Method and device for establishing radio frequency noninvasive blood glucose monitoring electromagnetic sensing model, storage medium and equipment

Abstract

The invention discloses an establishment method and device of a radio frequency noninvasive blood glucose monitoring electromagnetic sensing model, a storage medium and equipment. The method comprises the following steps: preliminarily dividing human tissues into static tissues and dynamic tissues; the radio frequency sensing signals are transmitted to the static tissue and the dynamic tissue respectively, response characteristics of the radio frequency sensing signals are obtained, clustering is carried out according to the response characteristics, and a plurality of types of human tissue are obtained; carrying out voxel reconstruction on the plurality of types of human tissues by utilizing a sparse regularization graph variation method to obtain a blood glucose monitoring electromagnetic sensing model; based on a Cole-Cole formula modeling method, dielectric property analysis is carried out on the multiple types of human tissues, dielectric property assignment is carried out on all voxel points of the blood glucose monitoring electromagnetic sensing model according to analysis results, and a response function of the human tissues to blood glucose sensing signals is established. The radio frequency blood glucose monitoring electromagnetic sensing model established by the method is more practical, and a theoretical and simulation model is provided for subsequent more accurate blood glucose monitoring.

Description

Technical Field

[0001] This invention belongs to the field of medical and health technology. Specifically, it relates to a method, apparatus, computer-readable storage medium, and computer equipment for establishing an electromagnetic sensing model for radiofrequency non-invasive blood glucose monitoring. Background Technology

[0002] Non-invasive blood glucose monitoring methods can be categorized into optical analysis methods (such as near-infrared spectroscopy, mid-infrared spectroscopy, Raman spectroscopy, optical correlation tomography, and polarization rotation), iontophoresis, energy metabolism, physiological sensing, and radio frequency sensing. Compared to invasive and minimally invasive blood glucose monitoring methods, non-invasive methods offer advantages such as painlessness, comfort, convenience, speed, and low cost, making them uniquely advantageous for continuous blood glucose monitoring. However, current non-invasive blood glucose monitoring methods face significant challenges in measurement accuracy. With the development of radio frequency sensing technology, radio frequency-based non-invasive blood glucose monitoring technology has become an important direction and research hotspot. The principle of radio frequency-based non-invasive blood glucose monitoring is as follows: When a radio frequency signal is incident on human tissue, it propagates through the skin, subcutaneous fat, and blood, and its propagation characteristics are affected by the tissue's dielectric constant, conductivity, and polarization behavior. Changes in blood glucose concentration cause alterations in blood ion concentration, molecular polarizability, and the binding state of water molecules, resulting in subtle but measurable changes in the blood. When radio frequency signals propagate through tissues, their reflected or transmitted signals exhibit characteristic changes in response to variations in blood glucose levels. By analyzing these changes, non-invasive estimation of blood glucose concentration can be achieved. A similar implementation scheme is proposed below.

[0003] (1) The paper "A noninvasive measurement of blood glucose concentration by UWB microwave spectrum" established an electromagnetic simulation model of blood glucose in three tissue layers: skin, fat, and blood, and studied the changes in signal S-parameters caused by different glucose concentrations under ultra-wideband frequency band using time-frequency characteristics. (2) The paper "Simulating the effects of skin thickness and fingerprints to highlight problems with non-invasive RF blood glucose sensing from fingerertips" established a fingertip electromagnetic model including skin, fat, blood, and bone, studied the characteristics of the resonant frequency change of RF sensor at different frequencies, and explored the influence of different fingerprint structures on the resonant frequency shift of RF sensor. (3) The paper "Thumb positioning analysis of new elliptical-shaped microwave sensors for non-invasive glucose monitoring" modeled the electromagnetic model of blood glucose monitoring as skin-fat-muscle-blood, and used this model to observe the influence of frequency shift under different blood dielectric constants. (4) The paper “Analysis and Design of a Microwave Coplanar Sensor for Non-Invasive Blood Glucose Measurements” established a sensing model including skin, blood, bone, fat and muscle based on the biological tissue characteristics of the blood glucose monitoring site - finger. It studied the influence of sensors with different topologies on the stability and repeatability of blood glucose measurement results, and thus determined the design scheme of using a coplanar structure to design the resonator.

