EIT image reconstruction method, equipment and program product

Abstract

The invention relates to the field of intelligent medical treatment, in particular to an EIT image reconstruction method and device and a program product. Comprising the following steps: acquiring EIT measurement voltage change data; constructing a reconstruction matrix R, and calculating the measurement voltage change data through the reconstruction matrix R to obtain a reconstruction result; wherein the reconstruction matrix R is an error calculation for minimizing an expected conductivity change and a reconstruction conductivity change; the expected conductivity change is obtained by acquiring conductivity change data, constructing a modification matrix D and modifying the conductivity change data through the modification matrix D; the modification matrix D modifies the conductivity change through consensus indexes of the EIT image, and the consensus indexes comprise one or more of AR, PE, RNG, RES and SD. The application has good clinical value.

Description

Technical Field

[0001] This application relates to the field of intelligent healthcare, specifically to an EIT image reconstruction method, device, program product, and computer-readable storage medium. Background Technology

[0002] Electrical impedance tomography (EIT) reconstructs the internal conductivity distribution by measuring voltage changes using surface electrodes. With its advantages of being non-invasive, radiation-free, real-time, and portable, it holds significant potential in areas such as lung function monitoring, brain imaging, and breast cancer screening. However, current image reconstruction methods face serious challenges, including the inherent pathological nature of the inverse problem leading to unstable solutions and sensitivity to noise; simplifications of the forward problem model (such as neglecting electrode contact impedance and tissue anisotropy) introducing errors; inaccuracies in boundary shape and electrode position significantly degrading image quality; and limited measurement data (limited to boundary voltage only) resulting in low spatial resolution and severe artifacts. In particular, 2D EIT assumes that the measured domain is uniform and infinitely extended in the third dimension (z-axis), neglecting the actual three-dimensional diffusion of current and boundary effects. This leads to significant distortion, blurring, and quantization errors when reconstructing real three-dimensional targets (such as the human torso and organs), severely limiting its clinical applicability in scenarios with complex anatomical structures or requiring precise three-dimensional localization. Summary of the Invention

[0003] To address the above problems, this invention provides an EIT image reconstruction method, specifically including:

[0004] Acquire EIT measurement voltage change data;

[0005] A reconstruction matrix R is constructed, and the reconstruction result is obtained by calculating the measured voltage change data through the reconstruction matrix R.

[0006] Wherein, the reconstruction matrix R is the error calculation that minimizes the change in desired conductivity and the change in reconstructed conductivity;

[0007] The desired conductivity change is obtained by first acquiring conductivity change data and constructing a modification matrix D, and then modifying the conductivity change data using the modification matrix D. The modification matrix D modifies the conductivity change using consensus indicators of the EIT image. The consensus indicators include one or more of the following: AR, PE, RNG, RES, and SD.

[0008] The modification matrix D is the desired modification of the EIT measurement unit through the point spread function. The EIT measurement unit is a tissue unit at different locations in the measurement site.

[0009] Optionally, the point diffusion function divides the organizational units at different locations into regions, including a first region and a second region. The first region is the region where the organizational units at different locations are less than or equal to the diffusion radius from the diffusion center, and the second region is the region where the organizational units at different locations are greater than the diffusion radius from the diffusion center. The first region and the second region are then modified as desired.

[0010] Optionally, the formula for constructing the modification matrix D is:

[0011]

[0012]

[0013] in, It is a point spread function, p0 is The diffusion center, p represents the center position of the FEM element in the finite element model, and r is diffusion radius, The matrix represents the modification matrix, where i and j represent the electrodes of the EIT, s is a fuzzy control parameter, and b represents the control parameter for the positive compensation amplitude.

[0014] The modification matrix is ​​modified by RNG through... Positive compensation of RNG yields the results of changes in conductivity after RNG modification;

[0015] Optionally, let the compensation range of RNG be a, and the formula for calculating the compensation range is:

[0016]

[0017] Where b represents the control parameter for the positive compensation amplitude;

[0018] Optionally, the modification of RNG also includes RNG suppression, which is performed on the result of RNG modification conductivity change by a single-step feedback strategy to obtain the result of suppressed RNG modification conductivity change.

[0019] Optionally, the single-step feedback strategy process is as follows:

[0020] S1, let the modification matrix Calculations are performed when the compensation range limit approaches zero. Obtain the first modification matrix, and construct the first reconstruction matrix based on the first modification matrix;

[0021] S2. Calculate a of the first reconstruction matrix. t a t It is a normalized amplitude with reference to the maximum positive AR;

[0022] S3, let the modification matrix The second modification matrix is ​​calculated, and the result of suppressing the change in RNG modified conductivity is obtained based on the second modification matrix.

[0023] Optionally, a reconstruction matrix is ​​obtained by constructing a reconstruction matrix based on the second modification matrix.

[0024] The consensus metric AR also includes sensitivity compensation, which obtains uniform AR and modifies conductivity changes through uniform AP using a modification matrix.

[0025] Optionally, a uniform AR can be obtained by constructing a sensitivity compensation matrix to perform sensitivity compensation on the AP. The formula for constructing the sensitivity compensation matrix is ​​as follows:

[0026]

[0027] Where c is the control parameter for sensitivity compensation, i is the unit number, and j is the SMC number of the excitation measurement channel. express, express, express, This represents the unit sensitivity vector.

[0028] The control parameter c for sensitivity compensation is used to control the loss in the internal region, which is the region that is farther from the edge than the diffusion radius, and the diffusion radius is the radius of the diffusion function of the modification matrix.

[0029] Optionally, the control parameters for sensitivity compensation are calculated by training the reconstruction magnitude of the internal region;

[0030] Optionally, the process of reconstructing the magnitude of the training internal region is as follows:

[0031] Obtain the sensitivity vector of each unit in the internal region;

[0032] Sort the unit sensitivity vectors and select n units from the sorted unit sensitivity vectors as training points, where n is a natural number greater than 1.

[0033] A perturbation is generated centered on each training point, and the measured voltage at each training point is calculated.

[0034] By setting the control parameters for initial sensitivity compensation, and calculating the reconstruction matrix based on the disturbance, measured voltage, and initial sensitivity compensation control parameters, the reconstructed conductivity change is obtained.

[0035] The maximum reconstruction amplitude of each disturbance is obtained based on the reconstructed conductivity change.

[0036] The control parameter c for sensitivity compensation is obtained by calculating the loss between the maximum and minimum values ​​of the maximum reconstruction amplitude;

[0037] Optionally, the loss is calculated by constructing a loss function between the maximum and minimum values ​​of the maximum reconstruction amplitude and the control parameter c of the sensitivity compensation, and then solving for the control parameter c of the sensitivity compensation.

[0038] Optionally, the loss function is expressed as:

[0039]

[0040] in, This indicates the maximum reconstruction magnitude.

[0041] The EIT voltage change data includes one or more of the following: two-dimensional EIT voltage change data, three-dimensional EIT voltage change data;

[0042] Optionally, the EIT measured voltage change data is three-dimensional EIT measured voltage change data, and the reconstruction result is obtained by calculating the three-dimensional EIT measured voltage change data and the reconstruction matrix R;

[0043] Optionally, by constructing a three-dimensional finite element model, EIT voltage signal acquisition is performed on the three-dimensional finite element model to obtain EIT measurement voltage change data;

[0044] Optionally, the measurement site for the EIT voltage change measurement includes one or more of the following: brain, lungs, and EIT voltage change data is obtained by performing EIT measurement on the measurement site.

