Multi-scale biomagnetic response dynamics modeling method, device, equipment and medium

Abstract

This application relates to the fields of biomedical engineering and precision medicine, and discloses a method, device, equipment, and medium for multi-scale biomagnetic response dynamics modeling. The method includes: simultaneously acquiring response data of biological samples at different scales under an extremely weak magnetic field environment; performing standardized preprocessing on the response data at each scale to obtain a multi-scale signal sequence; extracting conservative response modes from the multi-scale signal sequence based on cross-scale modal resonance analysis; extracting quantitative features from the conservative response modes to form a conservative feature vector; constructing and training a population-level hybrid structure state-space dynamics model; performing personalized calibration of the hybrid structure state-space dynamics model based on Bayesian inference and individual data, and generating a magnetic response fingerprint; calculating a risk score based on the magnetic response fingerprint and outputting the risk assessment result. This application achieves systematic modeling and individualized characterization of multi-scale biomagnetic responses.

Description

Technical Field

[0001] This disclosure relates to the fields of biomedical engineering and precision medicine, and more specifically, to a multi-scale biomagnetic response dynamics modeling method, device, equipment, and medium. Background Technology

[0002] With the development of non-invasive physical stimulation techniques, weak magnetic fields (WMFs) have shown great promise in areas such as neural modulation, cardiac rhythm regulation, and directed stem cell differentiation due to their advantages of being non-invasive, low-energy, and highly penetrating. Studies have shown that alternating or static magnetic fields in the range of 10n–100 μT can induce cross-scale physiological responses by influencing ion channel gating, free radical pair mechanisms, and the mitochondrial electron transport chain. However, current research faces three major bottlenecks: fragmented research scales, with most studies focusing on a single level; outdated feature extraction methods; and the tendency to overlook individual differences. Summary of the Invention

[0003] In view of the above situation, this application provides a multi-scale biomagnetic response dynamics modeling method, device, equipment and medium, which aims to solve the above problems or at least partially solve the above problems.

[0004] In a first aspect, embodiments of this application provide a multi-scale biomagnetic response dynamics modeling method, the method comprising:

[0005] In an extremely weak magnetic field environment, response data of biological samples at different scales are collected simultaneously. The different scales include at least one of the following: subcellular, single-cell, tissue and organ levels. The response data includes at least one of the following: electrophysiological signals, optical signals and morphodynamic information.

[0006] The response data at each scale are standardized and preprocessed to obtain a multi-scale signal sequence.

[0007] Based on cross-scale modal resonance analysis, conservative response modes are extracted from the multi-scale signal sequence;

[0008] Quantitative features are extracted from conservative response modes to form a conservative feature vector. The quantitative features include at least one of the following: response peak time, signal attenuation coefficient, characteristic frequency bandwidth, phase lock value, modal energy ratio, and cross-scale consistency index.

[0009] A population-level hybrid structure state-space dynamics model is constructed and trained, wherein the observed output of the hybrid structure state-space dynamics model is a conservative feature vector;

[0010] Based on Bayesian inference and individual data, the hybrid structure state-space dynamics model is individually calibrated, and a magnetic response fingerprint is generated.

[0011] The magnetic response fingerprint is compared with a preset health database and a disease database, respectively, and a risk score is calculated and the risk assessment result is output.

[0012] Secondly, embodiments of this application also provide a multi-scale biomagnetic response dynamics modeling device, the device comprising:

[0013] The acquisition module is used to simultaneously acquire response data of biological samples at different scales in an extremely weak magnetic field environment. The different scales include at least one of the following: subcellular, single-cell, tissue and organ levels. The response data includes at least one of the following: electrophysiological signals, optical signals and morphodynamic information.

[0014] The preprocessing module is used to perform standardized preprocessing on response data at various scales to obtain multi-scale signal sequences;

[0015] An extraction module is used to extract conservative response modes from the multi-scale signal sequence based on cross-scale modal resonance analysis; and to extract quantitative features from the conservative response modes to form a conservative feature vector, wherein the quantitative features include at least one of the following: response peak time, signal attenuation coefficient, characteristic frequency bandwidth, phase lock value, modal energy ratio, and cross-scale consistency index;

[0016] A construction module is used to construct and train a population-level hybrid structure state-space dynamics model, wherein the observed output of the hybrid structure state-space dynamics model is a conservative feature vector;

[0017] The generation module is used to perform personalized calibration on the hybrid structure state-space dynamics model based on Bayesian inference and individual data, and to generate magnetic response fingerprints.

[0018] The calculation module is used to compare the magnetic response fingerprint with a preset health database and a disease database, respectively, calculate a risk score, and output a risk assessment result.

[0019] Thirdly, embodiments of this application also provide an electronic device, including: a processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform the steps described in the first aspect.

[0020] Fourthly, embodiments of this application also provide a computer-readable storage medium that stores one or more programs, which, when executed by an electronic device including multiple applications, cause the electronic device to perform the steps described in the first aspect.

[0021] The above-mentioned technical solutions adopted in the embodiments of this application can achieve the following beneficial effects: by synchronously collecting response data at different scales and extracting conservative response modes based on cross-scale modal resonance analysis, the limitations of traditional research scale separation are overcome; the intrinsic dynamic parameters such as response peak time, signal attenuation coefficient, and phase lock value extracted from the conservative response modes not only retain the consistent conservative response information across scales but also have clear physiological significance, avoiding the mode aliasing problem that is prone to occur in traditional univariate decomposition methods, and providing high-quality input features for subsequent modeling; and by using conservative feature vectors as observation outputs, a hybrid structural state containing linear conduction terms and nonlinear control terms is constructed. The spatial model not only characterizes the basic conduction pathways of magnetic field response but also captures complex collaborative regulatory behaviors across scales, outperforming pure black-box or pure mechanistic models. By fusing individual conservative feature vectors with physiological baseline data through Bayesian inference, the population model is personalized and low-dimensional magnetic response fingerprints are generated. This preserves individual differences while maintaining repeatability and comparability, providing a quantitative basis for personalized medical assessment. By comparing the magnetic response fingerprints with health and disease databases to calculate quantitative risk scores, an objective and quantitative assessment tool is provided for early disease warning and precise intervention, significantly promoting the clinical translation and application of extremely weak magnetic field technology in the field of precision medicine. 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 of this application 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 the multi-scale biomagnetic response dynamics modeling method provided in this application embodiment is shown.

