Tumor irreversible electroporation ablation simulation prediction system

Abstract

The invention discloses a tumor irreversible electroporation ablation simulation prediction system, which comprises the following modules: a simulation model geometric structure establishment module, which is used for determining related parameters of a tumor irreversible electroporation ablation operation and a catheter electrode structure, and establishing a geometric structure of a simulation model; the geometric model and numerical model setting module is used for setting a geometric model and a numerical model; respectively constructing a two-electrode plate type three-dimensional finite element multi-physical field simulation model and a three-electrode plate type three-dimensional finite element multi-physical field simulation model for irreversible electroporation ablation of tumors; the data set establishment module is used for respectively performing simulation calculation on the two numerical models constructed by the geometric model and the numerical model setting module; the model training module is used for training the respective data sets of the two numerical models obtained by the data set establishing module by using a random forest regression algorithm; and the simulation prediction module is used for predicting the tumor ablation condition and the damage condition of surrounding normal tissues. The system provided by the invention can accurately predict the tumor ablation condition and the damage condition of surrounding normal tissues.

Description

Technical Field

[0001] The present invention belongs to the technical field of machine learning, and specifically relates to a simulation prediction system for irreversible electroporation ablation of tumors. Background Art

[0002] Malignant tumors are major diseases that seriously threaten human health and are the most serious public health challenges currently faced globally. Malignant tumors of the liver, biliary tract, and pancreas are the most common tumor types in the digestive system. For the treatment of tumors, irreversible electroporation (IRE) is a new type of tumor electroablation technology. Its principle is to use electrical pulses to act on the surface of cancer cells, irreversibly damaging the phospholipid bilayer, forming defects on the cell membrane surface. When the electrical pulses reach a certain level, the cell membrane defects cannot be reversed, ultimately causing target cell necrosis. IRE belongs to a non-thermal ablation technology, which can avoid irreversible damage to adjacent tissue structures and retain intracellular macromolecules and tissue scaffolds, etc. Existing clinical studies have also confirmed that the IRE technology has significant advantages in preserving organ function and is particularly suitable for local ablation of tumors in small and complex organs such as the pancreas. Therefore, the IRE technology has important clinical significance for the treatment of pancreatic tumors.

[0003] In order to further improve the efficiency and accuracy of surgical planning and achieve the treatment goal of "maximally ablating tumor tissue and minimally damaging normal tissue", it is essential to continuously simulate and analyze the ablation surgical process and results before surgery. In current research on tumor ablation simulation, researchers mostly choose COMSOL as a finite element solver to calculate and solve the Laplace equation of tissue electric field distribution, etc. COMSOL provides an integrated development environment that enables researchers to continuously try to establish complex models and solve them. Its functions are very powerful, mainly including multi-physics field modeling, coupling effect analysis, custom modeling and development, and real-time visualization and post-processing. However, this method also has problems. For example, for IRE ablation of pancreatic tumors, its simulation model has complex geometric shapes and physical processes. Establishing an accurate model in COMSOL requires professional knowledge and experience and has high requirements for computing resources, which makes it very difficult to establish and adjust the model. Moreover, during the surgical planning process, it is necessary to quickly obtain the IRE ablation results for decision-making, while the COMSOL simulation takes a relatively long time and cannot meet the real-time requirements. Summary of the Invention

[0004] Aiming at the problems that the existing physical simulation methods are time-consuming and complex to operate and are not suitable for the surgical planning scenario, the purpose of the present invention is to provide a simulation prediction system for irreversible electroporation ablation of tumors, which can quickly and accurately predict the results of irreversible electroporation ablation.

[0005] To achieve the above object, the present invention is implemented by the following technical solutions:

[0006] A simulation prediction model for irreversible electroporation ablation of tumors, comprising the following modules:

[0007] A simulation model geometric structure establishment module, configured to determine parameters related to irreversible electroporation ablation surgery of tumors and the structure of catheter electrodes, and establish the geometric structure of the simulation model;

[0008] A geometric model and numerical model setting module, configured to set the geometric model and the numerical model to simulate the coupling effect of the electric field and the bio-thermal field during irreversible electroporation of tumors; respectively construct three-dimensional finite element multi-physics field simulation models for irreversible electroporation ablation of tumors with two electrode plate types and three electrode plate types, and calculate the tumor ablation rate and the thermal damage volume ratio of normal tissues through numerical analysis;