[0004] Existing research generally considers establishing multi-layered electromagnetic models of the human body, simply dividing the blood glucose monitoring site / pathway into layers such as skin, fat, muscle, and blood. However, the aforementioned layered electromagnetic models for blood glucose monitoring differ significantly from the actual human body structure, and these modeling methods cannot provide theoretical guidance for improving the accuracy of radio frequency (RF)-based non-invasive blood glucose monitoring technology. Furthermore, current modeling studies mainly consider the impact of changes in blood glucose concentration on the sensor signal; research on the influence of other human tissues and the combined effects of different human tissues on RF sensor signals has not been reported. Finally, existing blood glucose monitoring models generally only consider the RF sensor signal response characteristics under static conditions, ignoring the influence of physiological and environmental changes on the sensor signal, thus limiting the accuracy of blood glucose monitoring models. Summary of the Invention

[0005] (I) The technical problem to be solved by the present invention

[0006] The technical problem solved by this invention is: how to establish a method for building an electromagnetic sensing model for radiofrequency non-invasive blood glucose monitoring that is more closely related to the actual human body, so that the electromagnetic sensing model can respond to both static and dynamic human tissues simultaneously.

[0007] (II) Technical Solution Adopted in this Invention

[0008] A method for establishing an electromagnetic sensing model for radiofrequency non-invasive blood glucose monitoring, the method comprising:

[0009] Human tissue segmentation stage: The human tissue is initially divided into static tissue and dynamic tissue; radio frequency sensing signals are transmitted to the static tissue and dynamic tissue respectively to obtain the response characteristics of the radio frequency sensing signals, and clustering is performed based on the response characteristics to obtain several types of human tissue.

[0010] Initial stage of electromagnetic sensing model construction: Voxel reconstruction of several types of human tissues was carried out using sparse regularized graph variational method to obtain the electromagnetic sensing model for blood glucose monitoring.

[0011] The dielectric property assignment and response construction stages of the model: Based on the Cole-Cole formula modeling method, dielectric property analysis was performed on several types of human tissues. Based on the analysis results, dielectric property values ​​were assigned to each voxel point of the blood glucose monitoring electromagnetic sensing model. The response function of human tissue to blood glucose sensing signal was established for the blood glucose monitoring electromagnetic sensing model after the dielectric property assignment.

[0012] Optionally, human tissues can be initially divided into static tissues and dynamic tissues, including:

[0013] Based on human anatomical information, the dielectric properties of human tissues at different frequencies, the degree of influence of physiological changes on dielectric properties, and the degree of influence of environmental changes on dielectric properties, human tissues are initially divided into static tissues and dynamic tissues.

[0014] Optionally, the response features include amplitude information, phase information, and group velocity information of the radio frequency transmitted signal and the radio frequency reflected signal; the clustering is Gaussian mixture clustering.

[0015] Optionally, a sparsely regularized graph variational method is used to reconstruct voxels of several types of human tissues to obtain an electromagnetic sensing model for blood glucose monitoring, including:

[0016] Spatial similarity between different regions in image sequences of human tissues is mined using a sparsely regularized graph variational method.

[0017] Model parameters for the electromagnetic sensing model for blood glucose monitoring are obtained based on spatial similarity.

[0018] Optionally, dielectric property analysis is performed on several types of human tissues based on the Cole-Cole formula modeling method, including:

[0019] When analyzing the dielectric properties of static human tissue, the influence factors of movement, respiration, temperature, and humidity changes are subtracted from the Cole-Cole formula.

[0020] When analyzing the dielectric properties of dynamic human tissues, the influence factors of motion, respiration, temperature, and humidity changes are incorporated into the Cole-Cole formula.

[0021] Optionally, the response function of human tissue to the blood glucose sensing signal is established based on the electromagnetic sensing model of blood glucose monitoring after assigning dielectric properties, including:

[0022] The propagation characteristics of radio frequency sensing signals in different directions of the blood glucose monitoring electromagnetic sensing model were obtained respectively;

[0023] Using the controlled variable method, the radio frequency response characteristics of the same human tissue at different frequencies and the radio frequency response characteristics of different human tissues at the same frequency were obtained based on the propagation characteristics.

[0024] The response function of human tissue to blood glucose sensing signals is established based on the characteristics of each radio frequency response.

[0025] This application also discloses a device for establishing an electromagnetic sensing model for radio frequency non-invasive blood glucose monitoring, the device comprising:

[0026] The human tissue segmentation module is configured to: initially segment human tissue into static tissue and dynamic tissue; transmit radio frequency sensing signals to the static tissue and dynamic tissue respectively, obtain the response characteristics of the radio frequency sensing signals, and perform clustering based on the response characteristics to obtain several types of human tissue.

[0027] The electromagnetic sensing model initial construction module is configured to: use a sparse regularized graph variational method to reconstruct voxels of several types of human tissues to obtain a blood glucose monitoring electromagnetic sensing model.

[0028] The dielectric property assignment and response construction module of the model is configured to: perform dielectric property analysis on several types of human tissues based on the Cole-Cole formula modeling method; assign dielectric property values ​​to each voxel point of the blood glucose monitoring electromagnetic sensing model according to the analysis results; and establish the response function of human tissue to blood glucose sensing signal for the blood glucose monitoring electromagnetic sensing model after the dielectric property assignment.

