Pulmonary nodule state continuous monitoring method based on EIT technology

Abstract

The invention belongs to the technical field of intelligent treatment auxiliary diagnosis, and provides a pulmonary nodule monitoring method based on an electrical impedance tomography (EIT) technology. Firstly, a large number of boundary voltage values are obtained through a wearable electrode and related equipment thereof, secondly, a posterior probability function is utilized to describe an image of chest fault pulmonary nodule electrical impedance distribution, and a solving problem of the posterior probability function is converted into a solving likelihood function and a prior probability function through a Bayesian formula. Modeling is carried out for bioelectricity noise and electrical impedance disturbance noise in a human body, a likelihood function of the model is constructed, then pulmonary nodule growth law analysis based on pathological analysis is carried out, a prior probability function of the model is constructed, and thus pulmonary nodule state change information is recovered from large sample data based on a Bayesian learning framework. And the spatial resolution of pulmonary nodule image reconstruction is improved, so that continuous monitoring of the pulmonary nodule growth state is further realized.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent treatment-assisted diagnosis, and in particular, designs a method for continuously monitoring the status of lung nodules based on EIT technology. Background Art

[0002] Electrical impedance imaging originated in the United States in the 1970s. After 2000, electrical impedance imaging technology has also developed rapidly in my country. So far, my country's research on electrical impedance imaging has mainly focused on the improvement of methods and the application of technology. Electrical impedance imaging technology has a very wide range of applications in lung monitoring, especially in the dynamic monitoring and evaluation of lung lesions, including gas distribution and ventilation in the lungs, etc.

[0003] Lung nodules are an early manifestation of lung cancer. Rapid growth in a short period of time will most likely turn into lung cancer. There is a strong correlation between lung nodules and lung cancer, but there is currently no technical means to continuously observe the transformation of lung nodules into lung cancer. At present, the methods of lung nodule detection are generally divided into invasive and non-invasive. The invasive method is suitable for larger lung nodules. For smaller lung nodules (maximum diameter less than 1 cm), since their pathological development has not been fully determined, the diagnostic results of this method cannot be used as the final diagnostic results; non-invasive methods are represented by X-rays and chest CT. Although X-rays and chest CT can intuitively show the size and three-dimensional information of lung nodules projected in the front direction of the human body, they cannot obtain the diameter information of lung nodules in the front and back directions of the human body. In addition, chest CT and X-rays are harmful to the body and may cause cancer. They cannot be used frequently. At the same time, the equipment is large and expensive, so it is impossible to continuously detect the growth dynamics of lung nodules. The method for continuously monitoring the growth status of lung nodules based on lung EIT proposed in the present invention is a non-destructive monitoring technology that realizes imaging by measuring the electrical impedance distribution of the human lungs. By controlling the intensity of the current applied to the human body, harmless measurement of the human body can be achieved; the EIT image of the human lungs can obtain three-dimensional information of the lung nodules, thereby realizing continuous monitoring of the growth status of the lung nodules.

[0004] Since the human body is a dynamic area with strong interference, and the lung nodule area is relatively small, the effective information in the detection signal is very weak. The current electrical impedance image reconstruction methods cannot meet the monitoring requirements. Therefore, the present invention proposes a method for reconstructing lung nodule EIT images based on a Bayesian learning framework to improve the resolution of lung nodule EIT reconstructed images. Summary of the invention

[0005] In view of this, the present invention proposes for the first time to use electrical impedance imaging to continuously monitor lung nodules, collect a large number of data samples for the lung nodules that need to be monitored through wearable methods, establish a monitoring prior model based on the pathological laws of lung nodules, and then use large samples to establish an electrical impedance image reconstruction theory with statistical laws to achieve the monitoring of changes in the status of lung nodules.

