Close-range radiotherapy plan optimization method and device based on cumulative dose target

Abstract

The invention discloses a close-range radiotherapy plan optimization method and device based on a cumulative dose target. The method comprises the following steps: S1, importing current treatment data; s2, leading in a preorder treatment image and performing deformation registration to the current image; s3, mapping the preorder treatment dose to the current treatment image to serve as the background dose of the current treatment; s4, converting the dose data mapped to the current treatment image by the preorder treatment into a bioequivalent dose; s5, on the basis of the sketching data of the current treatment image and with reference to the clinical endangered organ dose limit value requirement, key area generation is carried out on each endangered organ; s6, based on the preorder treatment dosage, generating a corresponding organ-endangered high-risk dosage area and a target area insufficient area according to each organ-endangered dosage limit value and the target area prescription dosage; and S7, optimizing the treatment plan, and optimizing the treatment plan by taking the preorder treated dose and the sum of the planned dose to be optimized at the time to meet the clinical requirement as a target.

Description

Technical Field

[0001] The present invention relates to the technical field of radiotherapy plan optimization, and in particular to a method and device for optimizing a brachytherapy plan based on a cumulative dose target. Background Art

[0002] In recent years, brachytherapy (BT) has received widespread attention and application as an effective tumor therapy. Brachytherapy is also called internal irradiation, which corresponds to external irradiation (teleradiation). This technology irradiates the patient's tumor site by implanting a packaged radioactive source directly into the patient's tumor site through an applicator or a delivery catheter. Since the intensity of the radioactive source used for brachytherapy is relatively low and the effective treatment distance of the radiation is relatively short, most of its energy is absorbed by the tumor tissue. According to the inverse square law, the dose distribution around the radiation source decreases with the square of the distance from the radiation source. Therefore, in clinical applications, brachytherapy is usually not used alone, but as an adjuvant treatment for external irradiation, so as to give a higher dose at a specific site, thereby improving the local control rate of the tumor and effectively improving the patient's quality of life.

[0003] For example, the standard treatment regimen for cervical cancer typically includes the combination of external beam radiotherapy (EBRT) and high-dose-rate brachytherapy (HDR BT). HDR BT is typically administered in four to six treatment sessions. The application method and patient anatomy can vary significantly between treatments. Therefore, a CT (or MRI) scan is performed before each treatment to develop a treatment plan. Treatment plans are independently optimized and evaluated for each image, without considering previous treatments. However, while this approach can achieve optimal local dose distribution, it cannot achieve optimal overall dose distribution. To ensure local tumor control and protect surrounding healthy tissue, precise dose accumulation is crucial. Research and application of precise dose stacking techniques can significantly improve treatment precision, thereby enhancing patient survival and quality of life. The combined use of external beam radiotherapy and brachytherapy has become the standard modality for cervical cancer radiotherapy. In this case, the total radiation dose is composed of the external beam radiation dose and the brachytherapy dose. The total dose to the target volume and organs at risk (OARs) obtained directly through dose superposition is based on the assumption that the dose distribution area of ​​each dose evaluation indicator is consistent during each treatment fraction. However, this method fails to fully consider the heterogeneity of organ deformation and dose distribution, and thus cannot reflect the true total dose. Therefore, it cannot fully utilize the dosimetric advantages of combined EBRT+BT radiotherapy. In addition, it is not convenient to evaluate the total dose to the target volume and organs at risk (OARs).

[0004] In recent years, deformable registration has been widely used to assess the cumulative dose in radiotherapy for cervical cancer and has become an important tool. However, during radiotherapy, dose accumulation in the pelvic region is particularly challenging due to tumor shrinkage, changes in bladder and rectal filling during fractionated treatment, and the need to implant an applicator during afterloading brachytherapy. Because tumor shrinkage, organ filling changes, intestinal gas, and the presence of BT applicators and vaginal packing lead to large and complex deformations, the accuracy of deformable registration still faces great challenges. At the same time, the current cumulative dose is mainly used for evaluation and is not considered as a base dose in the optimization design of current and subsequent treatment plans. The independent optimization of fractionated plans makes it impossible to coordinate the cumulative dose effects of previous EBRT and previous BT fractions, which in turn causes the dose hotspots of endangered organs such as the bladder and rectum to overlap, thereby increasing the risk of grade 3-5 urinary / gastrointestinal toxicity (the 5-year incidence rate is as high as 6.8%-8.5%). In addition, the pelvic anatomy undergoes complex deformation due to tumor shrinkage and organ filling. The registration errors of these changes (such as the Hausdorff distance can reach 12.06 mm) further increase the difficulty of cumulative dose calculation.