[0029] This application also discloses a computer-readable storage medium storing a program for establishing an electromagnetic sensing model for radio frequency non-invasive blood glucose monitoring. When the program for establishing the electromagnetic sensing model for radio frequency non-invasive blood glucose monitoring is executed by a processor, the above-mentioned method for establishing the electromagnetic sensing model for radio frequency non-invasive blood glucose monitoring is implemented.

[0030] This application also discloses a computer device, which includes a computer-readable storage medium, a processor, and a program for establishing a radio frequency non-invasive blood glucose monitoring electromagnetic sensing model stored in the computer-readable storage medium. When the program for establishing the radio frequency non-invasive blood glucose monitoring electromagnetic sensing model is executed by the processor, the above-described method for establishing the radio frequency non-invasive blood glucose monitoring electromagnetic sensing model is implemented.

[0031] (III) Beneficial Effects

[0032] The present invention discloses a method, apparatus, storage medium, and device for establishing an electromagnetic sensing model for radiofrequency non-invasive blood glucose monitoring, which has the following technical advantages compared with existing methods:

[0033] Based on the static and dynamic response characteristics of different human tissues using radio frequency (RF) methods, Gaussian mixture clustering was employed to classify human tissues. Building upon this, a voxel-based electromagnetic sensing model for RF blood glucose monitoring was established. This model comprehensively considered the dynamic changes in the dielectric properties of human tissues caused by physiological signals and the detection environment. The electromagnetic effects of voxels from different human tissues on the RF sensing signal, as well as the coupling mechanism between different voxels, were explored to improve the response mechanism of the RF sensing signal. This provides a theoretical and simulation model for improving the accuracy of RF-based non-invasive blood glucose monitoring technology. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating the method for establishing an electromagnetic sensing model for radio frequency non-invasive blood glucose monitoring according to Embodiment 1 of the present invention.

[0035] Figure 2 This is a schematic diagram of human tissue segmentation and classification based on Gaussian mixture clustering for the electromagnetic sensing model of radiofrequency non-invasive blood glucose monitoring according to Embodiment 1 of the present invention.

[0036] Figure 3 This is a schematic diagram of the electromagnetic sensing model for radio frequency non-invasive blood glucose monitoring according to Embodiment 1 of the present invention.

[0037] Figure 4 This is a schematic diagram of the response mechanism of voxels and radio frequency sensing signals of different human tissues in Embodiment 1 of the present invention.

[0038] Figure 5 This is a schematic diagram of the non-invasive blood glucose monitoring device according to Embodiment 2 of the present invention.

[0039] Figure 6 This is a schematic diagram of a computer device according to Embodiment 4 of the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0041] Before describing the various embodiments of this application in detail, the technical concept of this application is briefly described first: Current radiofrequency non-invasive blood glucose monitoring solutions often use electromagnetic sensing models that are difficult to accurately distinguish between different human tissues, failing to consider the impact of physiological and environmental changes, thus limiting the accuracy of the blood glucose monitoring model. Therefore, the key improvement of the method for establishing a radiofrequency non-invasive blood glucose monitoring electromagnetic sensing model provided in this application lies in first dividing human tissues into static and dynamic tissues, clustering them according to the response characteristics of the radiofrequency sensing signal, establishing the blood glucose monitoring electromagnetic sensing model using the voxel method, and performing dielectric property analysis based on the Cole-Cole formula modeling method to assign dielectric property values ​​to the model, thereby ultimately obtaining the response function of the blood glucose sensing signal. The radiofrequency blood glucose monitoring electromagnetic sensing model established by this method is more realistic, considers dynamic response factors, and is beneficial for providing a theoretical and simulation model for more accurate blood glucose monitoring in the future. The specific principles of the method for establishing a radiofrequency non-invasive blood glucose monitoring electromagnetic sensing model of this application are described below with reference to more embodiments.

[0042] Specifically, such as Figure 1As shown, the method for establishing the electromagnetic sensing model for radiofrequency non-invasive blood glucose monitoring in this embodiment includes the following steps:

[0043] Step S10: Human tissue segmentation stage: The human tissue is initially divided into static tissue and dynamic tissue; radio frequency sensing signals are transmitted to the static tissue and dynamic tissue respectively to obtain the response characteristics of the radio frequency sensing signals, and clustering is performed according to the response characteristics to obtain several types of human tissue.

[0044] Step S20: Initial construction stage of electromagnetic sensing model: Voxel reconstruction of several types of human tissues is carried out using sparse regularized graph variational method to obtain the electromagnetic sensing model for blood glucose monitoring.

[0045] Step S30: Dielectric property assignment and response construction stage of the model: Based on the Cole-Cole formula modeling method, dielectric property analysis is performed on several types of human tissues. Based on the analysis results, dielectric property values ​​are assigned to each voxel point of the blood glucose monitoring electromagnetic sensing model. The response function of human tissue to blood glucose sensing signal is established for the blood glucose monitoring electromagnetic sensing model after the dielectric property is assigned.

