Method and device for detecting encephaledema based on optomagnetic fusion

Abstract

The invention discloses a method and device for detecting encephaledema based on optomagnetic fusion, and relates to the field of biological detection.The method comprises the steps that to-be-processed optical information and to-be-processed magnetic information of the head of a human body are input into a preset brain tissue parameter detection model to obtain the encephaledema level of the head of the human body; the to-be-processed light information is emergent light information after the head of the human body is irradiated with multi-wavelength near-infrared light; the to-be-processed magnetic information is magnetic induction information corresponding to an excitation magnetic field generated around the head of the human body; the preset brain tissue parameter detection model is obtained by training a sparse automatic encoder deep learning network by adopting a photomagnetic information sample data set; the optomagnetic information sample data set is obtained by determining tissue parameters of each stage of encephaledema according to physiological and physical characteristics of the encephaledema, then performing magnetic field analog simulation and light field analog simulation at the same time, and calculating by adopting a finite element analysis method. The brain edema monitoring system and method can realize real-time and lossless continuous monitoring of brain edema.

Description

Technical Field

[0001] The present application relates to the field of biological detection, and in particular to a method and device for detecting brain edema based on optical-magnetic fusion. Background Art

[0002] At present, the commonly used methods for monitoring brain edema in clinical practice are mainly divided into invasive monitoring and non-invasive monitoring. Among them, the non-invasive monitoring mainly includes transcranial Doppler ultrasound, flash visual evoked potential, magnetic induction tomography, imaging monitoring, etc. Transcranial Doppler ultrasound is a non-invasive detection technology, a doppler examination method that uses a device that generates sound waves as energy. It is simple and easy to use and can be detected at the bedside; but it requires the examiner's skilled techniques and patience. Due to the density of the skull and the depth of detection, 5% to 15% of patients cannot find the basilar artery through the temporal window. The principle of flash visual evoked potential f-VEP is that neurons and neuronal fibers obtain energy from the blood circulation to meet their excitement and conduction; since f-VEP intracranial pressure detection itself is an electrophysiological examination, brain edema is generally accompanied by hypoxia, which can cause a significant decrease in the f-VEP waveform and unstable latency; hypoglycemia can also cause unstable f-VEP waveform. Imaging tests mainly include computed tomography, positron emission tomography, single photon computed tomography and functional magnetic resonance imaging; they are safe and non-invasive, and can accurately and intuitively display the conditions of various parts of brain tissue. However, imaging tests are not real-time, have high prices, and are complex to operate, and the equipment is expensive and large in size. Summary of the invention

[0003] The purpose of this application is to provide a method and device for detecting cerebral edema based on optical-magnetic fusion, which can realize real-time, non-destructive and continuous monitoring of cerebral edema.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a method for detecting cerebral edema based on optical-magnetic fusion, comprising:

[0006] Obtaining the light information to be processed and the magnetic information to be processed of the human head; the light information to be processed is the outgoing light information of the human head after irradiating the human head with multi-wavelength near-infrared light; the magnetic information to be processed is the magnetic induction information corresponding to the excitation magnetic field generated around the human head;

[0007] The optical information to be processed and the magnetic information to be processed are input into a preset brain tissue parameter detection model to obtain the brain edema level of the human head; wherein the preset brain tissue parameter detection model is obtained by training a sparse autoencoder deep learning network using an optical-magnetic information sample data set; the optical-magnetic information sample data set is obtained by determining the tissue parameters of each stage of brain edema according to the physiological and physical characteristics of brain edema, and then simultaneously performing magnetic field simulation and light field simulation, and calculating using the finite element analysis method.

[0008] In a second aspect, the present application provides a device for detecting cerebral edema based on optical-magnetic fusion, including a light source component, a magnetic detection component, a detection component and a main control component;

[0009] The light source component is used to: emit multi-wavelength near-infrared light to the human head according to the first instruction issued by the main control component;

[0010] The magnetic detection component is used to: generate an excitation magnetic field around the human head according to a second instruction issued by the main control component;

[0011] The detection component is used to: receive the emitted light from the human head and determine the corresponding light information to be processed; detect the eddy current magnetic induction signal corresponding to the excitation magnetic field to obtain the corresponding magnetic information to be processed;

[0012] The main control component is also used to: adopt a method for detecting brain edema based on optical-magnetic fusion, and determine the level of brain edema in the human head according to the magnetic information to be processed and the optical information to be processed.

[0013] According to the specific embodiments provided by the present application, the present application has the following technical effects: the present application provides a method and device for detecting brain edema based on optical-magnetic fusion, determines the tissue parameters of each stage of brain edema according to the physiological and physical characteristics of brain edema, and then simultaneously performs magnetic field simulation and light field simulation, and uses finite element analysis to calculate and obtain a sample data set of optical-magnetic information, realizes optical-magnetic feature fusion, and provides a data basis for subsequent processing of real-time detection using optical information and magnetic information at the same time; uses the optical-magnetic information sample data set to train the sparse autoencoder deep learning network to obtain a preset brain tissue parameter detection model, and then uses the trained model to simultaneously process the optical information to be processed and the magnetic information to be processed, and obtains the brain edema level of the human head, thereby realizing real-time and accurate judgment of the patient's brain edema. In addition, the detection method adopted by the present application realizes non-destructive detection, without the need to use expensive equipment, and the cost is greatly reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0015] Figure 1 A schematic diagram of a flow chart of a method for detecting cerebral edema based on optical-magnetic fusion provided in one embodiment of the present application;

[0016] Figure 2 A schematic diagram of the principle of obtaining the optical information to be processed of the human brain tissue model provided in one embodiment of the present application;

[0017] Figure 3 A schematic diagram of magnetic induction imaging provided in one embodiment of the present application;

[0018] Figure 4 An image of the distribution of excitation coils provided in one embodiment of the present application;

[0019] Figure 5 A schematic diagram of a COMSOL model of optical and magnetic physical fields provided in an embodiment of the present application;

[0020] Figure 6 A deep learning network structure diagram provided for an embodiment of the present application;

[0021] Figure 7 A schematic diagram of a device for detecting cerebral edema based on optical-magnetic fusion provided in one embodiment of the present application;

[0022] Figure 8 A flow chart of phase-locked demodulation based on a reference signal weighted-cumulative accumulation strategy provided in one embodiment of the present application.

