A tumor irreversible electroporation ablation procedure parameter optimization system

Abstract

The application discloses a tumor irreversible electroporation ablation operation parameter optimization system, comprising an initialization population module, an individual grade division module, a first generation sub-population generation module, a new population determination module, a new parent population establishment module, a new parent population update module, a merging module and a result selection module. The NSGA-II algorithm in the genetic algorithm is adopted to optimize the operation parameters, and a threshold setting is added. The system can assist doctors in formulating the best operation parameter scheme, so as to achieve the operation target of maximizing killing of tumor tissues, minimizing damage to normal tissues and minimizing the number of pulses.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of machine learning, and particularly relates to a tumor irreversible electroporation ablation operation parameter optimization system. BACKGROUND

[0002] Malignant tumor is a major disease that seriously threatens human health and is the most serious public health challenge currently faced by the world. Liver, biliary tract and pancreatic malignant tumors are the most common tumor types in the digestive system. In the current clinical diagnosis and treatment of malignant tumors, the most commonly used method is surgical resection, which is a radical treatment for solid tumors. However, due to the easy occurrence of local or distant invasion and metastasis of malignant tumors, the optimal surgical resection opportunity is often missed. Physical tumor ablation is another way to treat pancreatic cancer, which uses external energy to act on solid tumors to eradicate or destroy tumor tissue. It mainly includes three types of thermal ablation, cryoablation and electric ablation, but thermal ablation has the disadvantage of not being suitable for the vicinity of heat-sensitive organs, and cryoablation is prone to incomplete tumor ablation, leading to tumor recurrence. Irreversible electroporation technology is a new tumor electric ablation technology, which uses electric pulses to act on the surface of cancer cells to irreversibly damage the phospholipid bilayer, causing defects on the cell membrane surface. When the electric pulse reaches a certain level, the cell membrane defect cannot be reversed, ultimately causing the target cells to necrose. It belongs to non-thermal ablation technology and can avoid irreversible damage to adjacent tissue structures, preserving intracellular macromolecules and tissue scaffolds. Existing clinical studies have also confirmed that irreversible electroporation technology has a significant advantage in preserving organ function, especially for local ablation of small and weak organ tumors such as the pancreas.

[0003] For high-voltage pulse ablation surgery of tumors, how to achieve the surgical goal of "maximizing tumor tissue killing and minimizing normal tissue damage" has always been a hot issue. In order to achieve the best surgical effect, reasonable preoperative planning is crucial. However, the preoperative planning of ablation surgery at the present stage mainly relies on the clinical experience of doctors and is heavily dependent on the professional skill level of doctors, which is not conducive to developing the best ablation surgery plan and is prone to situations where cancer cells cannot be completely killed or "normal tissue is injured". In the irreversible electroporation ablation process, if the surgical parameters are not reasonable, it may result in too small an ablation range, incomplete killing of tumor tissue, thereby increasing the risk of tumor recurrence, or it may result in too large an ablation range, damaging the normal tissue around the tumor, which will increase the risk of bleeding and infection and cause other complications. At the same time, according to research, too many pulses will exacerbate cell electrolyte disorder and cause greater burden on patients, and the more the number of pulses, the longer the ablation treatment time will be, which may increase the risk of surgical trauma, infection and other complications. SUMMARY

[0004] In view of the fact that preoperative setting of operation parameters still depends on subjective experience of doctors, the present application aims to provide a tumor irreversible electroporation ablation operation parameter optimization system which can assist doctors in formulating an optimal operation parameter scheme to achieve the operation goal of maximizing killing of tumor tissues, minimizing damage to normal tissues and minimizing the number of pulses.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] A tumor irreversible electroporation ablation operation parameter optimization system comprises the following modules:

[0007] (1) An initialization population module for constructing a fitness function, determining an operation parameter optimization range, randomly initializing a population and randomly generating a certain number of individuals as an initialized population.

[0008] (2) An individual level division module for using a non-dominated sorting method to divide the initialized population obtained by the initialization population module into a plurality of different levels of Pareto frontiers and divide the individual levels in the initial population.