[0046] In one or more embodiments, human tissues are initially classified into static and dynamic tissues based on human anatomical information, the dielectric properties of human tissues at different frequencies, the degree of influence of physiological changes on dielectric properties, and the degree of influence of environmental changes on dielectric properties. For example, anatomical information of different parts of the human body is obtained through relevant literature, the differences in the distribution of human tissues in different parts are analyzed, and sites with relatively stable tissue distribution and high similarity in tissue distribution among different individuals are selected as sites for blood glucose monitoring.

[0047] Specifically, based on human anatomical information, the blood glucose monitoring component is divided into multiple non-homogeneous media, including the epidermis, dermis, subcutaneous tissue, tissue fluid, fat, and blood. Building upon this, the dielectric properties (relative permittivity, conductivity, etc.) of different human tissues at different frequencies are obtained through existing literature. Considering the influence of physiological changes (exercise, respiration, etc.) and environmental changes (temperature, humidity, etc.) on the dielectric properties of human tissues, human tissues are initially classified into static response tissues and dynamic response tissues. Static response tissues refer to those whose dielectric properties are less affected by external interference and whose radio frequency sensing signal response characteristics are stable over a period of time, such as fat and muscle. If the human tissue is easily affected by exercise, respiration, temperature, humidity, or changes in blood glucose concentration (such as tissue fluid, blood, etc.), then the tissue is classified as a dynamic response tissue.

[0048] Furthermore, such as Figure 2As shown, by incident radio frequency (RF) sensing signals onto the aforementioned human tissues, the response characteristics of the RF sensing signals are obtained. These response characteristics include amplitude, phase, and group velocity information of the transmitted and reflected RF signals. Based on these response characteristics, a Gaussian mixture clustering method is used to cluster both static and dynamic human tissues, grouping tissues with similar response characteristics into the same category, thereby achieving the segmentation and classification of the human body.

[0049] Gaussian mixture clustering is a linear combination of a finite number of independent multivariate Gaussian distribution models. Assume that the radio frequency response characteristics of a certain human tissue at frequency f are x. k (k=1,2,3,...K), then the Gaussian mixture clustering method can be expressed as equation (1):

[0050]

[0051] In the Gaussian mixture clustering model, there are three parameters that need to be evaluated: μ, α, and S. μ is the expectation of the model, α represents the distribution weights of each RF response characteristic, and S represents the variance of the model. Therefore, equation (1) can be rewritten as:

[0052]

[0053] To determine the values ​​of the three parameters mentioned above, the maximum likelihood method is used for parameter estimation. First, μ, α, and S are initialized, and the posterior probability is calculated as shown in equation (3):

[0054]

[0055] The maximum likelihood values ​​of μ, α, and S are calculated separately, as shown in Equation (4), until the algorithm converges. Based on the calculation results of the Gaussian mixture model above, clustering of human tissues with similar radio frequency response characteristics is achieved.

[0056]

[0057] In one or more embodiments, a sparsely regularized graph variational method is used to reconstruct voxels of several types of human tissues to obtain a blood glucose monitoring electromagnetic sensing model, including: using the sparsely regularized graph variational method to mine the spatial similarity between different regions in the image sequence of human tissues; and obtaining the model parameters of the blood glucose monitoring electromagnetic sensing model based on the spatial similarity.

[0058] For example, considering the significant differences in structure, thickness, and distribution of different human tissues, a voxel reconstruction of human tissues is achieved using a sparsely regularized graph variational method. This represents human tissue as a model composed of multiple ordered voxels, thereby establishing a voxel-based electromagnetic sensing model for radio frequency blood glucose monitoring. In the graph variational method, spatial similarity between different regions in an image sequence is mined, establishing connections between similar regions throughout the image, independent of spatial distance. This overcomes the limitation of only considering the similarity between voxels and their neighboring regions, while preserving the boundary details and fine structure of different human tissues. The graph variational method requires establishing a smoothing model to reconstruct the signal from noisy samples. Solving this smoothing model can be viewed as a convex optimization problem. Considering the high computational cost and slow speed of traditional optimization methods, a combined forward-backward primal dual method is adopted, combining gradient descent and the proximity operator to solve for the non-differentiable terms in the smoothing model.

[0059] In graph variational methods, a graph is a data representation method used to describe the geometric structure in the data. Graph regularization requires the establishment of a smoothing model to reconstruct the signal from noisy samples. Graph variational regularization can be expressed as shown in equation (5):

[0060]

[0061] Among them, W i.j The weight matrix of the graph. This indicates that the traversal is based on ζ. i The i-th element represents all related voxels, and α represents the average distance of the related voxels to other voxels.