[0024] Figure 2 A flowchart of a multi-scale biomagnetic response dynamics modeling method provided in another embodiment of this application is shown;

[0025] Figure 3 This paper illustrates a flowchart of the model construction process for the hybrid structure state-space model provided in an embodiment of this application.

[0026] Figure 4 A flowchart of a multi-scale biomagnetic response dynamics modeling method provided in another embodiment of this application is shown;

[0027] Figure 5 A flowchart of a multi-scale biomagnetic response dynamics modeling method provided in another embodiment of this application is shown;

[0028] Figure 6The system structure diagram of the multi-scale biomagnetic response dynamics modeling method provided in the embodiments of this application is shown.

[0029] Figure 7 A structural diagram of the multi-scale biomagnetic response dynamics modeling device provided in an embodiment of this application is shown;

[0030] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such use can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the term "comprising" and its variations should be interpreted as open-ended terms meaning "including but not limited to."

[0033] As described in the background section, existing research faces three major bottlenecks. Specifically: 1. Fragmented research scales: Most studies focus on a single level—subcellular level focusing on ROS generation or Ca2+ fluctuations; single-cell level recording action potential firing; organ level using fMRI / MEG to observe brain network changes; lack of a unified framework to correlate signals at different levels, resulting in an inability to reveal the overall conduction pathway of magnetic field effects. 2. Outdated feature extraction methods: Traditional intrinsic mode decomposition is only applicable to univariate signals, making it difficult to handle multi-scale heterogeneous data and prone to modality aliasing problems, resulting in a lack of cross-scale consistency in feature extraction. 3. Lack of modeling of individual differences: Population average models mask individual specificity, failing to provide personalized assessment criteria for clinical practice, thus hindering the transformation of extremely weak magnetic field technology into precision medicine.

[0034] Therefore, there is an urgent need to develop a new method that can simultaneously acquire multi-scale responses, extract conservative dynamic features, construct a unified dynamic model, and generate individualized representations.

[0035] Figure 1 This paper illustrates a flowchart of the multi-scale biomagnetic response dynamics modeling method provided in an embodiment of this application. Figure 1It can be seen that this application includes at least steps S101-S107:

[0036] Step S101: Under an extremely weak magnetic field environment, simultaneously collect response data of biological samples at different scales.

[0037] Specifically, a controllable, extremely weak magnetic field stimulation environment is constructed. The core equipment is a triaxial controllable Helmholtz coil system, capable of generating sinusoidal alternating magnetic field (AMF) or pulsed magnetic field (PMF). The magnetic field strength range is set to 10 nT–100 μT, with fine-tuning in 0.1 nT increments via a controller. The frequency range is set to 0.1 Hz–100 Hz, with typical frequency bands of 1 Hz, 10 Hz, and 40 Hz. Furthermore, the coil system requires gradient compensation to ensure magnetic field uniformity in the acquisition area, with gradient field compensation reduced to <1 nT / m to eliminate interference from spatial non-uniformity.

[0038] Furthermore, the different scales include at least one of the following: subcellular, single-cell, tissue, and organ levels. Under magnetic field stimulation, response data at different biological scales are simultaneously acquired using the following acquisition devices. Specifically, for the subcellular scale, the detection target is mitochondrial membrane potential. , For oscillation and ROS levels, fluorescent probes (TMRE, Fluo-4 AM, DCFH-DA) and confocal microscopy were configured with excitation wavelengths of 488 nm / 561 nm and a frame rate of 30 fps. For single-cell scale, the detection targets were action potential sequences, firing frequencies, and synaptic morphological changes, using a multi-channel microelectrode array (MEA, 64 channels) and a super-resolution microscope (STED) with a sampling rate of 20 kHz and a spatial resolution of 50 nm. For tissue scale, the detection targets were local field potentials (LFP) and calcium imaging signal distribution, using a multiphoton microscope and an LFP electrode array with a field of view depth of 500 μm and a scanning frequency of 2 Hz. For organ scale, the detection targets were brain functional connectivity, blood oxygen metabolism, and cardiac magnetic activity, using fMRI / fNIRS and OPM-MEG / SERF-MCG with a temporal resolution of 0.5 s.

[0039] In addition, a time synchronization mechanism needs to be set up to ensure that multiple devices collect data synchronously. For example, GPS timing PPS pulses and the IEEE 1588 precision time protocol can be used to control the triggering error of all devices to be less than 0.1ms. After synchronous acquisition is completed, all data is stored in a unified format, such as HDF5, and labeled with timestamps and scale tags.

[0040] The response data mentioned above includes at least one of the following: electrophysiological signals, optical signals, and morphodynamic information. Electrophysiological signals reflect changes in potential or magnetic fields generated by charge flow within the organism, such as action potential sequences recorded by multichannel microelectrode arrays at the single-cell scale, and local field potentials recorded by LFP electrode arrays at the tissue scale. Optical signals represent the conversion of specific biochemical processes into detectable light signals through fluorescent probe labeling, such as calcium ion concentration acquired by fluorescent probe signals at the subcellular scale, and calcium imaging signals acquired at the tissue scale. Morphodynamic information records changes in the structure, shape, position, or movement of biological samples under stimulation, such as organelle dynamics at the subcellular scale, cell structure dynamics at the single-cell scale, and microstructure dynamics and connections at the tissue scale.