[0009] A data set establishment module, configured to perform simulation calculations on the two numerical models constructed by the geometric model and numerical model setting module respectively, obtain simulation results under different surgical parameter settings in different electrode structure models, collect and organize the simulation data of the two numerical models as their respective data sets, and divide the training set and test set of each simulation data respectively;

[0010] A model training module, configured to use the data sets of the two numerical models obtained by the data set establishment module, and respectively use the random forest regression algorithm for training and optimize the hyperparameters to obtain the trained simulation prediction models for irreversible electroporation ablation of tumors corresponding to the two different electrode structures;

[0011] A simulation prediction module, configured to use the trained simulation prediction model for irreversible electroporation ablation of tumors to predict the tumor ablation situation and the damage situation of surrounding normal tissues corresponding to a certain surgical parameter setting.

[0012] Compared with the prior art, the present invention has the following technical effects:

[0013] In view of the problems that physical simulation is time-consuming and complex in operation and not applicable to the surgical planning scenario, the present invention proposes an irreversible electroporation ablation damage estimation and prediction model for pancreatic tumors. By constructing a three-dimensional finite element multi-physics field simulation model of irreversible electroporation ablation of pancreatic tumors, after performing simulation calculations on the numerical model respectively, the corresponding simulation data sets are collected and sorted out, and then a damage estimation and prediction model is constructed using the random forest regression algorithm, which is used to quickly and accurately predict the corresponding ablation results under certain surgical parameter settings. Specifically, the present invention uses the random forest regression algorithm to explore the relationships and laws between the five surgical parameters of electrode spacing, electrode length, voltage amplitude, pulse width, and pulse number and the two ablation overall effect indexes of thermal damage volume ratio and tumor ablation rate. Through comparative experiments with several commonly used algorithms in machine learning, the results show that the prediction effect using the random forest regression algorithm is better than several other machine learning algorithms, achieving the purpose of quickly and accurately predicting the ablation results.

[0014] Meanwhile, the method of the present invention can better solve the characteristics of slow running speed and complex operation of COMSOL in the clinical process, greatly improve the efficiency of clinical surgical decision-making, and the accuracy is highly similar to the simulation results, achieving the purpose of high efficiency and precision. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of the method of the present invention.

[0016] Figure 2 It is a schematic diagram of the geometric model;

[0017] Figure 3 It is an ablation example diagram of the two-electrode plate type model of the present invention;

[0018] Figure 4 It is an ablation example diagram of the three-electrode plate type model of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0019] The following further elaborates on the specific content of the present invention in conjunction with the embodiments.

[0020] In the surgical planning stage, currently, COMSOL simulation is mostly used to obtain the IRE ablation results of tumors. However, this method requires professional modeling knowledge and experience, has high requirements for computing resources, is difficult to establish and adjust the model, and the COMSOL simulation takes a relatively long time and cannot meet the real-time requirements. Therefore, based on the constructed three-dimensional multi-physics field simulation model, this embodiment combines the random forest regression algorithm in machine learning to construct a simulation prediction model for irreversible electroporation ablation of tumors.

[0021] The simulation prediction system for irreversible electroporation ablation of tumors given by the present invention includes the following modules:

[0022] (1) Simulation model geometric structure establishment module, used to determine the parameters related to irreversible electroporation ablation surgery of tumors and the catheter electrode structure, and establish the geometric structure of the simulation model. Among them:

[0023] The parameters related to ablation surgery include electrode parameters and pulse parameters. The electrode parameters include electrode spacing and electrode length, and the pulse parameters include voltage amplitude, pulse width, and number of pulses; the catheter electrode structure includes two-electrode plate type and three-electrode plate type.

[0024] The catheter electrode includes two-electrode plate type electrode and three-electrode plate type electrode (as Figure 2 shown). The two-electrode plate type catheter electrode includes a positive electrode and a negative electrode, and the three-electrode plate type includes a positive electrode and two negative electrodes. The electrode settings of both types of catheter electrodes are parallel to the pancreatic duct tissue, used to simulate the situation of ablation after the catheter electrode enters the pancreatic duct through the natural body cavity, and both types of catheter electrodes are symmetrically placed with respect to the tumor tissue.

[0025] The geometric structure of the simulation model includes biological tissue and electrode structure. The biological tissue includes pancreatic duct tissue and tumor tissue. The pancreatic duct tissue is a horizontally arranged tube with an inner diameter of 0.2 cm, an outer diameter of 0.25 cm, and a length of 10 cm. The tumor tissue is an ellipsoid model with semi-axes a, b, and c values of 5 mm, 6 mm, and 6 mm respectively; the electrode structure is a catheter electrode.