[0005] Clinical research projects are currently exploring cumulative dose-based brachytherapy plan optimization. Qi Fu et al. retrospectively analyzed nine patients with cervical cancer who received EBRT followed by high-dose-rate BT. For each BT treatment, deformable registration technology was used to accumulate the dose distributions of the previous EBRT and BT. These cumulative dose distributions were then imported into a customized commercial BT treatment planning system as the precursor dose for adaptive dose optimization. This study compared the differences in key target and organ-at-risk (OAR) dosimetric parameters between adaptive BT (ABT) and conventional BT (CBT) planning methods. The results showed that this approach can reduce OAR doses and improve the total dose distribution of combined radiotherapy, demonstrating promising clinical application prospects. Typically, the optimization target for CBT plans can be set based on the recommended dose limits for a single BT fraction. However, due to the presence of background dose, the optimization target for ABT plans should consider more constraints, including the dose limits for the individual BT fractions and the cumulative dose limits. Specifically, this includes measures to address target cold spots and hot spots in the target volume, as well as dose limits that approach or exceed the OARs during the preceding BT treatments.

[0006] In addition, in recent years, a number of automated optimization algorithms have emerged to improve the efficiency of single-time plans. For example, patent CN202210067145A uses a particle swarm-genetic hybrid algorithm to automatically optimize the dwell time and dose parameters, thereby shortening the manual parameter adjustment time and reducing plan differences; at the same time, multi-objective Bayesian optimization can generate a Pareto optimal solution within 1 minute, accelerating the adjustment of penalty weights. However, in the context of existing technologies, further research and development of more effective algorithms and methods are still needed to achieve more efficient optimization of brachytherapy.

[0007] In general, the existing technology still has the following shortcomings:

[0008] (1) The current cumulative dose is mainly used for evaluation and is not considered as a baseline dose in the optimization design of current and subsequent treatment plans. Independent optimization of fractionation plans makes it impossible to coordinate the cumulative dose effects of previous EBRT and previous BT fractions, resulting in overlapping dose hotspots in organs at risk, such as the bladder and rectum, thereby increasing the risk of organ-threatening toxicity.

[0009] (2) Current treatment plan optimization uses automated optimization algorithms, such as particle swarm optimization-genetic algorithm (PSO-GA), which mainly focus on optimizing the physical dose of a single fraction and fail to effectively incorporate the cumulative biological effect (Equivalent Dose in 2Gy fractions, EQD2). This is a significant flaw because the response of tumor cells to radiotherapy depends not only on the dose of a single treatment, but also on the long-term cumulative dose and its biological effect. The assessment of biological effects helps to more accurately predict treatment outcomes and side effects. Therefore, ignoring the cumulative biological effect may lead to an incomplete treatment plan, affect the local control effect of the tumor, and increase the treatment risk for patients.

[0010] (3) Optimization targets are set based on the recommended dose limits for BT therapy. Currently, these targets are targeted at the entire target volume or the entire organ at risk. However, in practice, the areas where optimization targets are difficult to meet are mainly located in the vicinity of the target volume and the organ at risk. If optimization targets can be set based on more precise regions, the efficiency of plan optimization will be improved, and a satisfactory plan will be more easily obtained.

[0011] (4) Currently, commercial TPS only has plan optimization functions for single-fraction treatment. Plan optimization based on cumulative dose targets, in addition to considering the dose limit of a single treatment, also requires setting constraints for the dose limits determined previously, including the dose limit of the current BT treatment and the cumulative dose limit. In particular, it requires countermeasures for target dose cold spots, hot spots, and organs at risk that are close to or exceed the dose limit in the previous BT treatment. These are not considered in current plan optimization technologies. Summary of the Invention

[0012] In response to the above technical problems, the present invention proposes a method and device for optimizing brachytherapy plans based on cumulative dose targets. The present invention significantly improves the accuracy of target doses and the protection of organs at risk by realizing the dynamic conversion of previous treatment doses to optimization constraints. The key area constraint generation method based on the characteristics of brachytherapy is reflected in the overall optimization objective function as an independent objective function, so that the efficiency of plan optimization can be improved through weight adjustment, and the ideal treatment plan can be quickly obtained. The weighted dynamic constraint mechanism ensures that the constraint strength is dynamically adjusted when the anatomy changes, thereby avoiding invalid optimization and ensuring that the target dose meets the standard. The multi-objective deep coupling optimization engine effectively combines single optimization with cumulative targets, improves optimization efficiency, and reduces the risk of toxicity to organs at risk such as the rectum. Overall, the present invention provides a more efficient, accurate and safe solution for brachytherapy.