[0009] (3) A first generation sub-population generation module for performing selection, crossover and mutation operations on the individuals in the population obtained by the individual level division module to generate a first generation sub-population Q0.

[0010] (4) A new population determination module for merging the parent population P0 and the sub-population Q0 obtained by the first generation sub-population generation module into a new population R0.

[0011] (5) A new parent population establishment module for using a fast non-dominated sorting method to divide the individuals in the new population R0 obtained by the new population determination module into different order Pareto frontiers, using an elite strategy, adding the individual levels in sequence to a new parent population according to a dominance level order until the individual levels cannot all be added to the new parent population;

[0012] (6) A new parent population updating module for calculating the crowding distance of the individuals in the level in which the individuals obtained by the new parent population establishment module cannot all be added to the new parent population, adding the individuals with larger crowding distances to the new parent population until a specified number of population individuals is reached.

[0013] (7) A merging module for performing selection, crossover and mutation operations on the individuals in the new parent population obtained by the new parent population updating module to generate a sub-population, merging the parent population and the sub-population as a new population to replace the new population R0 obtained by the new population determination module and entering the next cycle, repeating the new parent population establishment module, the new parent population updating module and the merging module until the number of iterations reaches a specified number.

[0014] (8) a result selection module configured to compare the Pareto frontiers obtained in all sub-iterations with a set threshold value, and output the first 10 Pareto frontiers that meet the threshold value as the optimization results of the surgical parameter settings for the surgeon to select according to clinical practice.

[0015] Further, in the initialization population module, the fitness function is set according to the tumor irreversible electroporation ablation operation target, including maximizing the tumor ablation rate, minimizing the normal tissue damage and minimizing the number of pulses applied in the operation process;

[0016] The surgical 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 optimization range of the surgical parameters is represented by lb and rb, lb is a set of minimum values of the values of each surgical parameter, and rb is a set of maximum values of the values of each surgical parameter, the elements in the set are the minimum or maximum values of electrode spacing, electrode length, voltage amplitude, pulse width and pulse number in turn;

[0017] The initialization method uses random initialization, specifically: a 200x5 matrix is constructed, the elements in the matrix are randomly generated decimal numbers between 0 and 1, the matrix is multiplied by two matrices obtained by subtracting lb and rb from the matrix respectively, and finally added with lb to obtain an initialization population P0 with a population size of 200;

[0018] Further, in the individual level division module, in the determination of the first layer of non-dominated frontiers, each individual is compared with all other individuals in turn to determine the dominance relationship, and in the determination of the second layer of non-dominated frontiers, the individuals other than those belonging to the first layer of non-dominated frontiers are again subjected to dominance relationship determination, and the above process is repeated to obtain the levels of all individuals;

[0019] Further, in the first generation sub-population generation module, the selection operation uses a binary tournament selection method, two individuals are randomly selected from the population each time, and the better individual is selected into the sub-population according to the fitness value, and the operation is repeated multiple times until the size of the sub-population is the same as that of the parent population;

[0020] The mutation operation uses a polynomial mutation method, and the mutation form is:

[0021] x 1j (t)=x 1j (t)+Δ j

[0022] In the formula,

[0023] Wherein, u is a random number between 0 and 1, The distribution factor is set to 1;

[0024] The crossover operation uses an analog binary crossover operator, assuming two parent individuals and Two offspring individuals generated using the crossover operator and can be calculated by the following formula:

[0025]

[0026] In the formula,

[0027] where u is a random number between 0 and 1, The distribution factor is set to 1.

