A volumetric modulation intensity modulation optimization method and system based on multi-task evolution

Abstract

The present invention provides a volumetric rotation intensity modulation optimization method and system based on multi-task evolution, including: inputting parameters, importing radiotherapy data; constructing a constrained multi-objective optimization model, taking the motion constraints of the multi-leaf collimator leaves as the constraints of the optimization problem, taking the prescription dose requirements of the target area and surrounding organs at risk as the optimization goals, and converting it into a constrained multi-objective optimization problem containing two optimization goals through modeling; constructing a multi-task evolution framework; guiding the constrained multi-objective optimization based on the multi-task evolution method: screening out several volumetric rotation intensity modulation treatment plans optimized by the multi-task evolution algorithm, each treatment plan automatically outlines the target area and organs at risk in the treatment plan, visualizes the dose of the organ and compares it with the clinical requirements for doctors to choose. The present invention guides the constrained multi-objective optimization based on the multi-task evolution method, thereby obtaining multiple treatment plans that meet the leaf motion constraints and meet the clinical dose requirements for doctors to choose.

Description

Technical Field

[0001] The present invention relates to the technical field of volumetric rotation intensity modulation therapy, and in particular to a volumetric rotation intensity modulation optimization method based on multi-task evolution and a system thereof. Background Art

[0002] There are currently three main ways to treat cancer, namely radiotherapy, surgery and chemotherapy. Volumetric modulation of radiation is a new type of radiotherapy technology that is widely used due to its high treatment efficiency. During treatment, the linear accelerator rotates around the human body for irradiation, and the leaves of the multi-leaf collimator need to move to block the rays as the accelerator rotates to form a corresponding irradiation field. Through multiple irradiation fields of different intensities and shapes, a highly conformal dose distribution is provided to kill tumor cells while avoiding irradiation of surrounding normal cells. Compared with traditional radiotherapy techniques, volumetric modulation of radiation uses less treatment time, so it can reduce the side effects caused by motion errors, scattered doses and radiation during treatment.

[0003] However, volumetric modulated radiation optimization is a complex optimization problem, which involves a large number of optimization parameters and physical constraints such as the movement speed of the multi-leaf collimator leaves. Many scholars have studied the optimization of volumetric modulated radiation at home and abroad, such as heuristic algorithms such as simulated annealing and genetic algorithms, and non-heuristic algorithms such as gradient descent. However, the volumetric modulated radiation problem contains thousands of variables related to shape and intensity, which makes it difficult for the algorithm to converge. In addition, due to the physical limitations of the multi-leaf collimator leaves, the generated plan may be invalid and require a lot of computing resources to repair. The effectiveness and efficiency of existing algorithms are still unsatisfactory, and there is still much room for improvement in the research on plan diversity and the final clinical radiotherapy effect. Summary of the invention

[0004] To solve the above problems, the present invention aims to propose a volumetric rotation intensity modulation optimization method and system based on multi-task evolution. By guiding the constrained multi-objective optimization based on the multi-task evolution method, multiple treatment plans that meet the blade motion constraints and meet the clinical dose requirements are obtained for doctors to choose.

[0005] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0006] A volumetric rotation intensity modulation optimization method based on multi-task evolution includes the following steps:

[0007] S1: Input linear accelerator speed , the maximum movement speed of the leaves of the multi-leaf collimator , and calculate the maximum distance that the blade can move when the machine rotates 1°, import the radiotherapy data, including the contribution matrix D, the doctor's prescription dose constraints for the target area and the prescription dose constraints for the surrounding organs at risk;

[0008] S2: Construct a constrained multi-objective optimization model, take the motion constraints of the multi-leaf collimator leaves as the constraints of the optimization problem, take the prescription dose requirements of the target area and surrounding organs at risk as the optimization objectives, and transform it into a constrained multi-objective optimization problem with two optimization objectives through modeling, including the motion constraints CV of the multi-leaf collimator leaves and the maximum dose , minimum dose , Maximum dose volume constraint and minimum dose volume constraints ;

[0009] S3: Construct a multi-task evolutionary framework, taking the complex original optimization problem as the main task and constructing a low-dimensional related auxiliary task. The two tasks are optimized by the main population and the auxiliary population respectively, and knowledge transfer is performed during the evolution process with a preset probability;

[0010] S4: Guided constrained multi-objective optimization based on multi-task evolutionary method;

[0011] S5: Screen out several volumetric intensity modulated therapy plans optimized by a multi-task evolutionary algorithm. Each treatment plan automatically outlines the target area and organs at risk in the treatment plan, visualizes the organ dose and compares it with clinical requirements for doctors to choose.

