A non-uniform lattice distribution optimization method based on local dose constraints

Abstract

The application provides a non-equidistant lattice distribution optimization method based on local dose constraints, comprising generating an initial lattice based on patient images and a delineated target region; pre-calculating and storing dose kernels based on a Monte Carlo algorithm; calling the dose kernels to perform fast dose optimization and calculation; establishing an independent peripheral subspace dose evaluation region and a local dose constraint condition for each lattice target region; establishing a multi-objective function for lattice target region position optimization; dynamically adjusting the lattice position and interval by using a fast simulated annealing method or a gradient descent method until a target region position distribution and a dose distribution satisfying the dose multi-objective function and the constraint condition are generated; and the application realizes accurate control of the peripheral valley dose of each lattice by dynamically adjusting the lattice interval and position and combining patient-specific anatomical structures, thereby solving the problems of uneven target region dose distribution and insufficient valley dose control in traditional equidistant lattice radiotherapy.

Description

Technical Field

[0001] This invention relates to the field of radiotherapy technology, and in particular to a method for optimizing non-equidistant lattice distribution based on local dose constraints. Background Technology

[0002] Spatial fractionation radiotherapy is considered a revolutionary treatment technique for large-volume or advanced tumors. It mainly involves creating a highly non-uniform dose distribution within the tumor's geovitreal volume (GTV) through a single high-dose irradiation, thereby effectively controlling the progression of malignant tumors while reducing radiotherapy damage to normal tissues. Its safety and efficacy have been proven.

[0003] Immunological mechanisms of spatially fractionated radiotherapy suggest that the regularly spaced high- and low-dose dose distribution may synergistically enhance anti-tumor immune responses. Based on this theory, it is generally required that the peak dose within the target lattice be sufficiently high to induce a tumor-specific immune response, while the trough dose between the target lattice is sufficiently low to protect the function of tumor microvessels and allow the circulation of cytokines / chemokines / or immunogenic factors, thereby synergistically inducing anti-tumor immunity. This dose distribution characteristic is typically quantified using the peak-to-trough dose ratio (PVDR). Therefore, current spatially fractionated radiotherapy planning usually uses the PVDR as the primary dosimetric optimization parameter to achieve this characteristic dose distribution.

[0004] Currently, there are two main ways to achieve this peak and trough dose interval distribution pattern:

[0005] (1) Based on the volume of tumor GTV, determine the size and uniform spacing of the lattice target area, manually / automatically delineate the equally spaced lattice target areas, and use the peak-to-valley dose ratio as the optimization target parameter to achieve the design of the target dose and treatment plan.

[0006] (2) The spatial fractionation radiotherapy plan is optimized by using the size of the lattice target area, uniform spacing, and peak-to-valley dose ratio as the main parameters.

[0007] However, based on a pre-determined, equally spaced lattice target region, the focus is usually on planned dosimetric optimization. However, since the lattice size and spacing are fixed, the achievable optimal dose target is limited, preventing the peak-to-valley dose ratio from reaching its ideal state and thus reducing tumor control rates. Even when lattice size and spacing parameters are introduced during the planning optimization process to achieve further dose optimization, with a large number of lattices, due to the dose superposition effect, this equally spaced lattice target region layout still results in higher valley doses in the central region of the PTV compared to the edge regions. This may increase the probability of damaging the circulatory function of microvessels around the lattice target region, thereby reducing the immune effect. Furthermore, while the peak-to-valley dose ratio is currently a good quantitative evaluation indicator, it is the ratio of the dosimetric statistics of two large regions (lattice target region and lattice target region), lacking a reflection of the subspatial dose distribution characteristics around each lattice target region. Therefore, it cannot guarantee that each lattice target region receives sufficient valley doses to enhance the synergistic immune effect of each subspace. Summary of the Invention

[0008] To address the shortcomings of existing technologies, this invention provides a non-equidistant lattice distribution optimization method based on local dose constraints. This invention achieves precise control of the valley dose around each lattice by dynamically adjusting the spacing and position of each lattice and combining it with the patient's specific anatomical structure.