[0013] In order to achieve the above-mentioned purpose, the technical solution of the present invention provides a method for optimizing brachytherapy plans based on cumulative dose targets, which includes the following steps: S1: importing the current treatment data, including the current treatment image and the corresponding delineation data; S2: deformation registration of the previous image, importing the previous treatment image and the corresponding delineation data, and performing deformation registration with the current treatment image; S3: mapping the previous treatment dose, based on the deformation registration result of the previous treatment image and the current treatment image, mapping the previous treatment dose to the current treatment image as the background dose of this treatment; S4: biological equivalent dose conversion, converting the previous treatment dose to the current treatment image; The dose data of the sequential treatment mapped to the current treatment image is converted into a biological equivalent dose; S5: key area constraint generation, based on the delineation data of the current treatment image and referring to the clinical dose limit requirements of the organs at risk, the key area is generated for each organ at risk; S6: predecessor dose constraint generation, based on the previous treatment dose, for each organ at risk dose limit and target area prescription dose, the corresponding organ at risk high-risk dose area and target area underdose area are generated; S7: treatment plan optimization, with the goal of meeting clinical needs with the sum of the previous treatment dose and the planned dose to be optimized at the current time.

[0014] Furthermore, step S5 specifically includes: S51: uniformly expanding the target area contour and performing an intersection operation with the endangered organ to obtain an area A where the endangered organ is close to the target area, wherein the distance of the uniform expansion of the target area depends on the minimum evaluation volume; S52: in order to avoid constraint conflicts of the endangered organ adjacent to the target area, calculating the part of area A whose distance from the target area boundary is less than a threshold τ, and removing it from area A; S53: if the volume of area A after removal is smaller than the minimum evaluation volume, supplementing some areas in order of distance from far to near until the requirement is met, and finally outputting area A as the key area of ​​the corresponding endangered organ.

[0015] Furthermore, in step S6, based on the previous treatment dose, the previous high-risk dose assessment value of each organ at risk is calculated for each organ at risk dose limit, and then the isodose line area corresponding to the previous high-risk dose assessment value is generated one by one, and the intersection with each organ at risk is calculated. The obtained area is the high-risk dose area corresponding to the organ at risk.

[0016] Furthermore, in step S6, a target dose underdose assessment value is calculated for the target prescription dose, and then an isodose line area corresponding to the dose underdose assessment value is generated, and the target contour is subtracted from the corresponding isodose line area to obtain the target underdose area.

[0017] Furthermore, in step S6, the calculation method of the pre-sequence high-risk dose assessment value of each organ at risk is: the total BT treatment dose limit corresponding to the organ at risk / total BT treatment fractions*BT treated fractions*organ risk assessment coefficient; and the calculation method of the target area dose underdose assessment value is: the total BT treatment dose corresponding to the target area / total BT treatment fractions*BT treated fractions*target area risk assessment coefficient.

[0018] Furthermore, in step S7, the key area constraints, the preceding dose constraints, the target dose target defined by the clinical user, and the dose limit target for organs at risk are combined to form an overall optimization objective function, and hybrid objective plan optimization is performed.

[0019] Furthermore, in step S7, the objective function is set to:

[0020] F(d)=f1(d)+f2(d)+f3(d)+f4(d)+f5(d)

[0021] Among them, f1(d) represents the contribution of the critical region constraint of the organ at risk to the objective function, and its expression is:

[0022]

[0023]

[0024] N OAR,H Indicates the total number of critical areas of organs at risk, represents the penalty function of the i-th critical area of ​​the organ at risk, Indicates the number of total dose points in the area, is the dose threshold of the i-th key area set clinically, d j (x) represents the contribution of all dwell points to the jth dose point, represents the weight corresponding to the key area of ​​the i-th organ at risk,

[0025] f2(d) represents the contribution of the underdose area in the target area's pre-dose to the objective function, and its expression is:

[0026]

[0027] represents the penalty function for the i-th under-measured area within the target area, Indicates the number of total dose points in the area, is the expected cumulative prescription dose of the target area, D cum,j Denotes the cumulative dose at the jth point, D cum,j (x) = D prev,j (x)+d j (x), represents the weight corresponding to the underdose area in the previous dose of the i-th target area,

[0028] f3(d) represents the contribution of the OARs preceding excess dose region to the objective function, and its expression is:

[0029]

[0030] Similarly, is a quadratic penalty function, represents the total number of dose points in the excess area of ​​the i-th organ at risk, is the total tolerated dose to the organ at risk, D cum,j Denotes the cumulative dose at the jth point, D cum,j (x) = D prev,j (x)+d j (x), represents the weight corresponding to the excess area in the preceding dose of the i-th organ at risk,

[0031] f4(d) and f5(d) are the contributions of the target area and critical organ radiation dose to the objective function during the treatment.