[0023] Figure numerals: 10 - near-infrared light source, 11 - sampling point, 12 - bandage, 13 - magnetic induction coil, 14 - wearable component, 15 - detector. DETAILED DESCRIPTION

[0024] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0025] This application uses MIT (Magnetic induction tomography) and NIRS (Near Infrared Spectrum Instrument) technologies, supplemented by artificial intelligence, to research and develop multimodal wearable intelligent brain edema monitoring technology, to achieve real-time, non-destructive and continuous monitoring of brain edema caused by brain injury, thereby providing an important basis for timely rescue and plan formulation for patients with brain edema.

[0026] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0027] In an exemplary embodiment, Figure 1 As shown, the present application provides a method for detecting brain edema based on optical-magnetic fusion, which is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the application example of the present application, the method includes the following steps 101 to 102.

[0028] Step 101, obtaining the light information to be processed and the magnetic information to be processed of the human head; the light information to be processed is the outgoing light information of the human head after irradiating the human head with multi-wavelength near-infrared light; the magnetic information to be processed is the magnetic induction information corresponding to the excitation magnetic field generated around the human head; wherein, the light information to be processed includes the light signal peak value, average light intensity value and outgoing light intensity value of the outgoing light of the human head, and the magnetic information to be processed includes the conductivity, phase difference and amplitude of the induced magnetic field.

[0029] In an application example, between step 101 and step 102, the method further includes: cross-stacking and arranging the time domain sequence of the optical information to be processed of different modalities and the time domain sequence of the magnetic information to be processed to form a 2D array, thereby obtaining a multi-source physiological parameter fusion matrix. Specifically, the time domain sequences of near-infrared light parameters (i.e., optical information to be processed) and magnetic induction signals (i.e., magnetic information to be processed) of different modalities are vertically stacked to form a 2D array to fuse optical and magnetic signals, such as the top-down arrangement form: optical signal peak value, conductivity, average light intensity value, phase difference, exiting light intensity value and amplitude, thereby realizing cross-stacking of multimodal data. The above-mentioned fusion technology can simultaneously capture the dynamic changes of data over time and the spatial dependence on the modality. In the row dimension, the independent information of the optical information and the magnetic information can be obtained, and in the column dimension, the information related to the optical information and the magnetic information can be obtained.

[0030] Step 102, inputting the optical information to be processed and the magnetic information to be processed into a preset brain tissue parameter detection model to obtain the brain edema level of the human head; wherein the preset brain tissue parameter detection model is obtained by training a sparse autoencoder deep learning network using an optical-magnetic information sample data set; the optical-magnetic information sample data set is obtained by determining tissue parameters of each stage of brain edema according to the physiological and physical characteristics of brain edema, and then simultaneously performing magnetic field simulation and light field simulation, and calculating using finite element analysis.

[0031] Specifically, the process of determining the optical and magnetic information sample data set can be implemented using COMSOL software, and the following steps are performed in the COMSOL software:

[0032] (1) Based on the scanned image of the human head and the physiological and physical characteristics of brain edema, a human brain tissue model is constructed; the human brain tissue model includes a multi-layer head structure tissue and edema tissue embedded in the multi-layer head structure tissue.

[0033] Specifically, a real human head image is obtained by CT (Computed Tomography) / Magnetic Resonance Imaging (MRI) scanning of the human head, and then the human head is divided into the following types according to the human brain anatomy: Figure 2 The present application specifically uses COMSOL software to construct a human brain tissue model, which is divided into a two-dimensional multi-layer cranial optical tissue model with five layers of head structure tissue and edema tissue. Moreover, the edema tissue can be segmented by COMSOL software to obtain edema that is consistent with the shape and size of real intracranial edema.

[0034] (2) Scaling the irregular boundaries of the edema tissue in the human brain tissue model to obtain edema tissue at different stages. In the process of constructing the sample set, it is necessary to obtain edema tissues of different shapes and at different stages. Therefore, the present application establishes sample data of irregular edema tissue by scaling the boundaries of irregular edema. Specifically: first, the boundaries of irregular edema are extracted, and after obtaining the edema boundary data, the boundaries are enlarged and reduced according to a step size of 0.01 mm (relevant technical personnel can adjust as needed), and finally the sample data of edema tissue at different stages established by different edema size data is obtained.

[0035] (3) For edema tissue at any stage, magnetic field simulation and light field simulation are simultaneously performed on the corresponding human brain tissue model, and the human brain tissue model is segmented into finite element human brain tissue meshes. The process of segmenting the human brain tissue model into finite element human brain tissue meshes includes: using COMSOL software to determine the segmentation size based on the magnetic field simulation and the light field simulation; then segmenting the human brain tissue model in the form of triangular units based on the segmentation size.