[0041] Step S102: Standardize and preprocess the response data at each scale to obtain a multi-scale signal sequence.

[0042] The preprocessing includes detrending, bandpass filtering, and standardization.

[0043] Step S103: Based on cross-scale modal resonance analysis, extract conservative response modes from multi-scale signal sequences.

[0044] Step S104: Extract quantitative features from the conservative response modes to form a conservative feature vector.

[0045] The quantitative characteristics include at least one of the following: peak response time, signal attenuation coefficient, characteristic frequency bandwidth, phase lock value, modal energy ratio, and cross-scale consistency index.

[0046] Specifically, for each conservative response mode obtained in step S103 Calculate a set of intrinsic dynamic parameters that characterize its dynamic properties to form the feature descriptor of this mode. :

[0047] Peak response time The time elapsed from the initial point of magnetic field stimulation until the amplitude of this mode (obtained through Hilbert transform) reaches its first maximum value;

[0048] Signal attenuation coefficient : Apply a single exponential function to the modal envelope decay segment after the stimulus peak. The attenuation rate constant is obtained by fitting the data.

[0049] Characteristic frequency bandwidth Calculate the full width at half maximum (FWHM) of the instantaneous frequency distribution (Hilbert spectrum) of this mode, which reflects the stability of the oscillation frequency;

[0050] Phase Lock Value : Calculate the statistic of synchronicity between the instantaneous phase of this mode and the phase of the external periodic magnetic field stimulus, with a value ranging from 0 to 1;

[0051] Modal energy ratio The energy of this mode within the time window The percentage of total energy of all conservative modes;

[0052] All Feature descriptor of a conservative mode By assembling them in sequence, a conservative eigenvector that comprehensively characterizes this magnetic response experiment is formed. (For example, if M=3, and each descriptor is 6-dimensional, then D=18). This feature vector It is a static, high-dimensional generalized feature that will serve as the a priori adjustment basis for personalized calibration in the subsequent step S106.

[0053] Step S105: Construct and train a population-level hybrid structure state-space dynamics model.

[0054] The observed output of the hybrid structure state-space dynamics model is a conservative eigenvector.

[0055] Step S106: Based on Bayesian inference and individual data, perform personalized calibration on the hybrid structure state-space dynamics model and generate a magnetic response fingerprint.

[0056] The Bayesian inference framework is a probabilistic statistical inference paradigm based on Bayes' theorem. Its core idea is to use new observational data to update the probability distribution of a certain model parameter. In this embodiment, the group model parameters are used as priors, and individual physiological baseline data are used as new evidence. The updated posterior distribution of the model parameters belonging to that individual is calculated.

[0057] Individual data represents static physiological and clinical background information that is relevant to the individual and does not change with the current magnetic field experiment. Examples include age, sex, genetic markers, clinical scores, medication use, and lifestyle habits. In this embodiment, these are used as covariates to adjust prior information, so that the prior probability is no longer a fixed population average, but a dynamic starting point that changes with individual characteristics.

[0058] A magnetic response fingerprint (MRF) is a low-dimensional, unique, repeatable, and comparable numerical vector obtained by encoding and reducing the dimensionality of a personalized, calibrated model parameter set. Magnetic response fingerprints are unique, similar to biological fingerprints, and can distinguish between different individuals.

[0059] Step S107: Compare the magnetic response fingerprint with the preset health database and disease database respectively, calculate the risk score and output the risk assessment result.

[0060] By comparing magnetic response fingerprints with health and disease databases, the degree of deviation of an individual from healthy or diseased groups can be determined, thus avoiding the bias that may arise from a single reference standard. Furthermore, health and disease databases can be continuously updated and optimized with the addition of new cases, enabling risk assessment to evolve continuously and ensuring its accuracy.

[0061] from Figure 1 As can be seen from the method shown, this application overcomes the limitations of traditional research scale fragmentation by simultaneously collecting response data at different scales and extracting conservative response modes based on cross-scale modal resonance analysis. The intrinsic dynamic parameters extracted from the conservative response modes, such as response peak time, signal attenuation coefficient, and phase lock value, retain both consistent conservative response information across scales and clear physiological significance, avoiding the mode aliasing problem easily encountered in traditional univariate decomposition methods, and providing high-quality input features for subsequent modeling. Using conservative feature vectors as observation outputs, a hybrid structure state-space model containing linear conduction terms and nonlinear control terms is constructed, which characterizes... The fundamental conduction pathways of magnetic field response can capture complex collaborative regulatory behaviors across scales, outperforming pure black-box models or pure mechanistic models. By fusing individual conservative feature vectors with physiological baseline data through Bayesian inference, the population model is personalized and low-dimensional magnetic response fingerprints are generated. This preserves individual differences while maintaining repeatability and comparability, providing a quantitative basis for personalized medical assessment. By comparing magnetic response fingerprints with health and disease databases to calculate quantitative risk scores, an objective and quantitative assessment tool is provided for early warning and precise intervention of diseases, significantly promoting the clinical translation and application of extremely weak magnetic field technology in the field of precision medicine.

[0062] In some embodiments of this application, in the above method, in step S102, preprocessed data is obtained by performing detrending, bandpass filtering, and standardization on the response data at each scale. Specifically, detrending removes long-term trends or baseline drift from the signal, causing it to fluctuate around zero. Bandpass filtering preserves frequency components related to the physiological response while filtering out high-frequency noise and DC offset. Standardization includes calculating the mean and standard values.

[0063] Further based on the original response data mean and standard values Perform standard value transformation. Specifically, the preprocessed data is obtained based on the following formula. : .

[0064] The preprocessed data at each scale are further multiplied by preset scale weights and then summed to obtain the fused data. The preset scale weights are determined based on the signal-to-noise ratio and physiological importance. For example, considering the subcellular scale as the initial response source point, the weight of subcellular is set to 0.3; considering the single-cell scale as a key relay node, the weight of single cells is set to 0.25; considering the tissue scale as a regional integration platform, the weight of tissue is set to 0.25; and considering the organ scale as macroscopic output performance, the weight of organ is set to 0.2.

