Virtual grid arrangement method and system based on proton radiotherapy beam direction optimization

Abstract

The application discloses a virtual Grid arrangement method and system based on proton radiotherapy beam direction optimization. The method comprises the following steps: acquiring tumor target region contour data of a patient, and organ at risk and / or planning risk volume contour data, constructing structure binary images, internal shrinkage three-dimensional images and external expansion three-dimensional images; in a plurality of candidate beam directions, a beam direction robustness score is constructed according to water equivalent path length penetrating the target region, and a target beam direction is determined; the target beam direction is taken as the central axis direction of a virtual Grid cylinder, an initial Grid center coordinate set is generated in a two-dimensional beam eye view plane perpendicular to the target beam direction; a continuous available interval is searched along the target beam direction for the optimized center coordinate, and a finite length virtual Grid cylinder is generated; and a virtual Grid arrangement result is determined and output according to a Grid volume proportion, a minimum safety distance, a cylinder separation degree, a weighted normal tissue overlapping volume and a robust risk penalty term.

Description

Technical Field

[0001] This invention relates to the field of radiotherapy planning and image-guided optimization technology, and in particular to a virtual grid arrangement method and system based on proton radiotherapy beam direction optimization. Background Technology

[0002] Spatially fractionated radiotherapy, by constructing a non-uniform dose structure within the tumor with alternating high-dose peaks and low-dose troughs, offers a dosimetric strategy distinct from conventional uniform irradiation in treatment scenarios involving large tumors, radiation-resistant lesions, and those requiring rapid tumor and symptom reduction. Compared to photon grid or lattice radiotherapy, proton grid radiotherapy combines the physical advantages of peak-trough dose structure with the rapid drop-off at the distal end of the Bragg peak, theoretically making it more effective in maintaining intra-target spatial heterogeneity while reducing the dose received by normal tissues outside the target area.

[0003] However, proton grid radiotherapy planning is not simply a geometric arrangement problem, but a comprehensive optimization problem involving the coupling of beam orientation selection and grid structure generation. On the one hand, the actual range of the proton beam is highly sensitive to positioning errors, anatomical changes, respiratory motion, and stopping ability conversion errors, and the stability of the water equivalent path length corresponding to different beam orientations varies significantly. If the beam orientation is not properly selected, the predetermined high-dose peak region may shift during actual treatment, thereby weakening the stability of the peak-valley structure and increasing the risk of radiation exposure to distal normal tissues. On the other hand, after the beam orientation is determined, how to automatically generate a patient-specific virtual grid structure while taking into account constraints such as target area boundary contraction, safe avoidance of endangered organs, minimum spacing between cylinders, and minimum axial length still relies heavily on manual planning experience in current procedures.

[0004] Existing methods for automatically arranging virtual grids primarily focus on geometric layout within a preset fixed coordinate axis, or on selecting beam directions based on experience and then manually adjusting the grid structure locally. These methods share several shortcomings: first, they fail to incorporate patient-specific beam direction robustness evaluation into the critical decision-making process before grid geometry generation; second, they do not establish a clear geometric coupling relationship between the selected beam direction and the common axis of the grid cylinder; and third, they lack a unified evaluation mechanism for organ protection, geometric feasibility, and beam direction robustness. Therefore, a standardized method and system are urgently needed that can robustly screen candidate beam directions before automatically arranging the virtual grid under the constraint of the selected target beam direction. Summary of the Invention

[0005] The purpose of this invention is to provide a virtual grid layout method and system based on proton radiotherapy beam direction optimization. By establishing an integrated process of robust screening of candidate beam directions, grid layout under target beam direction constraints, and comprehensive evaluation of multiple indicators, this invention solves the technical problems in existing proton grid planning, such as beam direction selection relying on experience, disconnect between grid layout and beam direction information, and difficulty in unifying organ protection and geometric feasibility.

[0006] To achieve the above objectives, the first aspect of this disclosure provides a virtual grid arrangement method based on proton radiotherapy beam direction optimization, comprising: S1, acquire tumor target volume contour data, organ at risk OAR and / or planned risk volume PRV contour data of the patient in the lesion area; S2, construct a first structural binary image of the gross tumor volume based on the tumor target area contour data, and perform inward shrinking processing on the first structural binary image to obtain an inward shrinking three-dimensional image; S3, construct a second structural binary image based on the organ at risk (OAR) and / or the planned risk volume (PRV) contour data, and perform outward expansion or safety avoidance processing on the second structural binary image to obtain an outward expanded three-dimensional image; S4. Among multiple candidate beam directions, a robustness score for the candidate beam direction is constructed based on the statistical distribution of the water equivalent path length on the beam path through the tumor target area, and the target beam direction is determined based on the robustness score. S5, using the target beam direction as the common central axis direction of all virtual Grid cylinders, map the inward 3D image and the outward 3D image onto a 2D beam eye view plane perpendicular to the target beam direction to construct a 2D effective projection area; S6. Within the two-dimensional effective projection area, an initial set of grid cell center coordinates is generated based on the preset minimum distance constraint between grid cells, and candidate center coordinate schemes are formed through overall translation, overall rotation and local perturbation. S7, search for a continuous available interval along the target beam direction for each candidate center coordinate, and generate a finite-length virtual Grid cylinder parallel to the target beam direction based on the continuous available interval with a length not less than the minimum length threshold. S8. The candidate center coordinate scheme is evaluated based on the Grid volume ratio corresponding to the finite length virtual Grid cylinder, the minimum safe distance from the organ at risk, the cylinder separation degree, the weighted normal tissue overlap volume and / or the bundle robust risk penalty term, and the target three-dimensional virtual Grid target area is determined. S9. Based on the beam direction parameters corresponding to the target beam direction and the target three-dimensional virtual grid target area, generate and output the virtual grid layout result for guiding external radiotherapy.

[0007] Preferably, the step of constructing a robustness score for candidate beam directions based on the statistical distribution of water equivalent path lengths along the beam path traversing the tumor target region among multiple candidate beam directions, and determining the target beam direction based on the robustness score, includes: Within a preset range of the frame angle, discrete sampling is performed with a first angle step size to obtain multiple candidate beam directions; The water equivalent path length of each candidate beam direction through the tumor target region is determined, and a beam direction robustness score is constructed based on the variance, standard deviation and / or quantile difference of the water equivalent path length. The top N candidate beam directions with the best beam robustness scores are determined as the optimized center angles, and fine sampling is performed in the neighborhood of the optimized center angles with a second angle step size smaller than the first angle step size to determine the target beam direction.