[0012] Furthermore, the movement distance constraint of the multi-leaf collimator leaves is calculated based on the maximum movement speed of the leaves and the machine speed. The contribution matrix D represents the dose deposition of each voxel in the patient's body by a beam of unit intensity under each irradiation field in the treatment plan. The prescription dose constraints of the target area and surrounding organs at risk are given by clinicians.

[0013] Further, the direction of the beam is divided: the coplanar beam angle is initialized as , divide several equally spaced beam directions according to the contribution matrix;

[0014] (1)

[0015] In the formula represents the minimum interval angle of beam division, Represents a single irradiation field;

[0016] Assuming that the patient's body is divided into several voxels and the irradiation field is divided into several small grids, calculate the dose of a single voxel v in volumetric rotation intensity modulation The method is:

[0017] (2)

[0018] In the formula, represents the number of irradiation fields, represents the set of all grids in the ith irradiation field, represents the dose contribution of the g-th grid to voxel v under unit conditions, represents the radiation intensity of the i-th irradiation field, Indicates whether the g-th grid of the i-th irradiation field is blocked by the leaves of the multi-leaf collimator.

[0019] Furthermore, the constrained multi-objective optimization model is constructed as follows: according to the relationship between the objective functions, the prescription dose constraint function is divided into two categories, and combined with the leaf position constraint, it is converted into a constrained multi-objective optimization model with two objectives:

[0020] Min F (3)

[0021] (4)

[0022]

[0023] (5)

[0024] (6)

[0025] Where X represents a solution, which is composed of all leaf position variables and irradiation field intensity variables; PTV represents the set of all target volumes, and OAR represents the set of all organs at risk. Only when the CV value is 0, the solution X is a feasible solution. represents the number of rows of leaves in each irradiation field, represents the rth row of leaves in the mth irradiation field, It represents the rotation angle of the machine when reaching the next irradiation field. CLM represents the maximum distance that the blade can move when the machine rotates 1°. Indicates the maximum distance the blade can move when the machine moves to the next irradiation field; Function Contains Functions and function, where The function is used to constrain the minimum dose of each voxel in the region to be no less than the predefined dose. The function is used to constrain the proportion of voxels in the region that are below the clinically set dose level; Function Contains Functions and function, where The function is used to constrain the maximum dose of each voxel in the region to be no higher than the predefined dose. The function is used to constrain the proportion of voxels in the region that are above the clinically set dose level.

[0026] Furthermore, the maximum dose , minimum dose , Maximum dose volume constraint , minimum dose volume constraint The objective function is as follows:

[0027] = (7)

[0028] = (8)

[0029] MinDVH =

[0030] (9)

[0031] MaxDVH =

[0032] (10)

[0033] In the formula The minimum dose for each voxel in region V is constrained to be no less than the predefined dose. , It is used to constrain the maximum dose of each voxel in the region V to be no higher than the predefined dose. ,MinDVH The proportion of tumor voxels used to constrain region V to receive a dose lower than the clinically set dose level is represents the minimum dose at ratio p, MaxDVH represents the proportion of voxels in the constraint region V that receive doses higher than the clinically set dose level, where Indicated in proportion The maximum dose.

[0034] Furthermore, the construction of the multi-task evolutionary framework specifically refers to: taking the complex original task as the main task, generating simple auxiliary tasks by deleting some variables at equal intervals, and restoring the auxiliary tasks to the main task through linear interpolation; using the potential information of the auxiliary tasks to promote the evolution of the main task, and the main task also helps the auxiliary task to search for the missed feasible domain, so as to achieve the common evolution of the two tasks; the two tasks are optimized by the main population and the auxiliary population respectively, and the populations promote each other through knowledge transfer.

[0035] Furthermore, the multi-task evolutionary method-based constrained multi-objective optimization specifically includes the following steps:

[0036] Step 1: When performing knowledge transfer, compare the non-dominated relationship of feasible solutions of the main population and the auxiliary population, and construct a solution set to be transferred;

[0037] Step 2: The solution set to be migrated guides the population to perform segment crossover and mutation;

[0038] Step 3: The solution of each population is considered as a feasible solution according to the constraint value equal to 0 in S3. The feasible solution selects offspring according to the fitness size. The fitness size of the feasible solution is obtained by the two optimization objectives transformed in step S3. The solution with a small fitness value enters the next generation;

[0039] Step 4: If the number of iterations reaches the set maximum number of iterations, the optimization ends; otherwise, go to step 1 to continue the optimization.