[0065] Specifically, the following formula is used for signal fusion: ;in, Indicates the fused signal. Indicates weight, This indicates the preprocessed data.

[0066] In the embodiments of this application, by preprocessing and fusing multi-source heterogeneous data, an indispensable and high-quality data foundation is provided for the subsequent extraction of conservative features with cross-scale consistency and the construction of reliable individualized models.

[0067] In some embodiments of this application, in the above method, the feature extraction process of step S103 includes steps S1031-S1033, such as... Figure 2 As shown:

[0068] Step S1031: Decompose each preprocessed single-scale signal sequence using ensemble empirical mode decomposition to obtain the intrinsic mode function components of each scale.

[0069] For each preprocessed single-scale signal The Ensemble Empirical Mode Decomposition (EEMD) algorithm is applied independently. EEMD effectively overcomes the mode mixing problem by adding white noise to assist the analysis, adaptively decomposing the signal at each scale into a set of eigenmode functions. and a residual term, in which Indicates the modal order (from high frequency to low frequency).

[0070] Step S1032: Calculate the cross-correlation coefficients between the intrinsic mode function components whose center frequencies differ within a preset range at different scales, and select mode combinations that appear at at least three scales and whose average cross-correlation coefficients between pairs of elements are greater than a preset threshold, as cross-scale resonance modes.

[0071] Set a frequency tolerance window (e.g., ±0.5 Hz) and a time window for conditions that occur at at least three biological scales and whose average cross-correlation coefficients between any two conditions are greater than a preset threshold (e.g., the preset threshold is 100 Hz). The mode set of ) is considered to constitute a cross-scale resonant mode.

[0072] The cross-correlation coefficient can be the Pearson correlation coefficient, etc.

[0073] Step S1033: Take a weighted average of the intrinsic mode function components of each scale contained in each cross-scale resonance mode to generate a conservative response mode.

[0074] For each selected cross-scale resonant mode group, its constituent scales are... IMF The components are weighted and averaged, with the weights determined based on the signal-to-noise ratio of the original signal to which each component belongs, ultimately generating a conservative response mode. ,in , This represents the total number of modes selected. This represents the core dynamic response mode of biological systems to extremely weak magnetic field stimulation, which involves coordinated oscillations at multiple tissue levels.

[0075] In this embodiment, by performing ensemble empirical mode decomposition on signals at each scale and calculating cross-correlation coefficients among modes with similar frequencies, mode combinations that appear simultaneously at at least three scales and whose correlations meet a threshold are selected as cross-scale resonant modes. Then, a weighted average is used to generate conservative response modes. This method effectively overcomes the limitation of traditional single-scale decomposition in capturing cross-scale collaborative information. Through the dual constraints of frequency tolerance and correlation, it accurately identifies the core oscillation modes that run from the subcellular to the organ level, avoiding mode aliasing and spurious correlations. At the same time, the weighted fusion retains the contributions of each scale, providing robust feature inputs with clear physiological significance and cross-scale consistency for subsequent dynamic modeling, thus laying a solid foundation for revealing the overall conduction path of magnetic field effects.

[0076] In some embodiments of this application, the hybrid structure state-space dynamics model in step S105 of the above method is constructed as follows:

[0077]

[0078] in, Represents a d-dimensional potential state vector. Indicates an external magnetic field input. Indicates a conservative response mode. Represents a linear parameter matrix, Indicates by parameters The defined two-layer fully connected neural network, This represents the Gaussian noise term.

[0079] Using a large amount of data collected from a group of healthy volunteers Train this model. The expectation-maximization algorithm is used to jointly estimate the latent state sequence. and all model parameters In the E-step of EM, the posterior distribution of the states is estimated using an extended Kalman filter and smoother; in the M-step, the model parameters are updated based on the estimated states. The optimal latent state dimension is determined by cross-validation or the information criterion (AIC / BIC). After training, a population baseline model is obtained. .

[0080] Specifically, since the model contains latent state variables that cannot be directly observed, directly maximizing the likelihood function is difficult to solve. Therefore, an expectation-maximization algorithm is used for iterative estimation. In each expectation step, the current model parameters are fixed, and the nonlinear state-space model is approximated using an extended Kalman filter. The posterior distribution and covariance of the latent state vector over the entire time series are recursively estimated and smoothed. In the subsequent maximization step, the latent state estimates obtained in the expectation step are fixed, and the expected value of the log-likelihood function of the complete data is used as the objective. At this point, the parameter update of the linear part of the model can be transformed into a constrained linear regression problem to obtain an analytical solution, while the parameters of the nonlinear neural network part are optimized using gradient descent. The noise covariance matrix can also be re-estimated based on the residuals. This expectation step and maximization step are iterated repeatedly until the change in the log-likelihood function value of the model converges to a preset threshold, thus obtaining a set of optimal parameters that maximize the probability of the observed data. Finally, to determine the optimal potential state vector dimension as a key hyperparameter in the model and avoid overfitting or underfitting, this application adopts a model selection criterion for quantitative evaluation: complete hybrid structure state space dynamics models are trained on a pre-defined set of dimension candidates, and their respective Akaike information criterion values ​​or Bayesian information criterion values ​​are calculated. Finally, the dimension that minimizes the criterion value is selected as the optimal structure of the model.

[0081] like Figure 3 As shown, the hybrid state-space model constructed in this application innovatively integrates a linear dynamics module with a nonlinear neural network module to balance the physical interpretability of the basic conduction path in biomagnetic responses with the complex nonlinear characteristics of cross-scale collaborative regulation. The model's state-space equations consist of two superimposed parts: the linear part characterizes the basic dynamics of magnetic field stimulation conduction along the hierarchy in the biological system; the nonlinear part is implemented through a two-layer fully connected neural network to capture collaborative regulation, feedback loops, and nonlinear coupling effects across different scales. The model's observation equations map the latent state to an observable output, i.e., the predicted response mode waveform, thereby achieving accurate reconstruction of multi-scale magnetic response signals.