[0008] Preferably, the step of constructing a bundle robustness score based on the variance, standard deviation, and / or quantile difference of the water equivalent path length includes: For any candidate beam direction θ and the projection point y in the beam eye view plane, the water equivalent path length Lω(θ,y) in the uncertain scenario ω is calculated using the following formula: L ω (θ,y)=∫ℓ(θ,y;ω)ρ(x)dl Where ρ(x) represents the relative stopping power or water equivalent density at voxel x, and ℓ(θ,y;ω) represents the central ray in scene ω with candidate beam direction θ and passing through projection point y; since the fluctuation of water equivalent path length in different scenes ω reflects the sensitivity of the candidate beam direction to range uncertainty; Calculate the robust fluctuation term R WEPL (θ,y): R WEPL (θ,y)=Q M ({L ω})-Q N ({L ω}) Among them, Q M and Q N To avoid relying solely on range fluctuations and ignoring proximal tissue heterogeneity and distal organ risk in candidate beam orientation selection, a proximal media heterogeneity term R is introduced, representing the quantiles of M percent and N percent of the cross-scenario samples, respectively. het (θ,y) and the risk item R for distal organs at risk dist(θ,y), and establish the robust risk map R(θ,y) for the candidate beam direction using the following formula: R(θ,y)=αR WEPL (θ,y)+βR het (θ,y)+γR dist (θ,y) For a candidate beam direction θ, the distribution characteristics of the robust risk map R(θ,y) within its feasible projection area are statistically analyzed, and a beam direction scoring function J is constructed by combining the effective projection area ratio. angle (θ): J angle (θ)=Q 75 ({R(θ,y)|y∈F θ})-λ|F θ | / |P θ (Ω L )| Among them, F θ P represents the feasible projection region that satisfies the basic safety conditions under the candidate beam direction θ. θ (Ω L ) represents Ω L In the two-dimensional projection region under the candidate beam direction θ, select J angle The candidate beam direction with the smallest (θ) is taken as the target beam direction θ. * .

[0009] Preferably, the step of mapping the inward-expanding 3D image and the outward-expanding 3D image onto a 2D beam-eye view plane perpendicular to the target beam direction, using the target beam direction as the common central axis of all virtual Grid cylinders, to construct a 2D effective projection area includes: Perform three-dimensional morphological erosion on the candidate arrangement region within the target corresponding to the indented three-dimensional image; Two-dimensional expansion or outward expansion is performed on the organ at risk (OAR) and / or the planned risk volume (PRV) in a plane perpendicular to the direction of the target beam; The robust three-dimensional feasible region is obtained by performing a difference operation on the processed intra-target candidate deployment region and the endangered organ avoidance region. The robust three-dimensional feasible region is projected along the target beam direction to obtain the two-dimensional effective projection region.

[0010] Preferably, generating the initial set of grid cell center coordinates within the two-dimensional effective projection area based on a preset minimum distance constraint between grid cells includes: When generating the initial set of grid cell center coordinates within the two-dimensional effective projection area, a hexagonal close-packing rule satisfying the minimum central axis spacing constraint is adopted, where the distance between any adjacent center coordinates is not less than the preset minimum central axis spacing d.min .

[0011] Preferably, when generating the initial set of grid cell center coordinates within the two-dimensional effective projection area, a hexagonal close-packing rule satisfying the minimum central axis spacing constraint is adopted, wherein the distance between any adjacent center coordinates is not less than the preset minimum central axis spacing d. min include: Apply a global translation and a global rotation to the initial set of center coordinates of the Grid cells to obtain a set of candidate center coordinates after rigid body transformation; Apply a local perturbation q to each center coordinate in the candidate center coordinate set. i And constrain the local disturbance q i The amplitude does not exceed the preset upper limit q max Furthermore, the center coordinates after local perturbation still lie within the two-dimensional effective projection area, and the distance between any two center coordinates is not less than the minimum central axis spacing d. min .

[0012] Preferably, the step of searching for a continuous available interval along the target beam direction for each candidate center coordinate, and generating a finite-length virtual Grid cylinder parallel to the target beam direction based on a continuous available interval with a length not less than a minimum length threshold, includes: When searching for a continuous usable interval along the target beam direction for each candidate center coordinate, only those intervals with a length not less than the minimum length threshold L are retained. min A continuous available interval; The near and far endpoints of the finite-length virtual Grid cylinders are determined based on the start and end positions of the continuous available intervals; the central axis of each generated finite-length virtual Grid cylinder is parallel to the direction of the target beam.

[0013] Preferably, the step of evaluating the candidate center coordinate scheme based on the Grid volume ratio corresponding to the finite-length virtual Grid cylinder, the minimum safe distance from the organ at risk, the cylinder separation degree, the weighted normal tissue overlap volume, and / or the bundle-oriented robust risk penalty term, and determining the target three-dimensional virtual Grid target area includes: The weighted normal tissue overlap volume V n,weighted Calculate using the following formula: V n,weighted =Σ k=1 K w k ·|G i ∩Ω O,k | Among them, w k G represents the weighting coefficient for the k-th organ at risk or planned risk volume. iLet Ω represent the i-th virtual grid cylinder of finite length. O,k This represents the set of voxels corresponding to the k-th volume of the organ at risk or the planned risk.

[0014] Preferably, in step S8, a lexicographical sorting rule or a comprehensive objective function J is used. pos The comprehensive objective function J is evaluated. pos Calculate using the following formula: J pos =λ1η+λ2δ+λ3σ-λ4V n,weighted -λ5P risk Among them, P risk The cumulative penalty term for the candidate center coordinates on the bundle-oriented robust risk map is represented by λ1 to λ5, which are non-negative weighting coefficients. η is the grid volume percentage, δ is the minimum safe distance, σ is the cylinder separation degree, and V is the weighting coefficient. n,weighted For weighted normal tissue overlap volume, P risk To constrain robust risk penalty items.

[0015] The second aspect of this disclosure provides a virtual grid layout system based on proton radiotherapy beam direction optimization, comprising: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method of any one of the first aspects.