[0032]

[0033] is the prescribed dose for the i-th target area during treatment, is the prescribed dose for the i-th critical organ treated at that time, is the penalty function of the i-th target area, is the penalty function of the i-th critical organ represents the total number of dose points in the i-th target area, represents the total number of dose points of the i-th organ at risk, d j (x) represents the contribution of all dwell points to the jth dose point.

[0034] Furthermore, in step S7, a hybrid strategy of particle swarm optimization (PSO) and genetic algorithm (GA) is used to optimize the target according to the following steps:

[0035] T1 particle swarm initialization: Each particle represents a set of residence time vectors, and each particle P is defined k represents a set of radioactive source dwell time vectors t = [t1, t2, ..., tn], where n is the number of dwell points;

[0036] T2 genetic operation embedding: crossover and mutation are performed probabilistically in iterations;

[0037] T3 fitness evaluation: Combined with the objective function, each particle P k The residence time t k , calculate the comprehensive objective function value:

[0038] F(d)=f1(d)+f2(d)+f3(d)+f4(d)+f5(d)

[0039] At this time, d=D(t k ) is the dose distribution corresponding to the current particle,

[0040] And perform fitness mapping:

[0041] Furthermore, in step T2, in each PSO iteration, the genetic operation is triggered with probability:

[0042] (1) For crossover operation:

[0043] Select parent generation: select particles with high fitness,

[0044] Single-point crossover: Randomly select a split point j∈[1,n-1] and generate offspring:

[0045] t child1 =[t a1 ,...,t aj ,t b(j+1) ,...,t bn ]

[0046] Elite retention: offspring replace the particle with the lowest fitness in the current population;

[0047] (2) Mutation operation:

[0048] Gaussian mutation: for randomly selected particles P k Add noise to a dimension i:

[0049]

[0050] Boundary correction: If t kIf the boundary is crossed, the dose is clamped to [t min , max ] where t min , max is a time range constraint set according to the physical limitations of the device.

[0051] The technical scheme of the present application also provides a cumulative dose target-based brachytherapy plan optimization device, which comprises the following modules: a treatment data import module for importing current treatment data, including current treatment images and corresponding delineation data; a deformation registration module for pre-sequence image deformation registration, importing pre-sequence treatment images and corresponding delineation data, and performing deformation registration with the current treatment images; a dose mapping module for pre-sequence treatment dose mapping, mapping the pre-sequence treatment dose to the current treatment images based on the deformation registration result of the pre-sequence treatment images and the current treatment images, as the background dose of the current treatment; a biological equivalent dose conversion module for biological equivalent dose conversion, converting the dose data of the pre-sequence treatment mapped to the current treatment images into biological equivalent dose; a key region constraint generation module for key region constraint generation, generating a key region for each critical organ based on the delineation data of the current treatment images and referring to the dose limit requirements of the clinical critical organs; a pre-sequence dose constraint generation module for pre-sequence dose constraint generation, generating a high-risk dose region of the corresponding critical organ and a target region underdose region based on the pre-sequence treatment dose and the dose limit of each critical organ and the target region prescription dose; and a treatment plan optimization module for treatment plan optimization, optimizing the treatment plan to meet the clinical requirements of the sum of the pre-sequence treatment dose and the optimized plan dose. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0053] Figure 1 is a flowchart of the cumulative dose target-based brachytherapy plan optimization method of the present application. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described clearly and completely below. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0055] The key points of the present invention include:

[0056] (1) Dynamic transformation of cumulative dose into optimization constraints

[0057] The present invention converts the previous treatment dose into a spatial constraint. First, the organ deformation registration technology is used to accurately map the previous dose to the current anatomical space. Secondly, based on the biological equivalent dose (EQD2) model, the dose of multiple treatments is standardized, the α / β of the target area is set to 10, and the organs at risk (OARs) is 3. Finally, the automatically generated three-dimensional dose map identifies the underdose area of ​​the target area and the overdose area of ​​the organs at risk, so as to make subsequent optimization decisions. The implementation of this technology solves the problem that the historical dose cannot be directly referenced due to the deformation of the pelvic organs, and enables the cumulative dose to play a role in the optimization decision, rather than just being evaluated in the later stage. After the historical dose is accurately converted, it can be used as a spatial constraint, effectively avoiding the problem of hot spot overlap of organs at risk, and at the same time, targeted dose supplementation is carried out for the underdose area of ​​the target area, thereby improving the control rate of the tumor.

[0058] (2) Key area constraint generation method based on the characteristics of brachytherapy

[0059] This approach innovatively incorporates the characteristics of brachytherapy before optimization, automatically generating localized regions of organs at risk adjacent to the target volume as key constraint regions. During this generation, regions with obvious conflicting constraints are excluded, while areas critical for dose control are retained as key constraint regions. This is incorporated as an independent objective function within the overall optimization objective function, allowing for improved plan optimization efficiency through weight adjustment, allowing for rapid achievement of an ideal treatment plan.