[0082] During training, the Expectation-Maximization (EM) algorithm and Extended Kalman Filter (EKF) are jointly used to iteratively optimize the model parameters. EKF recursively estimates the latent state given the parameters, while EM updates the linear parameter matrix and neural network weights based on the state estimate. These two methods alternate until convergence. To determine the optimal model structure, the AIC and BIC criteria are introduced. By plotting the curves of the criterion values ​​versus complexity, a balance is struck between fitting accuracy and model simplicity, avoiding overfitting. After completing the training of the population model, Bayesian inference is used to personalize the model parameters for individual data, outputting individualized posterior parameter distributions. These distributions are then visually displayed as distribution plots, visually illustrating the differences in key dynamic parameters among different individuals, providing a quantitative basis for subsequent generation of magnetic response fingerprints and disease risk assessment.

[0083] The hybrid structure state-space model constructed in this application integrates linear dynamics and nonlinear neural network modules, and its state-space equation is the sum of linear and nonlinear terms. The model input consists of multi-scale response data under stimulation in an extremely weak magnetic field environment. The observed output, i.e., the predicted response mode waveform, is obtained through state updates. The training process employs the EM algorithm combined with Extended Kalman Filter (EKF) for iterative optimization to achieve parameter estimation. In model selection, the AIC / BIC criterion is used to determine the optimal structure; after personalized calibration, an individualized parameter distribution is output, thereby accurately characterizing the magnetic response properties of different individuals.

[0084] In the embodiments of this application, not only are the estimated model parameters ensured to explain the observed multi-scale magnetic response characteristics to the greatest extent, but the model structure itself is also optimized through objective criteria, laying a solid mathematical model foundation with statistical optimality for the subsequent generation of reliable and comparable individualized magnetic response fingerprints.

[0085] In some embodiments of this application, step S106 in the above method can be specifically implemented as follows: Figure 4 Steps S1061-S1064 shown:

[0086] S1061: Establish a hierarchical Bayesian model.

[0087] The prior distribution of individual model parameters is centered on the population parameters, and is adjusted by incorporating the individual's conservative feature vector and physiological baseline data.

[0088] For a specific individual We assume its model parameters Obey a group parameter The prior distribution is centered, but the width or location of this prior can be determined by the conservative feature vector of the individual. and physiological baseline data (e.g., age, genotype) through a regression function Adjustments are made. That is: .

[0089] The prior distribution of model parameters in hybrid structure state-space dynamics models is not fixed, but dynamically adjusted using an individual's physiological baseline data through a mapping function. For example, for an individual carrying the APOE4 gene, the mean of the prior distribution of parameters related to the rate of neuronal excitability recovery may be adjusted to more closely approximate the average level of known Alzheimer's disease patients. This allows the model to incorporate the individual's risk background information before it even sees the individual's experimental data.

[0090] S1062: Define the likelihood function for generating observation data based on individual model parameters and extremely weak magnetic field environment.

[0091] Given individual parameters and magnetic field stimulation The model generates observation data. The probability, i.e., the likelihood function .

[0092] By substituting candidate personalized parameters into the hybrid structure state-space dynamics model, running the model generates a set of predicted conservative feature vector sequences. The likelihood function quantifies the degree of matching between this predicted sequence and the actual measured feature vector data of the individual, and is typically modeled as a Gaussian distribution centered on the predicted values. The higher the likelihood value, the more accurately the model under the current parameters can reproduce the true magnetic response dynamics of the individual.

[0093] S1063: Based on Bayes' theorem, the prior distribution and likelihood function are integrated, and the variational inference algorithm is used to calculate the posterior distribution of individual model parameters. The mean of the posterior distribution is then used as the personalized calibrated model parameters.

[0094] According to Bayes' theorem, the posterior distribution of the parameters after knowing the individual data is proportional to the product of the likelihood function and the prior distribution: , Represents an individual dynamic parameter set (wait), This represents the multi-scale experimental data of this individual. Let represent the baseline covariates. The posterior distribution is efficiently approximated using variational inference algorithms, and the mean of this posterior distribution is used. As a parameter of the individual's personalized model.

[0095] Due to the complexity of the model, the accurate posterior distribution is difficult to calculate directly. Therefore, this application employs variational inference, an efficient approximation method. The principle is to introduce a simple family of distributions defined by a small number of parameters, and continuously adjust these parameters through an optimization algorithm to make this simple distribution as close as possible to the true but complex posterior distribution. After optimization, the mean of this simple distribution can be used as a point estimate of the personalized parameters, while its variance reflects the uncertainty of the estimate.

[0096] S1064: Using a pre-trained deep autoencoder, the personalized calibrated model parameters are encoded into a low-dimensional latent space to generate a magnetic response fingerprint.

[0097] Personalized parameters for all individuals This constitutes a high-dimensional space. To obtain a compact and comparable identifier, a pre-trained deep autoencoder is used to map this high-dimensional parameter space to a low-dimensional, continuous latent space. This autoencoder is pre-trained unsupervised on a large population dataset (containing healthy individuals and various disease states) to learn the manifold structure of the parameter space. The bottleneck layer of the encoder outputs an 8-dimensional vector. This is the magnetic response fingerprint of the individual. This fingerprint has high individual uniqueness and state repeatability.

[0098] In this embodiment, the adjusted prior distribution is incorporated into the individual baseline risk, and the individual experimental data is transformed into an accurate posterior parameter estimate through Bayesian update. Finally, the data is refined into a powerful, digital magnetic response fingerprint using coding technology, providing a core basis for precision medicine decision-making.