[0009] The technical solution of this invention is: a non-equidistant lattice distribution optimization method based on local dose constraints, comprising the following steps:

[0010] S1) Generate an initial lattice based on patient images and the delineated target area;

[0011] S2) Pre-calculate and store the dose kernel based on the Monte Carlo algorithm;

[0012] S3) Call the dose kernel for rapid dose optimization and calculation;

[0013] S4) Establish an independent peripheral subspace dose evaluation region and local dose constraint conditions for each lattice target region;

[0014] S5) Establish and calculate a multi-objective function for optimizing the position of the lattice target region;

[0015] S6) By using rapid simulated annealing or gradient descent, the lattice position and spacing are dynamically adjusted until a target position distribution and dose distribution that satisfy the dose multi-objective function in step S5) and the constraints in step S4) are generated.

[0016] Preferably, in step S1), the generation of the initial lattice includes the following steps:

[0017] S11) Establish a tumor volume morphology model based on patient images and delineated target areas, and calculate tumor volume morphology parameters;

[0018] S12) Construct the initial lattice position distribution based on the tumor volume morphology model and tumor volume morphology parameters.

[0019] Preferably, in step S11), the three-dimensional surface mesh of the tumor is extracted as a tumor volume morphology model using the patient's CT / MRI images and the delineated set of anatomical structures, and the tumor spatial domain is defined as Ω.

[0020] Preferably, in step S11), the tumor volume morphological parameters include the local curvature of the target area, the distance between the target area and adjacent sensitive organs, and the volume density distribution.

[0021] The local curvature of the target area is quantified by Gaussian curvature K to determine the surface roughness, where K > 0.2 indicates a high curvature region.

[0022] The distance d between the target area and the nearest sensitive organ is the Euclidean distance from each point on the tumor surface to the nearest sensitive organ;

[0023] The volume density distribution ρ∈(0,1) represents the division of the tumor into 1mm segments. 3 Voxels are used to count the percentage of tumor tissue within a voxel.

[0024] Preferably, in step S12), the initial lattice arrangement rule is as follows:

[0025] a) In high curvature regions, the lattice density increases linearly with respect to curvature K, and the spacing s satisfies:

[0026] s = s base -λ K ·K;

[0027] In the formula, s base The system default lattice spacing; λ K Indicates the spacing adjustment constant;

[0028] b) On the side adjacent to the sensitive organ, if d <d safe The spacing s is increased to:

[0029] s = s base +μ·(d safe -d);

[0030] In the formula, d safe The safe distance; μ represents the safe distance adjustment constant;

[0031] c) The tumor core region has uniformly distributed lattice points with a volume density distribution ρ > 0.8; spacing

[0032] s core =s base .

[0033] Preferably, in step S2), the dose kernel K is calculated and simulated in advance using the Monte Carlo algorithm to generate and store the dose kernel matrix corresponding to each field angle for subsequent rapid dose calculation.

[0034] Preferably, in step S3, the dose kernel is invoked for rapid dose optimization and calculation, specifically as follows:

[0035] For a given voxel, the total dose D i This is equal to all the doses of nuclear K that contribute to it. ij With the corresponding weight W j Product, that is:

[0036] D i =∑ j K ij W j ;

[0037] In the formula, j represents the field number; K ij It is a sparse matrix, with non-zero dose convolution kernels only on the path of the shooting sub-beams, and the dose at each point comes from the cumulative dose contribution of each shooting sub-beam within a finite surrounding range.

[0038] Preferably, in step S4), establishing an independent peripheral subspace dose evaluation region for each lattice target region depends on a three-dimensional Volonoj map, and the segmentation rule of the three-dimensional Volonoj map is as follows:

[0039] a) Using the initial lattice point set Using the seed points, a Voronoi diagram is generated, dividing the tumor spatial domain Ω into N non-overlapping Voronoi diagram units. satisfy:

[0040]

[0041] In the formula, x represents the boundary coordinates of the Voronoi diagram;

[0042] b) Each Voronoitu unit V i center point c i As an optimized lattice target.