[0006] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0007] This aspect provides a method for monitoring pulmonary nodules based on electrical impedance imaging technology, comprising the following steps:

[0008] Step S1: Use the CT image to accurately segment the lung nodules to be monitored, and obtain the initial position state of the lung nodule ROI object field area based on the segmented lung nodule image. Use a non-uniform grid to finely mesh the ROI area and coarsely mesh the non-ROI area to achieve high-resolution image reconstruction of the ROI object field, reducing the pathological nature of the image reconstruction calculation. The electrical impedance distribution of the ROI object field space area can be expressed as: Z ROI (x, y, z, t), where (x, y, z)∈P ROI , t∈R;

[0009] Step S2: Wearable device measurement data. A large number of monitoring data samples are obtained through wearable electrodes and related equipment. The measurement equipment includes a measurement electrode array (102) worn around the lung organ (101), and a sinusoidal constant current excitation (103) is applied between the boundary electrodes to measure the voltage data, and the quasi-steady field assumption is satisfied; after the boundary electrode obtains the data, the collected data is amplified by the amplifier (104), and the data is wirelessly transmitted to the computer (105) through the Bluetooth module, and the boundary voltage value is obtained and saved in the host computer. Finally, the image of the chest tomographic impedance distribution is obtained using the image reconstruction algorithm. When a sinusoidal AC excitation is applied to the boundary of the object field, the electric field is instantaneously established at each point in the field, and the time taken for the current to travel from one point to another in the field is ignored. When a sinusoidal excitation current is injected, the boundary voltage and impedance distribution are uniquely determined by the nonlinear function F:

[0010]

[0011] where (x, y, z)∈R 3 , t∈R, Γ is the chest surface, for convenience, Z is used to represent Z(x, y, z, t), express Z0 represents Z(x, y, z, t0);

[0012] Step S3: Calculation of sensitivity function. The sensitivity function is a linear approximate mapping of boundary voltage change and impedance change when the impedance distribution is Z0:

[0013] in

[0014] The sensitivity function is solved using Maxwell's equations for the electromagnetic field:

[0015]

[0016] Where D is the electric displacement, ρ is the charge density, E is the electric field intensity, B is the magnetic induction intensity, H is the magnetic field intensity, J is the current density, μ, ∈, ρ correspond to the dielectric constant, magnetic permeability and electrical conductivity of the medium respectively. Under the given boundary conditions:

[0017]

[0018] The sensitivity function can be solved:

[0019]

[0020] Step S4: Establish a Bayesian learning framework to realize EIT image reconstruction. Through Bayesian learning, the distribution of electrical impedance is expressed as the solution of posterior probability distribution:

[0021]

[0022] In the actual measurement process, the boundary voltage and the electrical impedance distribution are both discrete. Assume that there are N measured voltages and M electrical impedance distribution values:

[0023]

[0024] ΔZ 1×N = {Z0, Z1, ... Z N-1}

[0025] Build a relationship:

[0026]

[0027] in, The initial distribution of electrical impedance can be obtained through CT images. In the Bayesian learning model of the present invention, Z0 and J Z0 is regarded as a constant, and according to the Bayesian formula we can get:

[0028]

[0029] Re-express the above equation using the likelihood function and assume full probability:

[0030]

[0031] and log(p(ΔZ M×1 )) is the likelihood function and prior probability function to be solved in the subsequent steps, which are mainly established by the bioelectric noise and human body electrical impedance disturbance noise in the human body and the growth law of lung nodules; λ is the weight coefficient for adjusting the constraints of the prior probability function on the reconstruction results;

[0032] Step S5: Modeling human bioelectric noise. The form of bioelectric signals is very complex, and there are large differences between different people. It is difficult to obtain an accurate mathematical model through theoretical deduction. The present invention estimates the likelihood function of bioelectric noise by means of experimental data. First, a hidden Markov model is used to model bioelectric noise. Assume that there are N bio Noise sources:

[0033]

[0034] The noise source has a great influence on the measuring electrode. The bioelectric noise can be expressed as:

[0035]

[0036] in, To describe the parameters of the model. Without applying AC excitation, the observed data of the Markov model are as follows:

[0037]

[0038] The observed data can be used to calculate the parameter P bio Estimates.