[0045] The EIT voltage change data is obtained through an excitation measurement mode, which includes one or more of the following:

[0046] Adjacent excitation measurement mode, phase-to-phase excitation measurement mode, opposing excitation measurement mode, maximum excitation measurement mode;

[0047] Optionally, the maximum excitation measurement mode is a mode in which the maximum independent excitation measurement channel is selected from all excitation measurement channels of the EIT electrode and excitation measurement is performed.

[0048] Optionally, the maximum independence means that the number of times the EIT electrode is excited is not redundant.

[0049] The purpose of this invention is to provide a computer program product that includes a computer program or instructions, which are executed by a processor to implement the above-described EIT image reconstruction method.

[0050] The purpose of this invention is to provide a computer device comprising a memory, a processor, and a computer program or instructions stored in the memory, wherein the computer program or instructions are executed by the processor to implement the above-described EIT image reconstruction method.

[0051] The purpose of this invention is to provide a computer-readable storage medium having a computer program or instructions stored thereon, which are executed by a processor to implement the above-described EIT image reconstruction method.

[0052] Advantages of this invention:

[0053] 1. Regarding the excitation measurement mode, the expansion from 2D to 3D means a larger imaging area at the measurement site and an increase in areas far from the electrodes. This leads to a weakened sensitivity to changes in conductivity within the measurement area. In terms of image reconstruction algorithms, the increased imaging area and the non-uniform 3D electrode distribution further deteriorate the uniformity of the imaging amplitude response (AR). Therefore, this invention constructs a sensitivity matrix to address the non-uniform amplitude response, optimizes the reconstruction matrix in image reconstruction, and improves AR uniformity.

[0054] 2. In the process of image reconstruction, the modification matrix is ​​a key parameter for reconstruction. This invention proposes a paradigm expression for the expected image matrix D based on the Mexican hat function, which is more conducive to parameter control in image reconstruction and improves the sensitivity and accuracy of reconstruction.

[0055] 3. Changes in conductivity are more complex, and opposite changes in conductivity often indicate lesions of different natures, which has an important impact on clinical judgment. Therefore, EIT inhibition of RNG is necessary. This invention achieves RNG inhibition through a single-step feedback strategy. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a schematic diagram of the EIT image reconstruction method provided in an embodiment of the present invention;

[0058] Figure 2 A schematic diagram of the EIT image reconstruction system provided in an embodiment of the present invention;

[0059] Figure 3 A schematic diagram of an EIT image reconstruction device provided in an embodiment of the present invention;

[0060] Figure 4The two-dimensional and one-dimensional distribution diagrams of the D paradigm provided in the embodiments of the present invention show that the volume or area under the curve of the two types of functions (a) and (b) are equal. (a) Sigmoid function (the maximum value of the point diffusion center is 0.999); (b) Improved Mexican Hat function (fuzzy control parameter s=4, the larger s is, the closer the graph is to the step shape, positive compensation amplitude a=0.05);

[0061] Figure 5 The 16-9-7 three-ring head electrode model provided in this embodiment of the invention. Electrodes E1-E16 form a 16-electrode ring R1, electrodes E17-E25 form a 9-electrode ring R2, and electrodes E26-E32 form a 7-electrode ring R3. (a) and (b) are the left view and top view of the electrode model, respectively.

[0062] Figure 6 A four-layer numerical model of the head is provided for an embodiment of the present invention. (a) Model geometry: The head model comprises four layers: scalp, skull, cerebrospinal fluid, and brain tissue. Excitation and measurement are achieved through a double-layer structure of Ag / AgCl electrodes and conductive paste. (b) Model mesh: The mesh contains 284,546 tetrahedral elements and 52,916 nodes.

[0063] Figure 7 The influence of NF on the AR uniformity of GREIT and AUEIT provided in the embodiments of the present invention. (a) Based on the unit sensitivity vector, 9 training points are selected as the training set to determine the sensitivity compensation coefficient c of AUEIT. Two perturbations are generated in the internal and edge regions and tested. The right side is a slice of the 3D imaging result on the R1 electrode plane. The values ​​in the image are the maximum reconstruction amplitudes of the two perturbations. The volume of the spherical perturbation used for training or testing in the figure is 15 ml; (b) The ratio of the AR of the 9 training samples corresponding to NF to the minimum AR among them.

[0064] Figure 8 This invention provides a comparison of consensus indices and the distribution of brain RNG and AR in GREIT-S, GREIT-M, and AUEIT brain imaging under a uniform conductivity background. (a) PE, RES, SD, and RNG are the mean values ​​of all perturbation-related indices, and AR is the coefficient of variation of AR corresponding to all perturbations; (b) and (c) are RNG distribution plots for GREIT-S and GREIT-M. The darker the color of the scatter point and the larger its volume, the larger the RNG at the location of the scatter point. In (d), the horizontal and vertical coordinates of the scatter points are the RNG (denoted as RNG) of the perturbated GREIT-S and GREIT-M reconstructed images, respectively. S and RNG M The color mapping of the scatter plots is based on RNG. M / RNG SThe natural logarithm of RNG-M / RNG-S is such that the larger the RNG-M / RNG-S ratio, the redder the scatter points, and vice versa. AV is the average value of RNG. The plotting logic of (e)(f)(g) is consistent with that of (b)(c)(d), reflecting the AR distribution of GREIT-S and AUEIT. CV in (g) is the AR variation coefficient.

[0065] Figure 9 This invention provides an example of AR distribution and simulation experiment of GREIT-S and AUEIT under the background of skull conductivity. (a) and (b) are AR distribution plots of GREIT-S and AUEIT. The darker the color of the scatter point and the larger the volume, the greater the AR at the location of the scatter point. In (c), the horizontal and vertical coordinates of the scatter points are the AR of the perturbed GREIT-S and AUEIT reconstructed images (denoted as ARGR and ARAU), respectively. The color mapping of the scatter points is based on the natural logarithm of ARAU / ARGR. The larger the ARAU / ARGR, the redder the scatter point, and vice versa. CV is the coefficient of variation. (d) shows 9 brain perturbation samples set on the major and minor axes of the R1 electrode plane. The volume of the spherical perturbation is 15ml. The perturbation is achieved by setting the conductivity of brain and hemorrhagic tissue, respectively. (e) shows the conductivity of head tissue, with a corresponding current excitation frequency of 50kHz. (f) shows the 3D reconstructed images of the 9 samples of GREIT-S and AUEIT and the corresponding image slices of the R1 electrode plane. The values ​​in the image slices are the normalized maximum reconstruction amplitude.

[0066] Figure 10 The GREIT-S and AUEIT images provided for embodiments of the present invention are 3D reconstructed images of 9 perturbation samples and corresponding image slices of the R1 electrode plane. The values ​​in the image slices are the normalized maximum reconstruction amplitude.