[0040] Furthermore, in step 1: the set of solutions to be migrated is determined by comparing the non-dominated situations of feasible solutions of the two populations; Constraint technology, when the solution constraint value is less than is considered as a feasible solution, and relaxing the constraints helps the population retain solutions with better evolutionary prospects; when comparing solutions reconciliation If one of the following conditions is met, then the solution is called untie :

[0041] 1) CV( )< , CV( )< ,

[0042] 2) CV( )= , CV( )= ,

[0043] 3) CV( )> , CV( )> , CV( ) <CV( )

[0044] in Representative, solution Pareto superior solution , set a , the update formula is as follows:

[0045] = (11)

[0046] (12)

[0047] The initial value is , The constraint violation ranking in the population is The constraint violation value of the solution is The value gradually decreases until the number of evolutions k reaches Generation, through After the generation, The value is zero; compare the non-dominated cases of feasible solutions of the two populations to determine the set of solutions to be migrated.

[0048] Furthermore, in step 2: the solution set to be migrated guides the population to perform knowledge migration in a fragment crossover manner; in the process of fragment crossover, the position variables of the same leaf in different irradiation fields are "bundled" together to perform crossover in a fragment manner;

[0049] In step 3: select according to the fitness of each solution in the population,

[0050] The fitness of solution X is obtained by adding the original fitness value and the density estimate. The solution with a small fitness value continues to enter the next generation. The specific formula is as follows:

[0051] (13)

[0052] In the formula Represents the original fitness value of solution X, representing The individual who dominates the solution X The sum of all individuals dominated, It represents the density estimate of the solution X, which represents the inverse of the distance between individual X and its kth nearest neighbor as the density estimate.

[0053] In order to achieve the above object, the present invention also provides a volumetric rotation intensity modulation optimization system based on multi-task evolution, comprising the following modules:

[0054] Data input module: used to input the linear accelerator speed , the maximum movement speed of the leaves of the multi-leaf collimator , and calculate the maximum distance that the blade can move when the machine rotates 1°, import the radiotherapy data, including the contribution matrix D, the doctor's prescription dose constraints for the target area and the prescription dose constraints for the surrounding organs at risk;

[0055] Constrained multi-objective optimization module: Construct a constrained multi-objective optimization model, take the motion constraints of the multi-leaf collimator leaves as the constraints of the optimization problem, take the prescription dose requirements of the target area and surrounding organs at risk as the optimization objectives, and transform it into a constrained multi-objective optimization problem with two optimization objectives through modeling, including the motion constraints CV of the multi-leaf collimator leaves and the maximum dose , minimum dose , Maximum dose volume constraint and minimum dose volume constraints ;

[0056] Multi-task evolution module: used to build a multi-task evolution framework, taking the complex original optimization problem as the main task and building a low-dimensional related auxiliary task. The two tasks are optimized by the main population and the auxiliary population respectively, and knowledge transfer is performed during the evolution process with a preset probability;

[0057] Population evolution module: guides constrained multi-objective optimization based on multi-task evolution method;

[0058] Data output module: used to screen out several volumetric rotation intensity modulated treatment plans optimized by multi-task evolutionary algorithms. Each treatment plan automatically outlines the target area and organs at risk in the treatment plan, visualizes the organ dose and compares it with clinical requirements for doctors to choose.

[0059] Beneficial effects: The present invention utilizes a multi-task evolutionary framework to improve the effectiveness of constrained multi-objective optimization; wherein, the multi-task framework constructs a low-dimensional related auxiliary task based on the complex original task; the complex original task is used as the main task, and an auxiliary task that is easy to evolve is constructed by deleting variables. The two tasks are optimized by the main population and the auxiliary population respectively, and the populations promote each other through knowledge transfer. By guiding the constrained multi-objective optimization based on the multi-task evolutionary method, multiple treatment plans that meet the blade motion constraints and meet the clinical dose requirements are obtained for doctors to choose. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] 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:

[0061] Figure 1 A flowchart of the steps of the volumetric rotation intensity modulation optimization method based on multi-task evolution according to an embodiment of the present invention;

[0062] Figure 2 The main algorithm flow chart of the volumetric rotation intensity modulation optimization method based on multi-task evolution described in an embodiment of the present invention;