[0039] Step S6: Modeling of electrical impedance disturbance noise. When the human body is quiet and relaxed, the distribution of electrical impedance is mainly determined by human physiological processes such as blood flow and metabolism. The change in electrical impedance distribution can be roughly divided into two parts:

[0040]

[0041] in and They represent the changes in the electrical impedance distribution outside the lung region of interest and in the region of interest. The electrical impedance disturbance noise is caused by In order to analyze the impedance disturbance noise, let You can get:

[0042]

[0043] It can be seen that the electrical impedance disturbance noise is mainly determined by the disturbance of M1 average electrical impedance values;

[0044] Step S7: Modeling the growth law of lung nodules. Modeling the growth law of lung nodules mainly involves establishing the prior probability distribution in the Bayesian theory image reconstruction model:

[0045]

[0046] The prior probability distribution represents the probability of the electrical impedance distribution of the ROI region after Δt when the electrical impedance distribution of the current ROI region is known.

[0047] The growth law modeling of lung nodules in the present invention is mainly based on pathological modeling and image analysis modeling. In pathological modeling, existing medical studies have shown that if the lung nodules are benign, their growth process will be relatively slow and can be approximated by a linear function:

[0048] V(t)=V0+kt

[0049] If the lung nodule is malignant, the volume of the lung nodule will grow exponentially according to the division rate of cancer cells, which can be approximated by the following formula:

[0050]

[0051] In the formula, DT is the doubling time, which can be measured through medical experiments.

[0052] In image analysis modeling, all data on the electrical impedance distribution of known lung nodules form a low-dimensional manifold in the feature space. For the current electrical impedance distribution of a known lung nodule, its position on the manifold is determined, and the probability that the lung nodule distribution changes to other lung nodule distributions is proportional to its distance on the manifold. In the modeling process, a large number of CT images containing lung nodules are collected, and the lung nodule images contained therein are segmented to obtain the CT image representation of the lung nodules:

[0053]

[0054] Design a mapping function f to transform the CT image representation of lung nodules into electrical impedance distribution and perform normalization:

[0055]

[0056] By designing a manifold learning algorithm, the electrical impedance distribution of pulmonary nodules is mapped to a low-dimensional space manifold M and a distance function d(·) is defined on the manifold M. The expression of the growth law of pulmonary nodules can be obtained:

[0057]

[0058] Step S8: Use Bayesian learning to solve the posterior probability distribution. Solve and optimize the posterior probability function through the likelihood function and the prior probability function, and use the large sample measurement data to establish the EIT image reconstruction model based on Bayesian learning, so as to further improve the spatial resolution of EIT image reconstruction for lung nodule monitoring; lung nodule regional segmentation. Perform detailed lung nodule regional segmentation on the reconstructed EIT image, which can accurately identify and extract the shape and size of the lung nodules, continuously monitor the growth status of the lung nodules, and timely understand the changes in the status of the lung nodules. It can provide early warning for patients at risk of cancer, achieve early detection and early treatment, and is of great significance to the study of the pathological evolution of lung nodules.

[0059] The method for continuously monitoring the status of lung nodules based on EIT technology described in the present invention has the following advantages:

[0060] This invention innovatively proposes a method for continuous monitoring of lung nodules based on EIT image reconstruction technology. It is the first time to use EIT to monitor lung nodules, and to use the prior knowledge of human body noise and lung nodule growth to perform a large sample image reconstruction model based on statistical learning. The initial state of lung nodules is obtained using CT images, and a large number of monitoring data samples are obtained with the help of wearable devices. The EIT image reconstruction is described using the posterior probability function, which is converted into the solution of the likelihood function and the prior probability function based on Bayesian learning. The likelihood function and the prior probability function are constructed for the human body bioelectric noise and electrical impedance disturbance noise model measured by EIT and the lung nodule growth model. Based on Bayesian learning, the information on the change of lung nodule status is recovered from large sample data, overcoming the difficulties of strong noise and weak signals, and improving the spatial resolution of image reconstruction, thereby achieving early warning of lung cancer.

