias.

[0072] In the embodiments of this disclosure, the magnetic topological charge density can be calculated based on the standardized multi-channel magnetocardiogram signal according to the differential form. (In the formula, The direction of the unit magnetic field. ijk The symbol for Levi-Civita. The direction of the unit magnetic field. , (These are the partial derivatives corresponding to the j-direction and the k-direction), and then a three-dimensional spatial integration is performed on the magnetic topological charge density to obtain the topological charge value. This value is an integer, which can accurately characterize the number and chirality of cardiac magnetic vortex structures. Ultimately, this topological charge value is used as the core feature to support subsequent cardiac fiber tract modeling and risk assessment.

[0073] This technique overcomes the limitations of traditional cardiac magnetic signal analysis, which only extracts classical shallow features such as magnetic field amplitude and frequency. By calculating and integrating the magnetic topological charge density to obtain the topological charge value, it can accurately capture the topological essence of the magnetic vortex structure in the cardiac electromagnetic field. This topological charge value, as a core feature, has topological invariance and is not affected by dynamic deformations such as cardiac contraction and relaxation. It can stably reflect the inherent topological defects of the cardiac electrical conduction system, solving the problem that traditional feature parameters are sensitive to changes in cardiac morphology. It can provide key data support with high discriminative power and physiological relevance for subsequent construction of cardiac fiber plexus topological geometric models and the detection of non-equilibrium phase transition dynamics.

[0074] Step 320: Construct a topological geometric model of the cardiac fiber tract, and extract topological invariants of the cardiac electrical conduction system based on the core features and the topological geometric model of the cardiac fiber tract.

[0075] For embodiments of this disclosure, step 320 may include the following steps:

[0076] Step 320-1: Using the three-dimensional manifold of the heart as the base space and the U (1) phase space as the fiber, construct the topological geometric model of the heart fiber bundle.

[0077] Among them, the three-dimensional manifold of the heart is a three-dimensional smooth manifold used to describe the anatomical structure and dynamic deformation characteristics of the heart. It can accurately characterize the irregular chamber geometry of the heart, the distribution of spiral muscle fibers, and the dynamic torsional characteristics during contraction and relaxation. The bottom space is the basic space that carries fibers in the fiber bundle theory. In this technical solution, the bottom space corresponds to the three-dimensional manifold of the heart and is the core basis for constructing the topological geometric model of the heart fiber bundle. The U (1) phase space is a one-dimensional compact Lie group composed of a circle group, which corresponds to the phase distribution law of the heart's electromagnetic field and can characterize the phase synchronization characteristics of the electrical activity of myocardial cells. The fiber is the space attached to each point of the bottom space in the fiber bundle theory. In this technical solution, the fiber is the U (1) phase space, which can reflect the phase state of the electromagnetic field at a specific anatomical location of the heart. The topological geometric model of the heart fiber bundle is a mathematical model of the heart's electrical conduction system constructed based on the fiber bundle theory. By using the three-dimensional manifold of the heart as the bottom space and the U (1) phase space as the fiber, it can characterize the deep topological essence of the heart's electrical conduction system.

[0078] In this embodiment of the disclosure, a three-dimensional manifold M representing the anatomical structure and dynamic deformation characteristics of the heart can be used as the base space, and a phase space U(1) corresponding to the phase distribution of the heart's electromagnetic field can be used as the fiber to construct a topological geometric model of the heart fiber bundle, with the main bundle structure being P(M,U(1)). This technical step can overcome the limitation of the traditional Euclidean geometric model that simplifies the heart to a regular ellipsoid. By constructing a topological geometric model of the heart fiber bundle, the spiral muscle fiber structure, irregular chamber geometry, and dynamic deformation characteristics of the heart can be accurately characterized.

[0079] Step 320-2: Calculate the connection form of the cardiac fiber tract topological geometric model based on the core features, perform external differential operation on the connection form to obtain the curvature tensor of the cardiac fiber tract topological geometric model, and calculate the first Chern class value representing the total number of cardiac topological charges and the Pontryagin class value representing the degree of fiber tract torsion based on the curvature tensor.

[0080] Among them, the connection form is the core geometric quantity describing the connection between fibers and the base space in the topological geometric model of the cardiac fiber tract. It can characterize the phase distribution law of the cardiac electromagnetic field, and its mathematical expression can accurately correlate the correspondence between magnetic flux and phase angle. The external differential operation is a basic operation in the field of differential geometry. It can be applied to differential forms such as the connection form. Through the operation, the curvature tensor that can characterize the degree of twisting of the fiber tract is obtained. It is the key mathematical means to connect the connection form and the curvature tensor. The curvature tensor is the core quantitative parameter of the topological geometric model of the cardiac fiber tract. It can intuitively reflect the degree of twisting and geometric characteristics of the fiber tract. It is the direct basis for calculating the first Chern class value. The first Chern class value is a topological invariant calculated based on the curvature tensor. Its value is an integer. It can accurately characterize the total number of cardiac topological charges. It is the core indicator reflecting the inherent topological properties of the cardiac electrical conduction system.

[0081] In this embodiment of the disclosure, based on the core features that characterize the magnetic field properties and topological essence of the heart, a topological geometric model of the cardiac fiber bundle with the three-dimensional manifold of the heart as the base space and the U(1) phase space as the fiber can be substituted to calculate the connection form of the model. (In the formula, For communication purposes, (For the differential elements in the corresponding directions); perform exterior differentiation on the connection form to obtain the curvature tensor F=dA, and then, based on the calculated curvature tensor, represent the first Chern class value of the total cardiac topological load. and Pontryagin class numerical values ​​characterizing the degree of fiber bundle twisting Both types of values ​​are topological invariants of the cardiac electrical conduction system, which can accurately reflect the inherent topological properties of the system.

[0082] This technique overcomes the limitations of traditional Euclidean geometry models in simplifying the complex structure of the heart. By calculating the connection forms and curvature tensors based on core features, it can accurately capture the geometric characteristics and degree of distortion of the cardiac fiber tract topological geometry model. The calculated first Chern class and Pontryagin class values, as topological invariants, remain stable during dynamic deformations such as cardiac contraction and relaxation, unaffected by changes in cardiac morphology. This solves the problem of traditional feature parameters being sensitive to dynamic deformations of the heart. It can reflect the inherent properties and structural complexity of the cardiac electrical conduction system from a topological perspective, providing core topological basis for subsequent non-equilibrium phase transition dynamics detection and quantification of the degree of chaos in the cardiac electrical conduction system.

[0083] Step 320-3: Treat the first Chen class values ​​and the Pontryagin class values ​​as topological invariants of the cardiac electrical conduction system.

[0084] Step 330: Conduct non-equilibrium phase transition dynamics detection on the cardiac fiber plexus topological geometric model to obtain phase transition characteristics that reflect the stability of the cardiac system.

[0085] For embodiments of this disclosure, step 330 may include the following steps:

[0086] Step 330-1: Extract thermodynamic data from the cardiac electrical conduction system characterized by the topological geometric model of the cardiac fiber plexus, and calculate the entropy generation rate of the cardiac electrical conduction system based on the stochastic thermodynamic analysis framework constructed from the thermodynamic data.

[0087] Thermodynamic data consists of quantitative parameters related to energy transfer and entropy change extracted from the cardiac electrical conduction system, with the core being thermodynamic flux characterizing the system's energy flow. Thermodynamic forces characterizing the degree to which a system deviates from equilibrium The stochastic thermodynamic analysis framework is a thermodynamic analysis system built for cardiac electrical conduction systems far from equilibrium. It decomposes the total entropy change of the system into entropy exchange terms with the environment and internal entropy generation terms, which are used to analyze the non-equilibrium dynamic behavior of the system. The entropy generation rate is a quantitative indicator describing the rate of entropy generation of irreversible processes inside the cardiac electrical conduction system, reflecting the degree to which the system deviates from equilibrium.

[0088] In the embodiments of this disclosure, thermodynamic flow can be extracted from the cardiac electrical conduction system characterized by a topological geometric model of the cardiac fiber bundle with the cardiac three-dimensional manifold as the base space and the U(1) phase space as the fiber. and thermodynamic forces Based on the thermodynamic data, a stochastic thermodynamic analysis framework is constructed, decomposing the total entropy change of the system into an entropy exchange term with the environment and an entropy generation term from internal irreversible processes. The internal entropy generation term is always greater than 0, according to the formula... The entropy generation rate of the cardiac electrical conduction system was calculated, and the detailed fluctuation theorem was verified by statistically analyzing the probabilities of forward and reverse entropy generation. Where ΔS is the total entropy change of the system, P(+ΔS) and P( ΔS) represents the probability of generating forward and reverse entropy, respectively, thereby enabling a precise analysis of the non-equilibrium thermodynamic characteristics of the cardiac electrical conduction system.