[0067] Figure 11 Comparison of uniformity and sensitivity distribution of Dual-M SMP and Skip-n SMP provided in this embodiment of the invention. (a) Coefficient of variation of sensitivity distribution in the plane containing the 16 electrode ring R1 of the SMP, with the numerical model set as a uniform conductivity distribution. (b) Excitation-measurement distribution chord plot and sensitivity distribution plot of the SMP. The vertex set of the chord plot represents the electrodes, and the edge set represents the excitation or measurement pairs in the SMP. The width of the vertices and edges is related to the number of corresponding electrodes and excitation-measurement pairs. The values ​​in the sensitivity distribution plot are the minimum sensitivity after normalization of the maximum sensitivity. From left to right and from top to bottom, they represent Dual-MSMP and Skip-0 to Skip-7 SMP, respectively. Detailed Implementation

[0068] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0069] In some of the processes described in the specification, claims, and accompanying drawings of this invention, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as S101, S102, etc., are merely used to distinguish different operations and do not represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0070] Figure 1 The schematic diagram of the EIT image reconstruction method provided in this embodiment of the invention specifically includes:

[0071] S1: Acquire EIT measurement voltage change data;

[0072] In one embodiment, the EIT voltage change data is obtained through an excitation measurement mode, which includes one or more of the following:

[0073] Adjacent excitation measurement mode, phase-to-phase excitation measurement mode, opposing excitation measurement mode, and maximum excitation measurement mode.

[0074] In one embodiment, the maximum excitation measurement mode is a mode in which the maximum independent excitation measurement channel is selected from all excitation measurement channels of the EIT electrode and excitation measurement is performed.

[0075] Optionally, the maximum independence means that the number of times the EIT electrode is excited is not redundant.

[0076] In one embodiment, the process of the maximum excitation measurement mode is as follows: N EIT electrodes, where N is a natural number greater than 1; construct a full-rank electrode matrix; generate an excitation measurement vector matrix based on the full-rank electrode matrix; calculate the maximum excitation measurement channel set based on the excitation measurement matrix; and obtain the maximum excitation measurement mode of the EIT electrodes based on the maximum excitation measurement channel set.

[0077] Optionally, the calculation involves eliminating elements from the excitation measurement matrix to obtain the maximum set of excitation measurement channels; optionally, the elimination includes one or more of the following:

[0078] In one embodiment, the construction process of the full-rank electrode matrix is ​​as follows: Gaussian elimination, LU decomposition, and column pivoting elimination;

[0079] Optionally, the elimination is Gaussian elimination.

[0080] The first column of the full-rank matrix is ​​generated using a random method;

[0081] The first excitation measurement vector matrix is ​​obtained by exciting and measuring the EIT electrode based on the first column full-rank matrix and through the first excitation measurement mode;

[0082] Calculate the rank of the first excitation vector matrix. When the rank of the first excitation vector matrix is ​​equal to the maximum number of independent excitation measurements of the first excitation measurement, the current first excitation vector matrix is ​​a column full-rank electrode matrix.

[0083] Optionally, the first excitation measurement mode includes any one of the following: adjacent excitation measurement, relative excitation measurement, and interphase excitation measurement.

[0084] In one embodiment, the excitation measurement vector is generated by calculating the excitation vector and measurement vector using a full-rank column electrode matrix, and the excitation vector and measurement vector are multiplied together to obtain the excitation measurement vector.

[0085] Optionally, an excitation measurement vector matrix is ​​obtained by generating excitation measurement vectors for N EIT electrodes based on the full-rank column electrode matrix;

[0086] Optionally, the maximum excitation measurement further includes calculating the number of independent excitation measurement channels, which is obtained by calculating the rank of the excitation measurement matrix.

[0087] In one embodiment, the maximum excitation measurement further includes maximum sensitivity calculation, which maximizes the measurement sensitivity of the change in conductivity at the measurement site when performing excitation measurement on the maximum excitation measurement mode, thereby obtaining the maximum sensitive excitation measurement mode.

[0088] Optionally, the maximum sensitive excitation measurement mode is obtained by calculating and sorting the sensitivity of the excitation measurement channels in the EIT electrode, and adding the excitation measurement channels with high sensitivity to the maximum excitation measurement mode.

[0089] Optionally, the calculation process for the maximum sensitive excitation measurement mode is as follows:

[0090] S101. Obtain all excitation measurement channels for N EIT electrodes;

[0091] S102. Calculate the sensitivity of each excitation measurement channel and sort the sensitivities.

[0092] S103. Based on sensitivity, the excitation measurement channels are sequentially added to the maximum excitation measurement mode to obtain the second excitation measurement mode, starting from the excitation measurement channel corresponding to the first sensitivity matrix.

[0093] S104. Calculate the excitation measurement vector matrix of the second excitation measurement mode;

[0094] S105. Based on the excitation measurement vector matrix, determine the independence of the added excitation measurement channels. If the determination result is independent, retain them; otherwise, remove the added excitation measurement channels.

[0095] Repeat steps S103-S105 to obtain the maximum sensitive excitation measurement mode.

[0096] The calculation of the measurement sensitivity to maximize the change in conductivity at the measurement site is obtained by calculating the change in measurement voltage of the excitation measurement channel, the change in conductivity at the measurement site, and the sensitivity of the excitation measurement channel.

[0097] Optionally, the calculation of maximizing the measurement sensitivity of the change in conductivity at the measurement site further includes correction. The measurement voltage change of the first excitation measurement mode is calculated to obtain the first measurement voltage change. The measurement voltage change of any excitation measurement channel is obtained by superimposing and taking the absolute value of the corresponding index of the first measurement change. The measurement sensitivity of maximizing the change in conductivity at the measurement site is calculated based on the measurement voltage change corrected by the excitation measurement channel, the change in conductivity at the measurement site, and the sensitivity of the excitation measurement channel.

[0098] In one specific embodiment, EIT widely employs the four-electrode method to measure voltage. For an N-electrode EIT system, a single excitation measurement can be described as follows:

[0099]

[0100] in, These are the electrode numbers for positive and negative excitation and positive and negative measurement, respectively. I is the excitation current, and v is the measurement voltage. This refers to the potential of the measuring electrode; the excitation electrode is not used as the measuring electrode. A single excitation measurement represents a stimulation and measurement channel (SMC), which is accessed through... A stimulation and measurement pattern (SMP) is a set of SMCs that conform to a specific rule. Skip-n mode is a commonly used SMP, where excitation occurs at n-intervals, and measurements are taken at adjacent electrodes. Due to the superposition theorem and reciprocity theorem, Skip-n mode often contains a large number of redundant SMCs. Taking Skip-0 as an example, this mode has N excitations, each with N-3 measurements, resulting in N(N-3) SMCs. Due to the superposition theorem, one of the N excitations is redundant. Due to the reciprocity theorem, except for the SMCs corresponding to the first and second excitations which are all independent, each subsequent excitation generates one additional redundant SMC. Therefore, the number M of independent SMCs is...

[0101]

[0102] Furthermore, other SMCs can be generated through SMC superposition and reciprocity in Skip-0 mode. Therefore, for an N-electrode EIT system, the maximum number of independent SMCs is... If an SMP has one and only If there are 1 independent SMC, then the SMP is the Max-I SMP.

[0103] To obtain Max-I SMP, this invention proposes a method for screening independent SMCs. One SMC, It consists of four independent electrode components, and N electrodes correspond to N independent components, thus constructing an electrode matrix with full column rank. A vector representing the SMC is generated through E. ,

[0104]

[0105]

[0106] in, This represents the independent component corresponding to electrode i. This represents the positional multiplication of vectors, also known as the Hadamard product. Equations (3) and (4) represent the excitation and measurement processes, satisfying the superposition theorem, and equation (5) represents the excitation and measurement processes, satisfying the reciprocity theorem. A suitable E ensures that sm maintains independence consistent with SMC, therefore the matrix... It can reflect the independent redundancy relationship of SMC in SMP. Among them, E is the key to ensuring that sm and SMC have consistent independence.