[0016] The technical solution adopted in this invention is as follows: First, acquire the contour data of the patient's tumor target area, the contour data of the organ at risk (OAR) and / or the planned risk volume (PRV), and construct a first structural binary image, a contracted three-dimensional image, a second structural binary image, and an expanded three-dimensional image of the gross tumor volume, thereby completing the basic modeling of the target area, the danger zone, and the candidate grid layout area. Second, perform coarse and fine two-level discrete sampling of the candidate beam directions within a preset gantry angle range, calculate the water equivalent path length distribution through the target area for each candidate beam direction, and construct a beam direction robustness scoring function based on the proximal medium inhomogeneity and the distal organ at risk risk, thereby determining the patient-specific target beam direction. Third, using the target beam direction as the common central axis direction of all grid cylinders, construct a robust three-dimensional feasible region and a two-dimensional effective projection region in the beam-eye view coordinate system perpendicular to the target beam direction, thereby transforming the original three-dimensional grid layout problem into a two-dimensional center coordinate optimization problem within the two-dimensional effective projection region. Then, within the two-dimensional effective projection area, an initial set of grid cell center coordinates is generated using hexagonal close-packing. Multiple candidate center coordinate schemes are then formed through overall translation, overall rotation, and local perturbation. Next, for each candidate center coordinate, a continuous usable interval is searched along the target beam direction, retaining only intervals with a length not less than a minimum length threshold to generate a finite-length virtual grid cylinder. Finally, based on the grid volume ratio, the minimum safe distance to the organ at risk, cylinder separation, weighted normal tissue overlap volume, and beam-direction robustness risk penalty, the candidate center coordinate schemes are evaluated using lexicographical order or a comprehensive objective function to determine the target three-dimensional virtual grid target area, and the target beam direction and corresponding virtual grid arrangement results are output. This process has at least the following beneficial effects: (1) First, the candidate beam direction is selected for patient-specific robustness screening, and then a virtual grid is arranged under the selected target beam direction. This avoids the problem of beam direction mismatch with geometric structure caused by first arranging and then adapting, and improves the physical rationality and clinical feasibility of the initial proton grid scheme.

[0017] (2) The central axis of all Grid cylinders is uniformly constrained to be parallel to the direction of the target beam, thereby transforming the complex three-dimensional arrangement problem into a two-dimensional center coordinate optimization problem under the beam eye view. Then, the cylinder length is recovered by searching along the axial direction of the target beam. This not only maintains the clarity of the algorithm expression, but also reduces the solution complexity and enhances the reproducibility of the method.

[0018] (3) Simultaneously considering the inward shrinkage of the target area boundary, the lateral avoidance of organs at risk, the minimum spacing between cylinders, the minimum axial length, and the weighted protection of normal tissue, a unified evaluation of beam robustness, geometric feasibility, and clinical safety is achieved, which can more realistically reflect the comprehensive advantages and disadvantages of the proton grid scheme in actual treatment scenarios.

[0019] (4) The output results include not only the target beam direction parameters, but also the finite-length virtual grid cylinder profile, central axis coordinate set, parameterized arrangement matrix and structure file, which can be used as standardized geometric input for subsequent fixed gantry proton grid planning optimization, which helps to reduce manual planning differences and improve consistency and clinical reproducibility among different operators. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the present disclosure and form part of the specification. They are used together with the following detailed description to explain the present disclosure, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart illustrating a virtual grid arrangement method based on proton radiotherapy beam direction optimization, as shown in the embodiments of the specification. Detailed Implementation

[0021] The present invention will be further described in detail below with reference to the accompanying drawings and preferred embodiments. It should be understood that the following embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the present invention. The core of the present invention does not lie in optimizing a single geometric parameter in isolation, but in establishing a complete technical chain: first, robustly screening candidate beam directions; then, constraining the common axis of the virtual Grid cylinder with the selected beam direction; and finally, completing the Grid structure generation and the screening of the candidate center coordinate schemes within the same evaluation framework. Without departing from the core idea of ​​the present invention, those skilled in the art can make equivalent substitutions or combinations of the steps, parameters, and evaluation indicators.

[0022] Specifically, this disclosure provides a virtual grid arrangement method based on proton radiotherapy beam direction optimization, including: S1, acquire medical imaging data of the patient in the lesion area, tumor target volume contour data, organ at risk OAR and / or planned risk volume PRV contour data; S2, construct a first structural binary image of the gross tumor volume based on the tumor target area contour data, and perform inward shrinking processing on the first structural binary image to obtain an inward shrinking three-dimensional image; S3, construct a second structural binary image based on the organ at risk (OAR) and / or the planned risk volume (PRV) contour data, and perform outward expansion or safety avoidance processing on the second structural binary image to obtain an outward expanded three-dimensional image; S4. Among multiple candidate beam directions, a robustness score for the candidate beam direction is constructed based on the statistical distribution of the water equivalent path length on the beam path through the tumor target area, and the target beam direction is determined based on the robustness score. S5, using the target beam direction as the common central axis direction of all virtual Grid cylinders, map the inward 3D image and the outward 3D image onto a 2D beam eye view plane perpendicular to the target beam direction to construct a 2D effective projection area; S6. Within the two-dimensional effective projection area, an initial set of grid cell center coordinates is generated based on the preset minimum distance constraint between grid cells, and candidate center coordinate schemes are formed through overall translation, overall rotation and local perturbation. S7, search for a continuous available interval along the target beam direction for each candidate center coordinate, and generate a finite-length virtual Grid cylinder parallel to the target beam direction based on the continuous available interval with a length not less than the minimum length threshold. S8. The candidate center coordinate scheme is evaluated based on the Grid volume ratio corresponding to the finite length virtual Grid cylinder, the minimum safe distance from the organ at risk, the cylinder separation degree, the weighted normal tissue overlap volume and / or the bundle robust risk penalty term, and the target three-dimensional virtual Grid target area is determined. S9. Based on the beam direction parameters corresponding to the target beam direction and the target three-dimensional virtual grid target area, generate and output the virtual grid layout result for guiding external radiotherapy.

[0023] Preferably, the step of constructing a robustness score for candidate beam directions based on the statistical distribution of water equivalent path lengths along the beam path traversing the tumor target region among multiple candidate beam directions, and determining the target beam direction based on the robustness score, includes: Within a preset range of the frame angle, discrete sampling is performed with a first angle step size to obtain multiple candidate beam directions; The water equivalent path length of each candidate beam direction through the tumor target region is determined, and a beam direction robustness score is constructed based on the variance, standard deviation and / or quantile difference of the water equivalent path length. The top N candidate beam directions with the best beam robustness scores are determined as the optimized center angles, and fine sampling is performed in the neighborhood of the optimized center angles with a second angle step size smaller than the first angle step size to determine the target beam direction.