[0089] This technique overcomes the theoretical limitations of traditional methods that treat the cardiac system as an equilibrium state. By constructing a stochastic thermodynamic analysis framework and combining it with the topological properties of the cardiac fiber bundle topological geometry model, it can accurately extract thermodynamic data of the cardiac electrical conduction system. The calculated entropy generation rate can effectively characterize the degree to which the system deviates from the equilibrium state. The detailed verification of the fluctuation theorem can further ensure the rigor and scientific nature of the thermodynamic analysis, providing a reliable thermodynamic basis for subsequent non-equilibrium phase transition dynamics analysis such as critical slowing detection and multi-stable state characteristic judgment.

[0090] Step 330-2: Based on thermodynamic data, analyze the dynamics of the order parameters of the cardiac electrical conduction system, detect the variation law of the order parameters near the critical point, and obtain the dynamic analysis results of the order parameters.

[0091] Among them, the order parameter is the core quantitative parameter used to characterize the phase transition state of the cardiac electrical conduction system, which can reflect the trend of the system transitioning from an ordered state to a disordered state. In this technical solution, it specifically refers to parameters such as quantum coherence. The order parameter dynamics is the evolution law of the order parameter over time, which can reflect the dynamic change characteristics of the cardiac electrical conduction system under non-equilibrium conditions. Its evolution equation can accurately describe the process of the system approaching the critical point. The critical point is the critical state point where the phase transition of the cardiac electrical conduction system occurs. At this time, the physical properties of the system will change abruptly, and the change law of the order parameter will show significant anomalies. The result of the order parameter dynamics analysis is a set of quantitative characteristics obtained by analyzing the evolution law of the order parameter. It contains key information such as the fluctuation characteristics of the order parameter near the critical point and the change of relaxation time, which is an important basis for judging the stability of the system.

[0092] In the embodiments of this disclosure, order parameters characterizing the phase transition state of the system can be determined based on thermodynamic data extracted from the cardiac electrical conduction system, and kinetic evolution equations for the order parameters can be established. (where ψ is the order parameter, and a and b are kinetic coefficients,) (For thermodynamic fluctuations), by analyzing the variation of parameters in this equation, we focus on the state changes of the system when the value of 'a' approaches 0, and combine dynamic scaling theory to fit the correlation between relaxation time and system state. The correlation formula is as follows: (In the formula, Let T be the relaxation time, and T be the current state parameter of the system. (where ν is the critical state parameter, z is the critical exponent, and z is the dynamic exponent) is used to detect the key laws such as the fluctuation characteristics and rate of change of the order parameter near the critical point, and finally obtain the dynamic analysis results of the order parameter that can reflect the non-equilibrium phase transition trend of the system.

[0093] This technique overcomes the limitations of traditional methods that only focus on the static state of the system. By constructing the dynamic evolution equation of the order parameter and combining it with thermodynamic data for analysis, it can accurately capture the dynamic characteristics of the cardiac electrical conduction system evolving toward the phase transition critical point. It clearly presents the abnormal change patterns of the order parameter near the critical point, providing key dynamic basis for subsequent identification of critical slowing parameters and judgment of the system's multi-steady-state characteristics. This can effectively improve the scientific nature of the stability assessment of the cardiac electrical conduction system, thereby enhancing the early identification capability and accuracy of arrhythmia and sudden death risk prediction.

[0094] Step 330-3: Based on the results of the order parameter dynamic analysis, calculate the relaxation time of the cardiac electrical conduction system, and identify the critical slowing parameter of the cardiac electrical conduction system by fitting the correlation between the relaxation time and the system state through dynamic scaling theory.

[0095] Among them, relaxation time is the time required for the cardiac electrical conduction system to recover to its original equilibrium state after deviating from steady state. It is a core parameter characterizing the dynamic characteristics of the system and exhibits significant divergence when the system approaches the phase transition critical point. Dynamic scaling theory is a theory describing the scaling relationship between various physical quantities and system state parameters when the system approaches the phase transition critical point. It can be used to reveal the universal laws of the system's critical behavior. Critical slowing parameters are a set of core parameters characterizing the degree of critical slowing of the cardiac electrical conduction system. They include the critical exponent ν and the dynamic exponent z, and their values ​​directly reflect the degree to which the system approaches the phase transition critical point.

[0096] In this embodiment of the disclosure, key time features of system state changes can be extracted from the sequence parameter dynamics analysis results of the cardiac electrical conduction system. The relaxation time τ required for the system to recover equilibrium after deviating from steady state can be calculated. Dynamic scaling theory is introduced to correlate and fit the relaxation time with the system state parameters. The core correlation formula is as follows: (In the formula, Let T be the relaxation time, and T be the current state parameter of the system. (where ν is the critical state parameter, z is the critical exponent, and z is the dynamic exponent) By fitting and calculating the relaxation time and system state parameter data under multiple different states, the specific values ​​of the critical exponent ν and the dynamic exponent z in the formula are determined, thereby completing the accurate identification of the critical slowing parameters of the cardiac electrical conduction system.

[0097] This technique overcomes the limitations of traditional methods that focus solely on static system characteristics. By calculating relaxation time and combining it with dynamic scaling theory to fit its correlation with system state, it can accurately capture the critical slowing phenomenon of the cardiac electrical conduction system near the phase transition critical point. The identified critical slowing parameters can quantitatively characterize the degree to which the system deviates from steady state, providing crucial dynamic basis for determining whether the system is about to undergo arrhythmia-related topological phase transitions. This effectively compensates for the lack of dynamic evolutionary characteristics in traditional risk assessment, significantly improving the scientific rigor and early identification capability of arrhythmia-related sudden death risk prediction.

[0098] Step 330-4: Based on the entropy generation rate, the results of the order parameter dynamic analysis, the relaxation time, and the critical slowing parameter, reconstruct the effective potential energy function of the cardiac electrical conduction system, calculate the potential well depth based on the effective potential energy function, and determine the multi-steady-state characteristics of the cardiac electrical conduction system based on the magnitude of the potential well depth.

[0099] Among them, the effective potential function is a mathematical function describing the energy distribution of the cardiac electrical conduction system under different states. It can reflect the energy characteristics of the system's steady state and is the core basis for judging the multi-steady-state characteristics of the system. The potential well depth is the difference between the maximum and minimum potential energy of the potential well in the effective potential function. It quantifies the difficulty of the system switching between steady states. The smaller the value, the easier it is for the system to switch between different states. Multi-steady-state characteristics are the property of the cardiac electrical conduction system that can switch between multiple different steady states. It is closely related to the occurrence of arrhythmias. Systems with significant multi-steady-state characteristics have a higher risk of malignant arrhythmias.

[0100] In the embodiments of this disclosure, a state variable and dynamic correlation model of the cardiac electrical conduction system can be constructed based on the entropy generation rate, which reflects the degree of deviation of the system from equilibrium, the order parameter dynamic analysis results, which characterize the phase transition evolution trend of the system, the relaxation time, which reflects the dynamic characteristics of the system, and the critical slowing parameter, which reflects the critical state of the system. The effective potential energy function calculation formula can then be substituted into this model. ,in, Let be the dynamic force function of the system. By solving this integral, the effective potential energy function of the system can be obtained. Then, the potential well depth can be calculated based on this effective potential energy function. ,in, The maximum value of the effective potential energy function. The minimum effective potential energy function is used to determine the multi-stable characteristics of the cardiac electrical conduction system. The smaller the potential well depth, the easier it is for the system to switch between different stable states, and the more significant the multi-stable characteristics are.

[0101] This technique overcomes the limitations of traditional methods that rely solely on single-dimensional parameter analysis of system stability. By organically integrating the thermodynamic characteristics of entropy generation rate, the evolutionary characteristics of order parameter kinetic analysis results, and the critical characteristics of relaxation time and critical slowing parameters, it enables precise characterization of the energy distribution of the cardiac electrical conduction system. The reconstructed effective potential energy function and calculated potential well depth can quantify the system's stable state switching capability from an energy perspective. The identified multi-steady-state characteristics are directly related to the risk of malignant arrhythmias in the cardiac electrical conduction system, compensating for the shortcomings of relying solely on topological or kinetic single-dimensional analysis. This provides crucial energy-level evidence for subsequent arrhythmia and sudden cardiac death risk assessment, significantly improving the comprehensiveness and accuracy of risk prediction.

[0102] Step 330-5: Integrate the entropy generation rate, critical slowing parameter, potential well depth, and corresponding multistable characteristics as phase transition features reflecting the stability of the cardiac electrical conduction system.

[0103] Step 340: Construct a Vietoris-Rips complex based on the location of the cardiac magnetosensor, extract features from the topological invariants using the persistent homology algorithm, and generate persistent landscape features that characterize the topological properties of the cardiac electrical conduction system.