[0043] Preferably, in step S4), the lattice target region T i For each lattice target point c i Centered on a radius of r peak In the spherical region, the dose of the lattice target region needs to reach the prescribed dose D. target ;

[0044] T i ={x∈Ω|‖|xc i ||≤r peak}

[0045] Preferably, in step S4), the lattice target region T i The region between the target lattice and its adjacent lattice is referred to as the subspace S surrounding the target lattice. ij That is, for any two adjacent lattice target points c i and c j Its Voronoi unit boundary is defined as subspace S ij The subspace S ij The expression is:

[0046] S ij ={x∈Ω|‖|xc i ||=||xc j ||}∩B((c i +c j ) / 2,R);

[0047] In the formula, B((c i +c j ) / 2,R) represents (c i +c j A sphere with center 0.5 / 2 and radius R.

[0048] Preferably, in step S4), if the subspace S ij Approaching sensitive organs, reshape them into cubes with sides of 2R to limit the dose assessment range; if subspace S ij Located at the core of the tumor, it maintains a spherical shape to enhance homogeneity.

[0049] Preferably, in step S4), the local dose constraint condition is: for each subspace S ij Maximum dose within satisfy:

[0050]

[0051] Peak-to-valley dose ratio (PVDR) ≤ 0.3;

[0052] Where η represents the expected valley dose coefficient;

[0053] The peak-to-valley dose ratio (PVDR) is the ratio of the global peak dose to the global valley dose; the global peak dose is the dose that covers 95% of the volume of all lattice target regions; the global valley dose is the dose that covers 95% of the volume of the valley dose region; the valley dose region is the area remaining after subtracting 3 mm of lattice target region expansion from the tumor volume.

[0054] Preferably, in step S5), the multi-objective function for optimizing the lattice target region is:

[0055]

[0056] In the formula, Peak dose target achievement item;

[0057] γ is the valley dose constraint term; γ·HI is the uniformity penalty term; ε is the sensitive organ protection term; ε·(PVDR-0.3) is the overall peak-to-valley ratio constraint term; N represents the total number of lattice target regions; α, β, γ, δ, and ε are the weighting system. D represents the peak dose of the i-th lattice target region; target Indicates the target prescription dose; Subspace S ij The maximum dose within; η represents the expected trough dose coefficient; HI represents the global dose uniformity index; M represents the number of sensitive organs; This represents the average dose to the k-th sensitive organ; This represents the maximum permissible dose to the k-th sensitive organ; PVDR is the peak-to-trough dose ratio.

[0058] Preferably, in step S6), if the rapid simulated annealing method is used, the lattice position orientation sampling strategy is as follows: after each dose optimization subroutine is completed, the T value for each lattice target area is calculated. i The dose D in the middle region of the adjacent lattice target region ij ;

[0059] If dose D ij If the dose is less than the target valley dose, this position is marked as 0 in the lattice position orientation sampling matrix; otherwise, it is marked as 1. In subsequent lattice position size sampling, only those marked as 1 in the lattice position orientation sampling matrix participate in resampling and subsequent optimization.

[0060] The location sampling size is chosen to be a long-range Cauchy-Lorentz distribution:

[0061] p(Δx)∝[(Δx) 2 +W(T) 2 ] -(n+1) / 2

[0062] In the formula, Δx is the step size of the lattice position change; n is the number of selectable lattice positions; W(T) is the temperature correlation width of the sampling distribution; and p(Δx) represents the probability corresponding to the step size Δx.

[0063] Preferably, in step S6), if the gradient descent method is used, the objective function with respect to the lattice position c is calculated using the adjoint field method. i gradient Used to update direction:

[0064]

[0065] In the formula, ∈ represents the step size, t represents the number of iterations, J represents the objective function, and the gradient represents the gradient. The objective function J is related to the lattice position c. i The partial derivatives are used to guide the direction of parameter updates;

[0066] Record the k most recent lattice movement directions and prohibit repeated searches of the same direction. If the objective function result does not improve after more than T iterations, trigger random perturbation to escape the local optimum.

[0067] The beneficial effects of this invention are as follows:

[0068] 1. This invention achieves precise control of the valley dose around each lattice by dynamically adjusting the lattice spacing and position, combined with the patient's specific anatomical structure, thereby solving the problems of uneven target dose distribution and insufficient valley dose control in traditional equal-spacing lattice radiotherapy.