[0024] Preferably, the step of constructing a bundle robustness score based on the variance, standard deviation, and / or quantile difference of the water equivalent path length includes: For any candidate beam direction θ and the projection point y in the beam eye view plane, the water equivalent path length Lω(θ,y) in the uncertain scenario ω is calculated using the following formula: L ω (θ,y)=∫ℓ(θ,y;ω)ρ(x)dl Where ρ(x) represents the relative stopping power or water equivalent density at voxel x, and ℓ(θ,y;ω) represents the central ray in scene ω with candidate beam direction θ and passing through projection point y; since the fluctuation of water equivalent path length in different scenes ω reflects the sensitivity of the candidate beam direction to range uncertainty; Calculate the robust fluctuation term R WEPL (θ,y): R WEPL (θ,y)=Q M ({L ω})-Q N ({L ω}) Among them, Q M and Q N To avoid relying solely on range fluctuations and ignoring proximal tissue heterogeneity and distal organ risk in candidate beam orientation selection, a proximal media heterogeneity term R is introduced, representing the quantiles of M percent and N percent of the cross-scenario samples, respectively. het (θ,y) and the risk item R for distal organs at risk dist (θ,y), and establish the robust risk map R(θ,y) for the candidate beam direction using the following formula: R(θ,y)=αR WEPL (θ,y)+βR het (θ,y)+γR dist (θ,y) For a candidate beam direction θ, the distribution characteristics of the robust risk map R(θ,y) within its feasible projection area are statistically analyzed, and a beam direction scoring function J is constructed by combining the effective projection area ratio. angle (θ): J angle (θ)=Q 75 ({R(θ,y)|y∈F θ})-λ|F θ | / |P θ (Ω L )| Among them, F θ P represents the feasible projection region that satisfies the basic safety conditions under the candidate beam direction θ. θ (Ω L ) represents Ω L In the two-dimensional projection region under the candidate beam direction θ, select J angle The candidate beam direction with the smallest (θ) is taken as the target beam direction θ. * .

[0025] Preferably, the step of mapping the inward-expanding 3D image and the outward-expanding 3D image onto a 2D beam-eye view plane perpendicular to the target beam direction, using the target beam direction as the common central axis of all virtual Grid cylinders, to construct a 2D effective projection area includes: Perform three-dimensional morphological erosion on the candidate arrangement region within the target corresponding to the indented three-dimensional image; Two-dimensional expansion or outward expansion is performed on the organ at risk (OAR) and / or the planned risk volume (PRV) in a plane perpendicular to the direction of the target beam; The robust three-dimensional feasible region is obtained by performing a difference operation on the processed intra-target candidate deployment region and the endangered organ avoidance region. The robust three-dimensional feasible region is projected along the target beam direction to obtain the two-dimensional effective projection region.

[0026] Preferably, generating the initial set of grid cell center coordinates within the two-dimensional effective projection area based on a preset minimum distance constraint between grid cells includes: When generating the initial set of grid cell center coordinates within the two-dimensional effective projection area, a hexagonal close-packing rule satisfying the minimum central axis spacing constraint is adopted, where the distance between any adjacent center coordinates is not less than the preset minimum central axis spacing d. min .

[0027] Preferably, when generating the initial set of grid cell center coordinates within the two-dimensional effective projection area, a hexagonal close-packing rule satisfying the minimum central axis spacing constraint is adopted, wherein the distance between any adjacent center coordinates is not less than the preset minimum central axis spacing d. min include: Apply a global translation and a global rotation to the initial set of center coordinates of the Grid cells to obtain a set of candidate center coordinates after rigid body transformation; Apply a local perturbation q to each center coordinate in the candidate center coordinate set. i And constrain the local disturbance q i The amplitude does not exceed the preset upper limit q max Furthermore, the center coordinates after local perturbation still lie within the two-dimensional effective projection area, and the distance between any two center coordinates is not less than the minimum central axis spacing d. min .

[0028] Preferably, the step of searching for a continuous available interval along the target beam direction for each candidate center coordinate, and generating a finite-length virtual Grid cylinder parallel to the target beam direction based on a continuous available interval with a length not less than a minimum length threshold, includes: When searching for a continuous usable interval along the target beam direction for each candidate center coordinate, only those intervals with a length not less than the minimum length threshold L are retained. min A continuous available interval; The near and far endpoints of the finite-length virtual Grid cylinders are determined based on the start and end positions of the continuous available intervals; the central axis of each generated finite-length virtual Grid cylinder is parallel to the direction of the target beam.

[0029] Preferably, the step of evaluating the candidate center coordinate scheme based on the Grid volume ratio corresponding to the finite-length virtual Grid cylinder, the minimum safe distance from the organ at risk, the cylinder separation degree, the weighted normal tissue overlap volume, and / or the bundle-oriented robust risk penalty term, and determining the target three-dimensional virtual Grid target area includes: The weighted normal tissue overlap volume V n,weighted Calculate using the following formula: V n,weighted =Σ k=1 K w k ·|G i ∩Ω O,k | Among them, w k G represents the weighting coefficient for the k-th organ at risk or planned risk volume. i Let Ω represent the i-th virtual grid cylinder of finite length. O,k This represents the set of voxels corresponding to the k-th volume of the organ at risk or the planned risk.

[0030] Preferably, in step S8, a lexicographical sorting rule or a comprehensive objective function J is used. pos The comprehensive objective function J is evaluated. pos Calculate using the following formula: J pos =λ1η+λ2δ+λ3σ-λ4V n,weighted -λ5P risk Among them, P risk The cumulative penalty term for the candidate center coordinates on the bundle-oriented robust risk map is represented by λ1 to λ5, which are non-negative weighting coefficients. η is the grid volume percentage, δ is the minimum safe distance, σ is the cylinder separation degree, and V is the weighting coefficient. n,weighted For weighted normal tissue overlap volume, P risk To constrain robust risk penalty items.