[0104] Among them, the Vietoris-Rips complex is a simple complex constructed based on the location point set of the cardiac magnetocardiogram sensor. It is the core structure for topological data analysis. By setting a distance threshold to filter the combination of point sets that satisfy the diameter condition, it can accurately characterize the spatial topological relationship of the sensor location. The persistent cohomology algorithm is an algorithm to analyze the evolution law of the topological structure with the change of distance threshold. By calculating the birth and death time of cohomology groups of different dimensions, it generates persistent barcodes, which can effectively capture the core topological features of the cardiac electrical conduction system. Topological invariants are quantitative parameters extracted from the topological geometric model of the cardiac fiber bundle that remain stable during the dynamic deformation of the heart. The core parameters include the first Chern class value representing the total topological charge of the heart and the Pontryagin class value representing the degree of twisting of the fiber bundle. Persistent landscape features are functional representations obtained by converting persistent barcodes. They are the key form of topological feature vectorization and can quantitatively characterize the topological properties of the cardiac electrical conduction system, providing high-quality input for subsequent topological neural network analysis.

[0105] In the embodiments of this disclosure, a Vietoris-Rips complex can be constructed based on the point set X formed by the locations of the magnetocardiogram sensors. (where σ is a subset of the point set X, and diam(σ) is the diameter of the subset σ,) (For the set distance threshold), the extracted topological invariants such as the first Chern class values ​​and Pontryagin class values ​​are incorporated into the structural weights of the Vietoris-Rips complex. The homology group of each dimension k is calculated as a function of the distance threshold using the persistent homology algorithm. The increasing evolutionary process yields the evolutionary relationships of homology groups, among which... < Based on this evolutionary relationship, persistent barcodes of various dimensions are generated, and then the persistent barcodes are converted into persistent landscape features, thereby completing the quantitative characterization of the topological properties of the cardiac electrical conduction system.

[0106] This technique overcomes the limitations of traditional geometric features in their sensitivity to dynamic cardiac deformation. By constructing Vietoris-Rips complexes and analyzing persistent cohomology algorithms, it can effectively capture the inherent topological characteristics of the cardiac electrical conduction system. The generated persistent landscape features can transform abstract topological information into computable quantitative features, solving the problem that topological invariants are difficult to directly input into neural networks for analysis. This provides highly discriminative and stable topological inputs for subsequent topological neural network-based quantitative analysis of chaos.

[0107] Step 350: Based on the myocardial region segmentation results of the cardiac electrical conduction system, construct a cardiac conduction map using myocardial regions as nodes and the electrical conduction connection strength between regions as edges, using a topological neural network.

[0108] Among them, the myocardial region segmentation result is based on cardiac structural and functional imaging data such as diffusion tensor imaging, dividing the myocardial tissue into multiple partitions with independent electrophysiological characteristics, each partition corresponding to a specific blood supply or functional area of ​​the heart; the electrical conduction connectivity strength characterizes the strength of electrical signal conduction between different myocardial regions, and is directly related to the density of myocardial fiber connections and electrophysiological synchronicity between regions; the topological neural network is a model that integrates topological data analysis and graph neural network technology, which can process the topological structure data of the cardiac conduction map and update node features through a message passing mechanism; the cardiac conduction map is a graph structure constructed with myocardial regions as nodes and electrical conduction connectivity strength between regions as edges, and is the core input carrier for topological neural network analysis of the cardiac electrical conduction system.

[0109] In this embodiment of the disclosure, based on the myocardial region division results of the cardiac electrical conduction system, each independent myocardial region can be used as a node of a topological neural network. Core features such as topological invariants and phase transition features corresponding to each node can be extracted and node feature vectors can be initialized. The electrical conduction connection strength between regions can be obtained by calculating the electrical signal conduction synchronicity and fiber connection density between different myocardial regions. This strength can be assigned as the edge weight between the corresponding nodes to construct a basic cardiac conduction graph structure.

[0110] This technique overcomes the limitations of traditional neural networks that treat the cardiac electrical conduction system as abstract data input and ignore its spatial topology. By constructing a cardiac conduction graph with myocardial regions as nodes and electrical conduction connection strength as edges, it can accurately reconstruct the physiological topological relationships of the cardiac electrical conduction system. The message passing and attention mechanisms of the topological neural network can adaptively capture key areas of electrical conduction abnormalities, effectively enhancing the representation ability of core features and solving the problem that traditional models have difficulty in characterizing the electrical conduction correlation between regions. It can provide physiologically accurate topological support for subsequent quantification of the degree of chaos in the cardiac electrical conduction system and generation of risk scores.

[0111] Step 360: Input persistent landscape features and phase transition features into a topological neural network based on the cardiac conduction map, and update the network node features through a message passing mechanism.

[0112] Among them, the message passing mechanism is the core algorithm for node feature updating in topological neural networks. By aggregating the features of neighboring nodes and edge weight information, it realizes the interactive updating of node features, which can effectively improve the representation ability of features.

[0113] In this embodiment of the disclosure, persistent landscape features that characterize the topological properties of the cardiac electrical conduction system and phase transition features that reflect the phase transition trend of the system can be fused. The fused feature vector is used as the initial input and loaded into a cardiac conduction graph constructed with myocardial regions as nodes and inter-regional electrical conduction connection strength as edges. The node feature vectors of the topological neural network are initialized, and the inter-regional electrical conduction connection strength is assigned as the edge weight between nodes. Then, the message passing mechanism of the topological neural network is used. (In the formula, is the updated feature vector of node v in the (l+1)th layer of the network; UPDATE() is the update function used to integrate the node's own features with the result of aggregating all neighborhood messages to generate the updated features of the node; Let N(v) be the feature vector of node v in the l-th layer network, and let N(v) be the set of neighboring nodes of node v. MESSAGE() is a message function used to generate a message from the neighboring nodes to the target node based on the target node features, neighboring node features, and edge features. Let be the feature vector of a neighboring node u in the l-th layer network; The edge feature vector between neighboring node u and node v (corresponding to the electrical conduction connection strength between myocardial region u and myocardial region i in this scheme) enables interactive updating of node features. Simultaneously, a topological attention mechanism is introduced to calculate the attention weights between nodes. (In the formula, In the topological attention mechanism, the attention weights of node v to its neighboring node u are used to measure the importance of node u's features when updating node v's features. The higher the weight, the greater the influence of the neighboring node's features on the target node. exp() is an exponential function used to map the output of LeakyReLU to the non-negative interval, ensuring the non-negativity of the attention weights and amplifying the differences exponentially to highlight the more important features of the neighboring nodes. LeakyReLU() is a modified linear unit activation function with leakage, used to introduce non-linear transformations to enhance the model's ability to fit complex features, while avoiding the neuron death problem that occurs with the ReLU function when there is a negative input. It is the transpose of the learnable attention vector, used to perform a linear transformation on the concatenated feature vector to generate an attention score, whose parameters are adaptively optimized during network training. The concatenation operation is performed on the feature vectors of node v and its neighboring node u after linear transformation. This operation fuses the feature information of the target node and its neighboring nodes, capturing the correlation between them. W is a learnable linear transformation weight matrix used to perform linear transformation on the node features, mapping the original node features to a higher-dimensional feature space, enhancing the expressive power of the features. Its parameters are adaptively optimized during network training. h v h is the original feature vector of node v in the current layer, carrying multi-dimensional information such as topology and phase transition of the myocardial region; u is the original feature vector of the neighboring node u in the current layer, carrying multi-dimensional information such as the topology and phase transition of the corresponding myocardial region; N(v) is the set of neighboring nodes of node v in the cardiac conduction graph, containing all myocardial region nodes that have electrical conduction connections with node v; h w The original feature vector of the neighboring node w in the current layer carries multi-dimensional information such as the topology and phase transition of the corresponding myocardial region. This completes the update of the node features of the topological neural network, resulting in high-dimensional node features that carry the dual characteristics of cardiac topology and phase transition.

[0114] This technology overcomes the limitations of traditional neural networks that only input single-dimensional features. By fusing the topological properties of persistent landscape features with the dynamic characteristics of phase transition features, it can integrate multi-dimensional information about the cardiac electrical conduction system. The topological neural network's structural design conforms to the physiological topological relationship of cardiac electrical conduction. The message passing mechanism and attention mechanism can adaptively capture key myocardial regions with abnormal electrical conduction, effectively enhancing the representational ability of node features and solving the problem that traditional models have difficulty in characterizing the electrical conduction correlation between regions.

[0115] Step 370: Based on the manifold characteristics of the cardiac fiber bundle topological geometry model, manifold convolution operation is used to optimize the updated network node features to ensure that the network output satisfies the covariance characteristics under the normalization transformation, thereby obtaining high-dimensional features for quantifying the degree of chaos.