[0061] The present invention adopts EIT to continuously monitor the status of lung nodules, and uses the electrical impedance distribution inside the human body for imaging, which does not involve invasive operations and is non-invasive; EIT provides real-time images of changes in the status of lung nodules, which is convenient for quickly understanding the dynamic changes of lung nodules and treating patients in time; compared with CT and X-rays, EIT has relatively low costs and relatively little harm to the human body, and is suitable for long-term and continuous monitoring and tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0063] Figure 1 It is a schematic diagram of the initial state of ROI obtained through the CT image of lung nodules in the present invention;

[0064] Figure 2A schematic diagram of a wearable electrode and related equipment for obtaining a boundary voltage value of a lung nodule according to the present invention;

[0065] Figure 3 It is a schematic diagram of the method for continuously monitoring the status of lung nodules based on EIT technology of the present invention;

[0066] Figure 4 It is a schematic diagram of the distribution of sensitivity matrix elements of the present invention;

[0067] Figure 5 Schematic diagram of the dimensionality reduction framework of manifold learning of lung nodule CT images of the present invention;

[0068] Figure 6 A schematic diagram of a flow chart for solving a mathematical model for image reconstruction of local high-precision electrical impedance imaging according to the present invention;

[0069] 101- lung organ; 102- measuring electrode; 103- sinusoidal constant current excitation; 104- amplifier; 105- computer. DETAILED DESCRIPTION

[0070] The present invention is further described in detail below through specific examples. The following examples are only descriptive and not restrictive, and the protection scope of the present invention cannot be limited by them. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0071] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0072] First, we need to use non-uniform grid division to obtain the initial state of the lung nodule ROI object field area. Figure 1 As shown in the figure, the blue circular area is the ROI object field, which refers to the spatial area containing the lung nodules to be monitored. The purpose of the image reconstruction algorithm is to improve the image resolution of the ROI object field area. lung , P sa Represent the spatial regions of the chest cavity, lungs, and lung nodules, respectively, and ROI Field P ROI The following conditions must be met: and The electrical impedance distribution in the ROI object field space is expressed as: Z ROI (x, y, z, t), where (x, y, z)∈P ROI , t∈R. The pulmonary nodule region to be monitored in the CT image is finely divided into non-uniform grids, and the initial position state of the pulmonary nodule is obtained based on the segmented image. Figure 1The ROI object field area in medium blue is segmented, and the non-ROI area is coarsely segmented, thereby achieving high-resolution image reconstruction and reducing pathological problems in reconstruction calculations.

[0073] Secondly, wearable electrode arrays and related equipment are used to measure data. The measurement equipment is as shown in the reference Figure 2 As shown, it includes wearing a measuring electrode array (102) around the lung organ (101), applying a sinusoidal constant current excitation (103) between the boundary electrodes to measure the voltage data, and satisfying the quasi-steady field assumption; after the boundary electrode obtains the data, the collected data is amplified and processed by the amplifier (104), and the data is wirelessly transmitted to the computer (105) through the Bluetooth module, and the boundary voltage value is obtained and saved in the host computer. When a sinusoidal AC excitation is applied to the boundary of the object field, the electric field is instantaneously established at each point in the field, and the time taken for the current to travel from one point to another in the field is ignored. When a sinusoidal excitation current is injected, the boundary voltage and electrical impedance distribution are uniquely determined by the nonlinear function F:

[0074]

[0075] where (x, y, z)∈R 3 , t∈R, Γ is the chest surface, for convenience, Z is used to represent Z(x, y, z, t), express Z0 represents Z(x, y, z, t0);

[0076] After the boundary voltage value is collected, the sensitivity function is calculated Since the electrical impedance distribution Z changes slowly, the formula can be simplified as shown below:

[0077] in

[0078] The sensitivity function is solved using Maxwell's equations for the electromagnetic field:

[0079]

[0080] Where D is the electric displacement, ρ is the charge density, E is the electric field intensity, B is the magnetic induction intensity, H is the magnetic field intensity, J is the current density, μ, ∈, ρ correspond to the dielectric constant, magnetic permeability and electrical conductivity of the medium respectively. Given the following boundary conditions:

[0081]

[0082] Finally, the formula of the sensitivity function is obtained:

[0083]