[0028] Further, in the merging operation of the new population determination module, the number of individuals in the offspring population and the parent population is consistent, and the number of individuals in the new population is the sum of the offspring population and the parent population. The number of initialized populations is 200, therefore, the number of individuals in the offspring population and the parent population is 200, and the number of individuals in the new population is 400;

[0029] Further, in the new parent population establishment module, the fast non-dominated sorting method operates as follows:

[0030] First, calculate the number of individuals dominated and dominating each individual in the new population R0 obtained by the new population determination module and form a set. Next, establish a set of individuals that are not dominated by other individuals, and set the rank of this set to 1. Then, form a set of remaining individuals dominated by the individual, and subtract 1 from the number of individuals dominated by each individual, if the number of individuals dominated after subtraction is 0, then it is one of the elements of the set of new individuals that are not dominated by other individuals. Finally, repeat the above operation to divide the sets of different ranks, i.e. to divide the Pareto front into different orders;

[0031] The elite strategy is to merge the parent population P t and the offspring population Q t into a new population R t , and then the individuals in the parent population P t and the offspring population Q t participate in competition. The individuals in the population are subjected to fast non-dominated sorting and crowding distance calculation, the individuals with low rank enter the next generation, then the crowding distance is calculated, the larger ones enter the next generation first, until the number of the next generation reaches a certain number;

[0032] Further, the crowding distance in the new parent population updating module is calculated by calculating the crowding distance of the individual in the same level with its two closest individuals, and then finding the optimal individual according to the crowding distance size, and the expression is as follows:

[0033]

[0034] In the formula, And f i k k

[0035] Further, the maximum number of iterations specified in the merging module is 200.

[0036] Further, the threshold condition in the result selection module is that the thermal damage volume ratio needs to be less than 20%, and the tumor ablation rate needs to be greater than 75%, and the top 10 Pareto solution set meeting the threshold condition can be selected as the output result.

[0037] Compared with the prior art, the present application has the following technical effects:

[0038] The present application aims at the problem that the current preoperative setting of surgical parameters still relies on the subjective experience of doctors, and in order to avoid the serious consequences of incomplete ablation or damage to adjacent tissues caused by improper setting of surgical parameters, the irreversible electroporation ablation surgery effect is optimized, and the best surgical plan needs to be made before the operation, and the most important thing is to determine the value of each surgical parameter. The system of the present application optimizes the surgical parameters by using the NSGA-II algorithm in the genetic algorithm, and sets the threshold value: the tumor ablation rate needs to be greater than 75%, and the thermal damage volume ratio needs to be less than 20%, which better screens the parameter combination most suitable for the operation condition, and the doctor can select a parameter combination from it as the IRE ablation surgical parameter value according to the clinical actual situation. In summary, the result obtained by the present application can assist the doctor to make the best surgical parameter plan, so as to achieve the surgical goal of maximizing the killing of tumor tissue, minimizing the damage to normal tissue and minimizing the number of pulses. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 It is the electrode structure geometric diagram in the test of the present application;

[0040] Figure 2 It is the parameter optimization result visualization schematic diagram of the two-electrode sheet type model in the test of the present application;

[0041] Figure 3 It is the parameter optimization result visualization schematic diagram of the three-electrode sheet type model in the test of the present application. DETAILED DESCRIPTION

[0042] The specific content of the present application is further explained in detail below in combination with examples.

[0043] Taking high-pressure pulse ablation surgery of pancreatic tumor as an example, how to achieve the surgical goal of “maximizing tumor tissue killing and minimizing normal tissue damage” has always been a hot issue. In order to achieve the best surgical effect, reasonable preoperative planning is crucial. However, the preoperative planning of ablation surgery at the present stage mainly depends on the clinical experience of doctors and is seriously dependent on the professional skill level of doctors, which is not conducive to developing the best ablation surgery plan and is prone to the situation of not being able to completely kill cancer cells or “injure” normal tissues. Therefore, the present embodiment adopts the NSGA-II algorithm with threshold to optimize the surgical parameters, obtains the parameter combination meeting the requirement of surgical effect, and assists the clinical doctors in decision-making, so as to achieve precise ablation of pancreatic tumor.

[0044] A parameter optimization method for irreversible electroporation ablation of pancreatic tumor based on NSGA-II algorithm with threshold, comprising the following steps:

[0045] (1) An initialization population module is used to construct an fitness function, determine the optimization range of surgical parameters, randomly initialize the population, and randomly generate a certain number of individuals as the initialized population.