[0063] Figure 3It is a process diagram of a knowledge transfer method based on segment crossover in a volumetric rotation intensity modulation optimization method based on multi-task evolution according to an embodiment of the present invention;

[0064] Figure 4 It is a schematic diagram of the structure of the volumetric modulation intensity optimization system based on multi-task evolution described in an embodiment of the present invention. DETAILED DESCRIPTION

[0065] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0066] Example 1

[0067] See also Figure 1-3 :A volumetric rotation intensity modulation optimization method based on multi-task evolution includes the following steps:

[0068] S1: Input linear accelerator speed , the maximum movement speed of the leaves of the multi-leaf collimator , and calculate the maximum distance that the blade can move when the machine rotates 1°, import the radiotherapy data, including the contribution matrix D, the doctor's prescription dose constraints for the target area and the prescription dose constraints for the surrounding organs at risk;

[0069] S2: Construct a constrained multi-objective optimization model, take the motion constraints of the multi-leaf collimator leaves as the constraints of the optimization problem, take the prescription dose requirements of the target area and surrounding organs at risk as the optimization objectives, and transform it into a constrained multi-objective optimization problem with two optimization objectives through modeling, including the motion constraints CV of the multi-leaf collimator leaves and the maximum dose , minimum dose , Maximum dose volume constraint and minimum dose volume constraints ;

[0070] S3: Construct a multi-task evolutionary framework, taking the complex original optimization problem as the main task and constructing a low-dimensional related auxiliary task. The two tasks are optimized by the main population and the auxiliary population respectively, and knowledge transfer is performed during the evolution process with a preset probability;

[0071] S4: Guided constrained multi-objective optimization based on multi-task evolutionary method;

[0072] S5: Screen out several volumetric intensity modulated therapy plans optimized by a multi-task evolutionary algorithm. Each treatment plan automatically outlines the target area and organs at risk in the treatment plan, visualizes the organ dose and compares it with clinical requirements for doctors to choose.

[0073] This embodiment uses a multi-task evolutionary framework to improve the effectiveness of constrained multi-objective optimization; wherein, the multi-task framework constructs a low-dimensional related auxiliary task based on the complex original task; the complex original task is used as the main task, and an auxiliary task that is easy to evolve is constructed by deleting variables. The two tasks are optimized by the main population and the auxiliary population respectively, and the populations promote each other through knowledge transfer. By guiding the constrained multi-objective optimization based on the multi-task evolutionary method, multiple treatment plans that meet the blade motion constraints and meet the clinical dose requirements are obtained for doctors to choose.

[0074] In a specific example, the movement distance constraint of the multi-leaf collimator leaves is calculated based on the maximum movement speed of the leaves and the machine rotation speed. The contribution matrix D represents the dose deposition of each voxel in the patient's body by a beam of unit intensity under each irradiation field in the treatment plan. The prescription dose constraints of the target area and surrounding organs at risk are given by clinicians.

[0075] It should be noted that the movement distance of the leaves of the multi-leaf collimator of this embodiment is constrained to be the maximum distance that the leaves can move within an irradiation field.

[0076] In a specific example, the beam direction is divided into: the initial coplanar beam angle is , divide several equally spaced beam directions according to the contribution matrix;

[0077] (1)

[0078] In the formula represents the minimum interval angle of beam division, Represents a single irradiation field;

[0079] Assuming that the patient's body is divided into several voxels and the irradiation field is divided into several small grids, calculate the dose of a single voxel v in volumetric rotation intensity modulation The method is:

[0080] (2)

[0081] In the formula, represents the number of irradiation fields, represents the set of all grids in the ith irradiation field, represents the dose contribution of the g-th grid to voxel v under unit conditions, represents the radiation intensity of the i-th irradiation field, Indicates whether the g-th grid of the i-th irradiation field is blocked by the leaves of the multi-leaf collimator.