[0084] Next, we establish the EIT image reconstruction algorithm, namely the Bayesian learning framework. Through Bayesian learning, the distribution of the electrical impedance is expressed as the solution of the posterior probability distribution:

[0085]

[0086] The boundary voltage and electrical impedance distribution are both discrete. Assume there are N measured voltages and M electrical impedance distribution values:

[0087]

[0088] ΔZ 1×N = {Z0, Z1, ... Z N-1}

[0089] Build a relationship:

[0090]

[0091] in, Z0 and J in Bayesian learning model Z0 is regarded as a constant, and according to the Bayesian formula we can get:

[0092]

[0093] Re-express the above equation using the likelihood function and assume full probability:

[0094]

[0095] and log(p(ΔZ M×1 )) is the likelihood function and prior probability function to be solved in the subsequent steps, which are mainly established by the bioelectric noise and human body electrical impedance disturbance noise in the human body and the growth law of lung nodules; λ is the weight coefficient for adjusting the constraints of the prior probability function on the reconstruction results, thus transforming the problem of solving the posterior probability distribution into solving the problem of solving the likelihood function and the prior probability function.

[0096] Solving the likelihood function requires modeling the bioelectric noise and electrical impedance disturbance noise of the human body. The bioelectric morphology of the human body is complex and varies significantly between individuals. Therefore, it is difficult to obtain an accurate mathematical model of the bioelectric noise of the human body through theoretical derivation. It is necessary to estimate the likelihood function of the bioelectric noise of the human body through the collection of experimental data. First, the hidden Markov model is used to model the bioelectric noise. There are N bio Noise sources:

[0097]

[0098] The influence of noise sources on the measuring electrodes is bioelectric noise, namely:

[0099]

[0100] in, To describe the parameters of the model. Without applying AC excitation, the observed data of the Markov model are as follows:

[0101]

[0102] The observed data can be used to calculate the parameter P bio Estimates.

[0103] exist Figure 3 middle and They represent the changes in the electrical impedance distribution outside the non-ROI area of ​​the lung and in the ROI area. When the human body remains still and relaxed, its electrical impedance distribution is mainly affected by physiological processes such as blood flow and metabolism. The changes in electrical impedance distribution can be divided into two parts:

[0104]

[0105] in and They represent the changes in the electrical impedance distribution outside the lung region of interest and in the region of interest. The electrical impedance disturbance noise is caused by In order to analyze the impedance disturbance noise, let You can get:

[0106]

[0107]

[0108] It can be seen that the electrical impedance disturbance noise is mainly determined by the disturbance of the M1 average electrical impedance values.

[0109] Solving the prior probability function is based on modeling the growth law of lung nodules:

[0110]

[0111] The probability of the impedance distribution of the ROI area after Δt is given when the impedance distribution of the current ROI area is known. The growth law modeling of pulmonary nodules is mainly based on pathological modeling and image analysis modeling. In pathological modeling, medical research shows that if the pulmonary nodules are benign, their growth process will be relatively slow and can be approximated by a linear function:

[0112] V(t)=V0+kt

[0113] If the lung nodule is malignant, the volume of the lung nodule will grow exponentially according to the division rate of cancer cells, which can be approximated by the following formula:

[0114]

[0115] In the formula, DT is the doubling time, which can be measured through medical experiments.

[0116] In the process of image analysis and modeling, as shown in the following figure: Figure 4 As shown in the figure, manifold learning dimensionality reduction is performed on the pulmonary nodule CT image. The known pulmonary nodule electrical impedance distribution data forms a low-dimensional manifold in the feature space. For the electrical impedance distribution of a known pulmonary nodule, its position on the manifold can be determined, and the probability of the pulmonary nodule transforming into other pulmonary nodule distributions is proportional to its distance on the manifold. In the modeling process, a large number of CT images containing pulmonary nodules are first collected, and the pulmonary nodule images contained therein are segmented to obtain the CT image representation of the pulmonary nodules:

[0117]

[0118] Design a mapping function f to transform the CT image representation of lung nodules into electrical impedance distribution and perform normalization:

[0119]