[0036] (4) Based on the finite element human brain tissue grid and the light field simulation, a diffusion equation is used to describe the propagation of near-infrared light in the human head to obtain output light sample information of the human head corresponding to near-infrared light of different wavelengths, different emission positions, and different receiving positions.

[0037] It should be noted that near-infrared light has good penetration into the skin and skull, and can effectively obtain the optical tissue parameters of the cranial cortex. When brain edema occurs, the intracranial volume and tissue density change, the light penetration rate decreases, and the source-detection distance is limited. A single light source and detector cannot provide enough information. Multi-wavelength light sources and multiple detectors can improve detection accuracy through effective combination and information dimension reduction. In the simulation of this step, relevant technicians can explore the impact of light sources of different wavelengths and the arrangement of detectors on the accuracy of brain edema monitoring through multiple simulations.

[0038] When exploring the effect of different light source and detector arrangements on the detection of brain edema, light sources of different wavelengths (775nm, 855nm) and detectors of different arrangements (including number and spacing) are placed on the finite element human brain tissue grid or human brain tissue model, and the best combination is screened by detecting the light intensity. An example of the exploration process is as follows: Figure 2 As shown in the figure, the light source and detector are located on the surface of the human brain tissue model and close to the human brain tissue model. The five detectors are located on the same side or around the light source. The straight-line distances between the detectors and the light source are 1 cm, 2 cm, 3 cm, 4 cm, and 5 cm, respectively. The source-detector spacing depends on the head circumference, but it is recommended that the spacing for adults should not exceed 3.5 cm. The COMSOL software is used to analyze the parameters such as the output light intensity, light signal peak, signal complexity, and brain tissue light intensity variation of the model at different levels of brain edema, and the corresponding relationship between the development level of brain edema and the optical parameters can be obtained.

[0039] In the light field simulation process, the relevant processing of the NIRS forward model needs to be considered. The positive problem of optical detection of biological tissues is to analyze the corresponding outgoing light information when studying the known optical parameters of biological tissues, and to understand the propagation process of light in biological tissues and their interactions. The theoretical basis of the NIRS forward problem is based on the diffusion equation. The near-infrared light incident from the light source first undergoes the absorption and scattering process of five layers of head structure tissue (including scalp, skull, cerebrospinal fluid, white matter and gray matter) inside the biological tissue. The skin will reflect the near-infrared light to the surface according to the absorption and scattering coefficient of the biological tissue. Finally, the detector set on the surface of the biological tissue is used to collect the near-infrared light outgoing light intensity information to complete the NIRS forward process.

[0040] In this application, the diffusion equation is used to approximately describe the transmission law of the excitation light in the human brain. Under the irradiation of near-infrared light of different wavelengths, the formula of the complex frequency domain coupling diffusion equation between the near-infrared light and the outgoing light through Laplace transformation is as follows.

[0041]

[0042]

[0043]

[0044] Wherein, x and m represent the incident wavelength and the emitted wavelength of the near-infrared light, respectively. Since the incident wavelength and the emitted wavelength are usually very close, they can be considered to be approximately equal in practical applications. Based on this, the absorption coefficient μ of the excitation light is set in this application. ax , reduced scattering coefficient μ' sx and the fluorescence absorption coefficient μ am , reduced scattering coefficient μ' sm are equal. c is the propagation speed of light in biological tissues, ▽ represents the gradient operator, is the Laplace transform of the time-dependent photon density, and β represents the Laplace transform factor. represents the position of the light source, k x,m is the diffusion coefficient.

[0045] Re-use the Robin boundary condition: And Fick's law: The output optical flux of the body surface of the biological tissue is obtained.

[0046] Next, the finite element method, a common numerical solution to the diffusion equation, is used to solve the propagation model of the light source in the tissue. The establishment of the finite element equation mainly includes the variational method and the weighted residual method. Among them, the Galerkin finite element method has comprehensive advantages in terms of computing speed and mathematical processing performance for deriving finite element equations, so it is often used in practical calculations. The Galerkin method requires a finite element human brain tissue mesh based on segmented triangular units, that is, the definition domain is divided into E independent triangulation units that completely cover the tissue body, and the vertices of the triangulation units constitute N segmentation nodes. The calculation formula of the photon density function is as follows:

[0047] in, represents the domain polynomial shape function, represents the photon density at the finite element mesh node, and N is the number of mesh nodes. The calculation formula of the above photon density function is solved using the Calerkin finite element method to obtain the following equation: The coefficient matrices in this equation are shown below.

[0048]

[0049]

[0050] The extrapolated boundary conditions are used to solve the equation Q(i,t). The shape function value on is 0, so the equation becomes the formula: Aφ(β)=Q.

[0051] From the above content, we can know that based on the finite element solution method, the solution of the DE equation is transformed into a linear equation solution problem. First, the photon density distribution Ω of the excitation light is calculated. x The solution is then used to solve the diffusion equation corresponding to the emitted light using the photon density distribution, and finally the photon density distribution of the emitted light in the biological tissue is obtained. At this point, the light field simulation in the human brain tissue model is realized.

[0052] (5) Based on the finite element human brain tissue grid and the magnetic field simulation, the magnetic sample information corresponding to the human head is calculated; the magnetic sample information corresponding to the human head is specifically: the effect of brain edema on the magnetic field in the human head, and the magnetic information detected by the induction coil outside the human head. The output light sample information of the human head corresponding to the edema tissue at each stage and the magnetic sample information corresponding to the human head constitute an optical magnetic information sample data set, and the stage of the edema tissue is used as the sample label information.