[0082] In a specific example, the constrained multi-objective optimization model is constructed as follows: according to the relationship between the objective functions, the prescription dose constraint function is divided into two categories, and combined with the leaf position constraint, it is converted into a constrained multi-objective optimization model with two objectives:

[0083] Min F (3)

[0084] (4)

[0085]

[0086] (5)

[0087] (6)

[0088] Where X represents a solution, which is composed of all leaf position variables and irradiation field intensity variables; PTV represents the set of all target volumes, and OAR represents the set of all organs at risk. Only when the CV value is 0, the solution X is a feasible solution. represents the number of rows of leaves in each irradiation field, represents the rth row of leaves in the mth irradiation field, It represents the rotation angle of the machine when reaching the next irradiation field. CLM represents the maximum distance that the blade can move when the machine rotates 1°. Indicates the maximum distance the blade can move when the machine moves to the next irradiation field; Function Contains Functions and function, where The function is used to constrain the minimum dose of each voxel in the region to be no less than the predefined dose. The function is used to constrain the proportion of voxels in the region that are below the clinically set dose level; Function Contains Functions and function, where The function is used to constrain the maximum dose of each voxel in the region to be no higher than the predefined dose. The function is used to constrain the proportion of voxels in the region that are above the clinically set dose level.

[0089] In one embodiment, the maximum dose , minimum dose , Maximum dose volume constraint , minimum dose volume constraint The objective function is as follows:

[0090] = (7)

[0091] = (8)

[0092] MinDVH =

[0093] (9)

[0094] MaxDVH =

[0095] (10)

[0096] In the formula The minimum dose for each voxel in region V is constrained to be no less than a predefined dose. , It is used to constrain the maximum dose of each voxel in the region V not to exceed the predefined dose. ,MinDVH The proportion of tumor voxels used to constrain region V to receive a dose lower than the clinically set dose level is represents the minimum dose at ratio p, MaxDVH represents the proportion of voxels in the constraint region V that receive doses higher than the clinically set dose level, where Indicated in proportion The maximum dose.

[0097] In a specific example, the construction of a multi-task evolutionary framework specifically refers to: taking the complex original task as the main task, generating simple auxiliary tasks by deleting some variables at equal intervals, and restoring the auxiliary tasks to the main task through linear interpolation; the auxiliary tasks have fewer variables and relatively loose constraints, and are easy to optimize and find feasible solutions faster. In addition, the variables of the auxiliary task solutions are highly correlated with the variables of the main task solutions. The potential information of the auxiliary tasks is used to promote the evolution of the main task, and the main task also helps the auxiliary task search for missed feasible domains to achieve the common evolution of the two tasks; the two tasks are optimized using the main population and the auxiliary population respectively, and the populations promote each other through knowledge transfer.

[0098] In a specific example, the multi-task evolutionary method-based constrained multi-objective optimization specifically includes the following steps:

[0099] Step 1: When performing knowledge transfer, compare the non-dominated relationship of feasible solutions of the main population and the auxiliary population, and construct a solution set to be transferred;

[0100] Step 2: The solution set to be migrated guides the population to perform segment crossover and mutation;

[0101] Step 3: The solution of each population is considered as a feasible solution according to the constraint value equal to 0 in S3. The feasible solution selects offspring according to the fitness size. The fitness size of the feasible solution is obtained by the two optimization objectives transformed in step S3. The solution with a small fitness value enters the next generation;

[0102] Step 4: If the number of iterations reaches the set maximum number of iterations, the optimization ends; otherwise, go to step 1 to continue the optimization.

[0103] In a specific example, in step 1: determining the set of solutions to be migrated by comparing the non-dominated situations of feasible solutions of two populations; using Constraint technology, when the solution constraint value is less than is considered as a feasible solution, and relaxing the constraints helps the population retain solutions with better evolutionary prospects; when comparing solutions reconciliation If one of the following conditions is met, then the solution is called untie :

[0104] 1) CV( )< , CV( )< ,

[0105] 2) CV( )= , CV( )= ,

[0106] 3) CV( )> , CV( )> , CV( ) <CV( )

[0107] in Representative, solution Pareto superior solution , set a , the update formula is as follows:

[0108] = (11)

[0109] (12)

[0110] The initial value is , The constraint violation ranking in the population is The constraint violation value of the solution is The value gradually decreases until the number of evolutions k reaches Generation, through After the generation, The value is zero; compare the non-dominated cases of feasible solutions of the two populations to determine the set of solutions to be migrated.

[0111] This embodiment provides a method for determining a set of solutions to be migrated based on the non-dominated situation of feasible solutions of two populations; The constraint technology relaxes the solution constraints, and then compares the non-dominated relationship of feasible solutions of two populations, and takes all non-dominated solutions as the solution set to be migrated.