[0120] By designing a manifold learning algorithm, the electrical impedance distribution of pulmonary nodules is mapped to a low-dimensional space manifold M and a distance function d(·) is defined on the manifold M. The expression of the growth law of pulmonary nodules can be obtained:

[0121]

[0122] Finally, the posterior probability function is obtained by solving the prior probability function and the likelihood function, and the electrical impedance distribution of the pulmonary nodules is obtained based on Bayesian learning, as shown in Figure 5 This method further improves the spatial resolution of electrical impedance image reconstruction, thereby recovering information about changes in the state of pulmonary nodules from large sample data and achieving continuous monitoring of the growth state of pulmonary nodules.

[0123] The process diagram is as follows Figure 6As shown. The establishment of the mathematical model is mainly based on the local area Bayesian learning framework. By solving the mathematical model of the positive problem, a lung model is established, which includes ROI excitation measurement data and simulation solution of the positive problem to obtain the sensitivity matrix; the electrical impedance distribution of the lung nodules is transformed into the problem of solving the posterior probability distribution of Bayesian learning, which requires the likelihood function and prior probability function. By modeling the human bioelectric noise and electrical impedance disturbance noise and the growth law of lung nodules, the likelihood function and prior probability function are solved, and finally the Bayesian imaging framework of the local area is obtained.

[0124] The above description is only a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several modifications and improvements without departing from the inventive concept, which all belong to the protection scope of the present invention.

Claims

1. A lung nodule monitoring method based on electrical impedance tomography, the main method used is a local area imaging framework based on Bayesian learning, which includes non-uniform grid division of the lung area and calculation of sensitivity function, likelihood function calculation based on human bioelectric noise modeling and electrical impedance disturbance noise modeling, and prior probability function calculation based on lung nodule growth law modeling. It is characterized by: Step S1: Non-uniform grid division of the lung area. The ROI (Region of Interest) object field refers to the spatial region containing the lung nodules to be monitored, that is, the region that requires high-resolution image reconstruction. In order to meet the imaging accuracy of lung nodule morphology monitoring, a non-uniform grid is used to finely divide the ROI area, and a coarse grid is used in the non-ROI object field area, which reduces the pathological nature of the image reconstruction calculation. First, the lung nodules to be monitored are accurately segmented using the CT image, and the initial state of the lung nodule ROI object field is obtained based on the segmented lung nodule image; Step S2: Start the measurement, inject a sinusoidal constant current excitation on the boundary electrode, and satisfy the quasi-steady field assumption to obtain the boundary voltage value The corresponding electrical impedance distribution is represented by Z; Step s3: Calculate the sensitivity function The sensitivity function is a linear approximation mapping of the boundary voltage change and the impedance change when the impedance distribution is Z0. Since the impedance distribution Z changes slowly, the formula can be simplified as shown below: in The sensitivity function is solved using the Maxwell equations of the electromagnetic field, given the following boundary conditions: Finally, the formula of the sensitivity function is obtained: Step S4: Establish a Bayesian decision framework for EIT image reconstruction. In actual measurement, the boundary voltage and electrical impedance distribution are both discrete. Suppose there are N measured voltages and M electrical impedance distribution values: ΔZ 1×N ={Z0,Z1,……Z N-1 } Able to establish a relationship: The Bayesian framework is used to realize EIT image reconstruction, and the distribution of electrical impedance is expressed as the solution of posterior probability: The electrical impedance distribution can be obtained using CT images, so in the Bayesian framework, Z0 and J z0 can be treated as a constant, according to Bayes' formula: Assuming full probability: is the likelihood function, which is mainly established by the bioelectric noise in the human body and the human body impedance disturbance noise, where log(p(ΔZ M×1 )) is the prior probability function, which is determined by the growth law of lung nodules; λ is the weight coefficient that adjusts the prior probability function to constrain the reconstruction results. The establishment of the likelihood function and the prior probability function will be introduced in detail in the following steps; Step S5: Modeling of human bioelectric noise. The likelihood function is calculated based on the modeling of human bioelectric noise and electrical impedance disturbance noise. The form of human bioelectricity is very complex, and there are large differences in electrical signals between different people. It is difficult to obtain an accurate mathematical model of electrical noise through theoretical deduction. Therefore, it is necessary to estimate the likelihood function of human bioelectric noise by collecting experimental data. First, the hidden Markov model is used to model the bioelectric noise. bio Noise sources: The influence of noise sources on the measuring electrodes is bioelectric noise, namely: in, To describe the parameters of the model. Without applying AC excitation, the observed data of the Markov model are as follows: The observed data can be used to calculate the parameter P bio estimates; Step S6: Modeling of electrical impedance disturbance noise. When the human body remains quiet and relaxed, the disturbance of electrical impedance distribution is mainly determined by human physiological processes such as blood flow and metabolism. The impedance distribution change can be divided into two parts: is the change in electrical impedance distribution outside the lung region of interest, is the change in the electrical impedance distribution of the lung region of interest. Therefore, in order to analyze the impedance disturbance noise, we can make You can get: It can be seen that the electrical impedance disturbance noise is mainly determined by the disturbance of M1 average electrical impedance values; Step S7: Modeling the growth law of lung nodules. Modeling the growth law of lung nodules is mainly for the image reconstruction model based on Bayesian theory. The prior probability of the model can be expressed as: The probability of the electrical impedance distribution in the ROI area after Δt, given the current electrical impedance distribution in the ROI area. For the current electrical impedance distribution of a known lung nodule, first determine its position on the manifold. The probability that the lung nodule distribution changes to other lung nodule distributions is proportional to its distance on the manifold. Collect a large number of CT images containing lung nodules, segment the lung nodules contained therein, and obtain CT images of lung nodules: Design a mapping function f to transform the CT image of the lung nodule into an electrical impedance distribution and perform normalization: Design a manifold learning algorithm to map the electrical impedance distribution of pulmonary nodules to a low-dimensional space manifold M. Define the distance function d(·) on the manifold M, and you will get: Step S8: By establishing the likelihood function and prior probability function of the model, solving and optimizing the posterior probability distribution, the probability distribution of the EIT image reconstruction impedance is obtained, and finally the image reconstruction result of the lung nodule EIT is obtained, thereby further improving the resolution of the lung nodule EIT image; the reconstructed EIT image is preprocessed to enhance the image quality, the lung nodule object field area of ​​the EIT image is segmented, the size, shape and other features of the lung nodules are extracted, and the growth status of the lung nodules is monitored.