[0099] In some embodiments of this application, step S107 in the above method can be specifically implemented as follows: Figure 5 Steps S1071-S1073 shown:

[0100] S1071: Calculate the first Mahalanobis distance from the magnetic response fingerprint to the health database and the second Mahalanobis distance to the disease database.

[0101] S1072: Calculate the risk score based on the first Mahalanobis distance and the second Mahalanobis distance.

[0102]

[0103] in, Indicates the first i Risk score of magnetic response fingerprint, This represents the first Mahalanobis distance from the magnetic response fingerprint to the health database. This represents the second Mahalanobis distance from the magnetic response fingerprint to the disease database. This represents the weighting coefficient.

[0104] S1073: If the risk score is greater than the preset threshold, the patient is determined to be a high-risk individual, and the risk assessment result is output.

[0105] The risk assessment results include risk scores, visual radar charts of risk levels, and structured reports.

[0106] In addition, recommendations for regular testing can be provided for individuals whose risk scores are below a preset threshold.

[0107] In this embodiment, complex multidimensional biometric data is transformed into a clear and actionable clinical risk indicator through rigorous statistical distance calculation and weighted decision-making, thus realizing a complete technical closed loop of data collection, modeling, assessment, and reporting.

[0108] The method proposed in this application is illustrated below with a specific example: pre-Alzheimer's disease risk assessment.

[0109] 1. Experimental Design

[0110] Control group: 10 healthy older adults (65–75 years, MMSE) );

[0111] Experimental group: 10 patients with mild cognitive impairment (MCI) (MMSE: 24–27);

[0112] Magnetic field parameters: 10 Hz sinusoidal field, intensity 50 It lasts for 10 minutes.

[0113] Data acquisition equipment:

[0114] Subcellular: Mitochondria derived from PBMCs + TMRE staining;

[0115] Single cell: iPSC-induced neuron + 64-channel MEA;

[0116] Tissue: Postmortem brain slice calcium imaging;

[0117] Organ: OPM-MEG records default network dynamics.

[0118] 2. Data Processing and Modeling

[0119] One dominant conservative mode was extracted; the hybrid structure state-space dynamics model was successfully used to fit the response trajectory; Bayesian inference generated an 8-dimensional MRF vector for each person.

[0120] 3. Risk assessment results

[0121] Calculate the Mahalanobis distance between the MRF of each MCI patient and the healthy population. and distance from AD library ;

[0122] Risk Score: Threshold Those exceeding this limit are considered high-risk, as shown in Table 1.

[0123] Table 1

[0124]

[0125] Overall compliance rate: 8 / 10 = 80%, sensitivity 83%, specificity 75%.

[0126] This application achieves systematic modeling and individualized characterization of multi-scale biomagnetic responses, breaking through the limitations of fragmented existing research, providing core technical support for precision medicine in extremely weak magnetic fields, and has significant scientific research value and broad prospects for clinical translation.

[0127] Based on the above method, this application embodiment also provides a system structure diagram of applying a multi-scale biomagnetic response dynamics modeling method, such as... Figure 6 As shown, it includes an interactive display module, a multi-scale data acquisition module, a cross-scale feature extraction module, a dynamic model construction module, and a data storage and analysis module.

[0128] The interactive display module includes a visualization interface, displaying features such as four-scale signal overlay maps, IMF decomposition results, MRF radar charts, and risk heatmaps. The multi-scale data acquisition module includes a multi-scale data acquisition unit, integrating an MEA system, fluorescence detection unit, ultra-high resolution microscope, and data synchronization controller. It has a built-in PTP clock synchronization controller to ensure time deviation ≤0.1ms and supports real-time HDF5 writing. The cross-scale feature extraction module includes a cross-scale feature extraction unit, with a built-in MRMD algorithm engine, adaptive weighting unit, and conservative feature output. The dynamic model construction module includes a dynamic model construction unit, with a built-in HSSDM modeling unit supporting automatic modeling and manual editing; it generates magnetic response fingerprints through a Bayesian inference unit and a personalized calibration unit. The data storage and analysis module includes a data storage and analysis unit, using a MySQL + MinIO architecture to store raw data and MRF; it performs risk prediction through a fingerprint comparison engine and generates risk assessment reports.

[0129] In this embodiment, four-scale integrated synchronous acquisition is achieved for the first time, breaking the fragmented pattern of traditional research and providing complete data support for revealing the cross-scale transmission mechanism of magnetic field effects. The introduction of scale weighting and cross-scale correlation constraints significantly improves the accuracy and robustness of conservative feature extraction. An HSSDM model is constructed, combining physical interpretability with data-driven flexibility, outperforming pure black-box deep learning models. MRF fingerprints are generated based on Bayesian inference, integrating individual baseline information to achieve truly personalized modeling. MRF can be used for risk screening of various diseases and its effectiveness has been verified in AD and arrhythmias. The system supports cloud upgrades and federated learning, enabling multi-center knowledge sharing while ensuring privacy.

[0130] In some embodiments of this application, a multi-scale biomagnetic response dynamics modeling device is provided, which corresponds one-to-one with the multi-scale biomagnetic response dynamics modeling method in the above embodiments. For example... Figure 7 As shown, the multi-scale biomagnetic response dynamics modeling device includes an acquisition module 101, a preprocessing module 102, an extraction module 103, a construction module 104, a generation module 105, and a calculation module 106.

[0131] Acquisition module 101 is used to simultaneously acquire response data of biological samples at different scales in an extremely weak magnetic field environment. The different scales include at least one of the following: subcellular, single-cell, tissue and organ levels. The response data includes at least one of the following: electrophysiological signals, optical signals and morphodynamic information.