[0112] In a specific example, in step 2: the solution set to be migrated guides the population to perform knowledge migration in a fragment crossover manner; in the process of fragment crossover, the position variables of the same leaf in different irradiation fields are "bundled" together and crossover is performed in a fragment manner to reduce the probability of generating a constraint violation solution; the specific operation process is as follows Figure 3 As shown, it represents the position of the same leaf under different irradiation fields. First, the solution of the auxiliary population is linearly interpolated. For the same leaf, the position variables of different irradiation fields are "bundled" together. The fragment size is 4 (dark part). Several fragments are randomly selected for crossover. The probability of violating physical constraints is reduced by crossover of fragments.

[0113] In step 3: select according to the fitness of each solution in the population,

[0114] The fitness of solution X is obtained by adding the original fitness value and the density estimate. The solution with a small fitness value continues to enter the next generation. The specific formula is as follows:

[0115] (13)

[0116] In the formula Represents the original fitness value of solution X, representing The individual who dominates the solution X The sum of all individuals dominated, represents the density estimate of the solution X, and represents the inverse of the distance between individual X and its kth nearest neighbor as the density estimate; formula (13) is used to ensure the convergence and diversity of the algorithm during optimization.

[0117] Example 2

[0118] To achieve the above purpose, see Figure 4 This embodiment also provides a volumetric rotation intensity modulation optimization system based on multi-task evolution, including the following modules:

[0119] Data input module: used to input the linear accelerator speed , the maximum movement speed of the leaves of the multi-leaf collimator , and calculate the maximum distance that the blade can move when the machine rotates 1°, import the radiotherapy data, including the contribution matrix D, the doctor's prescription dose constraints for the target area and the prescription dose constraints for the surrounding organs at risk;

[0120] Constrained multi-objective optimization module: Construct a constrained multi-objective optimization model, take the motion constraints of the multi-leaf collimator leaves as the constraints of the optimization problem, take the prescription dose requirements of the target area and surrounding organs at risk as the optimization objectives, and transform it into a constrained multi-objective optimization problem with two optimization objectives through modeling, including the motion constraints CV of the multi-leaf collimator leaves and the maximum dose , minimum dose , Maximum dose volume constraint and minimum dose volume constraints ;

[0121] Multi-task evolution module: used to build a multi-task evolution framework, taking the complex original optimization problem as the main task and building a low-dimensional related auxiliary task. The two tasks are optimized by the main population and the auxiliary population respectively, and knowledge transfer is performed during the evolution process with a preset probability;

[0122] Population evolution module: guides constrained multi-objective optimization based on multi-task evolution method;

[0123] Data output module: used to screen out several volumetric rotation intensity modulated treatment plans optimized by multi-task evolutionary algorithms. Each treatment plan automatically outlines the target area and organs at risk in the treatment plan, visualizes the organ dose and compares it with clinical requirements for doctors to choose.

[0124] The volumetric modulation intensity modulation optimization system based on multi-task evolution of this embodiment has the same advantages as the volumetric modulation intensity modulation optimization method based on multi-task evolution over the prior art, which will not be described in detail here.

[0125] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A volumetric modulation intensity modulation optimization method based on multi-task evolution, characterized in that: The process includes the following steps: S1: Input linear accelerator speed , the maximum movement speed of the leaves of the multi-leaf collimator , and calculate the maximum distance that the blade can move when the machine rotates 1°, import the radiotherapy data, including the contribution matrix D, the doctor's prescription dose constraints for the target area and the prescription dose constraints for the surrounding organs at risk; S2: Construct a constrained multi-objective optimization model, take the motion constraints of the multi-leaf collimator leaves as the constraints of the optimization problem, take the prescription dose requirements of the target area and surrounding organs at risk as the optimization objectives, and transform it into a constrained multi-objective optimization problem with two optimization objectives through modeling, including the motion constraints CV of the multi-leaf collimator leaves and the maximum dose , minimum dose , Maximum dose volume constraint and minimum dose volume constraints ; The constrained multi-objective optimization model is specifically constructed as follows: according to the relationship between the objective functions, the prescription dose constraint function is divided into two categories, and combined with the leaf position constraint, it is converted into a constrained multi-objective optimization model with two objectives: My F (3) （4） （5） （6） Where X represents a solution, which is composed of all leaf position variables and irradiation field intensity variables; PTV represents the set of all target volumes, and OAR represents the set of all organs at risk. Only when the CV value is 0, the solution X is a feasible solution. represents the number of rows of leaves in each irradiation field, represents the rth row of leaves in the mth irradiation field, It represents the rotation angle of the machine when reaching the next irradiation field. CLM represents the maximum distance that the blade can move when the machine rotates 1°. Indicates the maximum distance the blade can move when the machine moves to the next irradiation field; Function Contains Functions and function, where The function is used to constrain the minimum dose of each voxel in the region to be no less than the predefined dose. The function is used to constrain the proportion of voxels in the region that are below the clinically set dose level; Function Contains Functions and function, where The function is used to constrain the maximum dose of each voxel in the region to be no higher than the predefined dose. The function is used to constrain the proportion of voxels in the region that are above the clinically set dose level; S3: Construct a multi-task evolutionary framework, taking the complex original optimization problem as the main task and constructing a low-dimensional related auxiliary task. The two tasks are optimized by the main population and the auxiliary population respectively, and knowledge transfer is performed during the evolution process with a preset probability; S4: Guided constrained multi-objective optimization based on multi-task evolutionary method; S5: Screen out several volumetric intensity modulated therapy plans optimized by a multi-task evolutionary algorithm. Each treatment plan automatically outlines the target area and organs at risk in the treatment plan, visualizes the organ dose and compares it with clinical requirements for doctors to choose.