[0107] For any SMP, calculating the rank of the corresponding SM yields the number of independent SMCs in the SMP. Gaussian elimination of the SM yields the maximal set of independent SMCs, thus obtaining the Max-I SMP. It should be noted that the set of independent SMCs in an SMP corresponds to a maximal linearly independent set in the SM matrix; therefore, the maximal set of independent SMCs is not unique.

[0108] In one specific embodiment, EIT observes changes in the boundary voltage. It is estimated to be due to changes in the internal conductivity distribution. , and These are the number of channels in the SMP model and the number of elements in the finite element model (FEM), respectively. The differential EIT forward model is...

[0109]

[0110] in, is the sensitivity matrix (Jacobi matrix), and n is the measurement noise.

[0111] Brain electrophysiological testing (EIT) focuses on changes in brain conductivity. However, the skull significantly impedes electrical current, resulting in only a small portion of the current flowing through brain tissue during excitation measurements, greatly reducing the sensitivity to changes in brain conductivity. This invention defines the sensitivity S of the brain conductivity measurement function (SMC) to changes in brain conductivity.

[0112]

[0113] This is the sensitivity row vector corresponding to SMC. A change of one unit in electrical conductivity within the brain represents a change of zero in electrical conductivity outside the brain. c This refers to the measured voltage change of the SMC. To maximize S, this invention proposes a Max-I SMP (Dual-M SMP) for maximizing S, calculated as follows:

[0114] For an N-electrode EIT system, the total number of possible SMCs is: ,

[0115]

[0116] in and Let represent the order number. The superposition theorem makes the positive and negative excitations or positive and negative measurement electrodes equivalent after interchange. The reciprocity theorem makes the excitation pair and measurement pair equivalent after interchange. Therefore... Down to,

[0117]

[0118] in and Represents the number of combinations. Generates a full-excitation measurement mode SMP. total ,Include For each SMC, calculate the S of all SMCs and sort them in descending order based on S. Starting with the SMC with the largest S, add the SMCs sequentially to the preset SMP. Use the corresponding SM matrix to determine the independence of the newly added SMCs. If they are independent, keep them; if they are redundant, remove them. Continue until all independent SMCs are added to obtain the Dual-M SMP.

[0119] The calculation of S for all SMCs and the numerous SMC independence checks may raise concerns for operators regarding computational costs. To reduce computational costs, the present invention provides the following optimizations:

[0120] For the calculation of S, based on the adjacent excitation and adjacent measurement mode (the excitation electrode also participates in the measurement), SMP is used. ads This indicates the change in voltage y when the brain's electrical conductivity changes by one unit. ads . y of any SMC c All can be passed through SMP ads of y ads The corresponding index is generated by superimposing the indices, and the absolute value is the corresponding S.

[0121] To address the issue of a large number of SMC independence checks, multiple SMCs are added at once with a large step size. The rank of the SM matrix is ​​then calculated to determine if all independent SMCs have been included. If not, more SMCs are added until all independent SMCs are included. Finally, Gaussian elimination is used to filter the independent SMCs. The Gaussian elimination method employs column-major order scanning and a partial pivot selection strategy to ensure that this method is identical to the Dual-M SMP generated when using a single SMC for independence checks.

[0122] In one specific embodiment, the generation of a suitable E is as follows: First, a full-rank matrix E is generated randomly. The randomness of E avoids a fixed proportional relationship between zero elements or certain elements, which can greatly ensure that no new LC is introduced. Furthermore, this invention utilizes known maximum independent SMPs or matrices containing M. max The SMP of an independent SMC avoids these special cases. The method is as follows: using adjacent excitation and adjacent measurement modes, the corresponding SM is generated based on the randomly generated E and formulas (3)(4)(5). ad Matrix, when SM ad rank r ad Equal to M max When we determine that E is the appropriate value at this time, we can use E. a express.

[0123] via r ad Is it equal to M? max The judgment, Ea This ensures that the Hadamard product does not introduce new LC, as proven below: For any Max-I SMP, denoted by SMP1, SMP1 has M max Each independent SMC, through the superposition theorem and reciprocity theorem, SMP1 can represent all SMCs of adjacent excitation and adjacent measurement modes. Correspondingly, SM1 and SM ad via E a With the generation of formulas (3), (4), and (5), the superposition theorem and the reciprocity theorem are automatically satisfied. Therefore, SM1 can also express SM. ad , Right now And because of SM ad rank r ad Equal to M max Therefore, r = M max That is, SM1 has no additional LC.

[0124] S2: Construct a reconstruction matrix R, and calculate the reconstruction result from the measured voltage change data using the reconstruction matrix R;

[0125] Wherein, the reconstruction matrix R is the error calculation that minimizes the change in desired conductivity and the change in reconstructed conductivity;

[0126] The desired conductivity change is obtained by first acquiring conductivity change data and constructing a modification matrix D, and then modifying the conductivity change data using the modification matrix D. The modification matrix D modifies the conductivity change using consensus indicators of the EIT image. The consensus indicators include one or more of the following: AR, PE, RNG, RES, and SD.

[0127] In one embodiment, the modification matrix D is the desired modification of the EIT measurement unit by a point spread function, where the EIT measurement unit is a tissue unit at different locations in the measurement site.

[0128] In one embodiment, the point diffusion function divides the organizational units at different locations into regions, including a first region and a second region. The first region is the region where the organizational units at different locations are less than or equal to the diffusion radius from the diffusion center, and the second region is the region where the organizational units at different locations are greater than the diffusion radius from the diffusion center. The first region and the second region are then modified as desired.

[0129] In one embodiment, the formula for constructing the modification matrix D is:

[0130]

[0131]

[0132] in, It is a point spread function, p0 is The diffusion center, p represents the center position of the FEM element in the finite element model, and r is diffusion radius, The matrix represents the modification matrix, where i and j represent the electrodes of the EIT, s is a fuzzy control parameter, and b represents the control parameter for the positive compensation amplitude.

[0133] In one embodiment, the modification matrix is ​​modified by RNG through... Positive compensation of RNG yields the results of changes in conductivity after RNG modification;

[0134] Optionally, let the compensation range of RNG be a, and the formula for calculating the compensation range is:

[0135]

[0136] Where b represents the control parameter for the positive compensation amplitude.

[0137] In one embodiment, the RNG modification further includes RNG suppression, which is performed on the RNG-modified conductivity change result by a single-step feedback strategy to obtain the suppressed RNG-modified conductivity change result.

[0138] In one embodiment, the single-step feedback strategy process is as follows:

[0139] S1, let the modification matrix Calculations are performed when the compensation range limit approaches zero. Obtain the first modification matrix, and construct the first reconstruction matrix based on the first modification matrix;

[0140] S2. Calculate a of the first reconstruction matrix. t a t It is a normalized amplitude with reference to the maximum positive AR;

[0141] S3, let the modification matrix The second modification matrix is ​​calculated, and the result of suppressing the change in RNG modified conductivity is obtained based on the second modification matrix.