[0116] Among them, manifold properties are the inherent geometric attributes of the three-dimensional manifold of the heart in the bottom space of the cardiac fiber bundle topological geometry model. The core features include smoothness, curvature, and connectivity, which can accurately represent the anatomical structure and dynamic deformation of the heart. Manifold convolution is a convolution calculation method adapted to the manifold geometry. Through the heat kernel approximation of the convolution operation on the three-dimensional manifold of the heart, it can effectively capture the local feature correlations on the manifold and avoid the geometric adaptation defects of traditional Euclidean space convolution. The updated network node features are node feature vectors that carry multi-dimensional information such as the topology and dynamics of the myocardial region after the topological neural network is processed by the message passing mechanism and the attention mechanism. The canonical transformation is a symmetric transformation on the cardiac fiber bundle, which can describe the local coordinate system transformation law of the fiber bundle and has a specific transformation effect on core physical quantities such as connection form and wave function. Covariance property means that the network output results follow a specific transformation law under the canonical transformation, ensuring that the output features will not change their essential properties due to the local transformation of the coordinate system. High-dimensional features are the feature set obtained after manifold convolution optimization, which can accurately represent the deep characteristics of the cardiac electrical conduction system and are the core input data for subsequent quantification of the chaos degree of the system.

[0117] In this embodiment of the disclosure, based on the manifold characteristics of the cardiac fiber bundle topological geometry model, a manifold convolution operation defined by the heat kernel approximation can be introduced to optimize the updated network node features after the message passing and attention mechanism processing of the topological neural network. The manifold convolution operation formula is as follows: ,in and Let f and g be the spectral domain representations of the functions f and g, respectively. Let Λ be the eigenfunction of the Laplace-Beltrami operator, and Λ be the truncation threshold. Simultaneously, a canonical isovariant constraint on the fiber bundle is introduced to ensure that the network output satisfies covariant properties under the canonical transformation. The specific form of the canonical transformation is... (in the formula, Ψ(x) is the transformed wave function after the normalization transformation; g(x) is the normalization transformation function; Ψ(x) is the original wave function on the cardiac fiber plexus. This is the transformed communication format after standardization. For the primitive connection form on the cardiac fiber plexus; g is the U (1) canonical transformation function g(x) consistent with the wave function transform, where the explicit label of the spatial coordinate x is omitted; g 1 To normalize the inverse of the transformation function g; To normalize the transformation function g with respect to spatial coordinates The partial derivatives (which describe the rate of change of the gauge transformation in the spatial dimension) ultimately yield a high-dimensional feature that can accurately characterize the topological and dynamic properties of the cardiac electrical conduction system and is used to quantify the degree of chaos.

[0118] This technique overcomes the limitations of traditional Euclidean convolution, which cannot adapt to the complex manifold structure of the heart. Through manifold convolution, it can accurately capture the local features of the heart's three-dimensional manifold. At the same time, the introduced isovariant constraints can ensure the stability of the output high-dimensional features under dynamic deformation and coordinate system transformation of the heart, effectively preserving the deep topology and dynamic essence of the cardiac electrical conduction system. It solves the problem of traditional features being sensitive to geometric transformations, and can provide high-quality feature support for subsequent quantification of the chaos of the cardiac electrical conduction system, significantly improving the scientificity and accuracy of arrhythmia and sudden death risk assessment.

[0119] Step 380: Based on high-dimensional features, reconstruct the strange attractors of the cardiac electrical conduction system, calculate the maximum Lyapunov exponent and correlation dimension of the cardiac electrical conduction system, and use them as the quantitative analysis results of the degree of chaos in the cardiac electrical conduction system.

[0120] Among them, strange attractors are a collection of fractal structures formed in phase space when the cardiac electrical conduction system is in a chaotic state. They can intuitively reflect the nonlinear dynamic behavior of the system, and their morphology is directly related to the degree of chaos of the system. The maximum Lyapunov exponent is the core quantitative indicator for measuring the sensitivity of a chaotic system to initial conditions. A positive exponent indicates that the system is in a chaotic state, and the larger the value, the higher the degree of chaos of the system. The correlation dimension is a key parameter for describing the fractal characteristics of strange attractors. It can quantify the degree of fill of attractors in phase space and is an important dimension for characterizing the chaotic characteristics of the system. The result of the quantitative analysis of the degree of chaos is a set of quantitative parameters composed of the maximum Lyapunov exponent and the correlation dimension. It can accurately reflect the chaotic state of the cardiac electrical conduction system and provide a core basis for the subsequent generation of risk scores.

[0121] In this embodiment of the disclosure, based on high-dimensional features optimized by manifold convolution and satisfying covariance properties, a time-delay embedding method is used to reconstruct the strange attractor of the cardiac electrical conduction system. Its phase space reconstruction formula is: (In the formula, y(t) is the phase space vector reconstructed at time t; x(t) is the original one-dimensional time series, corresponding to the observed signal of the cardiac electrical conduction system in this scheme; t is the time variable; τ is the time delay; d e To embed the dimension, reconstruct the dimension of the phase space; t+(d e 1) τ is the time of the last sampling point, that is, the time elapsed from t (d e At a time delay of one step, ensure that the reconstructed vector contains d. e (Sampling points), the optimal embedding parameters are determined using the spurious nearest neighbor method, and then the maximum Lyapunov exponent of the cardiac electrical conduction system is calculated. (where t is the evolution time;) Let be the Euclidean norm (distance) between two initially adjacent phase space trajectories at time t. At the initial time t=0, the initial Euclidean norm (initial distance) between two adjacent phase space trajectories is calculated (usually taken as a minimum value to measure the small difference in the initial state), and the correlation dimension is also calculated. Finally, the obtained maximum Lyapunov exponent and correlation dimension are used as the quantitative analysis results of the degree of chaos in the cardiac electrical conduction system.

[0122] This technique overcomes the limitations of traditional linear analysis methods in characterizing the nonlinear properties of the cardiac electrical conduction system. By reconstructing strange attractors, it accurately captures the chaotic dynamics of the system. The calculated maximum Lyapunov exponent and correlation dimension can quantify the degree of chaos of the system from two dimensions: initial condition sensitivity and fractal structure. This can solve the problem that traditional risk assessment lacks deep dynamic characteristics, and can provide core quantitative evidence with clear physiological significance for the subsequent generation of arrhythmia and sudden death risk scores and early warning indicators.

[0123] Step 390: Based on the quantitative analysis results, generate risk scores and early warning indicators for sudden cardiac death due to arrhythmia for the examinee.

[0124] For embodiments of this disclosure, step 390 may include the following steps:

[0125] Step 390-1: Using a tensor fusion architecture, the quantitative analysis results are fused with topological invariants and phase transition features across modalities to output the fused high-dimensional risk features.

[0126] Among them, it is a mathematical framework for multimodal feature fusion, which can organize features of different dimensions and types into high-dimensional tensor structures, and realize deep interaction and information integration of cross-modal features through tensor decomposition. Cross-modal feature fusion is a process of integrating features from three different sources: chaos degree quantification analysis results, topological invariants, and phase transition features. It can effectively explore the potential correlation between different modal features and improve the feature representation ability. High-dimensional risk features are feature sets obtained after processing by the tensor fusion architecture. They carry the three dimensions of chaos, topology, and phase transition information of the cardiac electrical conduction system and are the core input data for subsequent generation of individualized risk scores.

[0127] In embodiments of this disclosure, a tensor fusion architecture can be employed to organize the quantitative analysis results characterizing the degree of chaos in the cardiac electrical conduction system, the topological invariants reflecting the system's inherent topological properties, and the phase transition characteristics reflecting the system's non-equilibrium evolution trend, thereby constructing a multimodal feature tensor. Where d1 is the feature dimension of the quantization analysis result, d2 is the feature dimension of the topological invariants, and d3 is the feature dimension of the phase transition characteristics. The tensor is then decomposed using the Tucker decomposition-fusion algorithm, with the core formula being: ,in The core tensor is used to capture cross-modal interaction information between different modal features. U(1), U(2), and U(3) are the modal transformation matrices corresponding to the quantitative analysis results, topological invariants, and phase transition features, respectively. ×1, ×2, and ×3 are the tensor product operations of the corresponding dimensions. Through this decomposition process, the deep fusion of different modal features is achieved, and finally, high-dimensional risk features carrying multi-dimensional information of the cardiac electrical conduction system are output.