[0026] In the simulation model geometric structure establishment module designed above, by determining the parameters related to irreversible electroporation surgery, the factors affecting the surgical effect can be made precise, which helps to more accurately predict the ablation result. Based on the catheter electrode structure of the surgery, the corresponding simulation geometric model can be established, laying a foundation for the subsequent coupling of the electric field and biological heat field.

[0027] (2) Geometric model and numerical model setting module, used to set the geometric model and numerical model (COMSOL Multiphysics software can be used), simulate the coupling of the electric field and biological heat field during the irreversible electroporation of tumors; respectively construct three-dimensional finite element multi-physics field simulation models for irreversible electroporation ablation of tumors of the two-electrode plate type and three-electrode plate type, and through numerical analysis, calculate the tumor ablation situation (tumor ablation rate) and the damage situation of surrounding normal tissues (thermal damage volume ratio) during the simulation ablation process.

[0028] Specifically, in the geometric model and numerical model setting module, the numerical analysis includes tumor electrical analysis and normal tissue thermal analysis.

[0029] Specifically, the tumor electrical analysis includes the calculation of the electric field distribution during the irreversible electroporation process, the setting of tissue conductivity, and the calculation of tumor cell electrical damage;

[0030] ① During the irreversible electroporation process, the electric field distribution is represented by the electrical conductivity of the tumor tissue, and is specifically calculated using the non-linear Laplace equation:

[0031]

[0032] In the formula, represents the difference operation, σ is the electrical conductivity of the tumor tissue, and the electrical conductivity and the field strength influence each other. is the intensity of the externally applied voltage;

[0033] ② For the tissue conductivity setting, the pulse electric field distribution is specifically calculated using the tissue dynamic conductivity. The expression of the tissue dynamic conductivity is:

[0034] σ(E) = σ init + (σ max - σ init ) · exp(-exp(a · (E - E IRE )))

[0035] In the formula, σ(E) represents the tissue dynamic conductivity, σ init represents the initial electrical conductivity of the tumor tissue (in the embodiment, it is a pancreatic tumor), taking 0.341 S / m; σ max represents the maximum electrical conductivity of the pancreas and tumor tissue, taking 0.6138 S / m; a is set to 0.002 m / V; E is the tissue electric field strength, and E IRE is the minimum field strength required for irreversible electroporation of the cell membrane;

[0036] ③ The situation of tumor cell electro-injury represents the completion of tumor ablation. The Peleg-Fermi model is used to calculate the probability of tumor cell electro-injury during the irreversible electroporation ablation process. The specific calculation process is as follows:

[0037] First, the probability of tumor cell survival after electro-ablation is expressed by the following formula:

[0038]

[0039] In the formula, S is the probability of tumor cell survival after electro-ablation, E is the field strength magnitude, A determines the curve turning slope; E c (n) is the critical value of cell death, and A(n) represents the change during cell necrosis. The specific expressions of the two functions are as follows:

[0040] E c (n) = E 0 · exp(-k 1 n)

[0041] A(n) = A 0 · exp(-k2 n)

[0042] In the formula, E 0 , A 0 , k 1 and k 2 is the regression coefficient, and n is the number of pulses. The specific value of each regression coefficient is related to the applied pulse width and cell type;

[0043] Secondly, the probability of tumor cell death due to the pulsed electric field is expressed as:

[0044] P=1-S

[0045] Secondly, the cell death probability P is used as the electrical damage probability of tumor cells, recorded as PEI; if the electrical damage probability value of a certain area of ​​tumor tissue is greater than 0.99, it is said that irreversible electrical damage has occurred in this area. At the same time, in order to more intuitively represent the electrical damage of the tumor, the ratio of the tumor volume with an electrical damage probability greater than 0.99 to the entire tumor volume is defined as the tumor ablation rate, and the tumor ablation rate is used as one of the indicators to reflect the ablation situation.

[0046] Specifically, normal tissue thermal analysis includes temperature distribution calculation and thermal damage calculation. The Pennes biothermal equation is used to calculate the temperature distribution of various parts of the tissue during ablation. The equation expression is as follows:

[0047]

[0048] Where k is thermal conductivity, T is temperature, ω b is the blood perfusion rate, c b is the heat capacity of blood, T b is the arterial blood temperature, q ′′′ is metabolic heat production, ρ is density, c p is the tissue heat capacity, Joule heat Q Joule It is the physical heat generated during the ablation process;

[0049] The Arrhenius formula is used to quantify the thermal damage of tissues. The degree of thermal damage to biological tissues is related to factors such as temperature and exposure time. k The rate constant for the transformation from the original state to the thermally damaged state at (t) is expressed as:

[0050]

[0051] Where Ω(t) is the dimensionless time-dependent accumulated thermal damage, E a is the required activation energy, R is the ideal gas constant, and T(t) is the temperature distribution over time.