[0031] The technical solution adopted in this invention is as follows: First, patient medical imaging data, tumor target area contour data, organ at-risk area (OAR) and / or planned risk volume (PRV) contour data are acquired. First structural binary images, inward-focused three-dimensional images, and second structural binary images and outward-focused three-dimensional images of the gross tumor volume are constructed to complete the basic modeling of the target area, danger zone, and candidate grid layout area. Second, candidate beam directions are subjected to coarse and fine two-level discrete sampling within a preset gantry angle range. The water equivalent path length distribution through the target area is calculated for each candidate beam direction. A beam direction robustness scoring function is constructed by combining proximal media inhomogeneity and distal organ at-risk risk, thereby determining the patient-specific target beam direction. Third, the target beam direction is used as the common central axis direction of all grid cylinders. A robust three-dimensional feasible region and a two-dimensional effective projection region are constructed in the beam-eye view coordinate system perpendicular to the target beam direction, thus transforming the original three-dimensional grid layout problem into a two-dimensional center coordinate optimization problem within the two-dimensional effective projection region. Then, within the two-dimensional effective projection area, an initial set of grid cell center coordinates is generated using hexagonal close-packing. Multiple candidate center coordinate schemes are then formed through overall translation, overall rotation, and local perturbation. Next, for each candidate center coordinate, a continuous usable interval is searched along the target beam direction, retaining only intervals with a length not less than a minimum length threshold to generate a finite-length virtual grid cylinder. Finally, based on the grid volume ratio, the minimum safe distance to the organ at risk, cylinder separation, weighted normal tissue overlap volume, and beam-direction robustness risk penalty, the candidate center coordinate schemes are evaluated using lexicographical order or a comprehensive objective function to determine the target three-dimensional virtual grid target area, and the target beam direction and corresponding virtual grid arrangement results are output. This process has at least the following beneficial effects: (1) First, the candidate beam direction is selected for patient-specific robustness screening, and then a virtual grid is arranged under the selected target beam direction. This avoids the problem of beam direction mismatch with geometric structure caused by first arranging and then adapting, and improves the physical rationality and clinical feasibility of the initial proton grid scheme.

[0032] (2) The central axis of all Grid cylinders is uniformly constrained to be parallel to the direction of the target beam, thereby transforming the complex three-dimensional arrangement problem into a two-dimensional center coordinate optimization problem under the beam eye view. Then, the cylinder length is recovered by searching along the axial direction of the target beam. This not only maintains the clarity of the algorithm expression, but also reduces the solution complexity and enhances the reproducibility of the method.

[0033] (3) Simultaneously considering the inward shrinkage of the target area boundary, the lateral avoidance of organs at risk, the minimum spacing between cylinders, the minimum axial length, and the weighted protection of normal tissue, a unified evaluation of beam robustness, geometric feasibility, and clinical safety is achieved, which can more realistically reflect the comprehensive advantages and disadvantages of the proton grid scheme in actual treatment scenarios.

[0034] (4) The output results include not only the target beam direction parameters, but also the finite-length virtual grid cylinder profile, central axis coordinate set, parameterized arrangement matrix and structure file, which can be used as standardized geometric input for subsequent fixed gantry proton grid planning optimization, which helps to reduce manual planning differences and improve consistency and clinical reproducibility among different operators.

[0035] The present disclosure will be described in detail through the following specific embodiments: For ease of description, a three-dimensional coordinate representation related to the target beam direction is established. Let the target beam direction be θ. * The corresponding unit direction vector is u * =u(θ * ); in the direction perpendicular to u * If we take two mutually orthogonal unit basis vectors e1 and e2 in the plane, then any spatial point x can be expressed as x = ξe1 + ηe2 + su * Where (ξ,η) is the lateral coordinate of the point in the two-dimensional beam eye view plane, and s is the longitudinal coordinate along the target beam direction. Based on this coordinate representation, the central axis direction of all virtual Grid cylinders is determined by u. * The unique determination, and the optimization of the cylinder position mainly manifests as the optimization of the center coordinates in the eye view plane and along u * The length of the direction is determined.

[0036] I. Data Preparation and Structural Modeling Let Ω be the three-dimensional voxel domain of the patient, and Ω be the set of voxels corresponding to the tumor target region. G The grid allows for the placement of target candidate regions of Ω. L The volume of the kth endangered organ or planned risk involved in the avoidance is denoted as Ω. O,k Where k = 1, 2, ..., K. For organs at risk participating in the weighted evaluation, a non-negative weight coefficient w is assigned. k This is used to reflect the relative importance of different organs in geometric avoidance. To ensure that the virtual grid cylinder is both located inside the target area and maintains a reasonable shrinkage margin with respect to the target area boundary, a volume adaptive shrinkage function m(V) is introduced to the target area boundary, where V represents the tumor target area volume and m(V) is a monotonically non-decreasing bounded function with respect to V.

[0037] m min ≤m(V)≤m max (1) Let the radius of the virtual grid cylinder be *a*, then the morphological erosion radius on the target side can be expressed as ρ. G =m(V)+a;Set the lateral avoidance radius ρ for the volume side of the organ at risk or the planned risk. O,k =m O,k +a, where m O,k Let m(V) be the minimum safe avoidance distance for the k-th endangered organ. The volume adaptive shrinkage function m(V) satisfies the value constraints shown in formula (1). Through the above definition, the subsequent construction of the robust three-dimensional feasible region can be unified with the target area geometry, cylinder radius, and endangered organ avoidance requirements.

[0038] II. Robust screening of candidate beam directions Within a preset range of the frame angle, coarse sampling is performed with a first angular step size Δθ1 to obtain the candidate beam direction set Θ={θ1,θ2,…,θ M For any candidate beam direction θ and a projection point y on the beam eye view plane, in the uncertainty scenario ω, the equivalent water path length L corresponding to the central ray path passing through this projection point is defined. ω (θ,y) is: L ω (θ,y)=∫ℓ(θ,y;ω)ρ(x)dl (2) Where ρ(x) represents the relative stopping power or water equivalent density at voxel x, and ℓ(θ,y;ω) represents the central ray with beam direction θ and passing through projection point y in scene ω. Since the fluctuation of the water equivalent path length under different scenes directly reflects the sensitivity of the beam direction to range uncertainty, the robust fluctuation term R can be further calculated according to formula (3) based on the water equivalent path length distribution obtained by formula (2). WEPL (θ,y).

[0039] R WEPL (θ,y)=Q 95 ({L ω})-Q 05 ({L ω}) (3) Among them, Q 95 and Q 05 These represent the 95th and 5th percentiles of the cross-scenario samples, respectively. To avoid relying solely on range fluctuations in beam-directed screening while ignoring proximal tissue heterogeneity and distal organ-at-risk risks, this invention further introduces a proximal media heterogeneity term R. het (θ,y) and the risk item R for distal organs at risk dist (θ,y), and based on this, construct the robust risk map R(θ,y) of the candidate beam direction according to formula (4).

[0040] R(θ,y)=αRWEPL (θ,y)+βR het (θ,y)+γR dist (θ,y) (4) For a given candidate beam direction θ, the distribution characteristics of the robust risk map R(θ,y) within its feasible projection area are statistically analyzed, and a beam direction scoring function J is constructed by combining the effective projection area ratio. angle (θ). Based on this, the beam direction scoring function J can be calculated according to formulas (4) and (5). angle (θ).