[0142] In one embodiment, the average RNG of the first reconstruction matrix is ​​calculated in S2 by a t The quantitative metric for quantified average RNG is a. t .

[0143] In one embodiment, a reconstruction matrix is ​​obtained by constructing a reconstruction matrix based on a second modification matrix.

[0144] In one embodiment, the AR in the consensus index further includes sensitivity compensation, through which uniform AR is obtained, and conductivity variation is modified by uniform AP through a modification matrix.

[0145] In one embodiment, a uniform AR is obtained by constructing a sensitivity compensation matrix to perform sensitivity compensation on the AP. The formula for constructing the sensitivity compensation matrix is ​​as follows:

[0146]

[0147] Where c is the control parameter for sensitivity compensation, i is the unit number, and j is the SMC number of the excitation measurement channel. express, express, express, This represents the unit sensitivity vector.

[0148] In one embodiment, the control parameter c for sensitivity compensation is used to control the loss in the internal region, which is a region whose distance from the edge is greater than the diffusion radius, and the diffusion radius is the radius of the diffusion function of the modification matrix.

[0149] In one embodiment, the control parameters for sensitivity compensation are calculated by training the reconstruction magnitude of the internal region.

[0150] In one embodiment, the process of reconstructing the magnitude of the internal training region is as follows:

[0151] Obtain the sensitivity vector of each unit in the internal region;

[0152] Sort the unit sensitivity vectors and select n units from the sorted unit sensitivity vectors as training points, where n is a natural number greater than 1.

[0153] A perturbation is generated centered on each training point, and the measured voltage at each training point is calculated.

[0154] By setting the control parameters for initial sensitivity compensation, and calculating the reconstruction matrix based on the disturbance, measured voltage, and initial sensitivity compensation control parameters, the reconstructed conductivity change is obtained.

[0155] The maximum reconstruction amplitude of each disturbance is obtained based on the reconstructed conductivity change.

[0156] The control parameter c for sensitivity compensation is obtained by calculating the loss between the maximum and minimum values ​​of the maximum reconstruction amplitude.

[0157] In one embodiment, the loss is calculated by constructing a loss function between the maximum and minimum values ​​of the maximum reconstruction amplitude and the control parameter c of sensitivity compensation, and then solving for the control parameter c of sensitivity compensation.

[0158] In one embodiment, the loss function is expressed as:

[0159]

[0160] in, This indicates the maximum reconstruction range.

[0161] In one embodiment, the EIT voltage change data includes one or more of the following: two-dimensional EIT voltage change data and three-dimensional EIT voltage change data.

[0162] In one embodiment, the EIT measured voltage change data is three-dimensional EIT measured voltage change data, and the reconstruction result is obtained by calculating the three-dimensional EIT measured voltage change data and the reconstruction matrix R.

[0163] In one embodiment, EIT voltage signal acquisition is performed on the three-dimensional finite element model to obtain EIT measurement voltage change data.

[0164] In one embodiment, the measurement site for the EIT voltage change includes one or more of the following: brain, lungs, and EIT voltage change data is obtained by performing EIT measurement on the measurement site.

[0165] In one embodiment, the EIT voltage change data is obtained through an excitation measurement mode, which includes one or more of the following:

[0166] Adjacent excitation measurement mode, phase-to-phase excitation measurement mode, opposing excitation measurement mode, and maximum excitation measurement mode.

[0167] In one embodiment, the maximum excitation measurement mode is a mode in which the maximum independent excitation measurement channel is selected from all excitation measurement channels of the EIT electrode and excitation measurement is performed.

[0168] In one embodiment, the maximum independence means that the number of times the EIT electrode is excited is not redundant.

[0169] In one specific embodiment, the GERIT algorithm is consistent with the idea of ​​Wiener filtering; it seeks an optimal filter or reconstruction matrix R. The change in conductivity recovered by the measured voltage change y , The change in conductivity is related to the expected change. The mean squared error is the smallest. This is the result of modifying the change in conductivity distribution x based on expert consensus on EIT imaging, using a matrix. Describe the modification process. Assume x follows a distribution. The noise n of SMC follows a distribution The calculation of R is as follows:

[0170]

[0171] R satisfies, ,Right now

[0172]

[0173] Because x and n are independent , , The expression for R is,

[0174]

[0175] In this invention, for the sake of universality, it is assumed that the elements of x and n are independent of each other, the conductivity of x is set to be uniform, and the unit volume is corrected to make its influence on R the same, i.e. , Let I be the identity matrix of the corresponding dimension. Formula (13) can be written as:

[0176]

[0177] Regularization parameters A larger R value means that more noise is taken into account, resulting in stronger suppression of n during imaging. However, the fidelity of other image parameters will decrease. These are key parameters for reconstruction.

[0178] Noise Figure (NF) is used as a tool The choice of NF is now widely accepted. NF is the signal-to-noise ratio of the filter R output. Compared with input signal-to-noise ratio The ratio,

[0179] The input signal is y, which is caused by a one-unit change in brain conductivity. , The measurement noise follows a normal distribution with a standard deviation of 0.5. As input noise , and The reconstruction results are respectively used as output signals and output noise .

[0180] It can be noted that in formula (14) This refers to the average measurement sensitivity S of SMP.SMP ,therefore, ( This can be eliminated during the calculation of NF, therefore let ),get,

[0181]

[0182] The larger the value of R, the stronger the suppression of n. Therefore, under stable noise levels, about Monotonically increasing. Given a suitable NF, an approximate solution to equation (15) can be obtained using the bisection method. At this time This is expected or acceptable. For Skip-0 SMP, NF = 0.5, at which point... It is recommended. When this invention applies Dual-M SMP, via S... SMP Correcting NF ensures that the algorithm comparison is performed here. The following will proceed.

[0183] In one specific embodiment, the paradigmatic expression of D and the single-step feedback RNG suppression method: R in the classical regularized reconstruction algorithm is,

[0184]

[0185] in, It is a diagonal matrix, and the diagonal elements are the weights of the corresponding SMCs. It is a regularization constraint on x. It is the regularization parameter. When , , When the equations are equal, equations (13) and (16) are equal (the difference between the two equations equals 0). It can be seen that the improvement of GREIT is mainly achieved through matrix D, so D is another key parameter for reconstruction.

[0186] D modifies x using consensus metrics from EIT images. These metrics, ranked by importance, are: 1) uniform AR, 2) small and uniform position error (PE), 3) small RNG, 4) small and uniform resolution (RES), and 5) small shape variability (SD). Bartłomiej Grychtol constructed a paradigmatic expression for D based on the Sigmoid function, which to some extent achieves the modification of consensus metrics. For further optimization, this invention constructs a new paradigmatic expression for D based on an improved Mexican Hat function.

[0187]

[0188] in, It is a point spread function, p0 is The diffusion center, p represents the center position of the FEM element. r is... diffusion radius, , hour, It reflects the characteristics of the desired image. s is a fuzz control parameter; the larger s is, the clearer the EIT image, but RNG will increase. hour, For positive compensation of RNG, the compensation magnitude is set as a. In formula (17), b is the control parameter for positive compensation. The relationship between a and b is:

[0189]

[0190] The Sigmoid function and the improved Mexican Hat function used to generate D, such as Figure 4 As shown, compared to the paradigmatic expression of D constructed based on the Sigmoid function, the Mexican Hat function has definite vertices and zeros, which is more conducive to parameter control.