[0053] It should be noted that the goal of magnetic induction imaging is to reconstruct the conductivity of the target object. Figure 3 As shown, a sinusoidal alternating current of a specific frequency is passed through the excitation coil, and the excitation coil will generate an original magnetic field B0 in space, and the target object (the human brain tissue model in this application) is affected by the original magnetic field in the imaging domain to generate eddy currents, thereby generating a secondary magnetic field ΔB. When the above two magnetic fields are superimposed, a new magnetic field B0+ΔB is formed. The new magnetic fields induced by different biological tissues at the same excitation frequency are different. For example, when edema occurs, the induced magnetic field is different from that of normal tissue, so the current measured by the detection coil will also be different. By analyzing the difference in the eigenvalues ​​carried by the detection signal and through subsequent signal processing, it is possible to detect whether the biological tissue is abnormal, thereby providing a basis for the doctor's diagnosis. In the research process of magnetic induction imaging, the study of direct problems and inverse problems is usually involved. When the distribution of the electrical conductivity of biological tissues and the frequency, amplitude and other information of the excitation signal are known, the process of solving the distribution of the magnetic field inside and outside the target object is called the direct problem study of magnetic induction imaging technology. The process of reconstructing the image using the magnetic field information obtained by the detection coil is called the inverse problem study of magnetic induction imaging technology. This application uses the MIT direct problem mathematical model in the magnetic field simulation process.

[0054] In the process of magnetic field simulation, for alternating magnetic fields, the laws of electromagnetic fields can be characterized by the basic Maxwell equations, which describe the interaction between electric and magnetic fields and their changes over time and space. They can be expressed in differential form or integral form. In this application, the differential form of Maxwell's equations is:

[0055]

[0056] Among them, H represents the magnetic field intensity, E is the electric field intensity, μ and ε represent the magnetic permeability and dielectric constant of the medium in the field, B represents the magnetic induction intensity, and D is the electric displacement vector. The differential form of Maxwell's equations can be further transformed into the following equation:

[0057]

[0058] Where ω is the angular frequency. Then the magnetic vector potential A is introduced, and the relationship between the magnetic vector potential, the magnetic field intensity and the magnetic induction intensity is:

[0059]

[0060] E=-jωA.

[0061] Substituting the relationship between magnetic vector potential, magnetic field intensity and magnetic induction intensity into the first equation of the transformed Maxwell equations, we have:

[0062]

[0063] From the above formula, we can see that E+jωA is irrotational, so it can be written as the gradient of the scalar potential U, then:

[0064]

[0065] In addition, in the process of detecting biological tissues, there is no charge generating a field in the imaging domain, so:

[0066]

[0067] Further deduction, the governing equation of the magnetic induction direct problem can be expressed as:

[0068]

[0069] By solving the governing equations of the magnetic induction direct problem, the induced electromotive force on the detection coil can be expressed by a formula including the magnetic vector potential A, as shown below:

[0070] U=-iω∮A·dl.

[0071] Where dl is the length unit of the detection coil. For biological tissue, its conductivity corresponds to the imaginary part of the phase shift and plays a leading role in image reconstruction. Therefore, the phase shift of the measured tissue can be expressed as:

[0072]

[0073] Among them, U0 is the empty field induced voltage when there is no object to be measured, and U1 is the induced voltage when the object to be measured exists. Corresponding to the present application, the object to be measured is a human brain tissue model. After determining the above-mentioned MIT imaging positive problem, a finite element human brain tissue mesh is used to implement a finite element-based solution. The finite element human brain tissue mesh includes a plurality of triangular units, and the conductivity information at the center point of each triangular unit represents the conductivity information of the entire triangular unit. At the same time, COMSOL software is used to generate phase difference and amplitude data under different conductivity distributions. Since the location of brain edema mainly occurs on the inner side of the skull, it is important to dissect the tissue where brain edema may occur in order to more clearly show the edge of the heterogeneous body, that is, the edema.

[0074] In addition, through research, it was found that the phase difference contains the location information of brain edema. When there are 100 detection coils, the trend of the phase difference is related to the location of brain edema. The peak of the phase difference appears near the detection coil where the brain edema occurs. Therefore, the conductivity distribution image when brain edema occurs can be reconstructed through the phase difference. Figure 4 For example, only through Figure 4When the conductivity distribution image is reconstructed using the unidirectional phase difference of the red excitation coil, the trend of the phase difference and the coordinates of the peak occurrence mainly represent the y-axis coordinates of the cerebral edema, and the amplitude of the phase difference represents the x-axis coordinates of the cerebral edema. However, since the amplitude of the phase difference is extremely small, noise will affect the positioning of the cerebral edema on the x-axis, causing the position of the heterogeneous body in the imaging result to shift, thereby increasing the centroid distance error. Even if the number of detection coils is increased, the effect is not significantly improved, and this phenomenon will seriously affect the doctor's clinical judgment. Therefore, if the number of detection coils is to be reduced, the noise resistance of the phase difference data needs to be improved. Based on this, the present application also provides a bidirectional detection method, that is, in the step of calculating the external magnetic sample information of the human head based on the finite element human brain tissue grid and the magnetic field simulation, when the magnetic sample information represents the phase difference, it includes:

[0075] (51) determining the current position of the excitation coil and the current position of the induction coil set in the magnetic field simulation; Figure 4 As shown, the red excitation coil and the red induction coil are the corresponding current positions.

[0076] (52) Based on the finite element human brain tissue grid, a magnetic field simulation is performed according to the current position of the excitation coil and the current position of the induction coil to obtain a first type of phase difference data.