[0046] The fitness function is set according to the surgical goal of irreversible electroporation ablation of pancreatic tumor, including maximizing tumor ablation rate, minimizing normal tissue damage and minimizing the number of pulses applied in the surgical process.

[0047] The surgical 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 optimization range of surgical parameters is represented by lb and rb, lb is the set of minimum values of each surgical parameter value, and rb is the set of maximum values of each surgical parameter value, the set lb is [500, 8, 8, 25, 0], the set rb is [2000, 10, 10, 100, 300], and the elements in the set are the minimum or maximum values of electrode spacing, electrode length, voltage amplitude, pulse width and pulse number in turn;

[0048] The initialized population uses a random initialization method, specifically: a 200x5 matrix is constructed, the elements in the matrix are randomly generated decimals between 0 and 1, the two matrices obtained by subtracting lb and rb from the decimals are multiplied, and finally lb is added to obtain an initialized population P0 with a population size of 200;

[0049] In this module, the fitness function is set to make the optimization target more clear and explicit, and the determination of the parameters related to the operation and the optimization range can make the optimization object more explicit. The random initialization method is used to generate the initial population, which is an important start for the subsequent population iteration and can help the algorithm show better performance.

[0050] The individual level division module is used to divide the initial population obtained by the initial population module into multiple different levels of Pareto frontiers using a non-dominated sorting method, and divide the individual levels in the initial population. Specifically, the following operations are included:

[0051] The non-dominated sorting method needs to compare each individual with all other individuals in turn to determine the dominance relationship when determining the first layer of non-dominated frontiers. When determining the second layer of non-dominated frontiers, the individuals other than those belonging to the first layer of non-dominated frontiers are judged again for dominance relationship, and the above process is repeated to obtain all individual levels.

[0052] In this module, the non-dominated sorting can obtain the Pareto frontiers and corresponding levels of each individual in the initial population, laying a foundation for generating the first generation of sub-population.

[0053] The first generation of sub-population generation module is used to select, cross and mutate the individuals in the population obtained by the individual level division module to generate the first generation of sub-population Q0.

[0054] The selection operation uses a binary tournament selection method, which randomly selects two individuals from the population each time, selects the better individual according to the fitness value, and repeats the operation multiple times until the size of the sub-population is the same as that of the parent population.

[0055] The mutation operation uses a polynomial mutation method, and the mutation form is:

[0056] x 1j (t)=x 1j (t)+Δ j

[0057] In the formula,

[0058] where u is a random number between 0 and 1, is a distribution factor, which is set to 1.

[0059] The crossover operation uses a simulated binary crossover operator. Assuming that two parent individuals and The two offspring individuals generated by the crossover operator are and are calculated by the following formula:

[0060]

[0061] In the formula,

[0062] Wherein, u is a random number between 0-1, The distribution factor is set to 1.

[0063] In this module, according to the fitness evaluation, the individuals with high fitness are selected for the next generation breeding to generate new individuals, and the crossover and mutation operations can increase the ability of algorithm to explore information, which is also the core operation of genetic algorithm.

[0064] The new population determination module is used to combine the parent population P0 and the child population Q0 obtained by the first generation sub-population generation module into a new population R0.

[0065] In the merging operation, the number of individuals of the child population and the parent population is consistent, and the number of individuals of the new population is the sum of the child population and the parent population, and the number of initialized population is 200, therefore, the number of individuals of the child population and the parent population is 200, and the number of individuals of the new population is 400;

[0066] In this module, by merging the parent population and the child population to compete for the new generation population, it is beneficial to retain excellent individuals to a greater extent and improve the performance of the surgical parameter optimization method;

[0067] The new parent population establishment module is used to divide the individuals in the new population determination module into different order Pareto frontiers by using the fast non-dominated sorting method, and adopt the elite strategy to add the individuals of each level to the new parent population in turn according to the dominance level order until the level individuals cannot be added to the new parent population.