[0060] (3) Dynamic weight constraint mechanism

[0061] This invention designs an anatomically sensitive weight allocator that automatically adjusts the strength of the cumulative dose constraint. For real-time weight adjustment, the system automatically outputs λ and μ values ​​based on different registration scenarios. This mechanism addresses the issue of fixed constraint weights leading to decreased plan quality when anatomy changes, and overcomes the limitation of traditional systems that cannot quantify historical dose reference values. When deformation is large, the system automatically weakens historical constraints to avoid ineffective optimization; at the same time, underdose areas in the target area can also be adjusted accordingly to ensure that the tumor target dose meets the standard.

[0062] (4) Multi-objective optimization engine

[0063] The present invention constructs a hybrid objective function that deeply couples the single dose target with the cumulative dose target. The hybrid optimization strategy includes PSO initialization of the residence time vector, genetic algorithm crossover mutation and multi-objective Bayesian optimization screening of Pareto solutions, and the number of iterations is limited to 50. This optimization engine solves the problem of separation between single optimization and cumulative targets and improves the optimization efficiency under complex constraints. By automatically taking into account the fractionation plan and the total fractionation limit, it can achieve the goal of achieving D-level optimization for organs at risk such as the rectum. 2cc The effect of the reduction is reduced, and the optimization speed of the hybrid algorithm is improved.

[0064] See also Figure 1 The technical solution of the present invention specifically includes the following steps:

[0065] S1, current treatment data import: current treatment image and outline data import;

[0066] S2, deformation registration of previous images: import the previous treatment images and delineation data, and perform deformation registration with the current treatment images;

[0067] S3, previous treatment dose mapping: Based on the deformable registration results of the previous treatment image and the current treatment image, the previous treatment dose is mapped to the current treatment image as the background dose of the current treatment;

[0068] S4, bioequivalent dose conversion: convert the dose data of the previous treatment to the current treatment image into the bioequivalent dose;

[0069] S5, generation of key area constraints: Based on the delineation data of the current image and referring to the clinical dose limit requirements of the organs at risk, key areas are generated for each organ at risk. The specific method is to perform the following operations on each organ at risk:

[0070] The target area is uniformly expanded and intersected with the OAR to obtain an area A close to the target area. The distance of the uniform expansion of the target area depends on the minimum assessment volume, which is the volume of the OAR dose limit assessment multiplied by a configurable magnification factor (range 1 to 5, generally set to 2 by default). For example, when evaluating D2cc, i.e., the dose received by a 2cc volume of OAR, the magnification factor is set to 2. Then, the minimum volume of the area A close to the target area obtained by intersecting the expanded target area with the OAR should be no less than 4cc.

[0071] To avoid constraint conflicts for organs at risk that are adjacent to the target (e.g., adjacent locations must meet the high dose constraint of the target on one side and the dose limit constraint of the organs at risk on the other side, and the difference between the two constraint dose values ​​is significant, which is obviously impossible to meet in the real world), calculate the part of region A whose distance to the target boundary is less than a threshold τ (e.g., τ = 1 mm) and remove it from region A;

[0072] If the volume of region A after removal is less than the minimum evaluation volume, then supplement part of the region in order from far to near until the minimum evaluation volume is met, and finally output region A as the key region of the OAR.

[0073] S6, Preceding dose constraint generation:

[0074] (1) Based on the preceding treatment dose, calculate the preceding high-risk dose evaluation value of each OAR for each OAR dose limit, and the evaluation value calculation method is: OAR corresponding BT total dose limit / BT total treatment fraction * BT treated fraction * organ risk evaluation coefficient (input parameter, generally 0.5-1.0), generate the isodose line region corresponding to the evaluation value one by one, and intersect with each OAR to obtain the high-risk dose region corresponding to the OAR.

[0075] (2) Based on the preceding treatment dose, calculate the target dose under-dose evaluation value for the target prescription dose, and the evaluation value calculation method is: target corresponding BT total dose / BT total treatment fraction * BT treated fraction * target risk evaluation coefficient (input parameter, generally 0.8-1.0), generate the isodose line region corresponding to the evaluation value, and subtract the target contour from the isodose line region to obtain the target under-dose region.

[0076] S7, treatment plan optimization: with the sum of the preceding treated dose and the plan dose to be optimized this time meeting the clinical requirements as the goal, the treatment plan is optimized. The key region constraint, the preceding dose constraint, and the target dose target and the OAR dose limit target defined by the clinician are combined to form the total optimization objective function, and mixed target plan optimization is performed.