[0077] (53) Rotate the excitation coil and the induction coil in the same direction by a preset angle to obtain an updated position of the excitation coil and an updated position of the induction coil. Figure 4 As shown, the corresponding updated position is obtained by rotating the excitation coil and the detection coil at the red position 90° clockwise to the blue position.

[0078] (54) Based on the finite element human brain tissue grid, a magnetic field simulation is performed according to the updated position of the excitation coil and the updated position of the induction coil to obtain a second type of phase difference data; the first type of phase difference data and the second type of phase difference data are phase differences corresponding to the positions of different excitation coils or different induction coils.

[0079] In the case where the number of detection coils remains unchanged, the x component of the magnetic field at the blue position of the head and the y component of the magnetic field at the red position are detected in the above steps (52) and (54), respectively, to obtain magnetic field information in two perpendicular directions. By comparison, it can be concluded that the coil at the blue position is more sensitive to the x component of the magnetic flux density of the magnetic field, and the coil at the red position is more sensitive to the y component of the magnetic field change. Therefore, in the process of constructing sample data, the present application obtains the phase difference corresponding to the two different positions, and then calculates the phase difference through the magnetic flux density information in two perpendicular directions, reducing the weight of the trend and amplitude characteristics in the single reverse phase difference information for locating the position of brain edema, thereby increasing the anti-noise performance of the phase difference information, and the centroid distance is smaller, so that the position of brain edema can be more accurately located.

[0080] In another specific application example, from the similarities between the light field simulation in the above step (4) and the magnetic field simulation in the above step (5), both simulations use COMSOL software, specifically COMSOL Multiphysics 6.1 software, which has the ability to simulate and analyze various physical fields, so that a human brain tissue model can be established by effectively simulating the occasional effects of magnetic fields, light fields and multi-physics fields, such as Figure 5 In the simulation process after the human brain tissue model is built, in addition to the contents recorded in the above steps (4) and (5), the following settings also need to be considered:

[0081] Set the geometric size, shape, boundary conditions and constraints for the two-dimensional human brain tissue model to ensure that the model meets the actual problem. In optical and magnetic analysis, some material physical properties can be directly selected from the material library provided by COMSOL software, such as selecting copper as the coil material and setting the air domain, etc.; material properties can be set manually, such as conductivity, absorption and scattering coefficient, refractive index, etc. When the established geometric model is divided into finite element grids, the finer the grid division, the higher the solution accuracy and the better the image quality, but the corresponding calculation time will also increase, so it is crucial to choose an appropriate division degree. Finally, solve and analyze the results, and derive the parameters such as the output light intensity value, light signal peak value, signal complexity, brain tissue light intensity variation, conductivity and phase amplitude when edema occurs.

[0082] When using COMSOL software to perform light field simulation and magnetic field simulation, the physical field interfaces selected in this application are the "Electromagnetic Waves, Frequency Domain" interface, the "Magnetic Field and Electric Field" interface, and the "Helmholtz Equation" interface. Since MIT usually ignores the propagation effect, according to the mathematical model of the forward problem, the generation of eddy currents is mainly related to the conductivity of the conductive medium, and the measured phase difference signal is related to the conductivity inside the conductive tissue. Therefore, this application ignores the influence of the dielectric constant of the conductive medium. In addition, the difference in magnetic permeability between tissues has little effect on the reconstructed image results. Therefore, MIT mainly detects the changes in conductivity caused by lesions during the research process. During the simulation process, the dielectric constant of brain tissue is set to 1C 2 / (N·M 2 ), the magnetic permeability is set to 1H / m. A circular area with a radius of 250mm is used to simulate an infinite air domain. A 50mm×80mm rectangle is used in the air domain to simulate the excitation coil, which is 160mm away from the center of the air domain. The number of coil turns is 20, the current is 1A, and the excitation frequency is 1MHz. The material of the coil is set to copper, and the conductivity is 3.526e 7 S / m. The detection coil is used to detect the phase data at multiple fixed detection points around the tissue. The position of the coil is movable, which improves the flexibility of the detection system. During the construction of the NIRS physical field simulation, the incident port is set to be the same as the exit port, for a total of 16 ports, evenly distributed on the human brain tissue model. The scattering boundary conditions are set, and the outermost air domain is set to a perfect matching layer to prevent the second reflection of the near-infrared light after emission from affecting the experimental data. The relevant parameter settings of the brain tissue are shown in Table 1 below. This application divides the simulation model by finite element subdivision, and obtains a total of 69406 grids with a maximum unit size of 3mm.

[0083] Table 1

[0084] Brain tissue name Refractive Index Absorption coefficient (1 / m) Scattering coefficient (1 / m) skin 1.4 1810 21 skull 1.55 1520 19 White matter 1.33 8650 17 Gray matter 1.39 2090 42 Cerebrospinal fluid 1.38 230 5

[0085] In another specific application example, the process of determining the type of specific data contained in the optical information to be processed and the magnetic information to be processed is as follows: Compared with the single near-infrared light or magnetic induction technology to monitor brain edema, the fusion of optical and magnetic signals can provide richer information. In order to improve the detection accuracy of the development level of brain edema, the average light intensity value at a fixed wavelength extracted by COMSOL software, the degree of variation of brain tissue light intensity, the maximum value of the light signal, the absorption coefficient, the scattering coefficient, the output light intensity value, the conductivity, the magnetic permeability, the phase difference, the amplitude and the impedance change are screened to determine whether the selected features are statistically significant. The KS test is used to determine whether the above indicators obey the normal distribution. The spearman method measures the correlation between the indicators before and after the occurrence of brain edema, and it is concluded that there is no significant difference in the four indicators of absorption, scattering coefficient, magnetic permeability and impedance before and after edema. Then the Relief method is used to select features with higher weights to avoid feature redundancy, and finally the features are screened out: optical signal peak value, average light intensity value, conductivity, phase difference, output light intensity value, amplitude.