[0062] As can be seen from equation (5), it is a convex optimization problem. Considering that traditional optimization methods have drawbacks such as large computational load and slow speed, it is proposed to adopt the original dual method of forward-backward combination, that is, to combine the gradient descent method and the proximity operator to solve the non-differentiable terms in the model, as shown in equation (6).

[0063]

[0064] in,

[0065] For example, such as Figure 3As shown, after the voxel-based electromagnetic sensing model for blood glucose monitoring is established, a penalized least squares model for the three-dimensional shape is established to evaluate the model. An augmented Lagrange multiplier iterative algorithm is then used to optimize the estimation model and obtain the estimated three-dimensional shape. In the voxel-based electromagnetic sensing model for blood glucose monitoring, an improved non-uniform mesh is used on both sides of the interface between different human tissues to ensure the second-order accuracy of the numerical simulation algorithm. Specifically, fine meshes are introduced for regions with high dielectric constants or regions closely related to changes in blood glucose sensing signals (such as blood vessels and blood), while coarse meshes are used in regions where electromagnetic energy is relatively dispersed, making the electromagnetic mesh division in the blood glucose monitoring model freely variable. By using the time-domain discrete Green's function method to describe the signal absorption between different voxel boundaries, the signal absorption changes between different voxel boundaries in the blood glucose monitoring electromagnetic sensing model can be obtained, thus realizing the calculation of electromagnetic propagation.

[0066] According to electromagnetic field and electromagnetic wave theory, the tangential components of the magnetic field and electric field on the surface of different media (voxels) can be considered linear, but the normal component of the electric field is discontinuous. Therefore, an improved non-uniform mesh is proposed to be established on both sides of the interface between the media to ensure the second-order accuracy of the numerical simulation algorithm. Specifically, fine meshes are introduced for regions with high dielectric constants or regions closely related to changes in blood glucose sensing signals (such as blood vessels and blood), while coarse meshes are used in regions where electromagnetic energy is relatively dispersed, making the division of the electromagnetic mesh in the blood glucose monitoring model freely variable. Furthermore, the discrete Green's function method in the time domain is proposed to describe the signal absorption between different voxel boundaries. Assuming the incident waveform is a discrete plane wave, it can be expressed as equation (7):

[0067]

[0068] Based on the wave operator matrix, equation (7) can be rewritten as:

[0069]

[0070] According to equation (8), the signal absorption changes between different voxel boundaries in the blood glucose monitoring electromagnetic sensing model can be obtained, and the electromagnetic propagation can be calculated.

[0071] In one or more embodiments, after the electromagnetic sensing model for blood glucose monitoring is established, it is necessary to assign dielectric property values ​​to voxel points of different human tissues in the model to clarify the response characteristics of different human tissues to radio frequency sensing signals. The static and dynamic dielectric properties of human tissues are characterized using the Cole-Cole formula. For the dielectric properties of static human tissues, the influence factors of movement, respiration, temperature, and humidity changes can be subtracted from the Cole-Cole formula. For the dielectric properties of dynamic human tissues, the influence factors of movement, respiration, temperature, and humidity need to be introduced into the Cole-Cole formula.

[0072] Specifically, the Cole-Cole model is a commonly used model for characterizing the dielectric properties of human tissues, as shown in equation (9). The real part of the model is the relative permittivity of human tissues, while the imaginary part is related to the conductivity of human tissues, i.e. Where N represents the order of the model, ε ∞ For the relative permittivity where the frequency approaches infinity, Δε n τ represents the difference in relative permittivity corresponding to different orders. n Let σ be the relaxation time, σ0 be the conductivity under an electrostatic field, and α be the conductivity. n is the distribution parameter, and its value range is [0,1].

[0073]

[0074] According to equation (9), the above model is mainly used to characterize the static dielectric properties of human tissues, neglecting the effects of movement, respiration, temperature, and humidity changes. To enable equation (9) to describe the dynamic dielectric properties of human tissues, it is proposed to improve equation (9). First, experiments will be conducted to study the variation patterns of the dielectric properties of human tissues under different movement, respiration, temperature, and humidity conditions. Based on these variation patterns, a polynomial fitting method will be used to fit Δε. n and τ n The formula is modified to represent it as an equation with motion, respiration, temperature and humidity as factors, as shown in equations (10) and (11).

[0075]

[0076] τ n (S,B,T,H)=1-ξ1S -1 -ζ1B -1 -η1 T -1 -υ1H -1 ...-ξ n S -n -ζ n B -n -η n T-n -υ n H -n (11)

[0077] Where, φ n γ n , ψ n ξ n , ζ n η n υ n The parameters are constant factors to be determined, and S, B, T, and H represent the movement rate, respiration rate, temperature, and humidity, respectively. The optimal constant factor values ​​will be determined by fitting the experimental results using a particle swarm optimization algorithm.