[0041] J angle (θ)=Q 75 ({R(θ,y)|y∈F θ})-λ|F θ | / |P θ (Ω L (5) Among them, F θ P represents the feasible projection region that satisfies the basic safety conditions under the candidate beam direction θ. θ (Ω L ) represents Ω L The two-dimensional projection region under the candidate beam direction θ. Select J... angle The candidate beam direction with the smallest (θ) is taken as the target beam direction θ. * Preferably, based on the coarse sampling results, further local fine sampling can be performed with a second angular step size Δθ2, where Δθ2 < Δθ1. Therefore, the target beam direction θ * The determination of is not based on empirical selection, but is automatically completed based on the bundle robustness evaluation results corresponding to formula (5).

[0042] III. Robust 3D Feasible Region Construction under Target Beam Direction Constraints In the target beam direction θ * Once confirmed, use u * =u(θ * The direction of the central axis is used as the unified central axis for all Grid cylinders. For each uncertainty scenario ω, first, three-dimensional morphological erosion is performed on the candidate region within the target, and then the volume of the organ at risk or the planned risk is adjusted in relation to u. * Perform two-dimensional dilation within the vertical plane and then perform a difference operation to obtain the scene-dependent three-dimensional feasible region Ω. F (ω) (θ * Among them, the corrosion radius ρ on the target area side. G The determination of can be completed by combining the contraction function m(V) in formula (1).

[0043] Ω F (ω) (θ* )=Erode(Ω L (ω) ,ρ G )∖⋃ k=1 K Dil ⊥u* (Ω O,k (ω) ,ρ O,k (6) To improve the stability of the layout results under disturbances in multiple scenarios, the intersection of the three-dimensional feasible regions under all scenarios is taken to obtain the robust three-dimensional feasible region Ω. F,rob (θ * That is, formula (7) is a robust representation obtained by taking the intersection of all uncertain scenarios based on formula (6).

[0044] Ω F,rob (θ * )=⋂ ω∈U Ω F (ω) (θ * (7) Here, U represents the set of uncertain scenarios. By adopting this intersection definition, only voxel regions that remain feasible under all considered perturbation scenarios will be retained, thus ensuring that the subsequently generated virtual Grid cylinder has better geometrical scenario robustness.

[0045] IV. Generation of Two-Dimensional Effective Projection Region, Hexagonal Closest Packing, and Candidate Center Coordinates The robust three-dimensional feasible region along u * Direction projection to u * An orthogonal two-dimensional eye-view plane yields the two-dimensional projection region Π(θ). * Therefore, formula (8) transforms the original three-dimensional feasible region into a two-dimensional projection representation under the eye view, providing a unified geometric basis for subsequent center coordinate optimization.

[0046] Π(θ * )={(ξ,η)|∃s,ξe1+ηe2+su * ∈Ω F,rob (θ * )} (8) For any projection point (ξ,η), if its path along u * The direction exists within the robust three-dimensional feasible domain with a length not less than L. min If the continuous available interval is defined, it is retained within the two-dimensional effective projection region Π. L,min (θ * ). In Π L,min (θ *The initial center coordinate set is generated by densely packing hexagons within the two-dimensional effective projection area to improve the filling efficiency of the irregular target area projection. Based on the two-dimensional effective projection area, the initial center coordinates can be generated according to formula (9).

[0047] c mn (0) =c0+m[h,0] T +n[h / 2,√3h / 2] T (9) Where m and n are integers, and h is the basic lattice spacing, satisfying h ≥ d min To accommodate different patient anatomical structures, a global rotation and translation are applied to the initial set of central coordinates to obtain a rigidly transformed set of central coordinates; subsequently, a local perturbation q is applied to each central coordinate. i This is to compensate for the local geometric mismatch caused by irregular boundaries and adjacent endangered organs. Subsequently, the coordinates of each center are subjected to overall rigid body transformation and local perturbation according to formula (10).

[0048] c i =R(φ)c i (0) +t+q i , ‖q i ||≤q max (10) Any two candidate center coordinates must satisfy the minimum center axis spacing constraint to prevent the virtual Grid cylinders from excessively clustering and destroying the preset peak-valley structure. Therefore, formula (11) is used to ensure that any two cylinders satisfy the minimum spacing constraint.

[0049] ||c i -c j ‖≥d min , i≠j (11) V. Axial search for column formation along the target beam direction For any candidate center coordinate c i =(ξ i ,η i ), along u * The direction searches for its corresponding set of available intervals S in the robust three-dimensional feasible region. i Formula (12) defines the available axial range along the target beam direction.

[0050] S i ={s|ξ i e1+η i e2+su * ∈Ω F,rob (θ * (12) S iDecomposed into several non-overlapping continuous intervals [s] i,ℓ - ,s i,ℓ + Only retain those with a length not less than L. min The interval is selected, and the interval with the largest length or the best overall evaluation is preferably selected as the axial effective interval of the i-th virtual Grid cylinder [s]. i - ,s i + Based on this, a virtual Grid cylinder G of finite length can be constructed further according to formula (13). i .

[0051] G i ={x|‖(Iu * {u *} T (xx) i,0 )‖≤a, s i - ≤{u *} T x≤s i +} (13) Each virtual Grid cylinder generated in this way has a central axis parallel to the target beam direction and automatically satisfies the available length constraints in both the near and far directions. This step ensures that the length of the Grid cylinder is no longer dependent on a fixed template, but is determined by both the target beam direction and the robust 3D feasible region.

[0052] VI. Evaluation of the candidate center coordinate scheme and determination of the target 3D virtual grid target area The union of all the remaining finite-length virtual Grid cylinders is denoted as the candidate virtual Grid structure G = ⋃ i=1 N G i To select the optimal scheme from the multiple candidate center coordinate schemes, the Grid volume proportion η, the minimum safety distance δ, and the cylinder separation degree σ are defined respectively. Accordingly, the Grid volume proportion, the minimum safety distance, and the cylinder separation degree can be calculated according to formula (14) and formula (15) respectively.

[0053] η=|G∩Ω L | / |Ω L | (14) δ=min i,k dist(G i ,Ω O,k ), σ=min i≠j dist(Axis(G i ),Axis(G j(15) Meanwhile, to reflect the differences in importance of different organs at risk, a weighted overlap volume of normal tissue V was introduced. n,weighted Based on this, the weighted overlapping volume of normal tissues for different organs at risk can be calculated according to formula (16), so that the evaluation of the candidate center coordinate scheme not only reflects the geometric occupancy, but also reflects the priority of protection of critical organs.