[0086] After obtaining sample data containing the above six features through COMSOL software, combined with the sparse autoencoder deep learning network, such as Figure 6 As shown in the figure, the model uses a four-layer SAE deep learning network algorithm, with the detected optical and magnetic features as input and the corresponding brain edema level as output. The two hidden layers in the sparse autoencoder deep learning network are 200 neurons and 60 neurons respectively. The activation function uses the Sigmoid function for linear to nonlinear transformation, and the learning rate is 0.1. The sparse autoencoder deep learning network model is used to realize the nonlinear mapping of optical and magnetic response information to the distribution of brain tissue characteristic parameters. At the same time, accuracy, precision, recall rate, F1 value, and AUC value are used to evaluate the performance of the model, which further improves the accuracy of the model judgment results.

[0087] Based on the above-mentioned six features finally selected and the sparse autoencoder deep learning network adopted, this application sets up three groups for modeling respectively, and the modeling parameters of different models are shown in Table 2. Among them, the NIRS model only uses light characteristic parameters (including light signal peak value, average light intensity value, and exit light intensity value) for modeling; the MIT model only uses magnetic characteristic parameters (conductivity, phase difference, and amplitude) for modeling; the NIRS-MIT model uses light characteristic parameters and magnetic characteristic parameters to merge together, that is, light signal peak value, average light intensity value, exit light intensity value, conductivity, phase difference, and amplitude are used as input for modeling. And the NIRS model is used as the control group, and the MIT model and the NIRS-MIT model are used as the experimental group to better study the parameters related to the level of development of cerebral edema.

[0088] Table 2

[0089]

[0090]

[0091] The results showed that the accuracy of cerebral edema detection using the fusion of optical and magnetic feature parameters was improved by about 9% compared to using only optical or magnetic feature parameters, further illustrating that the fusion of multiple physical field features is more effective than using a single detection method.

[0092] The brain edema level output by the sparse autoencoder deep learning network in this application includes three levels: no brain edema, mild brain edema, and severe brain edema. The label value of no brain edema is defined as 0, the label value of mild brain edema is defined as 1, and the label value of severe brain edema is defined as 2, so as to facilitate training in the deep learning network.

[0093] Based on the same inventive concept, the embodiment of the present application also provides a device for detecting brain edema based on optical-magnetic fusion, including a light source component, a magnetic detection component, a detection component and a main control component. The light source component is used to: emit multi-wavelength near-infrared light to the human head according to the first instruction issued by the main control component; the magnetic detection component is used to: generate an excitation magnetic field around the human head according to the second instruction issued by the main control component; the detection component is used to: receive the emitted light of the human head and determine the corresponding light information to be processed; detect the eddy current magnetic induction signal corresponding to the excitation magnetic field to obtain the corresponding magnetic information to be processed; the main control component is also used to: adopt the above-mentioned method for detecting brain edema based on optical-magnetic fusion, and determine the level of brain edema in the human head according to the magnetic information to be processed and the optical information to be processed.

[0094] In practical applications, such as Figure 7 As shown, the light source component is a near-infrared light source 10; the magnetic detection component includes a plurality of magnetic induction coils and a metal rod; the metal rod is specifically arranged on a bandage 12 behind the human head according to the wearable component being located behind the human head to enhance the magnetic induction signal.

[0095] In practical applications, the device for detecting cerebral edema based on optical-magnetic fusion further includes a wearable component 14; the light source component, the magnetic detection component and the detection component are all arranged on the wearable component 14; the light source component and the detection component are both located on the surface of the human head according to the wearable component 14. The magnetic induction coil 13 is located on the side of the human head according to the wearable component 14. The detection component includes a plurality of detectors 15; the plurality of detectors 15 are located on the same side or around the light source component, and the straight-line distances between different detectors 15 and the light source component are different; the detector 15 includes an integrated photodiode with a built-in transimpedance amplifier.

[0096] In another practical application, according to the absorption characteristics of biological tissue in the near-infrared band and the basic principles of NIRS measurement technology, and referring to the performance of wavelength combinations in eliminating the interactive effects of oxidized / reduced hemoglobin concentration changes in existing studies, the near-infrared light source 10 can emit near-infrared light of two wavelengths, 775nm and 855nm, to obtain two hemoglobin concentration changes reflecting brain activation information, and a crystal-format acquisition four-source four-detector array can be used, with a total of ten sampling points 11 to obtain optical data of head tissue, corresponding to Figure 7 In the figure, the diamond represents the sampling point 11. According to the basic principle of magnetic induction imaging technology for detecting brain edema, and referring to the changes in magnetic field information caused by the distribution of magnetic induction coils around the brain in existing studies, the developed MIT device contains a total of seven magnetic induction coils.

[0097] Furthermore, the near-infrared light source 10 can be configured with a light source array according to imaging requirements. Taking a four-source-four-detector lattice array as an example, a voltage reference chip, an operational amplifier, and a self-developed constant current drive circuit of a field effect tube are used as a driver, and a dual-wavelength LED light source wrapped in flexible skin-friendly foam is used to emit dual-wavelength near-infrared light, and square wave modulation of all LED light sources is achieved through analog switches. The outer end of the detector 15 is wrapped with flexible skin-friendly foam, and a sixteen-bit synchronous sampling analog-to-digital data acquisition system is combined to complete digital sampling, and the measurement data is transmitted to the main control component through the SPI communication protocol.