[0077] (1) Set the specific objective function as follows:

[0078] F(d) = f1(d) + f2(d) + f3(d) + f4(d) + f5(d)

[0079] Wherein:

[0080] f1(d), represents the contribution of the OAR key region constraint (the region is obtained by intersecting the outer expanded target and the OAR) to the objective function. For each key region, a quadratic penalty function is used to realize the constraint. Its expression is:

[0081]

[0082] N OAR,H : represents the total number of OAR key regions.

[0083] represents the penalty function of the i-th OAR key region, Indicates the number of total dose points in the area, is the clinically defined dose threshold for the i-th focal area (which may be more stringent than the limit for the entire OAR), d j (x) represents the contribution of all dwell points to the jth dose point.

[0084] Represents the weight corresponding to the key area of ​​the i-th organ at risk.

[0085] f2(d) represents the contribution of the underdose area in the target volume's preceding dose to the objective function. During the optimization process, a compensatory penalty is imposed on the underdose area:

[0086]

[0087] represents the penalty function for the i-th under-measured area within the target area, Indicates the number of total dose points in the area, is the expected cumulative prescription dose of the target area, D cum,j represents the cumulative dose at the jth point (including the previous dose and the current dose), D cum,j (x) = D prev,j (x)+d j (x).

[0088] Represents the weight corresponding to the underdose area in the previous dose of the i-th target area.

[0089] f3(d) represents the contribution of the OARs' pre-treatment excess dose region to the objective function. For each OAR, if a high-dose region of OARs occurs during the pre-treatment dose that exceeds the OAR limit, a penalty is applied. Furthermore, the focus here is on the cumulative dose throughout the treatment process, not just the dose at that particular treatment.

[0090]

[0091] Similarly, is a quadratic penalty function: represents the total number of dose points in the excess area of ​​the i-th organ at risk, is the total tolerated dose of the organ at risk. cum,j : Cumulative dose at point j (including previous dose and current dose), D cum,j (x) = D prev,j (x)+d j (x).

[0092] represents the weight corresponding to the excess area in the preceding dose of the i-th organ at risk.

[0093] f4(d), f5(d) are the contribution of the target region and the critical organ dose in the current treatment to the objective function

[0094]

[0095] where is the prescribed dose for the i-th target region in the current treatment. Where is the prescribed dose for the i-th critical organ in the current treatment. is the penalty function for the i-th target region, is the penalty function for the i-th critical organ denotes the total number of dose points for the i-th target region, denotes the total number of dose points for the i-th critical organ, d j (x) represents the contribution of all residence points to the j-th dose point.

[0096] (2) Multi-objective weight setting

[0097] In actual optimization, other factors such as the uniformity of the dose distribution may also need to be considered, but the above objective function has covered multiple main factors, and the weight (ω) of each part needs to be adjusted and set according to the clinical importance. For example:

[0098] The weight of the key area is usually high because it is a high-risk area.

[0099] The weight of the target region underdose compensation reflects the importance of cumulative dose deficiency.

[0100] The weight of the OARs cumulative dose overdose is usually high to avoid complications.

[0101] The weight of the current target region coverage.

[0102] The weight of the current OARs dose limit.

[0103] (3) Mixed objective optimization

[0104] By constructing a multi-objective function that combines clinical objectives and historical dose compensation objectives, a suitable optimization algorithm is used to optimize the treatment plan under comprehensive constraints to ensure the best clinical treatment effect. For example, the PSO-GA (Particle Swarm Optimization-Genetic Algorithm) hybrid strategy is used to efficiently solve the residence time vector t = [t1, t2,..., tn], and ensure that the cumulative dose D cum (x) = D prev (x) + d(x) meets the clinical needs:

[0105] A hybrid strategy of particle swarm optimization (PSO) and genetic algorithm (GA) is adopted for target optimization:

[0106] Particle swarm initialization: each particle represents a set of dwell time vectors; define each particle P k represents a set of radioactive source dwell time vectors t = [t1, t2,..., tn], where n is the number of dwell points.

[0107] Initialization rules:

[0108] Time range constraints: t k ∈ [t min ,t max ](set according to device physical limitations).

[0109] Random generation strategy: uniformly randomly initialize the particle swarm (population size N = 50-100) within the feasible domain.

[0110] Initial velocity v ki = 0 (conservative start).

[0111] Genetic operation embedding: perform crossover (single-point crossover rate = 0.8) and mutation (Gaussian mutation rate = 0.1) with probability in iterations;

[0112] In each PSO iteration, trigger genetic operations with probability:

[0113] Crossover operation:

[0114] Select parents: select particles with high fitness.