[0132] Preprocessing module 102 is used to perform standardized preprocessing on response data at various scales to obtain multi-scale signal sequences;

[0133] Extraction module 103 is used to extract conservative response modes from the multi-scale signal sequence based on cross-scale modal resonance analysis; extract quantitative features from the conservative response modes to form a conservative feature vector, wherein the quantitative features include at least one of the following: response peak time, signal attenuation coefficient, characteristic frequency bandwidth, phase lock value, modal energy ratio, and cross-scale consistency index;

[0134] Module 104 is used to construct and train a population-level hybrid structure state-space dynamics model, wherein the observation output of the hybrid structure state-space dynamics model is a conservative feature vector.

[0135] The generation module 105 is used to perform personalized calibration on the hybrid structure state-space dynamics model based on Bayesian inference and individual data, and generate a magnetic response fingerprint.

[0136] The calculation module 106 is used to compare the magnetic response fingerprint with a preset health database and a disease database respectively, calculate a risk score and output a risk assessment result.

[0137] In some embodiments of this application, in the above-described apparatus, the preprocessing module 102 is specifically used to perform detrending processing, bandpass filtering processing, and standardization processing on the response data at each scale to obtain preprocessed data. The standardization processing includes calculating the mean and standard value, and performing standard value conversion based on the original response data, mean, and standard value; multiplying the preprocessed data at each scale by a preset scale weight and then summing them to obtain fused data.

[0138] In some embodiments of this application, in the above-described apparatus, the extraction module 103 is specifically used to decompose each preprocessed single-scale signal sequence using ensemble empirical mode decomposition to obtain the intrinsic mode function components of each scale; calculate the cross-correlation coefficients between intrinsic mode function components whose center frequencies differ within a preset range between different scales; select mode combinations that appear at least on three scales and whose average cross-correlation coefficients between pairs of elements are greater than a preset threshold as cross-scale resonance modes; and perform a weighted average of the intrinsic mode function components of each scale contained in each cross-scale resonance mode to generate a conservative response mode.

[0139] In some embodiments of this application, the hybrid structure state-space dynamics model in the above-described apparatus includes:

[0140]

[0141] in, Represents a d-dimensional potential state vector. Indicates an external magnetic field input. Indicates a conservative response mode. Represents a linear parameter matrix, Indicates by parameters The defined two-layer fully connected neural network, This represents the Gaussian noise term.

[0142] In some embodiments of this application, the extraction module 103 in the above-described apparatus is specifically used to extract quantitative features based on the following methods:

[0143] The time elapsed from the initial point of magnetic field stimulation to the first maximum value of the conservative response mode amplitude;

[0144] The signal attenuation coefficient is obtained by exponentially fitting the modal envelope after the stimulus peak.

[0145] The characteristic frequency bandwidth is obtained by calculating the full width at half maximum (FWHM) of the instantaneous frequency distribution using the Hilbert spectrum;

[0146] Phase-locking value characterizing the synchronization between the instantaneous phase of a mode and the phase of an external magnetic field stimulus;

[0147] And the percentage of the energy of this mode relative to the total energy of all conservative modes.

[0148] In some embodiments of this application, in the above-described apparatus, the generation module 106 is specifically used to establish a hierarchical Bayesian model, wherein the prior distribution of individual model parameters is centered on the population parameters and is adjusted by incorporating the individual's conservative feature vector and physiological baseline data; based on the individual model parameters and the extremely weak magnetic field environment, a likelihood function for generating observation data of the model is defined; according to Bayes' theorem, the prior distribution and the likelihood function are fused, and the posterior distribution of the individual model parameters is calculated using a variational inference algorithm, and the mean of the posterior distribution is used as the personalized calibrated model parameters; using a pre-trained deep autoencoder, the personalized calibrated model parameters are encoded into a low-dimensional latent space to generate a magnetic response fingerprint.

[0149] In some embodiments of this application, in the above-described apparatus, the calculation module 107 is specifically used to calculate the first Mahalanobis distance from the magnetic response fingerprint to the health database and the second Mahalanobis distance to the disease database.

[0150] Calculate the risk score based on the first Mahalanobis distance and the second Mahalanobis distance;

[0151] If the risk score is greater than a preset threshold, the patient is determined to be a high-risk individual, and a risk assessment result is output, which includes a risk score, a visual radar chart of the risk level, and a structured report.

[0152] It should be noted that any of the above-mentioned multi-scale biomagnetic response dynamics modeling devices can be used to implement the aforementioned multi-scale biomagnetic response dynamics modeling methods, which will not be elaborated here.

[0153] Figure 8 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. Figure 8 As shown, at the hardware level, this electronic device includes a processor, and optionally also includes an internal bus, a network interface, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or it may include non-volatile memory, such as at least one disk drive. Of course, this electronic device may also include other hardware required for other business operations.

[0154] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 8 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0155] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.

[0156] The processor reads the corresponding computer program from non-volatile memory into main memory and then runs it, forming a multi-scale biomagnetic response dynamics modeling device at the logical level. The processor executes the program stored in the memory and specifically performs the aforementioned methods.

[0157] The processor may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0158] This electronic device can execute the multi-scale biomagnetic response dynamics modeling method provided in several embodiments of this application, and realize it as a multi-scale biomagnetic response dynamics modeling device. Figure 7 The functions of the embodiments shown are not described in detail here.

[0159] This application also proposes a computer-readable storage medium that stores one or more programs, the programs including instructions that, when executed by an electronic device including multiple applications, enable the electronic device to perform the multi-scale biomagnetic response dynamics modeling method provided in several embodiments of this application.