2. The multi-task evolution-based volumetric modulation optimization method according to claim 1, characterized in that: The movement distance constraint of the multi-leaf collimator leaves is calculated based on the maximum movement speed of the leaves and the machine rotation speed. The contribution matrix D represents the dose deposition of each voxel in the patient's body by a beam of unit intensity under each irradiation field in the treatment plan. The prescription dose constraints of the target area and surrounding organs at risk are given by clinicians.

3. The multi-task evolution-based volumetric modulation optimization method according to claim 2, characterized in that: Beam direction division: Initialize the coplanar beam angle to , divide several equally spaced beam directions according to the contribution matrix; （1） In the formula represents the minimum interval angle of beam division, Represents a single irradiation field; Assuming that the patient's body is divided into several voxels and the irradiation field is divided into several small grids, calculate the dose of a single voxel v in volumetric rotation intensity modulation The method is: （2） In the formula, represents the number of irradiation fields, represents the set of all grids in the ith irradiation field, represents the dose contribution of the g-th grid to voxel v under unit conditions, represents the radiation intensity of the i-th irradiation field, Indicates whether the g-th grid of the i-th irradiation field is blocked by the leaves of the multi-leaf collimator.

4. The multi-task evolution-based volumetric modulation optimization method according to claim 1, characterized in that: The maximum dose , minimum dose , Maximum dose volume constraint , minimum dose volume constraint The objective function is as follows: = (7) = (8) MyDVH = (9) MaxDVH = (10) In the formula The minimum dose for each voxel in region V is constrained to be no less than the predefined dose. , It is used to constrain the maximum dose of each voxel in the region V to be no higher than the predefined dose. ,MinDVH The proportion of tumor voxels used to constrain region V to receive a dose lower than the clinically set dose level is represents the minimum dose at ratio p, MaxDVH represents the proportion of voxels in the constraint region V that receive doses higher than the clinically set dose level, where Indicated in proportion The maximum dose.

5. The multi-task evolution-based volumetric modulation optimization method according to claim 1, characterized in that: The construction of the multi-task evolutionary framework specifically refers to: taking the complex original task as the main task, generating simple auxiliary tasks by deleting some variables at equal intervals, and restoring the auxiliary tasks to the main task through linear interpolation; using the potential information of the auxiliary tasks to promote the evolution of the main task, and the main task also helps the auxiliary task to search for the missed feasible domain, so as to achieve the common evolution of the two tasks; the two tasks are optimized by the main population and the auxiliary population respectively, and the populations promote each other through knowledge transfer.

6. The multi-task evolution-based volumetric modulation optimization method according to claim 1, characterized in that: The multi-objective optimization guided by the multi-task evolution method specifically includes the following steps: Step 1: When performing knowledge transfer, compare the non-dominated relationship of feasible solutions of the main population and the auxiliary population, and construct a solution set to be transferred; Step 2: The solution set to be migrated guides the population to perform segment crossover and mutation; Step 3: The solution of each population is considered as a feasible solution according to the constraint value equal to 0 in S3. The feasible solution selects offspring according to the fitness size. The fitness size of the feasible solution is obtained by the two optimization objectives transformed in step S3. The solution with a small fitness value enters the next generation; Step 4: If the number of iterations reaches the set maximum number of iterations, the optimization ends; otherwise, go to step 1 to continue the optimization.