[0191] Changes in brain electrical conductivity are more complex, and opposite changes in conductivity often indicate lesions of different natures, which has a significant impact on clinical judgment. Therefore, inhibiting RNG in brain EIT is necessary. To inhibit RNG, this invention proposes a single-step feedback strategy: Let Obtain the temporary reconstruction matrix R t Calculate R t The resulting average RNG, -a t (a t This is the normalized amplitude referenced to the maximum positive AR. Then, let... To obtain a new reconstruction matrix R, R is obtained while maintaining R t While improving clarity, it will also enhance RNG's suppression capabilities.

[0192] In one specific embodiment, AR uniformity is improved through sensitivity compensation:

[0193] Uniform AR (Area of ​​Resonance) is the most important consensus indicator for EIT (Enhanced Imaging Technology). This invention analyzes the causes of EIT AR non-uniformity from the perspective of regularization parameter selection. Definition , , express eigenvalues, calculate the reconstruction matrix ,

[0194]

[0195] It is the right inverse of J, assuming the noise is 0, that is ,

[0196]

[0197] This invention yielded the desired change in conductivity. However, the maximum number of independent numbers of J is only Much smaller than n elem Equations (20) and (21) are not well-posed. To make them well-posed, Greater than It is necessary. This invention defines... , ,

[0198]

[0199] Formula (23) shows that, The larger, by That is, the greater the sensitivity, the worse the AR uniformity.

[0200] EIT sensitivity distribution is obtained through the unit sensitivity vector s. describe,

[0201]

[0202] Where i is the cell number, j is the SMC number, and s represents the sum of squares of the measured voltage changes when the conductivity of the cell changes by one unit. Since the measured voltage originates from the boundary, the sensitivity distribution is characterized by strong sensitivity in the boundary region and weak sensitivity in the interior region. To improve uniformity, this invention introduces sensitivity compensation, constructing a sensitivity compensation matrix S. ,

[0203]

[0204] Where c is a control parameter for sensitivity compensation, generally... , A uniform consensus image reconstruction algorithm (AUEIT) based on S is constructed.

[0205]

[0206] The selection of c is an optimization process: considering the loss caused by the point spread function in the edge region, this invention selects n corresponding training points at uniform intervals based on the unit sensitivity vector s, after sorting them by sensitivity, and generates a perturbation centered on each training point in the inner region, i.e., the region where the distance from the edge is greater than the spread radius r. And calculate the corresponding measured voltage. Given an initial value c0, calculate R and This allows us to obtain the maximum reconstruction amplitude for each training perturbation. Construct the loss function.

[0207]

[0208] The loss(c) approximates to be a concave function with respect to c. A stable solution can be quickly obtained by applying common optimization algorithms. During the optimization process, the normal function (NF) remains unchanged.

[0209] In one specific embodiment, the 16-9-7 triple-ring head electrode model (TREM) is designed based on the classic EIT 16-electrode lead ring and EEG 10-20-electrode lead ring, surrounding the entire brain tissue for imaging changes in brain conductivity. It retains the 16-electrode interface to achieve dual-mode 2D and 3D imaging. Electrode positions are indicated by En, where n represents the electrode number; there are 32 in total, as detailed below:

[0210] 1) Draw a main line on the scalp surface, L, connecting the root of the nose to the occipital protuberance. NI The midpoint of the line is E25.

[0211] 2) Let L NI The length is 100%, along L NI The position 10% backward from the root of the nose is E1. An electrode position is set every 20% backward from E1, namely E1, E17, E25, E21 and E9. Continue backward for another 10% to the external occipital protuberance, which is E29.

[0212] 3) The L1 curve is the line connecting E1 and E9 and perpendicular to the midsagittal plane (L... NI The intersection of the plane (where it is located) and the scalp surface, E1-E16 are equidistantly distributed on L1 in a clockwise direction (view from the foot side to the head side), and form the first 16-electrode ring R1.

[0213] 4) Curve L2 is the intersection of the plane defined by E3, E17, and E15 with the scalp surface. The length of L2 is set to 100%. Electrode positions are set every 25% of the distance from E3 to the right, namely E3, E18, E17, E24, and E15. Similarly, curve L3 is the intersection of the plane defined by E5, E25, and E13 with the scalp surface, and L3 sequentially defines E5, E19, E25, E23, and E13; curve L4 is the intersection of the plane defined by E7, E21, and E11 with the scalp surface, and L4 sequentially defines E7, E20, E21, E22, and E11. Among them, E17-E25 form the second 9-electrode ring R2.

[0214] 5) E26-E32 were identified through clear anatomical locations. E26 and E32 are located at the left and right temples, respectively; E27 and E31 are located at the left and right mastoid processes of the temporal bones, respectively. The L5 curve is the intersection of the plane defined by E27, E29, and E31 with the scalp surface. The length of L5 is set to 100%. Electrode positions are placed every 25% of the distance from E27 to the right, in the following order: E27, E28, E29, E30, and E31. E26-E32 form the third 7-electrode ring R3.

[0215] The above constructs a 16-9-7 three-ring head electrode model, as follows: Figure 5 As shown.

[0216] In one specific embodiment, due to the complex conductivity distribution in the head, the head is abstracted into a four-layer structure—scalp layer, skull layer, cerebrospinal fluid layer, and brain tissue layer—to construct a numerical model. The geometric information of the four-layer structure comes from the EIDORS EIT toolkit. Electrodes with a double-layer structure (Ag / AgCl electrode and conductive paste) and a diameter of 1 cm are configured based on the 16-9-7 three-ring electrode model. The numerical model of the four-layer head structure constructed in this invention is as follows: Figure 6 As shown.

[0217] The skull's resistance to electric current is significant. To construct a physical model that more closely resembles the actual conductivity distribution of the human head, plaster casts with similar conductivity were used to simulate the skull. The water tank was divided into three layers by the skull: the outer layer representing the scalp, the inner layer representing the brain tissue, and the conductivity of the scalp and brain tissue layers simulated using an electrolyte solution. Electrodes were configured based on a 16-9-7 three-ring electrode model, made of copper, with 1cm diameter circles as electrode contact surfaces.

[0218] Based on the above numerical and physical models, the Dual-M SMP and AUEIT image reconstruction algorithms were evaluated.

[0219] In one specific embodiment, the evaluation of the image reconstruction method AUEIT of the present invention focuses on the impact of NF on AR uniformity. NF determines the key parameters of EIT. , This significantly affects AR uniformity. For Skip-0 SMP, NF = 0.5 is recommended. By comparing the sensitivity of Skip-0 SMP and Dual-M SMP, the measurement sensitivity of Dual-M SMP is approximately 50 times that of Skip-0 SMP. According to formula (15), the NF of 2D Dual-M SMP is corrected to 0.01. Experimental experience of this invention shows that 3D and 2D Dual-M SMP are consistent in the choice of NF, that is, when NF = 0.01, the SNR of 3D brain EIT images is... outThis is acceptable. A series of normalized normalized (NF) values ​​were set around 0.01, namely 0.001, 0.0025, 0.005, 0.01, 0.025, and 0.05, to investigate the effect of NF on AR uniformity. Figure 7 As shown, AR uniformity deteriorates with increasing NF, but compared to GREIT, AUEIT exhibits smaller AR differences within and at the edges, maintaining relatively stable uniformity.