2. When obtaining the boundary voltage value in step S2 of claim 1, a sinusoidal constant current excitation is injected on the boundary electrode, and the quasi-steady field assumption is satisfied. The quasi-steady field assumption is that when a sinusoidal AC excitation is applied to the boundary of the physical field, the electric field is established instantaneously at each point in the field, that is, the time taken by the current to travel from one point to another in the field is ignored. The boundary voltage and electrical impedance distribution can be uniquely determined by a nonlinear function F, and the function definition is as follows: where (x, y, z)∈R 3 , t∈R, Γ is the chest surface, for convenience, Z is used to represent Z(x, y, z, t), express Z0 represents Z(x, y, z, t0).

3. When calculating the sensitivity function in step S3 according to claim 1, the electrical impedance distribution Z changes slowly, so it can be obtained: O((Z-Z0) 2 =0 Maxwell's equations are used to solve it: Where D is the electric displacement, ρ is the charge density, E is the electric field intensity, B is the magnetic induction intensity, H is the magnetic field intensity, J is the current density, and μ, ∈, and ρ correspond to the dielectric constant, magnetic permeability, and electrical conductivity of the medium, respectively.

4. When analyzing the growth pattern of pulmonary nodules in step S7 according to claim 1, existing medical studies have shown that if the pulmonary nodules are benign, their growth process is relatively slow and can be approximated by a linear function, namely: V(t)=V0+kt If the lung nodule is malignant, the volume of the lung nodule will grow exponentially according to the division rate of cancer cells, which can be approximated by the following function: DT is the doubling time, which can be measured through medical experiments.

Images