[0115] Single-point crossover: randomly select a split point j ∈ [1, n-1], generate offspring:

[0116] t child1 = [t a1 ,...,t aj ,t b( j+1),...,t bn ]

[0117] Elite preservation: replace the particle with the lowest fitness in the current population with the offspring.

[0118] Mutation operation:

[0119] Gaussian mutation: add noise to a certain dimension i of the randomly selected particle Pk:

[0120]

[0121] Boundary correction: if t k goes out of bounds, clamp it to [t min ,t max].

[0122] Fitness evaluation: Evaluation is performed in combination with the objective function.

[0123] Objective function calculation: For the residence time tk of each particle Pk, calculate the comprehensive objective function value:

[0124] F(d)=f1(d)+f2(d)+f3(d)+f4(d)+f5(d)

[0125] At this time, d=D(t k ) is the dose distribution corresponding to the current particle.

[0126] Fitness Map:

[0127] (Minimizing F(d) translates to maximizing fitness).

[0128] Beneficial technical effects of the technical solution of the present invention:

[0129] The present invention significantly improves the accuracy of target dose and the protection of organs at risk by realizing the dynamic conversion of the previous treatment dose into the optimization constraint. The key area constraint generation method combined with the characteristics of brachytherapy is reflected in the overall optimization objective function as an independent objective function, so that the efficiency of plan optimization can be improved through weight adjustment, and the ideal treatment plan can be quickly obtained. The weighted dynamic constraint mechanism ensures that the constraint strength is dynamically adjusted when the anatomy changes, thereby avoiding invalid optimization and ensuring that the target dose meets the standard. The multi-objective deep coupling optimization engine effectively combines single optimization with cumulative objectives, improves optimization efficiency, and reduces the risk of toxicity to organs at risk such as the rectum. Overall, the present invention provides a more efficient, accurate and safe solution for brachytherapy.

[0130] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. A brachytherapy plan optimization method based on cumulative dose target, characterized in that: The steps include: S1: Import the current treatment data, including the current treatment images and corresponding delineation data; S2: Deformation registration of previous images: import the previous treatment images and corresponding delineation data, and perform deformation registration with the current treatment images; S3: Previous treatment dose mapping: Based on the deformable registration results of the previous treatment image and the current treatment image, the previous treatment dose is mapped to the current treatment image as the background dose of this treatment; S4: Bioequivalent dose conversion, which converts the dose data of the previous treatment mapped to the current treatment image into the bioequivalent dose; S5: Generate key area constraints based on the delineation data of the treatment image and refer to the clinical dose limit requirements of the organs at risk to generate key areas for each organ at risk; S6: Generate the previous dose constraint. Based on the previous treatment dose, the dose limit of each organ at risk and the target prescription dose are used to generate the corresponding high-risk dose area of ​​the organ at risk and the target underdose area. S7: Treatment plan optimization: The treatment plan is optimized with the goal of ensuring that the sum of the previously treated dose and the currently optimized planned dose meets clinical needs.

2. The method according to claim 1, characterized in that Step S5 specifically includes: S51: Uniformly expand the target area contour and perform an intersection operation with the organ at risk to obtain an area A of the organ at risk close to the target area, wherein the uniform expansion distance of the target area depends on the minimum assessment volume; S52: To avoid constraint conflicts of organs at risk that are adjacent to the target area, calculate the portion of region A whose distance from the target area boundary is less than a threshold τ and remove it from region A; S53: If the volume of region A after removal is smaller than the minimum assessment volume, some regions are supplemented in order of distance from far to near until the volume is satisfied, and finally region A is output as the key region of the corresponding organ at risk.

3. The method according to claim 1, characterized in that In step S6, based on the previous treatment dose, the previous high-risk dose assessment value of each organ at risk is calculated for each organ at risk dose limit, and then the isodose line area corresponding to the previous high-risk dose assessment value is generated one by one, and the intersection with each organ at risk is calculated. The obtained area is the high-risk dose area corresponding to the organ at risk.

4. The method according to claim 3, characterized in that In step S6, the target area dose underdose assessment value is calculated for the target area prescription dose, and then the isodose line area corresponding to the dose underdose assessment value is generated, and the target area contour is subtracted from the corresponding isodose line area to obtain the target area underdose area.

5. The method according to claim 4, characterized in that In step S6, the calculation method of the pre-sequence high-risk dose assessment value of each organ at risk is: the total BT treatment dose limit corresponding to the organ at risk / total BT treatment fractions * BT treatment fractions * organ risk assessment coefficient; and The calculation method for the target area dose underdose assessment value is: the total BT treatment dose corresponding to the target area / total BT treatment fractions*BT treatment fractions*target area risk assessment coefficient.