[0069] 2. This invention can obtain a better distribution of lattice target area and dose distribution. It takes into account the limiting factors of the surrounding OAR, ensures the limit of the valley dose in the subspace region around each lattice target area, solves the problem of high dose in the central region of the tumor caused by conventional equal-spacing distribution, and takes into account the influence of changes in tumor boundary curvature on the dose. Attached Figure Description

[0070] Figure 1 This is a schematic flowchart of the method of the present invention;

[0071] Figure 2 This is a schematic diagram of the non-uniform lattice distribution in the tumor spatial domain of the present invention;

[0072] Figure 3 This is a schematic diagram of the subspace shape dynamic adjustment strategy of the present invention;

[0073] Figure 4 This is a schematic diagram illustrating the optimal result of the non-equidistant distribution of the lattice target region in this invention. Detailed Implementation

[0074] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0075] like Figure 1 As shown, this embodiment provides a non-equidistant lattice distribution optimization method based on local dose constraints, including the following steps:

[0076] S1) Generate an initial lattice based on patient images and the delineated target area; including the following steps:

[0077] S11) Establish a tumor volume morphology model based on patient images and delineated target areas, and calculate tumor volume morphology parameters;

[0078] In this implementation, the three-dimensional surface mesh of the tumor is extracted as a tumor volume morphology model using the patient's CT / MRI images and the delineated set of anatomical structures, and the tumor spatial domain is defined as Ω.

[0079] Based on the tumor volume morphology model, tumor volume morphology parameters are calculated. In this embodiment, the tumor volume morphology parameters include the local curvature of the target area, the distance between the target area and adjacent sensitive organs (such as the bladder and rectum), and the volume density distribution.

[0080] The local curvature of the target area is quantified by Gaussian curvature K to determine the surface roughness, where K > 0.2 indicates a high curvature region.

[0081] The distance d between the target area and the nearest sensitive organ is the Euclidean distance from each point on the tumor surface to the nearest sensitive organ;

[0082] The volume density distribution ρ∈(0,1) represents the division of the tumor into 1mm segments. 3 Voxels are used to count the percentage of tumor tissue within a voxel.

[0083] S12) Construct the initial lattice position distribution based on the tumor volume morphology model and tumor volume morphology parameters;

[0084] In this embodiment, the initial lattice arrangement rule is as follows:

[0085] a) In high curvature regions, the lattice density increases linearly with respect to curvature K, and the spacing s satisfies:

[0086] s = s base -λ K ·K;

[0087] In the formula, s base The system default lattice spacing; λ K This represents the spacing adjustment constant; in this embodiment, s base =30mm, λ K =10mm;

[0088] b) On the side adjacent to the sensitive organ, if d <d safe The spacing s is increased to:

[0089] s = s base +μ·(d safe -d);

[0090] In the formula, d safe The safe distance is represented by μ; μ represents the safe distance adjustment constant; in this embodiment, d safe =20mm, μ=0.25;

[0091] c) The tumor core region has uniformly distributed lattice points with a volume density distribution ρ > 0.8; the center-to-center spacing s of the lattice target region. core =s base .

[0092] S2) Pre-calculate and store the dose kernel based on the Monte Carlo algorithm;

[0093] This embodiment uses the Monte Carlo algorithm to perform calculations and simulations, generating and storing the dose kernel matrix corresponding to each field angle for subsequent rapid dose calculations.

[0094] S3) Call the dose kernel for rapid dose optimization and calculation;

[0095] In this embodiment, dosage optimization can be performed using a commonly used optimization algorithm engine in current treatment planning systems;

[0096] The dosage calculation method shown is as follows: for a given voxel, the total dose D i This is equal to all the doses of nuclear K that contribute to it. ij With the corresponding weight W j Product, that is:

[0097] D i =∑ j K ij W j ;

[0098] In the formula, j represents the field number; K ij It is a sparse matrix, with non-zero dose convolution kernels only on the path of the shooting sub-beams, and the dose at each point comes from the cumulative dose contribution of each shooting sub-beam within a finite surrounding range.

