Intelligent generation and optimization system for Lattice radiotherapy globule target region based on VMAT

Abstract

The invention discloses a VMAT-based Lattice radiotherapy globule target region intelligent generation and optimization system, and belongs to the crossing field of space segmentation radiotherapy and medical image processing technologies. According to the method, a genetic algorithm is introduced, the problem of a local optimal solution is avoided through global search and optimization of intra-layer and inter-layer spacing parameters, the distance between the center of a new small ball and an existing small ball is quickly checked by further combining a K-D tree algorithm, it is ensured that spacing constraints are met, the calculation complexity is remarkably reduced, the calculation efficiency is improved, and the method is suitable for large-scale popularization and application. Meanwhile, a more reasonable small ball target area distribution structure can be realized; the small balls can be more reasonably distributed in the target area through the genetic algorithm and the spacing parameter optimized by the K-D tree, the uncovered area is reduced, the generation efficiency of the small ball target area is remarkably improved through the combination of the genetic algorithm and the K-D tree algorithm, the layout of the small ball target area is optimized, the target area coverage rate and the peak-valley dose ratio are improved, and the method is suitable for large-scale popularization and application. And a more efficient and optimized solution is provided for clinical radiotherapy.

Description

TECHNICAL FIELD

[0001] The present application relates to a VMAT-based Lattice radiotherapy small ball target intelligent generation and optimization system, belonging to the cross field of spatially fractionated radiation therapy and medical image processing technology. BACKGROUND

[0002] Spatially Fractionated Radiation Therapy (SFRT) as a unique radiotherapy method has a history of more than 100 years. Its principle is different from that of traditional radiotherapy, which treats tumors through uneven dose distribution, effectively controls tumors, and better protects surrounding normal tissues, especially for the treatment of large malignant tumors. In the 1990s, GRID therapy used MV photon beams to treat large or recurrent tumors in patients who had received radiotherapy before, and achieved good oncology results without observing significant toxicity. It is usually performed using a conventional linear accelerator with a GRID block, which can be independently customized or commercially manufactured, but there are technical challenges in using it and following the beam divergence.

[0003] To solve the problem of high radiation dose outside the clinical target volume in 2D GRID treatment, Wu et al. proposed the concept of Lattice Radiation Therapy (LRT), which is a 3D extension of 2D GRID technology, with vertices strictly controlled within the tumor volume. LRT is based on the clinical effectiveness of GRID therapy and new developments in physics and technology, and can create a three-dimensional high-dose region within the tumor volume, focusing on small ball target regions called vertices, which has been widely used in clinical practice and has improved local control without increasing toxicity.

[0004] However, in the radiotherapy plan design of VMAT-based spatially fractionated radiotherapy LRT, the placement of small ball target regions is very complex and time-consuming. Physicists need to manually place small ball target regions within the PTV, and operate according to specific delineation principles, but when the target region is large, irregular in shape, and needs to be placed with many small ball target regions, it is easy to fall into local traps and cannot be considered globally, resulting in unsatisfactory ball placement, and the placed small ball target regions may only meet local requirements, or the number of placed small ball target regions is reduced, thereby making the peak-to-valley dose ratio (PVDR) in the target region not meet the dose requirements. The entire process is often time-consuming, which seriously affects the efficiency and quality of radiotherapy plan design. SUMMARY

[0005] In order to improve the accuracy of small ball target region layout and thus improve the target region coverage rate and peak-to-valley dose ratio, the present application provides a VMAT-based Lattice radiotherapy small ball target intelligent generation and optimization system, the technical solution of which is as follows:

[0006] The first object of the present application is to provide a VMAT-based Lattice radiotherapy bead target intelligent generation and optimization method, comprising:

[0007] Step 1: Obtain patient DICOM CT images and Dicom RT Structure files, extract normal tissue OAR structure and target PTV structure;

[0008] Step 2: In the target PTV structure, initialize the population by using genetic algorithm, each individual contains a set of layer spacing and interlayer spacing parameters, and obtain the initial bead target candidate point;

[0009] Step 3: Check whether the candidate point meets the distance constraint by using K-D tree, and generate a bead target that meets the constraint condition;

[0010] Step 4: Calculate the fitness of the current target bead according to the number of beads, target coverage, distance variance and peak-to-valley dose ratio;

[0011] Step 5: Generate a new population by selection, crossover and mutation, repeat steps 3 and 4 until the preset iteration number is completed;

[0012] Step 6: Select the bead target distribution structure corresponding to the highest fitness value as the final optimization result.

[0013] Optionally, the distance constraint includes: layer-intra-bead target center spacing, layer-inter-bead target center spacing, and bead edge and target PTV edge spacing.

[0014] Optionally, the method further comprises: eroding the target PTV by using an erosion function to ensure that the distance constraint is met between the bead target edge and the target PTV edge.

[0015] The second object of the present application is to provide a VMAT-based Lattice radiotherapy bead target intelligent generation and optimization system, comprising:

[0016] A data import module is configured to import patient DICOM CT images and Dicom RT Structure files, and extract normal tissue OAR structure and target PTV structure;

[0017] A parameter setting module is configured to set the bead target distance constraint parameter and the bead target optimization mode parameter;

[0018] The small ball target area generation and optimization module selects whether to use a genetic algorithm to optimize the small ball target area according to the small ball target area optimization mode parameter, when the genetic algorithm is used, a genetic algorithm optimization module is called to optimize the small ball target area distribution structure, otherwise, a K-D tree optimization module is used for optimization;

[0019] The K-D tree optimization module verifies the distance constraint between the candidate small ball and the existing small ball in real time based on the three-dimensional space data structure;

[0020] The genetic algorithm optimization module optimizes the small ball target area distribution structure by using the genetic algorithm optimization method described above;

[0021] The small ball generation module generates a compliant small ball target area in the target area PTV structure according to the optimization parameters;

[0022] The result output module generates a DICOM RT Structure file containing the small ball contour, center coordinates and statistical information.