[0052] Afterwards, the thermal damage is calculated using the Arrhenius formula:

[0053]

[0054] where t init is the initial time point of the heating process, and then the tissue thermal damage probability P is calculated using the following equation:

[0055] P = 1 - exp(-Ω(t))

[0056] where the cell death probability P is the thermal damage probability of normal tissue cells, denoted as PTI. When the thermal damage probability value in a certain region is greater than 0.99, it can be regarded as irreversible thermal damage. At the same time, in order to more intuitively represent the thermal damage situation of normal tissue, the ratio of the volume of normal tissue with thermal damage to the volume of normal tissue without thermal damage is defined as the thermal damage volume ratio.

[0057] By constructing a three-dimensional finite element multi-physics simulation model for irreversible electroporation ablation of pancreatic tumors, it can simulate the tumor ablation situation and the damage situation of surrounding normal tissues during the ablation process, realize the pre-operative simulation of the ablation surgery process and results, and at the same time provide data support for the establishment of subsequent damage estimation models.

[0058] (3) The dataset establishment module is used to perform simulation calculations on the two numerical models constructed by the geometric model and numerical model setting module respectively, obtain the simulation results under different surgical parameter settings in different electrode structure models, collect and organize the simulation data of the two numerical models as their respective datasets, and divide the training set and test set of each simulation data respectively;

[0059] Specifically, the specific processing process of the dataset establishment module:

[0060] Step 1: Perform global definitions in the COMSOL Multiphysics software, define the parameter values of the ablation surgery, obtain the tumor cell electro-damage probability using the tumor cell electro-damage function calculation method in the geometric model and numerical model setting module, obtain the tissue thermal damage probability using the thermal damage calculation method in the geometric model and numerical model setting module, and set the material properties, set the probes in the simulation and determine the simulation results to be output.

[0061] Step 2: Establish the geometric structure of the simulation model according to the geometric structure of the simulation model construction module, which are the two-electrode plate type and three-electrode plate type simulation models. Set the positions and sizes of the pancreatic duct tissue, tumor tissue, and electrodes in this simulation model. At the same time, set the biological properties of each part, including the thermal conductivity, density, and relative permittivity of the corresponding tissues.

[0062] Step 3: Add a high-voltage pulsed electric field and a bio-thermal field. For the high-voltage pulsed electric field, set the electric potential and the boundary region. For the bio-thermal field, complete the relevant settings such as bio-heat and thermal damage. Analyze the interaction between multiple physical fields, select the grid density and accuracy level, and perform mesh generation.

[0063] Step 4: Set the research and solution processes in the COMSOL software. Use the numerical calculation functions in the geometric model and numerical model setting modules to calculate the tumor ablation rate and the thermal damage volume of normal tissues, and perform feature analysis on the visualization results and relevant output data of the tumor ablation and normal tissue damage conditions.

[0064] Step 5: After performing simulation calculations on the two-electrode and three-electrode numerical models constructed by the geometric model and numerical model setting modules respectively, collect and organize the obtained datasets. The surgical parameters in both datasets include the electrode spacing D, electrode length L, voltage amplitude U, pulse width W, and number of pulses N. The indicators reflecting the overall effect of the ablation results are the tumor ablation rate and the thermal damage volume ratio. The simulation dataset of the two-electrode type contains a total of 24,080 data, and the simulation dataset of the three-electrode type contains a total of 9,635 data. Both are divided into training sets and test sets in the ratio of 80% and 20% in sequence.

[0065] (4) Model training module: Use the datasets of the two numerical models obtained by the dataset establishment module, and train them respectively using the random forest regression algorithm, and optimize the hyperparameters to obtain the trained tumor irreversible electroporation ablation simulation prediction models corresponding to the two different electrode structures;

[0066] Specifically, in the model training module, the grid search method is specifically used to optimize the hyperparameters. The optimized parameters include the number of decision trees, the maximum depth of the tree, the minimum number of samples required to split the internal nodes, and the number of features to consider when finding the best split.