[0220] In one specific embodiment, the evaluation of AUEIT reconstructed images involves uniformly applying 500-600 perturbations to the brain. The quality of GREIT images reconstructed using EIT (EI) is evaluated using consensus metrics (AR, PE, RES, SD, RNG) for D-based reconstructions (GREIT-S), GREIT images reconstructed using D-based reconstructions (GREIT-M), and AUEIT reconstructed 3D images. Here, NF=0.01, and to ensure unbiased evaluation, the evaluation is first conducted against a background of uniform conductivity. Figure 8 As shown in (a), the positional error, resolution, and shape variation performance of GREIT-S, GREIT-M, and AUEIT are similar. Figure 8 (bd) shows the specific differences in RNG distribution between GREIT-M and GREIT-S. Compared to GREIT-S, the RNG of GREIT-M decreased from 0.531 to 0.335, a reduction of 37%. AUEIT will enhance the RNG to some extent, but due to the single-step feedback strategy, the RNG is still somewhat improved compared to GREIT-S. Figure 8 The figure (e.g.) shows the specific differences between AUEIT and GREIT-S in the AR distribution. Compared with GREIT-S, the coefficient of variation of AR for AUEIT decreased from 0.103 to 0.0685, a reduction of 33%.

[0221] In one specific embodiment, a simulation experiment was conducted to evaluate the uniformity of AR in the brain using AUEIT against a background of cranial conductivity: Using GREIT-S as a reference, a numerical model was used to comprehensively evaluate the uniformity of AR in the brain under a background of cranial conductivity. The perturbation settings were consistent with those in the previous embodiment. Figure 9 The (ac) data showed that, compared to GREIT-S, the coefficient of variation of AR in AUEIT decreased from 0.1822 to 0.1183, a reduction of 35%, demonstrating that AUEIT's AR uniformity was still superior to GREIT-S in the context of intracranial electrical conductivity. Figure 9 The (df) provides a more intuitive simulation experiment example.

[0222] In one specific embodiment, a physical experimental evaluation of AUEIT's AR uniformity in the context of cranial conductivity was conducted: using GREIT-S as a reference, the improvement of AUEIT's AR uniformity in the context of cranial conductivity was further evaluated through a physical model. Nine metallic perturbations were uniformly placed along the major and minor axes of the R1 electrode plane for the experiment. The experimental results are as follows... Figure 10 As shown, the coefficients of variation (COP) for the nine samples from GREIT-S and AUEIT were 0.104 and 0.0692, respectively. Compared to GREIT-S, AUEIT reduced the COP by 33%, which physically validates the improvement in COP uniformity achieved by AUEIT in the context of skull conductivity.

[0223] In one specific embodiment, the impact of Dual-M SMP on the uniformity of sensitivity: EIT exhibits a characteristic of strong sensitivity in boundary regions and weak sensitivity in interior regions. Dual-M SMP enhances brain measurement sensitivity, meaning more current flows through the brain, which helps reduce the sensitivity difference between interior and exterior regions, resulting in a more uniform sensitivity distribution and, consequently, improving the AR uniformity of EIT. This invention evaluates the impact of SMP on the uniformity of sensitivity using the coefficient of variation (CV) of unit sensitivity s. Figure 11 As shown in (a), the Dual-M SMP exhibits a more uniform sensitivity distribution compared to Skip-n SMP. Figure 11 As shown in (b), the sensitivity of Dual-M SMP is more concentrated, that is, the difference in sensitivity between the boundary region and the interior region is smaller.

[0224] In one specific embodiment, based on the aforementioned four-layer head finite element model, the 16-electrode SMP... total The measurement sensitivity calculation time is approximately 3.3 seconds for a 32-electrode SMP. total The measurement sensitivity calculation time is approximately 4.5 seconds. When SM is used for SMC independence determination, the time cost to obtain the Dual-M SMP corresponding to 16 electrodes is 0.1 seconds, and the time cost to obtain the Dual-M SMP corresponding to 32 electrodes is 22 seconds.

[0225] In one specific embodiment, the impact of SMP on solving the EIT inverse problem is compared by analyzing the singular values, condition number, and number of independent SMCs for J in Dual-M SMP and Skip-n SMP. Dual-M SMP has larger singular values, implying stronger numerical stability and robustness in the main solution direction. Dual-M SMP has the smallest condition number and the largest number of independent SMCs, indicating weaker ill-conditioning and ill-posedness. Therefore, Dual-M SMP is more advantageous for solving the EIT inverse problem.

[0226] In one specific embodiment, AR uniformity is the most important indicator of EIT images; however, AR non-uniformity is an inherent problem in EIT. AR non-uniformity is mainly manifested in the fact that the larger the regularization parameter, the greater the non-uniformity, primarily reflecting differences in unit sensitivity. To improve AR uniformity, this invention proposes a sensitivity compensation matrix, where stable compensation parameters can be confirmed with only a small amount of data. Experimental results show that, through sensitivity compensation, AUEIT reduces the coefficient of variation of brain AR by more than 30% compared to GREIT. AUEIT, based on the sensitivity matrix compensation, achieves adaptation in any imaging domain, exhibits better generalization, and provides a more stable improvement in AR uniformity.

[0227] In one specific embodiment, EIT image reconstruction, by penalizing the sum of squares of model parameters through L2 regularization, tends to generate smooth solutions, exhibiting characteristics of excessive suppression of high-frequency components. The lack of high-frequency components leads to the Gibbs phenomenon (RNG). To suppress RNG, this invention defines a new paradigm expression for D by improving the Mexican hat function and proposes a feedback strategy: based on the known amplitude of the RNG, a positive circular oscillation is described in D, achieving neutralization of the inverse RNG as the reconstructed image approximates D. This invention sets the fuzzy control parameter s=4 for the new paradigm and applies a single-step feedback strategy, resulting in a 33% reduction in brain RNG.

[0228] While AUEIT optimizes AR uniformity, its sensitivity compensation has limitations (the compensation matrix is ​​a diagonal matrix to account for the relationships between units). Although it improves the uniformity of the maximum AR, it also causes a certain degree of loss in the uniformity of the overall AR sum. Therefore, AUEIT is more suitable for small object detection scenarios. On the other hand, AUEIT may also enhance brain RNG while performing sensitivity compensation. This is because the essence of sensitivity compensation is to suppress AR in highly sensitive areas and enhance RNG in less sensitive areas. RNG in internal brain regions may also be enhanced. However, the introduction of the single-step feedback mechanism balances this out. Overall, AUEIT still improves RNG to some extent.

[0229] The present invention also discloses a computer program product or system, including a computer program that, when executed by a processor, implements the above-described EIT image reconstruction method steps.

[0230] Figure 2 The schematic diagram of the EIT image reconstruction system provided in this embodiment of the invention specifically includes:

[0231] Acquisition Unit: Acquires EIT measurement voltage change data;

[0232] Reconstruction Unit: Constructs a reconstruction matrix R, and the measured voltage change data is used to calculate the reconstruction result through the reconstruction matrix R;

[0233] Wherein, the reconstruction matrix R is the error calculation that minimizes the change in desired conductivity and the change in reconstructed conductivity;

[0234] The desired conductivity change is obtained by first acquiring conductivity change data and constructing a modification matrix D, and then modifying the conductivity change data using the modification matrix D. The modification matrix D modifies the conductivity change using consensus indicators of the EIT image. The consensus indicators include one or more of the following: AR, PE, RNG, RES, and SD.