[0128] This technical approach overcomes the limitations of traditional single-modal feature analysis, which suffers from fragmented information. By employing a tensor fusion architecture, it deeply integrates the dynamic characteristics of the chaos degree quantification analysis results, the structural characteristics of topological invariants, and the thermodynamic characteristics of phase transitions. The core tensor can accurately capture the potential correlations between different modal features, effectively solving the problems of insufficient single feature dimensions and information redundancy. The fused high-dimensional risk features can comprehensively cover the key characteristics of the cardiac electrical conduction system, providing high-quality feature support for the subsequent generation of individualized arrhythmia and sudden cardiac death risk scores, and significantly improving the scientific rigor, comprehensiveness, and accuracy of risk assessment.

[0129] Step 390-2: Based on high-dimensional risk characteristics, construct a state-space model of the cardiac electrical conduction system using a particle filtering algorithm to predict the risk trajectory change trend of the cardiac electrical conduction system.

[0130] Among them, the particle filtering algorithm is a state estimation algorithm based on Monte Carlo sampling, which is suitable for state inference of nonlinear and non-Gaussian systems. By approximating the system state distribution through a particle ensemble, it can accurately capture the dynamic evolution characteristics of the cardiac electrical conduction system. The state space model is a mathematical model describing the evolution of the state of the cardiac electrical conduction system over time. It includes state equations characterizing the internal dynamics of the system and observation equations relating the state to the observation, and can quantify the changing patterns of the system state. The cardiac electrical conduction system is the physiological system inside the heart responsible for the generation and conduction of electrical signals. Its functional state is directly related to the risk of arrhythmia and has nonlinear and non-equilibrium dynamic characteristics. The risk trajectory change trend is the evolution path of the risk state of the cardiac electrical conduction system over time, which is predicted based on the state space model. It can intuitively reflect the rate and trend of the system approaching the critical state, and provide dynamic basis for risk warning.

[0131] In this embodiment of the disclosure, high-dimensional risk characteristics that can carry multi-dimensional information about the cardiac electrical conduction system can be used as input. A state-space model of the system is constructed using a particle filtering algorithm, and the model includes state equations. and observation equations (where x)t Let θ be the system state vector at time t, and θ be the model parameters. t For state noise, y t Let v be the observation vector. t To reduce observation noise, a large number of particles are used to approximate the posterior distribution of the system state through a particle filtering algorithm. The particle set is iteratively optimized using a resampling and weight update mechanism. Based on the optimized particle set, the system state at each time step is estimated. Then, according to the evolution law of the state equation, the future risk trajectory change trend of the cardiac electrical conduction system is predicted, and the dynamic characteristics of the system approaching the critical state are clarified.

[0132] This technical approach can overcome the limitations of traditional static risk assessment. High-dimensional risk features can provide comprehensive basic information support for the model. The particle filtering algorithm can effectively adapt to the nonlinearity and non-Gaussianity of the cardiac electrical conduction system. The state-space model can accurately depict the dynamic evolution of the system state. The combination of these three can achieve dynamic prediction of risk trajectories, solving the problem that traditional methods cannot capture the changes in risk over time, and can significantly improve the foresight and accuracy of risk prediction.

[0133] Step 390-3: Based on the trend of risk trajectory changes, the t-SNE algorithm is used to reduce the high-dimensional risk features to two-dimensional space to generate an individualized risk feature profile of the examinee.

[0134] Among them, the t-SNE algorithm is a non-linear dimensionality reduction algorithm that maps high-dimensional features to low-dimensional space by preserving the local similarity and global structural correlation in high-dimensional data. It is suitable for the visualization and feature condensation of high-dimensional risk features. The individualized risk feature profile combines the dimensionality-reduced two-dimensional risk features with the risk trajectory change trend to form a visualized and quantitative risk representation specific to the examinee, which can intuitively present the individual's core risk features and evolutionary patterns.

[0135] In this embodiment of the disclosure, the risk trajectory change trend of the cardiac electrical conduction system predicted by the particle filter algorithm can be used as input, along with high-dimensional risk features carrying multi-dimensional information. The t-SNE algorithm is then employed for nonlinear dimensionality reduction, with its core loss function being... (where w) ij Let q represent the similarity between samples i and j in a high-dimensional space. ij (Similarity in low-dimensional space) By minimizing this loss function, high-dimensional risk features are mapped to two-dimensional space. At the same time, dynamic features in the trend of risk trajectory changes are integrated to generate an individualized risk feature profile that can accurately reflect the individual risk characteristics and evolutionary laws of the examinee.

[0136] This technical approach overcomes the limitations of high-dimensional risk features being difficult to interpret intuitively and differentiate individually. The t-SNE algorithm can effectively retain key correlation information in high-dimensional features, transforming complex multi-dimensional data into intuitive two-dimensional representations. Combined with the individualized risk feature profile generated by the dynamic trend of risk trajectories, it can clearly present the core differences and evolution direction of individual risks, solving the problems of abstract and unspecific results in traditional risk assessment.

[0137] Step 390-4: Based on individualized risk profiles, calculate the chaotic edge distance and topological vulnerability index of the cardiac electrical conduction system as early warning indicators.

[0138] Among them, the chaotic edge distance is the core indicator that quantifies the distance between the cardiac electrical conduction system and the chaotic critical state. The smaller the distance, the closer the system is to the chaotic transition point, and the higher the risk of malignant arrhythmias. The topological fragility index is a quantitative parameter calculated based on the topological characteristics of the cardiac electrical conduction system. It reflects the tolerance of the system's topological structure to electrical conduction abnormalities. The higher the index, the more fragile the topological structure, and the more likely it is to trigger reentrant arrhythmias. The early warning index is a set of quantitative parameters that can provide early warning of the risk of abnormalities in the cardiac electrical conduction system. The chaotic edge distance and the topological fragility index together constitute the core early warning index that has both dynamic evolution characteristics and topological characteristics.

[0139] In this embodiment of the disclosure, based on an individualized risk feature profile that integrates multi-dimensional information of the cardiac electrical conduction system, the system dynamic parameters and topological features contained in the profile can be extracted, and then the single-parameter chaotic edge distance formula can be used. (where μ is the current dynamic parameter of the system,) Calculate the basic distance value (the critical parameter for the system to undergo chaotic transition), and then apply the extended formula to the multidimensional parameter space. (Where B is the chaotic boundary) Optimize the calculation, and at the same time calculate the topological vulnerability index by combining the topological invariants and magnetic vortex structure features in the image. This index comprehensively considers the first Chern class value, the Pontryagin class value and the concentration of the magnetic topological charge distribution. Finally, the calculated chaotic edge distance and topological vulnerability index are used as the core early warning indicators that can provide early warning of the risk of arrhythmia and sudden death.

[0140] This technology can overcome the limitations of traditional early warning indicators that only reflect the state of a single dimension of the system. With the comprehensive information support provided by individualized risk feature profiles, the calculated chaotic edge distance can accurately quantify the dynamic risk evolution trend of the system, while the topological vulnerability index can capture the inherent topological defects of the system. The early warning indicator formed by the combination of the two has both dynamism and stability. It can not only reflect the current risk status of the system, but also predict future risk changes in advance, which can solve the problems of weak targeting and short early warning window of traditional early warning methods.

[0141] Step 390-5: Integrate early warning indicators and clinical factors to generate a risk score for sudden cardiac death due to arrhythmia in the examinee.

[0142] Clinical factors are clinical parameters directly related to the risk of sudden cardiac death due to arrhythmia, including indicators with clear clinical correlation such as family history of arrhythmia, underlying heart disease, age, blood pressure, and blood lipids. The risk score is an individualized quantitative result obtained by integrating early warning indicators and clinical factors. It can intuitively and accurately reflect the risk level of the examinee in the event of sudden cardiac death due to arrhythmia, providing a direct basis for clinical decision-making.

[0143] In this embodiment of the disclosure, chaotic edge distance, which characterizes the dynamic risk of the cardiac electrical conduction system, and topological fragility index, which reflects the tolerance of the topological structure, can be collected. At the same time, clinical factors such as the examinee's family history of arrhythmia, underlying heart disease, and age are integrated and quantified. A weighted fusion algorithm calibrated with clinical data is used to standardize and weight the early warning indicators and clinical factors to sum them, and finally generate an arrhythmia sudden death risk score that can comprehensively reflect the individual risk of the examinee.

[0144] This technology can overcome the limitations of traditional risk assessment that relies on only a single dimension of information. By integrating the deep pathophysiological characteristics of early warning indicators with known risk factors in clinical practice, it can achieve comprehensive coverage of risk information. The weighted fusion algorithm, calibrated with clinical data, can ensure the scientificity and reliability of the score. The generated risk score can capture both the inherent topological and dynamic abnormalities of the cardiac electrical conduction system and take into account key influencing factors in clinical scenarios.