[0098] In an application example, the main control component is based on Cortex TM -M4 core microprocessor as the main control unit, generates the timestamp information, square wave modulation model, demodulation signal, etc. required for the measurement process, and completes the software phase-locked demodulation process based on the reference signal weighted-cumulative accumulation strategy, such as Figure 8 As shown; finally, based on the efficient time division multiplexing strategy, DMA is used to realize the acquisition-transmission dual-thread data processing. In addition, in the signal acquisition process, in order to avoid interference, the optical and magnetic signals adopt the time-sharing detection method, and the detection accuracy is improved by optical signal delay measurement.

[0099] In summary, compared with the prior art, the present application has the following advantages:

[0100] (1) This application determines the tissue parameters of each stage of brain edema based on the physiological and physical characteristics of brain edema, and then uses the finite element method to analyze the influence of brain tissue optical parameters, magnetic response parameters and field stimulation parameters at different levels of brain edema on eddy current distribution, amplitude-frequency and phase-frequency response of the induced magnetic field. On this basis, research is conducted on the optimization of neural network modeling, and by improving the model and amplifying the artificial sample set, the network structure parameters and connection weights are continuously optimized to establish a neural network model for accurate positioning and continuous monitoring of brain edema; in addition, the sensitivity, robustness and generalization ability of the algorithm can be studied through phantom experiments and animal experiments.

[0101] Among them, in the process of finite element method analysis, the principles of biomedical photonics, electromagnetic induction and eddy current effect are used to design photoelectric detection schemes and high-frequency sinusoidal signal excitation schemes for excitation coils. The photoelectric detection scheme is mainly based on the principle of near-infrared detection, and the degree of tissue edema is obtained by judging the changes in the optical parameters of cranial and brain tissues. Electromagnetic induction imaging causes the excitation coil to generate an excitation magnetic field, which will induce eddy currents inside the imaging body. A high-precision magnetic flux density detection system is designed, and the eddy currents will generate a disturbing magnetic field; the signal is collected by the receiving coil, and the conductivity distribution inside the imaging body is calculated to judge the degree of edema.

[0102] (2) This application integrates the optical detection and magnetic detection systems into an integrated design, which complements the problems of shallow detection depth but high resolution of brain edema by near-infrared light detection and deep detection depth but poor resolution of brain edema by magnetic detection, proving that the fusion of optical and magnetic signal features is significantly better than the effect of using only a single method for brain edema detection. This application also constructs a multi-source physiological parameter fusion matrix, mapping the characteristic parameters obtained by NIRS and MIT into an easy-to-use dimensionless matrix, which can obtain independent information of near-infrared light and magnetic induction signals in the row dimension, and information on the correlation between near-infrared light and magnetic induction signals in the column dimension, which helps to grade the development level of brain edema from a more detailed perspective. This application also integrates a transmission module and a control module, which can obtain the optical parameters and magnetic response parameters of brain edema in real time, so as to carry out early warning and grading of brain edema lesions.

[0103] (3) The device of the present application has developed a low-density forehead band system suitable for monitoring brain function among groups in natural situations from the perspectives of measurement dynamics, environmental adaptability, transmission reliability, information integrity and multi-machine synchronization. Specifically, the present application uses wearable components to set sampling points and detectors in the forehead area, that is, the front of the scalp, and then can monitor relevant information through fewer sampling points and detectors. The device of the present application also breaks through the limitations of the synchronization of multi-terminal wireless transmission, develops an efficient time division multiplexing strategy, realizes dual-threaded acquisition-transmission operation, configures a high-precision timer to provide timestamp information, eliminates the time difference of asynchronous communication between the host and multiple terminals, and realizes high-precision transmission of measurement information.

[0104] (4) The device of the present application solves the problem of system portability and reconstruction accuracy. It develops a real-time online software phase-locked demodulation strategy based on the reference signal weighted-cumulative accumulation strategy based on the on-chip transimpedance amplification photodiode, and combines the source-probe array with the micro-hardware circuit design to fully obtain the time series information of the prefrontal cortex stress blood oxygen response, and establishes an advanced inversion framework that overcomes the OT quantification and DOT robustness, effectively improving the NIRS imaging quantification and spatial resolution under the premise of wearability.

[0105] (5) The coil and source-detector array module used in this application are small in size and do not affect the wearing comfort at all. In addition, by reducing the distance between the detection sensor and the human head, the detection distance is reduced and the accuracy of the detection signal is improved. At the same time, in order to reduce the magnetic field signal generated by the electronic components inside the helmet, the internal electronic components use wired transmission methods as much as possible, separate the necessary wireless transmission components from the detection sensor, and increase the magnetic field isolation material to isolate the magnetic field generated by the necessary wireless transmission components.

[0106] (6) This application uses COMSOL software to fuse the optical field and magnetic field, solving the problem of inconsistent finite element segmentation caused by the existing near-infrared light detection of brain edema using NIRFAST Slicer software. By simultaneously segmenting the optical and magnetic physical fields to find the appropriate segmentation size, the problem of balancing calculation time and accuracy is solved. At the same time, multi-channel acquisition is used to avoid errors caused by inconsistent signal acquisition time.