[0099] S4) Establish an independent peripheral subspace dose evaluation region and local dose constraint conditions for each lattice target region;

[0100] The three-dimensional Volonoi diagram is created based on the initial lattice position distribution, such as... Figure 2 As shown;

[0101] In this embodiment, the three-dimensional Voronoi diagram segmentation rule is as follows:

[0102] a) Using the initial lattice point set Using the seed points, a Voronoi diagram is generated, dividing the tumor spatial domain Ω into N non-overlapping Voronoi diagram units. satisfy:

[0103]

[0104] In the formula, x represents the boundary coordinates of the Voronoi diagram;

[0105] b) Each Voronoitu unit V i center point c i As an optimized lattice target.

[0106] The lattice target region T i For each lattice target point c i Centered on a radius of r peak spherical region

[0107] The dose in the lattice target region needs to reach the prescribed dose D. target ;

[0108] T i ={x∈Ω|‖|xc i ||≤r peak}

[0109] lattice target region T i The region between the target lattice and its adjacent lattice is referred to as the subspace S surrounding the target lattice. ij That is, for any two adjacent lattice target points c i and c j Its Voronoi unit boundary is defined as subspace S ij The subspace S ij The expression is:

[0110] S ij ={x∈Ω|‖|xc i ||=||xc j ||}∩B((c i +c j ) / 2,R);

[0111] In the formula, B((c i +c j ) / 2,R) represents (c i +c j A sphere with center 0.5 / 2 and radius R.

[0112] like Figure 3 As shown, if the subspace S ij Approaching sensitive organs, reshape them into cubes with sides of 2R to limit the dose assessment range, such as... Figure 3 As shown in (a); if the subspace S ij Located at the core of the tumor, maintaining a spherical shape to enhance homogeneity, such as Figure 3 As shown in (b).

[0113] In this embodiment, the local dose constraint condition is: for each subspace S ij Maximum dose within

[0114] satisfy:

[0115]

[0116] Peak-to-valley dose ratio (PVDR) ≤ 0.3;

[0117] Where η represents the expected valley dose coefficient;

[0118] The peak-to-valley dose ratio (PVDR) is the ratio of the global peak dose to the global valley dose; the global peak dose is the dose that covers 95% of the volume of all lattice target regions; the global valley dose is the dose that covers 95% of the volume of the valley dose region; the valley dose region is the area remaining after subtracting 3 mm of lattice target region expansion from the tumor volume.

[0119] S5) Establish and calculate a multi-objective function for optimizing the lattice target region location; quantify the difference between the current dose distribution and the expected dose distribution; the multi-objective function for optimizing the lattice target region is:

[0120]

[0121] In the formula, Peak dose target achievement item;

[0122] γ is the valley dose constraint term; γ·HI is the uniformity penalty term; ε is the sensitive organ protection term; ε·(PVDR-0.3) is the overall peak-to-valley ratio constraint term; α, β, γ, δ, and ε are the weighting system. D represents the peak dose of the i-th lattice target region; target η represents the target prescription dose; M represents the expected trough dose coefficient; N represents the number of sensitive organs; and N represents the total number of lattice target regions. Subspace S ij The maximum dose within; HI represents the global dose uniformity index; This represents the average dose to the k-th sensitive organ; This represents the maximum permissible dose to the k-th sensitive organ; PVDR is the peak-to-trough dose ratio.

[0123] S6) Dynamically adjust the lattice position and spacing by using rapid simulated annealing or gradient descent until a target area position distribution and dose distribution that satisfy the dose objective function described in step S5) and the constraint conditions described in step S4) are generated;

[0124] In this embodiment, if the rapid simulated annealing method is used, the lattice position orientation sampling strategy is as follows: after each dose optimization subroutine shown in step S3) is completed, the T value for each lattice target region is calculated. i The maximum dose D in the subspace region between the target lattice and the surrounding adjacent lattice target regions ij =max(D ij (k),k∈[1,N]),k represents the kth pixel in the subspace, and N represents the total number of pixels in the subspace;

[0125] If dose D ij Less than the target trough dose η·D target If the position is 0, then the position is marked as 0 in the lattice position orientation sampling matrix; otherwise, it is marked as 1. In subsequent lattice position size sampling, only those marked as 1 in the lattice position orientation sampling matrix participate in resampling and subsequent optimization.