[0145] In specific application scenarios, as a preferred approach, the cardiac electrical conduction system can be segmented into myocardial regions based on diffusion tensor imaging data that can characterize myocardial fiber structure, resulting in segmented regions that correspond one-to-one with nodes of a topological neural network. For each segmented region, a local Vietoris-Rips complex can be constructed. (where σ is a subset of the point set X, and diam(σ) is the diameter of the subset σ,) Using a set distance threshold, local persistent cohomology features are extracted through a persistent cohomology algorithm. These features are then fused with topological invariants such as the first Chern class values ​​and Pontryagin class values ​​to generate a regional topological feature vector. This regional topological feature vector, along with early warning indicators such as chaotic edge distance and topological fragility index, are then used as input parameters and substituted into the reaction-diffusion equation. (Where R is the risk level, D is the diffusion coefficient, and f(R) is the nonlinear response term), the propagation process of risk between different segmented regions is simulated, and the risk propagation rate and the risk level distribution of each segmented region are obtained. Combining the risk level distribution of each segmented region with the previously predicted risk trajectory change trend, the risk distribution in three-dimensional space is combined with the risk evolution process in the time dimension to generate a four-dimensional spatiotemporal topological risk map that can dynamically reflect the spatiotemporal characteristics of risk. Finally, through a mixed reality visualization interface, the three-dimensional holographic display of the four-dimensional spatiotemporal topological risk map is realized, which is convenient for intuitive observation and interactive analysis.

[0146] Accordingly, the implementation steps may further include: segmenting the myocardial region of the cardiac electrical conduction system based on diffusion tensor imaging data to obtain segmented regions that correspond one-to-one with the nodes of the topological neural network; calculating the local persistent homology features of each segmented region based on the topological characteristics of the cardiac fiber plexus topological geometric model, and generating a regional topological feature vector based on the local persistent homology features and topological invariants; using the regional topological feature vector and early warning indicators as input parameters, simulating the risk propagation process in different segmented regions of the cardiac electrical conduction system through the reaction-diffusion equation to obtain the risk propagation rate and the risk level distribution of each segmented region; combining the risk distribution in three-dimensional space with the risk evolution process in the time dimension based on the risk level distribution and risk trajectory change trend of each segmented region to generate a four-dimensional spatiotemporal topological risk map of the cardiac electrical conduction system; and realizing a three-dimensional holographic display of the four-dimensional spatiotemporal topological risk map through a mixed reality visualization interface.

[0147] Among them, diffusion tensor imaging data is image data that can reflect the microstructure and orientation of myocardial fibers, accurately characterizing the anatomical connectivity of myocardial tissue, and is the core data support for myocardial region segmentation; myocardial region segmentation, based on diffusion tensor imaging data, divides myocardial tissue into multiple partitions with independent topological and functional characteristics, providing basic units for local topological analysis and risk propagation simulation; segmented regions are individual functional partitions obtained after myocardial region segmentation, corresponding one-to-one with nodes of the topological neural network, used to carry local topological features and risk level information; local persistent cohomology features are topological features calculated for a single segmented region, extracted by constructing a local Vietoris-Rips complex and using a persistent cohomology algorithm, which can reflect the local topological structure characteristics within the region; the region topological feature vector is a fusion of local persistent cohomology features and topological invariants (first Chern class values, Pontryagin...). The high-dimensional vector formed by the numerical model can comprehensively characterize the topological properties of a single segmented region; the reaction-diffusion equation is a mathematical equation used to simulate the propagation of risk between different regions of the myocardium, which can quantify the spatial diffusion law and temporal evolution characteristics of risk; the risk propagation rate is a quantitative parameter obtained by solving the reaction-diffusion equation, which characterizes the speed of risk propagation between different segmented regions; the risk level distribution is the set of risk quantification levels corresponding to each segmented region, which can intuitively present the spatial risk differences of the cardiac electrical conduction system; the four-dimensional spatiotemporal topological risk map is a visualization map formed by integrating the three-dimensional spatial risk level distribution and the temporal dimension risk evolution trend, which can dynamically display the spatiotemporal change law of risk; the mixed reality visualization interface is an interactive interface that combines virtual reality and augmented reality technologies, which can realize the three-dimensional holographic display and interactive operation of the four-dimensional spatiotemporal topological risk map; the three-dimensional holographic display is a display method that presents the four-dimensional spatiotemporal topological risk map in a three-dimensional form through mixed reality devices, which is convenient for intuitive observation of the risk distribution and evolution of various regions of the heart.

[0148] This technology overcomes the limitations of traditional risk assessment, which can only statically present a single dimension of risk. By segmenting the myocardial region, it enables localized risk analysis. The fusion of local persistent cohomological features and topological invariants enhances the comprehensiveness of regional feature representation. The reaction-diffusion equation can accurately simulate the spatiotemporal propagation of risk. The generated four-dimensional spatiotemporal topological risk map can dynamically and three-dimensionally present the risk distribution and evolution trend of the cardiac electrical conduction system. The mixed reality holographic display method can solve the problem of the abstract and difficult-to-interpret traditional risk results.

[0149] To facilitate understanding of the technical solutions in this application, the following is combined with... Figure 4The implementation process of the technical solution in this application is fully described as follows: First, the multi-channel magnetic field signals of the subject under resting and excitation test states are acquired by the OPM-MCG array signal acquisition device. After quantum tomography and fidelity enhancement processing, the multi-channel magnetic field signals are converted into magnetic field differential form and standardized. Then, the core features that can characterize the magnetic field properties of the heart are extracted by magnetic topological charge calculation. Subsequently, the three-dimensional manifold of the heart is used as the base space, U (1) A topological geometric model of the cardiac fiber bundle is constructed using phase space as the fiber model. The connection forms and curvature tensors of the model are calculated, and topological invariants such as the first Chern class and Pontryagin class are extracted. Then, thermodynamic data are extracted from this topological geometric model, and a stochastic thermodynamic analysis framework is constructed to calculate the entropy generation rate and verify the detailed fluctuation theorem. Critical slowing parameters are identified through order parameter dynamics analysis, the effective potential function is reconstructed, and multi-steady-state characteristics are determined, yielding phase transition features reflecting the stability of the cardiac electrical conduction system. Subsequently, a Vietoris-Rips complex is constructed based on the location of the cardiac magnetosensor, and persistent landscape features are generated using a persistent cohomology algorithm. Simultaneously, a cardiac conduction graph is constructed with myocardial regions as nodes and inter-regional electrical conduction intensities as edges, integrating topological invariants with phase transition features. The input is a topological neural network fused with graph convolution and manifold convolution. After updating node features through message passing and optimizing the output high-dimensional features through manifold convolution, strange attractors are reconstructed and the maximum Lyapunov exponent and correlation dimension are calculated to complete the quantitative analysis of the degree of chaos in the cardiac electrical conduction system. Finally, tensor fusion and particle filtering algorithms are used to construct a state-space model to predict risk trajectories, generate individualized risk feature profiles, and extract early warning indicators such as chaotic edge distance and topological fragility index. Combined with clinical factors, a comprehensive risk score is generated. If visualization is required, the risk propagation process is simulated based on cardiac DTI segmentation regions through reaction-diffusion equations to generate a four-dimensional spatiotemporal topological risk map and display it through mixed reality holography. Otherwise, the risk score and early warning report are directly output.

[0150] In summary, the technical solution in this application, by acquiring multi-channel cardiac magnetic signals and performing fidelity enhancement processing, can accurately extract core features of the cardiac magnetic field such as quantum coherence and magnetic vortices, breaking through the limitations of signal processing under the traditional classical electromagnetic theory framework; by constructing a topological geometric model of the cardiac fiber bundle and extracting topological invariants, it can effectively characterize the topological essence of the cardiac electrical conduction system, solving the problem that the Euclidean geometric model cannot adapt to the dynamic deformation of the heart; by conducting non-equilibrium phase transition dynamics detection on this model to obtain phase transition features, and combining it with a topological neural network to quantify the degree of chaos in the cardiac electrical conduction system, it can achieve accurate quantification of the stability and chaotic features of the cardiac system; finally, it generates individualized risk scores and early warning indicators, and can further generate a four-dimensional spatiotemporal topological risk map, which can significantly extend the early warning time of risk, improve the individualization of prediction, reduce the false positive rate, and meet the actual needs of clinical practice for early and accurate identification of high-risk patients with arrhythmia and sudden cardiac death.

[0151] Furthermore, as Figure 1 and Figure 3 The specific implementation of the method shown in this embodiment provides a device for predicting the risk of sudden cardiac death due to arrhythmia, such as... Figure 5 As shown, the device may include: a processing module 51, an extraction module 52, a detection module 53, an input module 54, and a generation module 55.

[0152] Processing module 51 can be used to acquire multi-channel magnetic field signals of the subject, perform fidelity enhancement processing on the multi-channel magnetic field signals, and extract core features that can characterize the magnetic field properties of the heart.