[0068] The fast non-dominated sorting method first calculates the number of individuals dominated and dominated by each individual and forms a set. Then, a set of individuals not dominated by other individuals is established, and the level of this set is set to 1. Then, the remaining individuals dominated by the individual form a set, and the number of individuals dominated by each individual is reduced by 1, if the number of individuals dominated after reduction is 0, then it is one of the elements of the new set of individuals not dominated by other individuals. Finally, repeat the above operation to divide into different order Pareto frontiers as the processing object of the elite strategy;

[0069] The above-mentioned elite strategy is to combine the obtained parent population P t and the child population Q t into a new population R t , and then the parent population P t and the child population Q tThe individuals in the population participate in competition together. Non-dominated sorting calculation and crowding distance calculation are performed, and the individuals with low ranks enter the next generation. Then, the crowding distance is calculated, and the individuals with larger ranks enter the next generation until the number of the next generation reaches a certain number.

[0070] In this step, fast non-dominated sorting greatly reduces the calculation complexity, and the ranks of the individuals in the population are obtained. The elite strategy can accelerate the execution of the surgical parameter optimization method and help to retain better individuals.

[0071] Step 6: The crowding distance of the individuals in the ranks of the new parent population obtained in step 5 is calculated, and the individuals with larger crowding distances are added to the new parent population until the number of the population individuals reaches a specified number.

[0072] The crowding distance is calculated by calculating the crowding distance of the adjacent two individuals in the same rank, and then the optimal individual is found according to the crowding distance size, and the expression is as follows:

[0073]

[0074] In the formula, and are the target values of the adjacent two individuals of individual i in the kth target, indicates the crowding distance of individual i in the kth target f k .

[0075] In this step, the calculation of the crowding distance is a comparison method between the individuals in the population, which enables the solutions on the Pareto front to be more evenly distributed, and helps to maintain the diversity of the population.

[0076] Step 7: The individuals in the new parent population obtained in step 6 are selected, crossed and mutated to generate a child population, and the parent population and the child population are merged to enter the next cycle until the number of iterations reaches a specified number.

[0077] The specified number is 200 times.

[0078] Step 8: The Pareto front obtained in all iterations is compared with a set threshold, and the top 10 solution sets that meet the threshold requirement are selected as the optimization results of the surgical parameter setting. The optimization results can be selected by the doctor according to the clinical practice.

[0079] The set threshold is that the thermal injury volume ratio is less than 20% and the tumor ablation rate is greater than 75%. The top 10 Pareto solution sets that meet the threshold condition can be selected as the output results.

[0080] The present embodiment is implemented using Pycharm software, and is carried out under two irreversible electroporation ablation catheter electrode structures, including two electrode sheet type structures and three electrode sheet type structures, as shown in Figure 1 Fig. 1, wherein the pancreatic duct tissue is a tube with an inner diameter of 0.2 cm, an outer diameter of 0.25 cm, and a length of 10 cm arranged horizontally, the tumor tissue is an ellipsoid with a half axis a, a half axis b, and a half axis c of 5 mm, 6 mm, and 6 mm respectively, and the red cylinder represents the positive electrode and the yellow cylinder represents the negative electrode. As can be seen from the figure, both are catheter electrodes, and the electrodes are arranged parallel to the lumen tissue, simulating the ablation of the catheter electrode after entering the pancreatic duct through the natural cavity of the human body. The difference is that the two-electrode sheet type includes one positive electrode and one negative electrode, and the two electrodes are symmetrically placed about the ellipsoid representing the tumor, while the three-electrode sheet type includes one positive electrode and two negative electrodes, and the positive electrode is located at the middle position of the simulation model, and the two negative electrodes are symmetrically divided on both sides of the positive electrode.

[0081] For the two-electrode sheet type model, the results of the irreversible electroporation ablation surgery parameter optimization are shown in Table 1. In the table, the surgery parameters include electrode spacing, electrode length, voltage amplitude, pulse width, and pulse number, and the thermal damage volume and tumor ablation rate are the damage volume of the lumen tissue and the volume ratio of the tumor ablation during the irreversible electroporation process corresponding to the left parameters. At the same time, the output results are visualized, and the Pareto frontier solution is relatively dispersed, and more optimal solution sets can be found in the parameter space without being concentrated in a certain area, and the results are shown in Figure 2 .