6. The method according to claim 1, characterized in that In step S7, the key area constraints, the preceding dose constraints, the target dose target defined by the clinical user, and the dose limit target for organs at risk are combined to form an overall optimization objective function, and hybrid objective plan optimization is performed.

7. The method according to claim 6, characterized in that In step S7, the objective function is set as: F(d)=f1(d)+f2(d)+f3(d)+f4(d)+f5(d) Among them, f1(d) represents the contribution of the critical region constraint of the organ at risk to the objective function, and its expression is: N OAR,H Indicates the total number of critical areas of organs at risk, represents the penalty function of the i-th critical area of ​​the organ at risk, Indicates the number of total dose points in the area, is the dose threshold of the i-th key area set clinically, d j (x) represents the contribution of all dwell points to the jth dose point, represents the weight corresponding to the key area of ​​the i-th organ at risk, f2(d) represents the contribution of the underdose area in the target area's pre-dose to the objective function, and its expression is: represents the penalty function for the i-th under-measured area within the target area, Indicates the number of total dose points in the area, is the expected cumulative prescription dose of the target area, D cum,j Denotes the cumulative dose at the jth point, D cum,j (x) = D prev,j (x)+d j (x), represents the weight corresponding to the underdose area in the previous dose of the i-th target area, f3(d) represents the contribution of the OARs preceding excess dose region to the objective function, and its expression is: Similarly, is a quadratic penalty function, represents the total number of dose points in the excess area of ​​the i-th organ at risk, is the total tolerated dose to the organ at risk, D cum,j Denotes the cumulative dose at the jth point, D cum,j (x) = D prev,j (x)+d j (x), represents the weight corresponding to the excess area in the preceding dose of the i-th organ at risk, f4(d) and f5(d) are the contributions of the target area and critical organ radiation dose to the objective function during the treatment. is the prescribed dose for the i-th target area during treatment, is the prescribed dose for the i-th critical organ treated at that time, is the penalty function of the i-th target area, is the penalty function of the i-th critical organ represents the total number of dose points in the i-th target area, represents the total number of dose points of the i-th organ at risk, d j (x) represents the contribution of all dwell points to the jth dose point.

8. The method according to claim 7, characterized in that In step S7, a hybrid strategy of particle swarm optimization (PSO) and genetic algorithm (GA) is used to optimize the target according to the following steps: T1 particle swarm initialization: Each particle represents a set of residence time vectors, and each particle P is defined k represents a set of radioactive source dwell time vectors t = [t1, t2, ..., tn], where n is the number of dwell points; T2 genetic operation embedding: crossover and mutation are performed probabilistically in iterations; T3 fitness evaluation: Combined with the objective function, each particle P k The residence time t k , calculate the comprehensive objective function value: F(d)=f1(d)+f2(d)+f3(d)+f4(d)+f5(d) At this time, d=D(t k ) is the dose distribution corresponding to the current particle, And perform fitness mapping:

9. The method according to claim 8, characterized in that In step T2, in each PSO iteration, the genetic operation is triggered with probability: (1) For crossover operation: Select parent generation: select particles with high fitness, Single-point crossover: Randomly select a split point j∈[1,n-1] and generate offspring: t child1 =[t a1 ,...,t aj ,t b( j+1),...,t bn ] Elite retention: offspring replace the particle with the lowest fitness in the current population; (2) Mutation operation: Gaussian mutation: for randomly selected particles P k Add noise to a dimension i: Boundary correction: If t k If out of bounds, clamp to [t min ,t max ], where t min ,t max A time range constraint set based on the physical limitations of the device.

10. A brachytherapy plan optimization device based on cumulative dose target, characterized in that: Includes the following modules: Treatment data import module: used to import the current treatment data, including the current treatment images and corresponding delineation data; Deformable Registration Module: used for deformation registration of previous images, importing previous treatment images and corresponding delineation data, and performing deformation registration with the current treatment images; Dose mapping module: used for previous treatment dose mapping. Based on the deformable registration results of the previous treatment image and the current treatment image, the previous treatment dose is mapped to the current treatment image as the background dose of the current treatment. Bioequivalent dose conversion module: used to convert bioequivalent doses, converting the dose data of the previous treatment mapped to the current treatment image into bioequivalent doses; Key area constraint generation module: used for key area constraint generation. Based on the delineation data of the treatment image and referring to the clinical dose limit requirements of the organs at risk, key areas are generated for each organ at risk. Precursor dose constraint generation module: used to generate precursor dose constraints. Based on the precursor treatment dose, it generates high-risk dose areas for organs at risk and target underdose areas for each organ at risk dose limit and target volume prescription dose. Treatment plan optimization module: used for treatment plan optimization, with the goal of ensuring that the sum of the previous treatment dose and the planned dose to be optimized meets clinical needs.

Images