[0153] Extraction module 52 can be used to construct a cardiac fiber tract topological geometric model and extract topological invariants of the cardiac electrical conduction system based on core features and the cardiac fiber tract topological geometric model.

[0154] The detection module 53 can be used to perform non-equilibrium phase transition dynamics detection on the topological geometric model of the cardiac fiber bundle and obtain phase transition characteristics that reflect the stability of the cardiac electrical conduction system.

[0155] Input module 54 can be used to input topological invariants and phase transition features into a preset topological neural network to complete the quantitative analysis of the degree of chaos in the cardiac electrical conduction system;

[0156] The generation module 55 can be used to generate risk scores and early warning indicators for sudden cardiac death due to arrhythmia in examinees based on quantitative analysis results.

[0157] In some embodiments of this application, the processing module 51 can be specifically used to acquire multi-channel magnetocardiogram (MCG) signals of the subject in a resting state and a stimulated test state based on an OPM-MCG device array; estimate the decoherence channel characteristics of the multi-channel MCG signals using quantum process tomography; design a compensation pulse sequence based on the decoherence channel characteristics; use the compensation pulse sequence to enhance the multi-channel MCG signals with fidelity; convert the enhanced multi-channel MCG signals into a differential form of the magnetic field to complete the signal standardization process; calculate the magnetic topological charge density based on the standardized multi-channel MCG signals; perform an integral operation on the magnetic topological charge density to obtain the topological charge value characterizing the magnetic vortex structure of the heart, which serves as the core feature.

[0158] In some embodiments of this application, the extraction module 52 can be specifically used to construct a cardiac fiber bundle topological geometric model with the cardiac three-dimensional manifold as the base space and the U (1) phase space as the fiber; calculate the connection form of the cardiac fiber bundle topological geometric model based on the core features; perform external differential operation on the connection form to obtain the curvature tensor of the cardiac fiber bundle topological geometric model; calculate the first Chern class value representing the total number of cardiac topological charges and the Pontryagin class value representing the degree of fiber bundle torsion based on the curvature tensor; and use the first Chern class value and the Pontryagin class value as the topological invariants of the cardiac electrical conduction system.

[0159] In some embodiments of this application, the detection module 53 can be specifically used to extract thermodynamic data from the cardiac electrical conduction system characterized by the topological geometric model of the cardiac fiber tract; calculate the entropy generation rate of the cardiac electrical conduction system based on a stochastic thermodynamic analysis framework constructed from the thermodynamic data; analyze the order parameter dynamics of the cardiac electrical conduction system based on the thermodynamic data, detect the variation law of the order parameter near the critical point, and obtain the order parameter dynamics analysis results; calculate the relaxation time of the cardiac electrical conduction system based on the order parameter dynamics analysis results, fit the correlation between the relaxation time and the system state through dynamic scaling theory, and identify the critical slowing parameter of the cardiac electrical conduction system; reconstruct the effective potential energy function of the cardiac electrical conduction system based on the entropy generation rate, the order parameter dynamics analysis results, the relaxation time, and the critical slowing parameter; calculate the potential well depth based on the effective potential energy function; determine the multistable characteristics of the cardiac electrical conduction system based on the magnitude of the potential well depth; and integrate the entropy generation rate, the critical slowing parameter, the potential well depth, and the corresponding multistable characteristics as phase transition characteristics reflecting the stability of the cardiac electrical conduction system.

[0160] In some embodiments of this application, the input module 54 can be specifically used to construct a Vietoris-Rips complex based on the location of the cardiac magnetocardiogram sensor, extract features from topological invariants using a persistent cohomology algorithm, and generate persistent landscape features characterizing the topological properties of the cardiac electrical conduction system; based on the myocardial region division results of the cardiac electrical conduction system, construct a cardiac conduction graph of a topological neural network with myocardial regions as nodes and inter-regional electrical conduction connection strength as edges; input the persistent landscape features and phase transition features into the topological neural network based on the cardiac conduction graph, and update the network node features through a message passing mechanism; based on the manifold properties of the cardiac fiber bundle topological geometric model, optimize the updated network node features using manifold convolution operations to ensure that the network output satisfies covariance under normalization transformation, and obtain high-dimensional features for quantifying the degree of chaos; based on the high-dimensional features, reconstruct the strange attractor of the cardiac electrical conduction system, calculate the maximum Lyapunov exponent and correlation dimension of the cardiac electrical conduction system, and use them as the quantitative analysis results of the degree of chaos in the cardiac electrical conduction system.

[0161] In some embodiments of this application, the generation module 55 can be specifically used to employ a tensor fusion architecture to perform cross-modal feature fusion of quantitative analysis results with topological invariants and phase transition features, outputting fused high-dimensional risk features; based on the high-dimensional risk features, a state-space model of the cardiac electrical conduction system is constructed using a particle filtering algorithm to predict the risk trajectory change trend of the cardiac electrical conduction system; based on the risk trajectory change trend, the t-SNE algorithm is used to reduce the high-dimensional risk features to a two-dimensional space, generating an individualized risk feature profile of the examinee; based on the individualized risk feature profile, the chaotic edge distance and topological fragility index of the cardiac electrical conduction system are calculated as early warning indicators; and the early warning indicators are fused with clinical factors to generate a risk score for the examinee regarding arrhythmia and sudden death.

[0162] In some embodiments of this application, such as Figure 6 As shown, the device also includes: a display module 56;

[0163] The display module 56 can be used to segment the myocardial region of the cardiac electrical conduction system based on diffusion tensor imaging data, obtaining segmented regions that correspond one-to-one with the nodes of the topological neural network; based on the topological characteristics of the cardiac fiber tract topological geometry model, it calculates the local persistent cohomology features of each segmented region, and generates a regional topological feature vector based on the local persistent cohomology features and topological invariants; using the regional topological feature vector and early warning indicators as input parameters, it simulates the propagation process of risk in different segmented regions of the cardiac electrical conduction system through the reaction-diffusion equation, obtaining the risk propagation rate and the risk level distribution of each segmented region; based on the risk level distribution and risk trajectory change trend of each segmented region, it combines the risk distribution in three-dimensional space with the risk evolution process in the time dimension to generate a four-dimensional spatiotemporal topological risk map of the cardiac electrical conduction system; and realizes the three-dimensional holographic display of the four-dimensional spatiotemporal topological risk map through a mixed reality visualization interface.

[0164] It should be noted that other corresponding descriptions of the functional units involved in the arrhythmia-induced sudden cardiac death risk prediction device provided in this embodiment can be found in [reference needed]. Figure 1 and Figure 3 The corresponding description in [the document] will not be repeated here.

[0165] Based on the above, Figure 1 and Figure 3 Accordingly, this embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the above-described method. Figure 1 and Figure 3 The method shown is for predicting the risk of sudden cardiac death due to arrhythmia.

[0166] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause an electronic device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0167] Based on the above, Figure 1 and Figure 3 The method shown, and Figure 5 , 6 To achieve the above objectives, the present application also provides an electronic device, specifically a personal computer, tablet computer, server, or other network device, as shown in the virtual device embodiment. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to achieve the above-described objectives. Figure 1 and Figure 3 The method shown is for predicting the risk of sudden cardiac death due to arrhythmia.

[0168] Optionally, the aforementioned physical devices may also include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, input units such as a keyboard, etc., and optional user interfaces may also include USB interfaces, card reader interfaces, etc. The network interface may optionally include standard wired interfaces, wireless interfaces (such as Wi-Fi interfaces), etc.

[0169] Those skilled in the art will understand that the physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0170] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0171] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platform, or it can be implemented by hardware.

[0172] This invention, through the acquisition of multi-channel cardiac magnetic signals and the execution of fidelity enhancement processing, can accurately extract core features of the cardiac magnetic field, such as quantum coherence and magnetic vortices, overcoming the limitations of signal processing under the traditional classical electromagnetic theory framework. By constructing a topological geometric model of the cardiac fiber bundle and extracting topological invariants, it can effectively characterize the topological essence of the cardiac electrical conduction system, solving the problem that the Euclidean geometric model cannot adapt to the dynamic deformation of the heart. By conducting non-equilibrium phase transition dynamics detection on this model to obtain phase transition features, and combining it with a topological neural network to quantify the degree of chaos in the cardiac electrical conduction system, it can achieve accurate quantification of the stability and chaotic features of the cardiac system. Finally, it generates individualized risk scores and early warning indicators, and can further generate a four-dimensional spatiotemporal topological risk map, which can significantly extend the early warning time of risks, improve the individualization of predictions, reduce the false positive rate, and meet the actual clinical needs for early and accurate identification of high-risk patients with arrhythmia and sudden cardiac death.

[0173] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes shown in the drawings are not necessarily essential for implementing this application. Those skilled in the art will understand that the modules in the apparatus of the embodiment can be distributed within the apparatus of the embodiment as described, or can be modified to be located in one or more apparatuses different from this embodiment. The modules of the above-described embodiment can be combined into one module, or further divided into multiple sub-modules.