[0126] The location sampling size is chosen to be a long-range Cauchy-Lorentz distribution:

[0127] p(Δx)∝[(Δx) 2 +W(T) 2 ] -(n+1) / 2

[0128] In the formula, Δx is the step size of the lattice position change; n is the number of selectable lattice positions; W(T) is the temperature correlation width of the sampling distribution; and p(Δx) represents the probability corresponding to the step size Δx.

[0129] In this embodiment, if the gradient descent method is used, the objective function with respect to lattice position c is calculated using the adjoint field method. i gradient Used to update direction:

[0130]

[0131] In the formula, ∈ represents the step size, t represents the number of iterations, J represents the objective function, and the gradient represents the gradient. The objective function J is related to the lattice position c. i The partial derivatives are used to guide the direction of parameter updates;

[0132] Record the k most recent lattice movement directions and prohibit repeated searches of the same direction. If the objective function result does not improve after more than T iterations, trigger random perturbation to escape the local optimum.

[0133] In this embodiment, each iteration of the above optimization process generates a position optimization result for the lattice target region. Each iteration preserves the corresponding parameters and performs rapid convergence optimization of the dose target until the target dose meets the requirements. This not only satisfies the statistical peak-to-valley ratio constraint but also optimizes the dose for each subspace by constructing a multi-objective function, thereby achieving the valley dose requirement for each sub-detail and obtaining the optimal result for the non-equidistant distribution of the lattice target region. Figure 4 As shown, Figure 4(a) is a schematic diagram of the uniform distribution of the lattice target region in the existing method; while Figure 4 (b) is a schematic diagram of the non-uniform distribution of the lattice target region in this embodiment.

[0134] The embodiments and descriptions above are merely illustrative of the principles and preferred embodiments of the present invention. Various changes and modifications may be made to the present invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed.

Claims

1. A method for optimizing non-equidistant lattice distributions based on local dose constraints, characterized in that, Includes the following steps: S1) Generate an initial lattice based on patient images and the delineated target area; S2) Pre-calculate and store the dose kernel based on the Monte Carlo algorithm; S3) Call the dose kernel for rapid dose optimization and calculation; S4) Establish an independent peripheral subspace dose evaluation region and local dose constraint conditions for each lattice target region; S5) Establish and calculate the multi-objective function for optimizing the position of the lattice target region; S6) By using rapid simulated annealing or gradient descent, the lattice position and spacing are dynamically adjusted until a target position distribution and dose distribution that satisfy the dose multi-objective function in step S5) and the constraints in step S4) are generated. In step S4), for each lattice target region Establishing an independent peripheral subspatial dose evaluation region relies on a three-dimensional Voronoi diagram; the segmentation rules for the three-dimensional Voronoi diagram are as follows: a) Using the initial lattice point set Using seed points, a Voronoi diagram is generated, dividing the tumor spatial domain. Divided into N non-overlapping Voronoi diagram units ,satisfy: ； In the formula, Represents the boundary coordinates of the Voronoi diagram; b) Each Voronoitu unit center point As an optimized lattice target; In step S4), the local dose constraint condition is: for each subspace Maximum dose within satisfy: ； Peak-to-valley dose ratio ; in, Indicates the expected trough dose coefficient; The peak-to-valley dose ratio It is the ratio of global peak dose to global valley dose; the global peak dose is the dose covering 95% of the volume of all lattice target regions; the global valley dose is the dose covering 95% of the volume of the valley dose region; the valley dose region is the area remaining after subtracting 3 mm of lattice target region expansion from the tumor volume; In step S5), the multi-objective function for optimizing the lattice target region position is: In the formula, Peak dose target achievement item; For valley dose constraint term; This is a uniformity penalty term; This is a protection item for sensitive organs; This is the overall peak-to-valley ratio constraint term; Indicates the total number of lattice target regions; , , , , These are the weighting coefficients; This represents the peak dose of the i-th lattice target region; Indicates the target prescription dose; Subspace Maximum dose within; Indicates the expected trough dose coefficient; Indicates the global dose uniformity index; Indicates the number of sensitive organs; This represents the average dose to the k-th sensitive organ; This represents the maximum permissible dose for the k-th sensitive organ; This represents the peak-to-valley dose ratio.