[0067] The construction of the damage estimation model uses the random forest regression algorithm. Randomly extract samples from the training set to obtain multiple new sub-training sets, and train multiple CART regression trees with the sub-training sets. During the training process, first randomly select multiple features from all features for each node, and then select the optimal splitting point from the features to perform the division of the sub-tree. The final prediction result of each CART regression tree is the mean value of the leaf node reached by the sample point, and the final prediction result of the random forest is the mean value of the prediction results of all CART regression trees.

[0068] The random forest regression algorithm, as a commonly used method for solving prediction problems, is applied to the construction process of the irreversible electroporation ablation damage estimation model for pancreatic tumors. The grid search method is used to optimize the hyperparameters in the model, solving the problem of long time consumption caused by using physical simulation in the current pre-operative planning scenario, and being able to accurately and quickly predict the ablation results under certain surgical parameter conditions.

[0069] (5) The simulation prediction module is used to predict the tumor ablation situation and the damage situation of the surrounding normal tissues corresponding to certain surgical parameter settings by using the trained irreversible electroporation ablation simulation prediction model for tumors.

[0070] The COMSOL software used in this embodiment is COMSOL Multiphysics version 5.6. In the three-dimensional finite element multi-physics field simulation model of irreversible electroporation ablation of pancreatic tumors, the pancreatic duct tissue is a horizontally arranged tube with an inner diameter of 0.2 cm, an outer diameter of 0.25 cm, and a length of 10 cm. The tumor tissue is an ellipsoid with a semi-axis a, semi-axis b, and semi-axis c values of 5 mm, 6 mm, and 6 mm respectively. The biological properties of each part, including the thermal conductivity, density, and relative permittivity of the corresponding tissue, etc., are shown in Table 1 for specific values. The value ranges of each surgical parameter under two electrode settings are shown in Tables 2 and 3. In the process of constructing the irreversible electroporation ablation damage estimation model for pancreatic tumors, in the parameter settings of the initial random forest regression model, the number of decision trees is 120, the maximum depth of the tree is 15, the minimum number of samples required for each division is 2, the minimum number of samples in the leaf node is 1, and parallel computing is not performed.

[0071] Table 1: Related parameters of the multi-physics field coupling model

[0072] Tissue property Pancreas Tumor Electrode Isobaric heat capacity J / (kg·K) 3164 3164 1000 Relative permittivity 6090 6090 1 Thermal conductivity W / (m·K) 0.521 0.521 250 <![CDATA[Density kg / m 3 > 1086.5 1086.5 2700

[0073] Table 2: Parameter value situations of the two-electrode plate model

[0074] Parameter name Symbol Unit Range Electrode spacing D mm 8-14 Electrode length L mm 8-14 Voltage amplitude U V 500-2000 Number of pulses N pcs 0-300 Pulse width W μs 25-100

[0075] Table 3: Parameter value situations of the three-electrode plate model

[0076] Parameter name Symbol Unit Range Electrode spacing D mm 8-10 Electrode length L mm 8-10 Voltage amplitude U V 500-2000 Number of pulses N pcs 0-300 Pulse width W μs 25-100

[0077] For the three-dimensional finite element multi-physics field simulation model of irreversible electroporation ablation of pancreatic tumors constructed in this embodiment, in the two-electrode plate simulation model, when the distance between the two electrodes is 10 mm, the length of the two treatment electrodes is 8 mm, the pulse voltage is 1000 V, the width of each pulse is 100, and the number of applied pulses is 130, the front views of the visualized electric field distribution, thermal damage distribution, and electric damage distribution are as Figure 3As shown. In the figure reflecting the electric field distribution, the legend range is 0 - 500 V / cm. The greater the electric field intensity, the more reddish the area. It can be seen from the figure that the electric field intensity is greater in the area closer to the electrode. According to research, when the electric field intensity exceeds 500 V / cm, electroporation occurs in the cell membrane. Therefore, under the conditions of this surgical parameter setting, some tumor cells closer to the electrode are electrically broken down, resulting in cell necrosis. In the figure reflecting the thermal damage of normal tissues, the legend range is 0 - 1, representing the probability of thermal damage. The area with a thermal damage probability value greater than 0.99 indicates the area where thermal damage occurs in normal tissues, corresponding to the red area in the figure. It can be seen from the figure that the normal tissues located between the positive and negative electrodes and close to the tumor tissue are thermally damaged earlier than the cell tissues in other positions. In the figure reflecting the ablation of tumor tissues, the legend range is 0 - 1, representing the probability of electrical damage. The area with an electrical damage probability value greater than 0.99 indicates the area where electrical damage occurs in tumor tissues, corresponding to the red area in the figure. It can be seen from the figure that the tumor area close to the electrode will be electrically damaged earlier.