[0107] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0108] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for detecting brain edema based on optical-magnetic fusion, characterized in that: The method for detecting cerebral edema based on optical-magnetic fusion comprises: Obtaining the light information to be processed and the magnetic information to be processed of the human head; the light information to be processed is the outgoing light information of the human head after irradiating the human head with multi-wavelength near-infrared light; the magnetic information to be processed is the magnetic induction information corresponding to the excitation magnetic field generated around the human head; The optical information to be processed and the magnetic information to be processed are input into a preset brain tissue parameter detection model to obtain the brain edema level of the human head; wherein the preset brain tissue parameter detection model is obtained by training a sparse autoencoder deep learning network using an optical-magnetic information sample data set; the optical-magnetic information sample data set is obtained by determining the tissue parameters of each stage of brain edema according to the physiological and physical characteristics of brain edema, and then simultaneously performing magnetic field simulation and light field simulation, and calculating using the finite element analysis method.

2. The method for detecting brain edema based on optical-magnetic fusion according to claim 1, characterized in that: The optical information to be processed includes the optical signal peak value, average light intensity value and output light intensity value of the output light from the human head, and the magnetic information to be processed includes the conductivity, phase difference and amplitude of the induced magnetic field.

3. The method for detecting brain edema based on optical-magnetic fusion according to claim 1, characterized in that: The process of determining the optical and magnetic information sample data set includes: Based on the scanned image of the human head and the physiological and physical characteristics of brain edema, a human brain tissue model is constructed; the human brain tissue model includes a multi-layer head structure tissue and edema tissue embedded in the multi-layer head structure tissue; scaling the irregular boundaries of the edema tissue in the human brain tissue model to obtain edema tissue at different stages; For edema tissue at any stage, magnetic field simulation and light field simulation are simultaneously performed on the corresponding human brain tissue model, and the human brain tissue model is divided into finite element human brain tissue grids; Based on the finite element human brain tissue grid and the light field simulation, a diffusion equation is used to describe the propagation of near-infrared light in the human head, so as to obtain the outgoing light sample information of the human head corresponding to near-infrared light of different wavelengths, different emission positions, and different receiving positions; Based on the finite element human brain tissue grid and the magnetic field simulation, the magnetic sample information corresponding to the human head is calculated; the output light sample information of the human head corresponding to the edematous tissue in each stage and the magnetic sample information corresponding to the human head constitute an optical-magnetic information sample data set, and the stage of the edematous tissue is used as the sample label information.

4. The method for detecting brain edema based on optical-magnetic fusion according to claim 1, characterized in that: The process of determining the optical and magnetic information sample data set is implemented using COMSOL software.

5. The method for detecting brain edema based on optical-magnetic fusion according to claim 1, characterized in that: Before the step of inputting the optical information to be processed and the magnetic information to be processed into a preset brain tissue parameter detection model, the method further includes: The time domain sequence of the optical information to be processed of different modalities and the time domain sequence of the magnetic information to be processed are cross-stacked and arranged to form a 2D array, thereby obtaining a multi-source physiological parameter fusion matrix.

6. The method for detecting brain edema based on optical-magnetic fusion according to claim 3, characterized in that: The process of dividing the human brain tissue model into finite element human brain tissue grids includes: using COMSOL software to determine the division size based on the magnetic field simulation and the light field simulation; and then dividing the human brain tissue model in the form of triangular units based on the division size.

7. The method for detecting brain edema based on optical-magnetic fusion according to claim 3, characterized in that: In the step of calculating the magnetic sample information corresponding to the human head based on the finite element human brain tissue grid and the magnetic field simulation, when the magnetic sample information represents a phase difference, the step includes: Determining the current position of the excitation coil and the current position of the induction coil set in the magnetic field simulation; Based on the finite element human brain tissue grid, a magnetic field simulation is performed according to the current position of the excitation coil and the current position of the induction coil to obtain first-type phase difference data; Rotating the excitation coil and the induction coil in the same direction by a preset angle to obtain an updated position of the excitation coil and an updated position of the induction coil; Based on the finite element human brain tissue grid, magnetic field simulation is performed according to the updated position of the excitation coil and the updated position of the induction coil to obtain the second type of phase difference data; the first type of phase difference data and the second type of phase difference data are the phase differences corresponding to the positions of different excitation coils or different induction coils.

8. A device for detecting brain edema based on optical-magnetic fusion, characterized in that: The device for detecting cerebral edema based on optical-magnetic fusion comprises a light source component, a magnetic detection component, a detection component and a main control component; The light source component is used to: emit multi-wavelength near-infrared light to the human head according to the first instruction issued by the main control component; The magnetic detection component is used to: generate an excitation magnetic field around the human head according to a second instruction issued by the main control component; The detection component is used to: receive the emitted light from the human head and determine the corresponding light information to be processed; detect the eddy current magnetic induction signal corresponding to the excitation magnetic field to obtain the corresponding magnetic information to be processed; The main control component is also used to: adopt the method for detecting brain edema based on optical-magnetic fusion as described in any one of claims 1-7, and determine the level of brain edema in the human head according to the magnetic information to be processed and the optical information to be processed.

9. The device for detecting brain edema based on optical-magnetic fusion according to claim 8, characterized in that: The device for detecting cerebral edema based on optical-magnetic fusion further includes a wearable component; the light source component, the magnetic detection component and the detection component are all arranged on the wearable component; The light source assembly and the detection assembly are both located on the surface of the human head according to the wearable component; The detection assembly includes a plurality of detectors; the plurality of detectors are located on the same side or around the light source assembly, and the straight-line distances between different detectors and the light source assembly are different; The detector includes an integrated photodiode with a built-in transimpedance amplifier.

10. The device for detecting brain edema based on optical-magnetic fusion according to claim 9, characterized in that: The magnetic detection component includes multiple magnetic induction coils and a metal rod; wherein the metal rod is located behind the human head according to the wearable component; and the magnetic induction coil is located on the side of the human head according to the wearable component.

Images