[0054] V n,weighted =Σ k=1 K w k ·|G∩Ω O,k | (16) Preferably, hard constraints are first used to eliminate those that do not satisfy δ≥δ min , σ≥σ min And the length of each cylinder is not less than L min The candidate center coordinate schemes are then compared using lexicographical order or by a comprehensive objective function J. pos Sort the candidate center coordinate schemes. Finally, the candidate center coordinate schemes can be sorted according to formula (17).

[0055] J pos =λ1η+λ2δ+λ3σ-λ4V n,weighted -λ5P risk (17) Among them, P risk The cumulative penalty term for the candidate center coordinates on the bundle-direction robustness risk map is represented by λ1 to λ5, which are non-negative weight coefficients. The candidate center coordinate scheme with the optimal evaluation result is determined as the target 3D virtual Grid target area. Thus, this invention completes a closed-loop optimization process from bundle-direction robustness screening to geometrically feasible region constraints, and then to comprehensive evaluation of candidate Grid structures.

[0056] VII. Results Output and System Calls The output of this invention may include: target beam direction θ * , and u * The system provides a parallel, finite-length virtual grid cylindrical profile, a set of central axis coordinates, a parametric arrangement matrix, and a structure file directly accessible to the proton therapy planning system. This structure file can be further used for robust spot strength optimization at a fixed gantry angle, thus forming a complete proton grid therapy plan. Therefore, it not only provides a geometry construction method but also offers standardized and reusable initial inputs for subsequent dose optimization.

[0057] Example Taking a patient with a skull base tumor as an example, the patient's planned CT scan, GTV, brainstem, spinal cord, and optic nerve structures are first imported. The virtual Grid cylinder radius is set to a = 4 mm, and the minimum central axis spacing is set to d.min =12 mm, minimum length threshold L min =10 mm; coarse sampling is performed on the gantry angle with a step size of 5°, and fine sampling is performed again with a step size of 1° in the neighborhood of the candidate center angle. After robust scoring of the candidate beam direction using formulas (2) to (5), a target beam direction θ with small fluctuation in water equivalent path length and low risk of distal organ endangerment can be determined. * .

[0058] Subsequently, with θ * A beam-eye view coordinate system is established for the corresponding beam direction. The candidate region within the target is shrunken, and the corresponding PRVs of the brainstem, spinal cord, and optic nerve are subjected to lateral safety avoidance in a plane perpendicular to the beam direction. After projecting the obtained robust three-dimensional feasible region, an initial set of center coordinates is generated by the densest packing of hexagons within the two-dimensional effective projection region. Multiple candidate center coordinate schemes are obtained through global rotation, global translation, and local perturbation. The robust three-dimensional feasible region and the two-dimensional effective projection region correspond to formulas (6) to (8), respectively, and the generation and optimization of the initial center coordinates correspond to formulas (9) to (11).

[0059] For each candidate center coordinate along θ * After searching for a continuous available interval in the directional search, a finite-length virtual grid cylinder that meets the length threshold can be automatically generated. By incorporating brainstem and spinal cord with higher weighting coefficients into the weighted normal tissue evaluation, the algorithm can automatically eliminate candidate cylinders that are too close to critical organs at risk, and retain virtual grid schemes with higher geometric feasibility and better beam-direction robustness within the target area. The final output includes the target beam direction θ. * The set of all finite-length virtual Grid cylinder outlines and their central axis coordinates. The generation of finite-length virtual Grid cylinders and the evaluation of the candidate central coordinate schemes correspond to formulas (12) to (17), respectively.

[0060] This disclosure also provides a virtual grid layout system based on proton radiotherapy beam direction optimization, including: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of any of the methods described in the foregoing embodiments.

[0061] The above embodiments are merely preferred embodiments of the present invention. Any equivalent substitutions or modifications made by those skilled in the art to the formula form, evaluation indicators, search strategies, output formats, and interface methods with subsequent planning systems without departing from the spirit and scope of the present invention should fall within the scope of protection of the present invention.

Claims

1. A virtual grid layout method based on proton radiotherapy beam direction optimization, characterized in that, The method includes: S1, acquire tumor target volume contour data, organ at risk OAR and / or planned risk volume PRV contour data of the patient in the lesion area; S2, construct a first structural binary image of the gross tumor volume based on the tumor target area contour data, and perform inward shrinking processing on the first structural binary image to obtain an inward shrinking three-dimensional image; S3, construct a second structural binary image based on the organ at risk (OAR) and / or the planned risk volume (PRV) contour data, and perform outward expansion or safety avoidance processing on the second structural binary image to obtain an outward expanded three-dimensional image; S4. Among multiple candidate beam directions, a robustness score for the candidate beam direction is constructed based on the statistical distribution of the water equivalent path length on the beam path through the tumor target area, and the target beam direction is determined based on the robustness score. S5, using the target beam direction as the common central axis direction of all virtual Grid cylinders, map the inward 3D image and the outward 3D image onto a 2D beam eye view plane perpendicular to the target beam direction to construct a 2D effective projection area; S6. Within the two-dimensional effective projection area, an initial set of grid cell center coordinates is generated based on the preset minimum distance constraint between grid cells, and candidate center coordinate schemes are formed through overall translation, overall rotation and local perturbation. S7, search for a continuous available interval along the target beam direction for each candidate center coordinate, and generate a finite-length virtual Grid cylinder parallel to the target beam direction based on the continuous available interval with a length not less than the minimum length threshold. S8. The candidate center coordinate scheme is evaluated based on the Grid volume ratio corresponding to the finite length virtual Grid cylinder, the minimum safe distance from the organ at risk, the cylinder separation degree, the weighted normal tissue overlap volume and / or the bundle robust risk penalty term, and the target three-dimensional virtual Grid target area is determined. S9. Based on the beam direction parameters corresponding to the target beam direction and the target three-dimensional virtual grid target area, generate and output the virtual grid layout result for guiding external radiotherapy.

2. The method according to claim 1, characterized in that, The process of constructing a robustness score for candidate beam directions based on the statistical distribution of water equivalent path lengths along the beam path traversing the tumor target region, and determining the target beam direction based on the robustness score, includes: Within a preset range of the frame angle, discrete sampling is performed with a first angle step size to obtain multiple candidate beam directions; The water equivalent path length of each candidate beam direction through the tumor target region is determined, and a beam direction robustness score is constructed based on the variance, standard deviation and / or quantile difference of the water equivalent path length. The top N candidate beam directions with the best beam robustness scores are determined as the optimized center angles, and fine sampling is performed in the neighborhood of the optimized center angles with a second angle step size smaller than the first angle step size to determine the target beam direction.