7. The multi-task evolution-based volumetric modulation optimization method according to claim 6, characterized in that: In step 1: determine the set of solutions to be migrated by comparing the non-dominated situations of feasible solutions of the two populations; adopt Constraint technology, when the solution constraint value is less than is considered as a feasible solution, and relaxing the constraints helps the population retain solutions with better evolutionary prospects; when comparing solutions reconciliation If the dominant relationship of satisfies one of the following conditions, then the solution is called untie : 1) CV( )< , CV( )< , 2) CV( )= , CV( )= , 3) CV( )> , CV( )> , CV( )<CV( ) in Representative, solution Pareto superior solution , set a , the update formula is as follows: = (11) (12) The initial value is , The constraint violation ranking in the population is The constraint violation value of the solution is The value gradually decreases until the number of evolutions k reaches Generation, through After the generation, The value is zero; compare the non-dominated cases of feasible solutions of the two populations to determine the set of solutions to be migrated.

8. The multi-task evolution-based volumetric modulation optimization method according to claim 6, characterized in that: In step 2: the solution set to be migrated guides the population to perform knowledge migration in a fragment crossover manner; in the process of fragment crossover, the position variables of the same leaf in different irradiation fields are "bundled" together to perform crossover in a fragment manner; In step 3: selecting according to the fitness of each solution in the population; The fitness of solution X is obtained by adding the original fitness value and the density estimate. The solution with a small fitness value continues to enter the next generation. The specific formula is as follows: (13) In the formula Represents the original fitness value of solution X, representing The individual who dominates the solution X The sum of all individuals dominated, It represents the density estimate of the solution X, which represents the inverse of the distance between individual X and its kth nearest neighbor as the density estimate.

9. A volumetric modulation intensity modulation optimization system based on multi-task evolution, characterized in that: Includes the following modules: Data input module: used to input the linear accelerator speed , the maximum movement speed of the leaves of the multi-leaf collimator , and calculate the maximum distance that the blade can move when the machine rotates 1°, import the radiotherapy data, including the contribution matrix D, the doctor's prescription dose constraints for the target area and the prescription dose constraints for the surrounding organs at risk; Constrained multi-objective optimization module: Construct a constrained multi-objective optimization model, take the motion constraints of the multi-leaf collimator leaves as the constraints of the optimization problem, take the prescription dose requirements of the target area and surrounding organs at risk as the optimization objectives, and transform it into a constrained multi-objective optimization problem with two optimization objectives through modeling, including the motion constraints CV of the multi-leaf collimator leaves and the maximum dose , minimum dose , Maximum dose volume constraint and minimum dose volume constraints ; The constrained multi-objective optimization model is specifically constructed as follows: according to the relationship between the objective functions, the prescription dose constraint function is divided into two categories, and combined with the leaf position constraint, it is converted into a constrained multi-objective optimization model with two objectives: My F (3) （4） （5） （6） Where X represents a solution, which is composed of all leaf position variables and irradiation field intensity variables; PTV represents the set of all target volumes, and OAR represents the set of all organs at risk. Only when the CV value is 0, the solution X is a feasible solution. represents the number of rows of leaves in each irradiation field, represents the rth row of leaves in the mth irradiation field, It represents the rotation angle of the machine when reaching the next irradiation field. CLM represents the maximum distance that the blade can move when the machine rotates 1°. Indicates the maximum distance the blade can move when the machine moves to the next irradiation field; Function Contains Functions and function, where The function is used to constrain the minimum dose of each voxel in the region to be no less than the predefined dose. The function is used to constrain the proportion of voxels in the region that are below the clinically set dose level; Function Contains Functions and function, where The function is used to constrain the maximum dose of each voxel in the region to be no higher than the predefined dose. The function is used to constrain the proportion of voxels in the region that are above the clinically set dose level; Multi-task evolution module: used to build a multi-task evolution framework, taking the complex original optimization problem as the main task and building a low-dimensional related auxiliary task. The two tasks are optimized by the main population and the auxiliary population respectively, and knowledge transfer is performed during the evolution process with a preset probability; Population evolution module: guides constrained multi-objective optimization based on multi-task evolution method; Data output module: used to screen out several volumetric rotation intensity modulated treatment plans optimized by multi-task evolutionary algorithms. Each treatment plan automatically outlines the target area and organs at risk in the treatment plan, visualizes the organ dose and compares it with clinical requirements for doctors to choose.

Images