2. The non-equidistant lattice distribution optimization method based on local dose constraint according to claim 1, characterized in that: In step S1), the generation of the initial lattice includes the following steps: S11) Establish a tumor volume morphology model based on patient images and delineated target areas, and calculate tumor volume morphology parameters; S12) Construct the initial lattice position distribution based on the tumor volume morphology model and tumor volume morphology parameters.

3. The non-equidistant lattice distribution optimization method based on local dose constraint according to claim 2, characterized in that: In step S11), the three-dimensional surface mesh of the tumor is extracted as a tumor volume morphology model using the patient's CT / MRI images and the delineated anatomical structure set, and the tumor spatial domain is defined as... ; The tumor volume morphological parameters include the local curvature of the target area, the distance between the target area and adjacent sensitive organs, and the volume density distribution. The local curvature of the target area is obtained by Gaussian curvature. Quantifying surface roughness, among which, This is a region with high curvature; The distance between the target area and adjacent sensitive organs The distance from each point on the tumor surface to the nearest sensitive organ is Euclidean. The aforementioned volume density distribution To divide the tumor into 1 mm³ voxels, the percentage of tumor tissue within each voxel was calculated.

4. The non-equidistant lattice distribution optimization method based on local dose constraint according to claim 3, characterized in that: In step S12), the initial lattice arrangement rule is as follows: a) High curvature regions, according to curvature Linearly increase lattice density, spacing satisfy: ； In the formula, The system default lattice spacing; Indicates the spacing adjustment constant; b) On the side adjacent to sensitive organs, if ,spacing Expanded to: ； In the formula, To maintain a safe distance; This represents the safety distance adjustment constant; c) The tumor core region has uniformly distributed lattice points and a high volume density distribution. ;spacing = .

5. The non-equidistant lattice distribution optimization method based on local dose constraint according to claim 1, characterized in that: In step S4), the lattice target region For each lattice target point Centered on, with radius as In the spherical region, the dose of the lattice target region needs to reach the prescribed dose. ; ； lattice target region The region between the target lattice and its adjacent lattice is considered as the subspace surrounding the target lattice. That is, for any two adjacent lattice target points and The boundary of its Voronoi unit is defined as subspace. The subspace mentioned The expression is: ； In the formula, Indicated by Centered on, A sphere with radius . Noya Space Approaching the sensitive organ, change its shape to have a side length of The cube is used to limit the range of dose assessment; if subspace Located at the core of the tumor, it maintains a spherical shape to enhance homogeneity.

6. The non-equidistant lattice distribution optimization method based on local dose constraint according to claim 1, characterized in that: In step S6), if the rapid simulated annealing method is used, the lattice position orientation sampling strategy is as follows: after each dose optimization subroutine in step S3) is completed, the lattice target area is calculated again. Dose in the middle region of the adjacent lattice target region ; If dosage If the dose is less than the target valley dose, this position is marked as 0 in the lattice position orientation sampling matrix; otherwise, it is marked as 1. In subsequent lattice position size sampling, only those marked as 1 in the lattice position orientation sampling matrix participate in resampling and subsequent optimization. The location sampling size is chosen to be a long-range Cauchy-Lorentz distribution: In the formula, The step size for changing the lattice position; The number of selectable lattice positions; The temperature correlation width of the sampled distribution; This indicates changing the step size to The probability corresponding to the time.

7. The non-equidistant lattice distribution optimization method based on local dose constraint according to claim 6, characterized in that: In step S6), if the gradient descent method is used, the objective function with respect to the lattice position is calculated using the adjoint field method. gradient Used to update direction: ； In the formula, Step size, Indicates the number of iterations; Describe the objective function and gradient. The objective function J is the lattice position The partial derivatives are used to guide the direction of parameter updates.

Images