3. The method according to claim 2, characterized in that, The construction of the bundle robustness score based on the variance, standard deviation, and / or quantile difference of the equivalent water path length includes: For any candidate beam direction θ and the projection point y in the beam eye view plane, the water equivalent path length Lω(θ,y) in the uncertain scenario ω is calculated using the following formula: L ω (θ,y)=∫ℓ(θ,y;ω)ρ(x)dl Where ρ(x) represents the relative stopping power or water equivalent density at voxel x, and ℓ(θ,y;ω) represents the central ray in scene ω with candidate beam direction θ and passing through projection point y; since the fluctuation of water equivalent path length in different scenes ω reflects the sensitivity of the candidate beam direction to range uncertainty; Calculate the robust fluctuation term R WEPL (θ,y): R WEPL (θ,y)=Q M ({L ω })-Q N ({L ω }) Among them, Q M and Q N To avoid relying solely on range fluctuations and ignoring proximal tissue heterogeneity and distal organ risk in candidate beam orientation selection, a proximal media heterogeneity term R is introduced, representing the quantiles of M percent and N percent of the cross-scenario samples, respectively. het (θ,y) and the risk item R for distal organs at risk dist (θ,y), and establish the robust risk map R(θ,y) for the candidate beam direction using the following formula: R(θ,y)=αR WEPL (θ,y)+βR het (θ,y)+γR dist (θ,y) For a candidate beam direction θ, the distribution characteristics of the robust risk map R(θ,y) within its feasible projection area are statistically analyzed, and a beam direction scoring function J is constructed by combining the effective projection area ratio. angle (θ): J angle (θ)=Q 75 ({R(θ,y)|y∈F θ })-λ|F θ | / |P θ (Oh L )| Among them, F θ P represents the feasible projection region that satisfies the basic safety conditions under the candidate beam direction θ. θ (Ω L ) represents Ω L In the two-dimensional projection region under the candidate beam direction θ, select J angle The candidate beam direction with the smallest (θ) is taken as the target beam direction θ. * .

4. The method according to claim 1, characterized in that, The method of mapping the inward-expanding 3D image and the outward-expanding 3D image onto a 2D beam-eye view plane perpendicular to the target beam direction, using the target beam direction as the common central axis of all virtual Grid cylinders, to construct a 2D effective projection area includes: Perform three-dimensional morphological erosion on the candidate arrangement region within the target corresponding to the indented three-dimensional image; Two-dimensional expansion or outward expansion is performed on the organ at risk (OAR) and / or the planned risk volume (PRV) in a plane perpendicular to the direction of the target beam; The robust three-dimensional feasible region is obtained by performing a difference operation on the processed intra-target candidate deployment region and the endangered organ avoidance region. The robust three-dimensional feasible region is projected along the target beam direction to obtain the two-dimensional effective projection region.

5. The method according to claim 1, characterized in that, The step of generating an initial set of Grid cell center coordinates within the two-dimensional effective projection area based on a preset minimum distance constraint between Grid cells includes: When generating the initial set of grid cell center coordinates within the two-dimensional effective projection area, a hexagonal close-packing rule satisfying the minimum central axis spacing constraint is adopted, where the distance between any adjacent center coordinates is not less than the preset minimum central axis spacing d. min .

6. The method according to claim 5, characterized in that, When generating the initial set of Grid cell center coordinates within the two-dimensional effective projection area, a hexagonal close-packing rule that satisfies the minimum central axis spacing constraint is adopted, and the distance between any adjacent center coordinates is not less than the preset minimum central axis spacing d. min include: Apply a global translation and a global rotation to the initial set of center coordinates of the Grid cells to obtain a set of candidate center coordinates after rigid body transformation; Apply a local perturbation q to each center coordinate in the candidate center coordinate set. i And constrain the local disturbance q i The amplitude does not exceed the preset upper limit q max Furthermore, the center coordinates after local perturbation still lie within the two-dimensional effective projection area, and the distance between any two center coordinates is not less than the minimum central axis spacing d. min .

7. The method according to claim 1, characterized in that, The step of searching for a continuous available interval along the target beam direction for each candidate center coordinate, and generating a finite-length virtual Grid cylinder parallel to the target beam direction based on a continuous available interval with a length not less than a minimum length threshold, includes: When searching for a continuous usable interval along the target beam direction for each candidate center coordinate, only those intervals with a length not less than the minimum length threshold L are retained. min Continuous available intervals; The near and far endpoints of the finite-length virtual Grid cylinders are determined based on the start and end positions of the continuous available intervals; the central axis of each generated finite-length virtual Grid cylinder is parallel to the direction of the target beam.

8. The method according to claim 1, characterized in that, The evaluation of the candidate center coordinate scheme based on the Grid volume ratio corresponding to the finite-length virtual Grid cylinder, the minimum safe distance from the organ at risk, the cylinder separation degree, the weighted normal tissue overlap volume, and / or the bundle-oriented robust risk penalty term, to determine the target three-dimensional virtual Grid target area includes: The weighted normal tissue overlap volume V n,weighted Calculate using the following formula: V n,weighted =S k=1 K w k ·|G i ∩Oh O,k | Among them, w k G represents the weighting coefficient for the k-th organ at risk or planned risk volume. i Let Ω represent the i-th virtual grid cylinder of finite length. O,k This represents the set of voxels corresponding to the k-th volume of the organ at risk or the planned risk.

9. The method according to any one of claims 1-8, characterized in that, S8 employs either a lexicographical sorting rule or a comprehensive objective function J. pos The comprehensive objective function J is evaluated. pos Calculate using the following formula: J pos =λ1η+λ2δ+λ3σ-λ4V n,weighted -λ5P risk Among them, P risk The cumulative penalty term for the candidate center coordinates on the bundle-oriented robust risk map is represented by λ1 to λ5, which are non-negative weighting coefficients. η is the grid volume percentage, δ is the minimum safe distance, σ is the cylinder separation degree, and V is the weighting coefficient. n,weighted For weighted normal tissue overlap volume, P risk This is a robust risk penalty item.

10. A virtual grid layout system based on proton radiotherapy beam direction optimization, characterized in that, include: A memory on which computer programs are stored; A processor for executing the computer program in the memory to implement the steps of the method according to any one of claims 1 to 9.

Images