[0078] In the three - electrode - plate type simulation model, when the distance between adjacent electrodes is 10 mm, the length of each of the three treatment electrodes is 8 mm, the pulse voltage is 1500 V, the width of each pulse is 25, and the number of applied pulses is 120, the front view of the visualization of the corresponding electric field distribution, thermal damage distribution, and electrical damage distribution is as Figure 4 shown. In the figure reflecting the electric field distribution, the legend range is 0 - 500 V / cm. The greater the electric field intensity, the more reddish the area. It can be seen from the figure that the electric field intensity is greater in the area between or near the electrodes, indicating that some tumor cells between or near the electrodes are electrically broken down, resulting in cell necrosis. In the figure reflecting the thermal damage of normal tissues, the legend range is 0 - 1, representing the probability of thermal damage. It can be seen from the red area in the figure that the normal tissues located between the positive and negative electrodes are thermally damaged earlier than the cell tissues in other positions. In the figure reflecting the ablation of tumor tissues, the legend range is 0 - 1, representing the probability of electrical damage. It can be seen from the red area in the figure that the tumor area close to the electrode will be electrically damaged earlier and the ablation will be completed earlier.

[0079] To verify the effectiveness of the irreversible electroporation ablation damage estimation model for pancreatic tumors in this embodiment, the goodness of fit is used as an evaluation index to represent the prediction error of the model. The prediction error of the thermal damage volume ratio is shown in Table 4, and the prediction error of the tumor ablation rate is shown in Table 5. It can be seen from the table that for the two-electrode model, the frequencies of the error rates of predicting the thermal damage volume ratio and the tumor ablation rate using the random forest regression model below 1% are 4,795 groups and 4,625 groups respectively, accounting for about 99% and 96% of the test set data, indicating that the error rates are concentrated below 1% and the prediction errors are small. Similarly, for the three-electrode model, the frequencies of the error rates of predicting the thermal damage volume ratio and the tumor ablation rate using the random forest regression model below 1% are 1,900 groups and 1,709 groups respectively, accounting for about 99% and 87% of the test set data, indicating that the prediction effect is good. Therefore, the prediction effect of the damage estimation model constructed using the random forest regression algorithm is good and can be used to predict the ablation result.

[0080] Table 4: Experimental results of predicting the thermal damage volume ratio

[0081] Error rate 0～1％ 1％～5％ Above 5% Frequency of two-electrode type (group) 4795 21 0 Frequency of three-electrode type (group) 1900 23 4

[0082] Table 5: Experimental results of predicting the tumor ablation rate

[0083] Error rate 0～1％ 1％～5％ Above 5% Frequency of two-electrode type (group) 4625 150 41 Frequency of three-electrode type (group) 1709 179 39

[0084] To further verify the effectiveness of the system in this embodiment, three common machine learning algorithms, namely the decision tree algorithm, the support vector machine regression algorithm, and the linear regression algorithm, are compared with the random forest regression algorithm. The performance effects are shown in Tables 6 and 7. It can be seen from the table that for the two-electrode model, the R 2 value of the random forest regression algorithm for predicting the thermal damage volume ratio is 0.999, while the R 2 values of the decision tree algorithm, the support vector machine regression algorithm, and the linear regression algorithm are 0.999, 0.586, and 0.71 respectively. For predicting the tumor ablation rate, the prediction R 2 value of the random forest regression algorithm is 0.987, while the R 2 values of the decision tree algorithm, the support vector machine regression algorithm, and the linear regression algorithm are 0.986, 0.907, and 0.543 respectively. The random forest regression algorithm has the best effect. Generally speaking, for the two-electrode structure, the random forest regression algorithm has the best prediction effect on the IRE ablation result of pancreatic tumors. For the three-electrode model, the R 2 value of the random forest regression algorithm for predicting the thermal damage volume ratio is 0.999, while the R 2The values are 0.999, 0.812, and 0.754 respectively. For predicting the tumor ablation rate, the predicted R of the random forest regression algorithm 2 value is 0.991, while the R values of the decision tree algorithm, support vector machine regression algorithm, and linear regression algorithm 2 are 0.983, 0.906, and 0.596 respectively. The random forest regression algorithm has the best effect. Generally speaking, for the three-electrode structure, the random forest regression algorithm has the best prediction effect on the IRE ablation results of pancreatic tumors. Therefore, under this electrode structure, it is feasible to establish an IRE ablation damage estimation model based on the random forest regression algorithm in this paper and can achieve a good prediction effect.