[0082] Table 1 Parameter optimization results of two-electrode sheet type model

[0083]

[0084] As can be seen from Table 1, the surgery parameter combination obtained by using the NSGA-II algorithm with threshold meets the threshold requirement, but there is no parameter combination in which the tumor ablation rate reaches 100% and the thermal damage volume ratio is 0%. The doctor can select the best surgery parameter combination according to the clinical practice, for example, if the ablation surgery goal is more inclined to achieve the minimum damage to normal tissue, the doctor can select the parameter combination with serial number 10 in the table; if the ablation surgery goal is more inclined to achieve complete killing of tumor tissue, the doctor can select the parameter combination with serial number 1 in the table; if the ablation surgery goal is more inclined to achieve a smaller pulse number, the doctor can select the parameter combination with serial number 10 in the table.

[0085] For the three-electrode sheet structure model, the NSGA-II algorithm with threshold is used to optimize the parameters of the irreversible electroporation ablation of pancreatic tumor, and the algorithm output results are shown in Table 2, and the corresponding Pareto solution set and the visualization results of the frontier are as shown inFigure 3 The results are shown in Table 2.

[0086] Table 2 Parameter optimization results of three-electrode patch type model

[0087]

[0088] From Table 2, it can be seen that the parameter combination obtained by the irreversible electroporation ablation parameter optimization method of pancreatic tumor using the NSGA-II algorithm with threshold meets the threshold requirement, and there is no parameter combination in which the tumor ablation rate reaches 100% while the thermal damage volume ratio reaches 0%. However, the parameter combination with serial number 4 in the table has the largest tumor ablation rate and the smallest thermal damage volume ratio. Therefore, if the doctor pays more attention to the electrical thermal damage, the parameter combination with serial number 4 can be selected; if the doctor prefers to achieve a smaller number of pulses, the surgical parameter combination with serial number 9 or 10 in the table can be selected.

Claims

1. A tumor irreversible electroporation ablation procedure parameter optimization system, comprising: The method comprises the following modules: An initialization population module is configured to build a fitness function, determine a surgical parameter optimization range, and randomly initialize a population and generate a certain number of individuals as an initialized population; An individual level division module is configured to use a non-dominated sorting method to divide the initialized population obtained by the initialization population module into a plurality of different levels of Pareto frontiers, and divide the individual levels in the initial population; The first generation sub-population generating module is configured to perform selection, crossover and mutation operations on the individuals in the population obtained by the individual level division module to generate a first generation sub-population ; a new population determination module for merging the parent population and the offspring population into a new population and the offspring population ;​ The new parent population establishing module is used to divide the individuals in the new population obtained by the new population determining module into different order Pareto frontiers by using a fast non-dominated sorting method, and to add the individuals of each level into the new parent population in turn according to the order of the domination level by using an elite strategy until the individuals of each level cannot all be added into the new parent population. ​ A new parent population updating module is configured to calculate the crowding distance of the individuals in the level in which the individuals obtained by the new parent population establishment module cannot all be added to the new parent population, and add the individuals with larger crowding distances to the new parent population until the number of individuals in the new parent population reaches a predetermined number; The merging module is configured to perform selection, crossover and mutation operations on the individuals of the new parent population obtained by the new parent population updating module, to generate a child population, and to merge the parent population and the child population as a new population to replace the new population obtained by the new population determining module The next cycle is entered, and the new parent population establishing module, the new parent population updating module and the merging module are repeatedly executed until the iteration number reaches a specified number. A result selection module is configured to compare the Pareto frontiers obtained in all iterations with a set threshold, and output the first several Pareto frontiers that meet the threshold as the optimization results of the surgical parameter settings. In the initialization population module, the fitness function is set according to the tumor irreversible electroporation ablation operation target, and includes maximizing the tumor ablation rate, minimizing the normal tissue damage, and minimizing the number of pulses applied in the operation process. The surgical parameters include electrode parameters and pulse parameters, the electrode parameters include an electrode spacing and an electrode length, and the pulse parameters include a voltage amplitude, a pulse width, and a pulse number. The surgical parameter optimization range is represented by lb and rb, lb is a set of minimum values of the values of the surgical parameters, and rb is a set of maximum values of the values of the surgical parameters, and the elements in the sets are minimum values or maximum values of the electrode spacing, the electrode length, the voltage amplitude, the pulse width, and the pulse number in sequence. The initialization method adopts random initialization, specifically: a matrix of size is constructed, the elements of which are randomly generated decimals between 0 and 1, which are multiplied by two matrices obtained by subtracting rb and lb respectively, and finally added by lb to obtain an initialization population with a population size of 200 . ​ 2. The tumor irreversible electroporation ablation parameter optimization system of claim 1, wherein, In the individual level division module, in the determination of the first layer of non-dominated frontiers, each individual is compared with all other individuals in sequence to determine the dominance relationship, and in the determination of the second layer of non-dominated frontiers, the individuals other than the individuals in the first layer of non-dominated frontiers are again subjected to the dominance relationship determination, and the above process is repeated to obtain all individual levels.