[0174] The serial numbers in this application are for descriptive purposes only and do not represent the superiority or inferiority of any particular implementation scenario. The above disclosures are merely a few specific implementation scenarios of this application; however, this application is not limited thereto, and any variations conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A method for predicting the risk of sudden cardiac death due to arrhythmia, characterized in that, include: Multi-channel magnetic field signals of the subject were collected, and the multi-channel magnetic field signals were subjected to fidelity enhancement processing to extract core features that can characterize the magnetic field properties of the heart. A topological geometric model of the cardiac fiber tract is constructed, and topological invariants of the cardiac electrical conduction system are extracted based on the core features and the topological geometric model of the cardiac fiber tract. Non-equilibrium phase transition dynamics were detected on the cardiac fiber plexus topological geometric model to obtain phase transition characteristics reflecting the stability of the cardiac electrical conduction system; The topological invariants and phase transition features are input into a preset topological neural network to complete the quantitative analysis of the degree of chaos in the cardiac electrical conduction system. Based on the quantitative analysis results, risk scores and early warning indicators for sudden cardiac death due to arrhythmia were generated for the subjects.

2. The method according to claim 1, characterized in that, The multi-channel magnetocardiogram (MCC) signals of the subjects are collected, and fidelity enhancement processing is performed on the multi-channel MCC signals to extract core features that characterize the magnetic field properties of the heart, including: Multi-channel magnetocardiogram signals of subjects were acquired based on the OPM-MCG equipment array during resting and provocation test states. The decoherence channel characteristics of the multi-channel magnetic field signal are estimated using quantum process tomography. A compensation pulse sequence is designed based on the decoherence channel characteristics, and the compensation pulse sequence is used to enhance the multi-channel magnetic field signal with high fidelity. The enhanced multi-channel magnetocardiogram signal is converted into a differential form of the magnetic field, thus completing the signal standardization process; The magnetic topological charge density is calculated based on the standardized multi-channel cardiac magneto-particle signals. The magnetic topological charge density is then integrated to obtain the topological charge value that characterizes the magnetic vortex structure of the heart, which serves as the core feature.

3. The method according to claim 1, characterized in that, The construction of the cardiac fiber plexus topological geometric model, and the extraction of topological invariants of the cardiac electrical conduction system based on the core features and the cardiac fiber plexus topological geometric model, includes: A topological geometric model of the cardiac fiber bundle is constructed using the three-dimensional manifold of the heart as the base space and the U (1) phase space as the fiber. Based on the core features, the connection form of the cardiac fiber tract topological geometry model is calculated. The external differential operation is performed on the connection form to obtain the curvature tensor of the cardiac fiber tract topological geometry model. Based on the curvature tensor, the first Chern class value representing the total cardiac topological load and the Pontryagin class value representing the degree of fiber tract torsion are calculated. The first Chen class value and the Pontryagin class value are used as topological invariants of the cardiac electrical conduction system.

4. The method according to claim 1, characterized in that, Non-equilibrium phase transition dynamics were performed on the cardiac fiber plexus topological geometric model to obtain phase transition characteristics reflecting the stability of the cardiac system, including: Thermodynamic data are extracted from the cardiac electrical conduction system characterized by the cardiac fiber plexus topological geometry model. Based on the stochastic thermodynamic analysis framework constructed according to the thermodynamic data, the entropy generation rate of the cardiac electrical conduction system is calculated. Based on the thermodynamic data, the order parameter dynamics of the cardiac electrical conduction system are analyzed to detect the variation law of the order parameter near the critical point and obtain the order parameter dynamics analysis results. Based on the results of the order parameter dynamic analysis, the relaxation time of the cardiac electrical conduction system is calculated. The correlation between the relaxation time and the system state is fitted by dynamic scaling theory to identify the critical slowing parameter of the cardiac electrical conduction system. Based on the entropy generation rate, the order parameter dynamic analysis results, the relaxation time, and the critical slowing parameter, the effective potential energy function of the cardiac electrical conduction system is reconstructed, the potential well depth is calculated based on the effective potential energy function, and the multistable characteristics of the cardiac electrical conduction system are determined according to the magnitude of the potential well depth. The entropy generation rate, the critical slowing parameter, the potential well depth, and the corresponding multistable characteristics are integrated as phase transition features reflecting the stability of the cardiac electrical conduction system.

5. The method according to claim 1, characterized in that, The step of inputting the topological invariants and the phase transition features into a preset topological neural network to complete the quantitative analysis of the degree of chaos in the cardiac electrical conduction system includes: A Vietoris-Rips complex is constructed based on the location of the cardiac magnetosensor. The topological invariants are then extracted using a persistent cohomology algorithm to generate persistent landscape features that characterize the topological properties of the cardiac electrical conduction system. Based on the myocardial region segmentation results of the cardiac electrical conduction system, a cardiac conduction graph of a topological neural network is constructed with myocardial regions as nodes and the electrical conduction connection strength between regions as edges. The persistent landscape features and the phase transition features are input into a topological neural network based on the cardiac conduction map, and the network node features are updated through a message passing mechanism. Based on the manifold properties of the cardiac fiber bundle topological geometry model, manifold convolution operation is used to optimize the updated network node features to ensure that the network output satisfies the covariance property under normal transformation, thereby obtaining high-dimensional features for quantifying the degree of chaos. Based on the high-dimensional features, the strange attractor of the cardiac electrical conduction system is reconstructed, and the maximum Lyapunov exponent and correlation dimension of the cardiac electrical conduction system are calculated as the quantitative analysis results of the degree of chaos of the cardiac electrical conduction system.

6. The method according to claim 1, characterized in that, Based on the quantitative analysis results, the risk score and early warning indicators for sudden cardiac death due to arrhythmia are generated for the examinee, including: A tensor fusion architecture is adopted to fuse the quantitative analysis results with the topological invariants and the phase transition features across modal features, and output the fused high-dimensional risk features; Based on the high-dimensional risk characteristics, a state-space model of the cardiac electrical conduction system is constructed using a particle filtering algorithm to predict the risk trajectory change trend of the cardiac electrical conduction system. Based on the risk trajectory change trend, the t-SNE algorithm is used to reduce the high-dimensional risk features to two-dimensional space to generate an individualized risk feature profile of the examinee. Based on the individualized risk profile, the chaotic edge distance and topological vulnerability index of the cardiac electrical conduction system are calculated as early warning indicators; By integrating the aforementioned early warning indicators with clinical factors, a risk score for sudden cardiac death due to arrhythmia is generated for the examinee.

7. The method according to claim 6, characterized in that, After generating the risk score and early warning indicators for sudden cardiac death due to arrhythmia in the subject, the method further includes: Based on diffusion tensor imaging data, myocardial regions of the cardiac electrical conduction system are segmented to obtain segmented regions that correspond one-to-one with the nodes of the topological neural network. Based on the topological properties of the cardiac fiber plexus topological geometry model, the local persistent homology features of each segmented region are calculated, and a region topological feature vector is generated based on the local persistent homology features and the topological invariants. Using the regional topological feature vector and the early warning index as input parameters, the risk propagation process in different segmented regions of the cardiac electrical conduction system is simulated through the reaction-diffusion equation to obtain the risk propagation rate and the risk level distribution of each segmented region. Based on the risk level distribution of each segmented region and the risk trajectory change trend, the risk distribution in three-dimensional space is combined with the risk evolution process in the time dimension to generate a four-dimensional spatiotemporal topological risk map of the cardiac electrical conduction system. A three-dimensional holographic display of the four-dimensional spatiotemporal topological risk map is achieved through a mixed reality visualization interface.

8. A device for predicting the risk of sudden cardiac death due to arrhythmia, characterized in that, include: The processing module is used to acquire multi-channel magnetic-cardiogram signals from the subject, perform fidelity enhancement processing on the multi-channel magnetic-cardiogram signals, and extract core features that can characterize the magnetic field properties of the heart. The extraction module is used to construct a topological geometric model of the cardiac fiber tract and extract topological invariants of the cardiac electrical conduction system based on the core features and the topological geometric model of the cardiac fiber tract. The detection module is used to perform non-equilibrium phase transition dynamics detection on the cardiac fiber bundle topological geometry model to obtain phase transition characteristics that reflect the stability of the cardiac electrical conduction system. The input module is used to input the topological invariants and the phase transition features into a preset topological neural network to complete the quantitative analysis of the degree of chaos in the cardiac electrical conduction system. The generation module is used to generate risk scores and early warning indicators for sudden cardiac death due to arrhythmia for the subjects based on the quantitative analysis results.

9. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.

10. An electronic device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 7.

Images