[0085] Table 6: Comparative experimental results of the thermal damage volume ratio

[0086]

[0087]

[0088] Table 7: Comparative experimental results of the tumor ablation rate

[0089] Random forest regression algorithm Decision tree algorithm Support vector machine regression algorithm Linear regression algorithm <![CDATA[Two - electrode - type R 2 value]]> 0.987 0.986 0.907 0.543 <![CDATA[Three - electrode - type R 2 value]]> 0.991 0.983 0.906 0.596

Claims

1. A simulation prediction system for irreversible electroporation ablation of tumors, characterized in that: Includes the following modules: A simulation model geometry building module is used to determine the parameters and catheter electrode structure related to irreversible electroporation ablation surgery for tumors, and to build the geometry of the simulation model; The geometric model and numerical model setting module is used to set the geometric model and numerical model to simulate the coupling effect of the electric field and the biological thermal field during the irreversible electroporation of tumors; three-dimensional finite element multi-physics field simulation models of irreversible electroporation ablation of tumors with two electrodes and three electrodes are constructed respectively, and the tumor ablation rate and the thermal damage volume ratio of normal tissues are calculated through numerical analysis; The data set establishment module is used to simulate and calculate the two numerical models constructed by the geometric model and numerical model setting modules respectively, obtain the simulation results under different surgical parameter settings in different electrode structure models, collect and organize the simulation data of the two numerical models as their respective data sets, and divide the simulation data into training sets and test sets respectively; A model training module is used to use the data sets of the two numerical models obtained by the data set establishment module, respectively train them using the random forest regression algorithm, optimize the hyperparameters, and obtain the trained irreversible electroporation ablation simulation prediction models of tumors corresponding to two different electrode structures; The simulation prediction module is used to use the trained tumor irreversible electroporation ablation simulation prediction model to predict the corresponding tumor ablation situation and surrounding normal tissue damage under certain surgical parameter settings.

2. The irreversible electroporation ablation simulation prediction system for tumors according to claim 1, characterized in that: In the simulation model geometry building module, the ablation surgery related parameters include electrode parameters and pulse parameters, the electrode parameters include electrode spacing and electrode length, and the pulse parameters include voltage amplitude, pulse width and pulse number; the catheter electrode structure includes a two-electrode sheet type and a three-electrode sheet type; The catheter electrode includes a two-electrode sheet-type electrode and a three-electrode sheet-type electrode. The two-electrode sheet-type catheter electrode includes a positive electrode and a negative electrode, and the three-electrode sheet-type includes a positive electrode and two negative electrodes. The geometric structure of the simulation model includes biological tissue and electrode structure, the biological tissue includes pancreatic duct tissue and tumor tissue; the electrode structure is a catheter electrode.

3. The irreversible electroporation ablation simulation prediction system for tumors according to claim 1, characterized in that: In the geometric model and numerical model setting module, the numerical analysis includes tumor electrical analysis and normal tissue thermal analysis; wherein: Tumor electrical analysis, including calculation of electric field distribution during irreversible electroporation, setting of tissue conductivity, and calculation of electrical damage to tumor cells; Thermal analysis of normal tissue, including temperature distribution calculation and thermal damage calculation.

4. The irreversible electroporation ablation simulation prediction system for tumors according to claim 3, characterized in that: In tumor electrophysiology: ① The electric field distribution during irreversible electroporation is represented by the conductivity of the tumor tissue, which is specifically calculated using the nonlinear Laplace equation: In the formula, is a differential operation, σ is the conductivity of the tumor tissue, the conductivity and the field strength affect each other, is the strength of the external voltage; ② The tissue conductivity setting specifically uses the tissue dynamic conductivity to calculate the pulse electric field distribution. The expression of the tissue dynamic conductivity is: σ(E)=σ init +(σ max -σ init )·exp(-exp(a·(EE IRE ))) Where σ(E) represents the dynamic conductivity of the tissue, σ init represents the initial conductivity of the tumor tissue, which is 0.341S / m; σ max represents the maximum conductivity of pancreatic and tumor tissues, which is 0.6138 S / m; a is set to 0.002 m / V; E is the tissue electric field strength, E IRE The minimum field strength required for irreversible electroporation of the cell membrane; ③ The Peleg-Fermi model was used to calculate the probability of electrical damage to tumor cells during irreversible electroporation ablation. The specific calculation process is as follows: First, the probability of tumor cell survival after electrical ablation is expressed by the following formula: Where S is the probability of tumor cell survival after electrical ablation, E is the magnitude of the electric field, and A determines the slope of the curve turning point; E c (n) is the critical value of cell death, A(n) represents the changes in the process of cell necrosis, and the specific expressions of the two functions are as follows: E c (n)=E0·exp(-k1n) A(n)=A0·exp(-k2n) Where E0, A0, k1 and k2 are regression coefficients, and n is the number of pulses; Secondly, the probability of tumor cell death due to the pulsed electric field is expressed as: P=1-S Again, the cell death probability P is used as the electrical damage probability of tumor cells, denoted as PEI; if the electrical damage probability value of a certain area of ​​tumor tissue is greater than 0.99, it is said that irreversible electrical damage has occurred in this area.