[0078] For example, after establishing static and dynamic dielectric property models of human tissue based on the Cole-Cole formula, the response characteristics of radio frequency (RF) sensing signals on different voxels such as fat and muscle, as well as mixed voxel models, are explored. The RF response characteristics of the same human tissue at different frequencies can be studied using the controlled variable method; the RF response characteristics of different human tissues at the same frequency can also be studied. By using ridge regression, lasso regression, and other methods to establish the response function of human tissue to blood glucose sensing signals, the influence of different human tissues on RF sensing signals can be analyzed.

[0079] like Figure 4 As shown, after establishing the static and dynamic dielectric property models of human tissue based on the Cole-Cole model, the response characteristics of radio frequency sensing signals on different voxels such as fat and muscle, as well as mixed voxel models, are explored using the Cole-Cole model. Specifically, based on Maxwell's equations, the radio frequency sensing signal can be expressed as f(x,y,z,t)=f(iΔx,jΔy,kΔz,nΔt)=f n Given (i,j,k), taking the partial derivative of the RF sensing signal in the X direction, we have:

[0080]

[0081] According to equations (12) and (13), the propagation characteristics of the electric field and magnetic field of the radio frequency sensing signal in the X direction of the voxel model can be obtained respectively, as shown in equations (14) and (15).

[0082]

[0083] Similarly, the propagation characteristics of the radio frequency (RF) sensing signal in the Y and Z directions were obtained. By acquiring the propagation characteristics of the RF sensing signal in different directions, the RF response characteristics of the same human tissue at different frequencies were studied using the controlled variable method. The RF response characteristics of different human tissues at the same frequency were also investigated. Response functions of human tissues to blood glucose sensing signals were established using ridge regression and lasso regression methods, thereby analyzing the influence of different human tissues on the RF sensing signal.

[0084] In one or more embodiments, experiments based on human tissue-simulated materials and in vivo experiments are conducted to verify and optimize the electromagnetic sensing model for radio frequency blood glucose monitoring.

[0085] The electromagnetic sensing model for radiofrequency blood glucose monitoring was validated through both experiments using human tissue-based simulated materials and in vivo experiments. Human tissue-based simulated materials refer to materials such as agar, deionized water, sodium chloride, aluminum powder, vegetable oil, liquid surfactants, and formalin solution. By optimizing the formulation ratios of these materials, the dielectric properties of the prepared material are made as close as possible to those of real human tissue. The preparation method of the human tissue-based simulated material is as follows: First, by changing the ratio of one formulation while keeping other formulation ratios constant, the dielectric constant and conductivity of the prepared human tissue-based simulated material at different frequencies are obtained using an impedance analyzer / network analyzer and a dielectric measurement probe. Based on the obtained experimental data, the effect of changes in the formulation ratios on the dielectric properties of the material is studied using a particle swarm optimization algorithm, establishing a functional relationship between dielectric properties and formulation ratios. Based on this functional relationship, different human tissue simulation materials, such as skin, fat, muscle, blood, and bone, were prepared. The prepared human tissue simulation materials were measured and compared with the dielectric properties of corresponding real human tissues. The error between the two was evaluated, and the established functional relationship between dielectric properties and formulation ratios was optimized and corrected based on the evaluation results. The specific evaluation process is as follows: First, the input factors are determined. The input factors include the dielectric constant of each formulation, the fitting function of conductivity, and the ratio of each formulation, i.e., U = {u1, u2, u3, ..., u...} n ,...,u N Secondly, the output factor is determined. The output factor mainly describes the difference in dielectric properties between the prepared bio-tissue simulation material and the real human tissue material. The output factor can be expressed as V={v1,v2,...,v...} m ,...,v MThe elements in the output factor represent different errors. Specifically, v1 represents an error less than 5%, v2 represents an error less than 10%, v3 represents an error less than 15%, and so on. Next, an evaluation matrix is ​​established from the input influence factor to the output evaluation factor, resulting in a mapping matrix from U to F(V), as shown in equation (16):

[0086]

[0087] The mapping matrix derived from f is the evaluation matrix M1, as shown in equation (17):

[0088]

[0089] If the judgment result Then it is normalized. Finally, according to the maximum membership principle, the maximum r is... i The corresponding item is the result of the evaluation. If r i If the value is not greater than the set threshold, it indicates that the formula ratio is reasonable. Furthermore, research is being conducted on methods for printing biological tissues using 3D printing technology, combined with methods for preparing human tissue simulation materials, to create physical models of human tissue that are similar in structure and dielectric properties to real human tissue. For example... Figure 4 As shown, using the printed physical model of human tissue, the effects of changes in glucose concentration, human tissue thickness, and human tissue dielectric properties on radio frequency sensing signals were studied. Based on the measurement results, the electromagnetic sensing of radio frequency blood glucose monitoring was verified and corrected.