[0160] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0161] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0162] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0163] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0164] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0165] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0166] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0167] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0168] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0169] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method of multi-scale biomagnetic response dynamics modeling, characterized by, The method includes: In an extremely weak magnetic field environment, response data of biological samples are simultaneously collected at different scales, including at least one of the following: subcellular, single-cell, tissue and organ levels, and the response data includes at least one of the following: electrophysiological signals, optical signals and morphodynamic information. The response data at each scale are standardized and preprocessed to obtain a multi-scale signal sequence. Based on cross-scale modal resonance analysis, conservative response modes are extracted from the multi-scale signal sequence; Quantitative features are extracted from conservative response modes to form a conservative feature vector. The quantitative features include at least one of the following: response peak time, signal attenuation coefficient, characteristic frequency bandwidth, phase lock value, modal energy ratio, and cross-scale consistency index. A population-level hybrid structure state-space dynamics model is constructed and trained, wherein the observed output of the hybrid structure state-space dynamics model is a conservative feature vector; Based on Bayesian inference and individual data, the hybrid structure state-space dynamics model is individually calibrated, and a magnetic response fingerprint is generated. The magnetic response fingerprint is compared with a preset health database and a disease database, respectively, and a risk score is calculated and the risk assessment result is output.

2. The method according to claim 1, characterized in that, The standardization preprocessing includes detrending processing, bandpass filtering processing, and standardization processing. The standardization processing includes calculating the mean and standard values, and performing standard value transformation based on the original response data, the mean, and the standard values.

3. The method according to claim 1, characterized in that, The extraction of conservative response modes from the multi-scale signal sequence based on cross-scale modal resonance analysis includes: Each preprocessed single-scale signal sequence is decomposed using ensemble empirical mode decomposition to obtain the intrinsic mode function components at each scale. Calculate the cross-correlation coefficients between the intrinsic mode function components whose center frequencies differ within a preset range at different scales, and select mode combinations that appear at at least three scales and whose average cross-correlation coefficients between pairs of each other are greater than a preset threshold as cross-scale resonant modes. The conservative response mode is generated by weighted averaging of the intrinsic mode function components of each scale contained in each cross-scale resonance mode.

4. The method according to claim 1, characterized in that, The step of extracting quantitative features from conservative response modes to form a conservative feature vector includes: The time elapsed from the initial point of magnetic field stimulation to the first maximum value of the conservative response mode amplitude; The signal attenuation coefficient is obtained by exponentially fitting the modal envelope after the stimulus peak. The characteristic frequency bandwidth is obtained by calculating the full width at half maximum (FWHM) of the instantaneous frequency distribution using the Hilbert spectrum; Phase-locking value characterizing the synchronization between the instantaneous phase of a mode and the phase of an external magnetic field stimulus; And the percentage of the energy of this mode relative to the total energy of all conservative modes.

5. The method according to claim 1, characterized in that, The hybrid structure state-space dynamics model includes: in, Represents a d-dimensional potential state vector. Indicates an external magnetic field input. Indicates a conservative response mode. Represents a linear parameter matrix, Indicates by parameters The defined two-layer fully connected neural network, This represents the Gaussian noise term.

6. The method according to claim 1, characterized in that, The method, based on a Bayesian inference framework and combined with individual physiological baseline data, performs personalized calibration of the model parameters of the hybrid structure state-space dynamics model to generate a magnetic response fingerprint, including: A hierarchical Bayesian model is established, in which the prior distribution of individual model parameters is centered on the population parameters, and the conservative feature vectors and physiological baseline data of individuals are introduced for adjustment. Based on the individual model parameters and the extremely weak magnetic field environment, a likelihood function for generating observation data by the model is defined. According to Bayes' theorem, the prior distribution and likelihood function are integrated, and the posterior distribution of individual model parameters is calculated using a variational inference algorithm. The mean of the posterior distribution is then used as the personalized calibrated model parameters. Using a pre-trained deep autoencoder, the personalized calibrated model parameters are encoded into a low-dimensional latent space to generate a magnetic response fingerprint.

7. The method according to claim 1, characterized in that, The step of comparing the magnetic response fingerprint with a preset health database and a disease database, calculating a risk score, and outputting a risk assessment result includes: Calculate the first Mahalanobis distance from the magnetic response fingerprint to the health database, and the second Mahalanobis distance to the disease database; Calculate the risk score based on the first Mahalanobis distance and the second Mahalanobis distance; If the risk score is greater than a preset threshold, the patient is determined to be a high-risk individual, and a risk assessment result is output, which includes a risk score, a visual radar chart of the risk level, and a structured report.

8. A multi-scale biomagnetic response dynamics modeling device, characterized in that, The device includes: The acquisition module is used to simultaneously acquire response data of biological samples at different scales in an extremely weak magnetic field environment. The different scales include at least one of the following: subcellular, single-cell, tissue and organ levels. The response data includes at least one of the following: electrophysiological signals, optical signals and morphodynamic information. The preprocessing module is used to perform standardized preprocessing on response data at various scales to obtain multi-scale signal sequences; An extraction module is used to extract conservative response modes from the multi-scale signal sequence based on cross-scale modal resonance analysis; and to extract quantitative features from the conservative response modes to form a conservative feature vector, wherein the quantitative features include at least one of the following: response peak time, signal attenuation coefficient, characteristic frequency bandwidth, phase lock value, modal energy ratio, and cross-scale consistency index; A construction module is used to construct and train a population-level hybrid structure state-space dynamics model, wherein the observed output of the hybrid structure state-space dynamics model is a conservative feature vector; The generation module is used to perform personalized calibration on the hybrid structure state-space dynamics model based on Bayesian inference and individual data, and to generate magnetic response fingerprints. The calculation module is used to compare the magnetic response fingerprint with a preset health database and a disease database, respectively, calculate a risk score, and output a risk assessment result.

9. An electronic device, comprising: processor; as well as A memory configured to store computer-executable instructions, characterized in that, when executed, the executable instructions cause the processor to perform the steps of the multi-scale biomagnetic response dynamics modeling method as described in any one of claims 1-7.

10. A computer-readable storage medium storing one or more programs, characterized in that, When the one or more programs are executed by an electronic device comprising multiple applications, the electronic device performs the steps of the multi-scale biomagnetic response dynamics modeling method as described in any one of claims 1-7.

Images