5. The irreversible electroporation ablation simulation prediction system for tumors according to claim 3, characterized in that: Thermal analysis of normal tissue: The Pennes bioheat equation is used to calculate the temperature distribution of various parts of the tissue during ablation. The equation is as follows: Where k is thermal conductivity, T is temperature, ω b is the blood perfusion rate, c b is the heat capacity of blood, T b is the arterial blood temperature, q″′ is metabolic heat production, ρ is density, c p is the tissue heat capacity, Joule heat Q Joule It is the physical heat generated during the ablation process; The Arrhenius formula is used to quantify the thermal damage of tissues. The degree of thermal damage to biological tissues is related to factors such as temperature and exposure time. k The rate constant for the transformation from the original state to the thermally damaged state at (t) is expressed as: Where Ω(t) is the dimensionless time-dependent accumulated thermal damage, E a is the required activation energy, R is the ideal gas constant, and T(t) is the temperature distribution over time; Afterwards, the thermal damage is calculated using the Arrhenius formula: In the formula, t init is the initial time point of the heating process, after which the probability of tissue thermal damage P is calculated using the following equation: P = 1-exp(-Ω(t)) In the formula, the probability of cell death P is the probability of thermal damage to normal tissue cells, denoted as PTI. When the probability of thermal damage in a certain area is greater than 0.99, it can be regarded as irreversible thermal damage. The ratio of the volume of normal tissue that has undergone thermal damage to the volume that has not undergone thermal damage is defined as the thermal damage volume ratio.

6. The irreversible electroporation ablation simulation prediction system for tumors according to claim 1, characterized in that: The specific processing process of the data set establishment module: Step 1. Perform global definition in COMSOL Multiphysics software, define the values ​​of various parameters of ablation surgery, use the tumor cell electrical damage function calculation method in the geometric model and numerical model setting module to obtain the probability of tumor cell electrical damage, use the thermal damage calculation method in the geometric model and numerical model setting module to obtain the probability of tissue thermal damage, set material properties, set the probe in the simulation and determine the simulation results to be output; Step 2, according to the simulation model geometry building module, the geometry of the simulation model is constructed, which are two-electrode sheet type and three-electrode sheet type simulation models, respectively, in which the positions and sizes of the pancreatic duct tissue, tumor tissue and electrodes are set, and at the same time, the biological properties of each part are set, including the thermal conductivity, density and relative dielectric constant of the corresponding tissue; Step 3: Add high-voltage pulse electric field and biothermal field. For high-voltage pulse electric field, set the electric potential and boundary area; for biothermal field, complete the related settings such as bioheat and thermal damage; analyze the interaction between multiple physical fields, select the grid density and accuracy level, and perform grid division; Step 4: Setting up the research and solution process in COMSOL software, using the geometric model and numerical model setting modules to calculate the tumor ablation rate and normal tissue thermal damage volume, and performing feature analysis on the visualization results and related output data of the tumor ablation and normal tissue damage conditions; Step 5: After simulation calculations are performed on the two-electrode sheet type and three-electrode sheet type numerical models constructed by the geometric model and numerical model setting module, data sets are collected and sorted. The two data sets are divided into a training set and a test set.

7. The irreversible electroporation ablation simulation prediction system for tumors according to claim 6, characterized in that: In step 5, the surgery-related parameters in the two data sets include electrode spacing D, electrode length L, voltage amplitude U, pulse width W and pulse number N, and the indicators reflecting the overall effect of the ablation results are the tumor ablation rate and thermal damage volume ratio.

Images