[0090] To further verify the feasibility and accuracy of the electromagnetic sensing modeling method for radiofrequency blood glucose monitoring, in vivo experiments were conducted. Healthy individuals and diabetic patients were recruited, and an oral glucose tolerance test (OGTT) was performed. Continuous radiofrequency response signals were collected during the experiment. Simultaneously, reference blood glucose values ​​were obtained by collecting finger-prick blood from the volunteers to study the radiofrequency response characteristics at different blood glucose concentrations. Based on this, imaging images of the detection site were acquired, and a voxel-based electromagnetic sensing model for radiofrequency blood glucose monitoring was established using the aforementioned method. The response characteristics of the radiofrequency sensing signal were analyzed based on this model, and compared with experimental measurement results to explore the sources of experimental error and further optimize the modeling method for radiofrequency blood glucose monitoring electromagnetic sensing.

[0091] The method for establishing an electromagnetic sensing model for non-invasive blood glucose monitoring disclosed in this embodiment (hereinafter referred to as "this method") has the following beneficial effects:

[0092] (1) The established electromagnetic sensing model for radiofrequency blood glucose monitoring is more realistic: Existing technologies use layered tissue division, while this method, through voxelization, can restore the non-uniform distribution (such as muscle fiber orientation) and cross-over characteristics (such as capillary network) of tissue.

[0093] (2) Consideration of dynamic response factors: Existing technologies only focus on the dielectric properties of static human tissues, while this method considers the dielectric properties of both static and dynamic human tissues.

[0094] (3) Higher modeling efficiency: Existing technologies have not solved the problem of balancing model precision and complexity. This method achieves reasonable classification through Gaussian mixture clustering, reducing model complexity under the same precision.

[0095] like Figure 5 As shown in Embodiment 2 of this application, a device for establishing an electromagnetic sensing model for radiofrequency non-invasive blood glucose monitoring is also disclosed. The device includes a human tissue segmentation module 100, an electromagnetic sensing model initial construction module 200, and a model dielectric property assignment and response construction module 300. The human tissue segmentation module 100 is configured to: initially segment human tissue into static and dynamic tissues; transmit radiofrequency sensing signals to the static and dynamic tissues respectively, obtain the response characteristics of the radiofrequency sensing signals, and cluster them according to the response characteristics to obtain several classes of human tissues. The electromagnetic sensing model initial construction module 200 is configured to: use a sparse regularized graph variational method to perform voxel reconstruction on several classes of human tissues to obtain the blood glucose monitoring electromagnetic sensing model. The dielectric property assignment and response construction module 300 of the model is configured to: perform dielectric property analysis on several types of human tissues based on the Cole-Cole formula modeling method; assign dielectric property values ​​to each voxel point of the blood glucose monitoring electromagnetic sensing model according to the analysis results; and establish the response function of human tissue to blood glucose sensing signals for the blood glucose monitoring electromagnetic sensing model after the dielectric property assignment. The calculation process of the human tissue segmentation module 100, the electromagnetic sensing model initial construction module 200, and the dielectric property assignment and response construction module 300 can be referred to the description in the previous embodiment, and will not be repeated here.

[0096] Embodiment 3 of this application also discloses a computer-readable storage medium storing a program for establishing a radio frequency non-invasive blood glucose monitoring electromagnetic sensing model. When the program for establishing the radio frequency non-invasive blood glucose monitoring electromagnetic sensing model is executed by a processor, the above-mentioned method for establishing the radio frequency non-invasive blood glucose monitoring electromagnetic sensing model is implemented.

[0097] This third embodiment also discloses a computer device, at the hardware level, such as... Figure 6As shown, the computer device includes a processor 12, an internal bus 13, a network interface 14, and a computer-readable storage medium 11. The processor 12 reads the corresponding computer program from the computer-readable storage medium and runs it, forming a request processing device at the logical level. Of course, in addition to the software implementation, one or more embodiments of this specification do not exclude other implementation methods, such as logic devices or a combination of hardware and software, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices. The computer-readable storage medium 11 stores a program for establishing a radio frequency non-invasive blood glucose monitoring electromagnetic sensing model. When the processor executes the program for establishing the radio frequency non-invasive blood glucose monitoring electromagnetic sensing model, it implements the above-described method for establishing the radio frequency non-invasive blood glucose monitoring electromagnetic sensing model.

[0098] Computer-readable storage media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer-readable 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, disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0099] The specific embodiments of the present invention have been described in detail above. Although some embodiments have been shown and described, those skilled in the art should understand that modifications and improvements can be made to these embodiments without departing from the principles and spirit of the present invention as defined by the claims and their equivalents, and such modifications and improvements should also be within the protection scope of the present invention.