3. The tumor irreversible electroporation ablation parameter optimization system of claim 1, wherein, In the first generation sub-population generation module, the selection operation uses a binary tournament selection method, two individuals are randomly selected from the population each time, and the better individual is selected into the sub-population according to the fitness value, and the operation is repeated multiple times until the size of the sub-population is the same as that of the parent population. The mutation operation uses a polynomial mutation method, and the mutation form is as follows: In the formulae, wherein is a random number between 0 and 1, is a distribution factor, set to a value of 1; The crossover operation uses an analog binary crossover operator, assuming two parent individuals and Two offspring individuals generated using the crossover operator and can be calculated by the following equation: In the formulae, wherein is a random number between 0 and 1, is a distribution factor, set to a value of 1.

4. The tumor irreversible electroporation ablation parameter optimization system of claim 1, wherein, In the merging operation of the new population determination module, the number of individuals in the parent population and the sub-population is consistent, and the number of individuals in the new population is the sum of the number of individuals in the parent population and the sub-population.

5. The tumor irreversible electroporation ablation parameter optimization system of claim 1, wherein, In the new parent population establishment module, the fast non-dominated sorting method operates as follows: First, the new population determined by the new population calculation module Each individual is dominated and the number of dominated individuals and constitutes a set; then, the set of individuals not dominated by other individuals is established, and the level of this set is set to 1; then, the remaining individuals dominated by the individual form a set, and the number of dominated individuals of each individual is reduced by 1, if the number of dominated individuals after reduction is 0, it is one of the elements of the new set of individuals not dominated by other individuals; finally, repeat the above operation to divide the sets of different levels, that is, to divide the Pareto front of different orders; The elite strategy is to combine the parent population with the child population into a new population , and then the individuals in the parent population and the child population participate in competition; the individuals in the population are quickly non-dominated sorted and crowded distance calculated, the individuals with low ranks enter the next generation, then the crowded distance is calculated, the larger ones prefer to enter the next generation until the number of the next generation reaches a certain number.

6. The tumor irreversible electroporation ablation parameter optimization system of claim 1, wherein, The crowding distance in the new parent population updating module is calculated by calculating the crowding distance of the individual and the two individuals closest to the individual in the same level, and then the optimal individual is found according to the crowding distance, and the expression is as follows: wherein and are the target values of the individual i On the k target, the target value of the individual adjacent to the individual i On the k target , the crowding distance of the individual 7. The tumor irreversible electroporation ablation parameter optimization system of claim 1, wherein, The maximum number of iterations specified in the merging module is 200.

8. The tumor irreversible electroporation ablation parameter optimization system of claim 1, wherein, The threshold condition in the result selection module is that the volume of thermal damage is less than 20% of the volume of the tumor and the tumor ablation rate is greater than 75%. The top 10 Pareto solutions that meet the threshold condition are selected as the output results.

Images