[0235] Figure 3 An embodiment of the present invention provides a schematic diagram of a computer device, specifically including:

[0236] A memory and a processor; the memory is used to store program instructions; the processor is used to invoke the program instructions, when the program instructions are executed any of the above-described EIT image reconstruction methods.

[0237] The present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, represents any of the above-described EIT image reconstruction methods.

[0238] The verification results of this verification embodiment show that assigning inherent weights to indications can improve the performance of this method compared to the default settings. Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for example, the division of units is merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, indirect coupling or communication connection of devices or units, and may be electrical, mechanical, or other forms. The units described as separate components may or may not be physically separated; the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of this embodiment. Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated units described above can be implemented in hardware or as software functional units. Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. This program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0239] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0240] The computer device provided by the present invention has been described in detail above. For those skilled in the art, there will be changes in the specific implementation and application scope based on the ideas of the embodiments of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. An EIT image reconstruction method, characterized in that, include: Acquire EIT measurement voltage change data; A reconstruction matrix R is constructed, and the reconstruction result is obtained by calculating the measured voltage change data through the reconstruction matrix R. Wherein, the reconstruction matrix R is the error calculation that minimizes the change in desired conductivity and the change in reconstructed conductivity; The desired conductivity change is obtained by first acquiring conductivity change data and constructing a modification matrix D, and then modifying the conductivity change data using the modification matrix D. The modification matrix D modifies the conductivity change using consensus indicators of the EIT image. The consensus indicators include one or more of the following: AR, PE, RNG, RES, and SD.

2. The EIT image reconstruction method according to claim 1, characterized in that, The modification matrix D is the desired modification of the EIT measurement unit through the point spread function. The EIT measurement unit is a tissue unit at different locations in the measurement site. Optionally, the point diffusion function divides the organizational units at different locations into regions, including a first region and a second region. The first region is the region where the organizational units at different locations are less than or equal to the diffusion radius from the diffusion center, and the second region is the region where the organizational units at different locations are greater than the diffusion radius from the diffusion center. The first region and the second region are then modified as desired. Optionally, the formula for constructing the modification matrix D is: in, It is a point spread function, p0 is The diffusion center, p represents the center position of the FEM element in the finite element model, and r is diffusion radius, The matrix represents the modification matrix, where i and j represent the electrodes of the EIT, s is a fuzzy control parameter, and b represents the control parameter for the positive compensation amplitude.

3. The EIT image reconstruction method according to claim 2, characterized in that, The modification matrix is ​​modified by RNG through... Positive compensation of RNG yields the results of changes in conductivity after RNG modification; Optionally, let the compensation range of RNG be a, and the formula for calculating the compensation range is: Where b represents the control parameter for the positive compensation amplitude; Optionally, the modification of RNG also includes RNG suppression, which is performed on the result of RNG modification conductivity change by a single-step feedback strategy to obtain the result of suppressed RNG modification conductivity change. Optionally, the single-step feedback strategy process is as follows: S1, let the modification matrix Calculations are performed when the compensation range limit approaches zero. Obtain the first modification matrix, and construct the first reconstruction matrix based on the first modification matrix; S2. Calculate a of the first reconstruction matrix. t a t It is a normalized amplitude with reference to the maximum positive AR; S3, let the modification matrix The second modification matrix is ​​calculated, and the result of suppressing the change in RNG modified conductivity is obtained based on the second modification matrix. Optionally, a reconstruction matrix is ​​obtained by constructing a reconstruction matrix based on the second modification matrix.

4. The EIT image reconstruction method according to claim 1, characterized in that, The AR in the consensus index also includes sensitivity compensation, which obtains uniform AR and modifies the conductivity change through uniform AP by a modification matrix. Optionally, a uniform AR can be obtained by constructing a sensitivity compensation matrix to perform sensitivity compensation on the AP. The formula for constructing the sensitivity compensation matrix is ​​as follows: Where c is the control parameter for sensitivity compensation, i is the unit number, and j is the SMC number of the excitation measurement channel. express, express, express, This represents the unit sensitivity vector.

5. The EIT image reconstruction method according to claim 4, characterized in that, The control parameter c for sensitivity compensation is used to control the loss in the internal region, which is the region that is farther from the edge than the diffusion radius, and the diffusion radius is the radius of the diffusion function of the modification matrix. Optionally, the control parameters for sensitivity compensation are calculated by training the reconstruction magnitude of the internal region; Optionally, the process of reconstructing the magnitude of the training internal region is as follows: Obtain the sensitivity vector of each unit in the internal region; Sort the unit sensitivity vectors and select n units from the sorted unit sensitivity vectors as training points, where n is a natural number greater than 1. A perturbation is generated centered on each training point, and the measured voltage at each training point is calculated. By setting the control parameters for initial sensitivity compensation, and calculating the reconstruction matrix based on the disturbance, measured voltage, and initial sensitivity compensation control parameters, the reconstructed conductivity change is obtained. The maximum reconstruction amplitude of each disturbance is obtained based on the reconstructed conductivity change. The control parameter c for sensitivity compensation is obtained by calculating the loss between the maximum and minimum values ​​of the maximum reconstruction amplitude; Optionally, the loss is calculated by constructing a loss function between the maximum and minimum values ​​of the maximum reconstruction amplitude and the control parameter c of the sensitivity compensation, and then solving for the control parameter c of the sensitivity compensation. Optionally, the loss function is expressed as: in, This indicates the maximum reconstruction range.

6. The EIT image reconstruction method according to claim 1, characterized in that, The EIT voltage change data includes one or more of the following: two-dimensional EIT voltage change data, three-dimensional EIT voltage change data; Optionally, the EIT measured voltage change data is three-dimensional EIT measured voltage change data, and the reconstruction result is obtained by calculating the three-dimensional EIT measured voltage change data and the reconstruction matrix R; Optionally, by constructing a three-dimensional finite element model, EIT voltage signal acquisition is performed on the three-dimensional finite element model to obtain EIT measurement voltage change data; Optionally, the measurement site for the EIT voltage change measurement includes one or more of the following: brain, lungs, and EIT voltage change data is obtained by performing EIT measurement on the measurement site.

7. The EIT image reconstruction method according to claim 1, characterized in that, The EIT voltage change data is obtained through an excitation measurement mode, which includes one or more of the following: Adjacent excitation measurement mode, phase-to-phase excitation measurement mode, opposing excitation measurement mode, maximum excitation measurement mode; Optionally, the maximum excitation measurement mode is a mode in which the maximum independent excitation measurement channel is selected from all excitation measurement channels of the EIT electrode and excitation measurement is performed. Optionally, the maximum independence means that the number of times the EIT electrode is excited is not redundant.

8. A computer program product comprising a computer program or instructions, characterized in that, The computer program or instructions are executed by a processor to implement the EIT image reconstruction method according to any one of claims 1-7.

9. A computer device comprising a memory, a processor, and a computer program or instructions stored in the memory, characterized in that, The computer program or instructions are executed by a processor to implement the EIT image reconstruction method according to any one of claims 1-7.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, The computer program or instructions are executed by a processor to implement the EIT image reconstruction method according to any one of claims 1-7.

Images