Claims

1. A method for establishing an electromagnetic sensing model for radiofrequency non-invasive blood glucose monitoring, characterized in that, The establishing method comprises: The human tissue division stage: the human tissue is initially divided into static tissue and dynamic tissue; the radio frequency sensing signal is respectively transmitted to the static tissue and the dynamic tissue, the response characteristics of the radio frequency sensing signal are obtained, and clustering is performed according to the response characteristics, so that a plurality of types of human tissues are obtained; The electromagnetic sensing model initial establishment stage: the voxel reconstruction is performed on the plurality of types of human tissues by using a sparse regularization graph variation method, so that the blood glucose monitoring electromagnetic sensing model is obtained; The dielectric property assignment and response construction stage: the dielectric property analysis is respectively performed on the plurality of types of human tissues based on a Cole-Cole formula modeling method, the dielectric property assignment is performed on each voxel point of the blood glucose monitoring electromagnetic sensing model according to an analysis result, and the response function of the human tissue to the blood glucose sensing signal is established for the blood glucose monitoring electromagnetic sensing model after the dielectric property assignment.

2. The method of claim 1, wherein the method further comprises: The human tissue is initially divided into static tissue and dynamic tissue, which comprises: The human tissue is initially divided into static tissue and dynamic tissue according to human anatomy information, dielectric properties of human tissues at different frequencies, an influence degree of physiological changes on the dielectric properties, and an influence degree of environmental changes on the dielectric properties.

3. The method of claim 2, wherein, The response characteristics comprise amplitude information, phase information and group velocity information of the radio frequency transmission signal and the radio frequency reflection signal; and the clustering is Gaussian mixture clustering.

4. The method of claim 1, wherein the method further comprises: The voxel reconstruction is performed on the plurality of types of human tissues by using the sparse regularization graph variation method, so that the blood glucose monitoring electromagnetic sensing model is obtained, which comprises: The spatial similarity between different regions in the image sequence of the human tissue is mined by using the sparse regularization graph variation method; The model parameters of the blood glucose monitoring electromagnetic sensing model are obtained according to the spatial similarity.

5. The method of claim 1, wherein, The dielectric property analysis is respectively performed on the plurality of types of human tissues based on the Cole-Cole formula modeling method, which comprises: When the dielectric property analysis is performed on the static human tissue, the influence factors of motion, respiration, temperature and humidity changes are subtracted from the Cole-Cole formula; When the dielectric property analysis is performed on the dynamic human tissue, the influence factors of motion, respiration, temperature and humidity changes are added to the Cole-Cole formula.

6. The method of claim 1, wherein the method further comprises: The response function of the human tissue to the blood glucose sensing signal is established for the blood glucose monitoring electromagnetic sensing model after the dielectric property assignment, which comprises: The propagation characteristics of the radio frequency sensing signal in different directions of the blood glucose monitoring electromagnetic sensing model are respectively obtained; The radio frequency response characteristics of the same human tissue at different frequencies and the radio frequency response characteristics of different human tissues at the same frequency are obtained according to the propagation characteristics by using the control variable method; The response function of the human tissue to the blood glucose sensing signal is established according to each radio frequency response characteristic.

7. A device for establishing a model of an electromagnetic sensor for radio frequency non-invasive blood glucose monitoring, characterized in that The establishing device comprises: A human tissue division module, which is configured to: initially divide human tissue into static tissue and dynamic tissue; transmit radio frequency sensing signals to the static tissue and the dynamic tissue respectively, obtain response characteristics of the radio frequency sensing signals, and perform clustering according to the response characteristics to obtain a plurality of types of human tissues; An electromagnetic sensor model initial modeling module configured to: perform voxel reconstruction on several types of human tissues by using a graph variation method with sparse regularization, to obtain a blood glucose monitoring electromagnetic sensor model; A dielectric property assignment and response construction module of the model configured to: respectively perform dielectric property analysis on several types of human tissues based on a Cole-Cole formula modeling method, assign dielectric properties to each voxel of the blood glucose monitoring electromagnetic sensor model according to the analysis results, and construct a response function of human tissues to blood glucose sensing signals for the blood glucose monitoring electromagnetic sensor model after the dielectric property assignment.

8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a radio frequency non-invasive blood glucose monitoring electromagnetic sensor model establishment program, and the radio frequency non-invasive blood glucose monitoring electromagnetic sensor model establishment program, when executed by the processor, implements the radio frequency non-invasive blood glucose monitoring electromagnetic sensor model establishment method in any one of claims 1 to 6.

9. A computer device, comprising: The computer device includes a computer readable storage medium, a processor, and a radio frequency non-invasive blood glucose monitoring electromagnetic sensor model establishment program stored in the computer readable storage medium, and the radio frequency non-invasive blood glucose monitoring electromagnetic sensor model establishment program, when executed by the processor, implements the radio frequency non-invasive blood glucose monitoring electromagnetic sensor model establishment method in any one of claims 1 to 6.

Images