[0023] Optionally, the K-D tree optimization module includes the following steps:

[0024] The target area mask is eroded, the area of each layer is calculated, the layer with the maximum area is selected as the initial center layer and the geometric center is obtained;

[0025] Random sampling is performed based on the eroded mask to generate candidate points;

[0026] A small ball mask is generated with the candidate point as the center, and it is checked whether it is completely located within the eroded target area and does not overlap with the organ at risk;

[0027] The K-D tree algorithm is used to verify whether the candidate point and the existing small ball satisfy the distance constraint, and the candidate point that satisfies the constraint condition is retained;

[0028] After the effective small ball is generated, candidate points are supplemented in its vicinity layer, and when the number of candidate points is insufficient, they are supplemented from the high area layer;

[0029] The DICOM RT Structure file containing the small ball contour is output.

[0030] Optionally, the system further includes a preview visualization module.

[0031] Optionally, the system further includes a statistical information generation module for generating statistical information of the target area small ball, including the number of CT layers involved and the three-dimensional coordinates of the small ball center.

[0032] Optionally, in the small ball target area distance constraint parameter, the interval range of the small ball target area center in the layer is 5-7 cm, the interval range of the small ball target area center between layers is 3-5 cm, and the default interval distance between the small ball edge and the target area PTV edge is 3 mm.

[0033] A third object of the present application is to provide a VMAT-based Lattice radiotherapy bead target intelligent generation and optimization device, comprising a memory and a processor;

[0034] The memory is used for storing a computer program;

[0035] The processor is used for realizing the VMAT-based Lattice radiotherapy bead target intelligent generation and optimization method according to any one of the above when the computer program is executed.

[0036] A fourth object of the present application is to provide a computer readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, the VMAT-based Lattice radiotherapy bead target intelligent generation and optimization method according to any one of the above is realized.

[0037] The present application has the following advantages:

[0038] The present application significantly improves the generation efficiency and quality of the spatial segmentation radiotherapy bead target through the synergistic effect of the genetic algorithm and the K-D tree algorithm.

[0039] The present application introduces the genetic algorithm, finds the optimal bead distribution scheme by globally searching and optimizing the intra-layer and inter-layer spacing parameters, avoids the problem of local optimal solution, further combines the K-D tree algorithm, quickly checks the distance between the new bead center and the existing beads through efficient space division and nearest neighbor search, ensures to meet the spacing constraint, significantly reduces the computational complexity, and improves the calculation efficiency.

[0040] The present application introduces the genetic algorithm, globally searches the optimal spacing parameter combination through simulating natural selection and genetic mechanism, maximizes the number of beads, ensures the uniform distribution of beads and meets the spacing constraint, and the K-D tree algorithm quickly verifies whether the bead target meets the distance constraint during the bead generation process, ensures high calculation efficiency, and realizes more reasonable bead target distribution structure.

[0041] The spacing parameters optimized by the genetic algorithm enable the beads to be more reasonably distributed in the target area, reduce the uncovered area, thereby improving the coverage rate of the target area, and at the same time, the K-D tree algorithm ensures that the placement of each bead meets the clinical requirements through fast distance checking, further improves the coverage rate of the target area; in addition, the optimized bead target distribution structure can better control the distribution of high-dose and low-dose regions, thereby indirectly improving the peak-to-valley dose ratio PVDR. Through global optimization of the genetic algorithm and efficient distance checking of the K-D tree, the present application can realize more optimal bead distribution, improve the treatment effect, and at the same time reduce the damage to the surrounding normal tissues.

[0042] In summary, the present application combines genetic algorithm and K-D tree algorithm, which not only significantly improves the generation efficiency of the small ball target region, but also optimizes the layout of the small ball target region, improves the target region coverage and peak-to-valley dose ratio, and provides a more efficient and optimized solution for clinical radiotherapy. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0044] Figure 1 is a flow chart of the intelligent generation and optimization method of the Lattice radiotherapy small ball target region based on VMAT of the present application.

[0045] Figure 2 is an operation interface diagram of the intelligent generation and optimization system of the Lattice radiotherapy small ball target region based on VMAT of the present application.

[0046] Figure 3 is a schematic diagram of the intelligent generation of the small ball target region from different perspectives.

[0047] Figure 4 is a visualization interface diagram of the small ball target region generation of the present application.

[0048] Figure 5 is a small ball target region dose distribution and dose volume histogram generated by the system of the present application.

[0049] Figure 6 is a small ball target region peak-to-valley dose ratio (PVDR) result diagram generated by the system of the present application.

[0050] Figure 7 is a small ball target region peak-to-valley dose ratio (PVDR) result diagram generated by the system of the present application.

[0051] Figure 8 is a small ball target region peak-to-valley dose ratio (PVDR) result diagram generated by the system of the present application.

[0052] Figure 9 is a small ball target region treatment plan Portal Dosimetry dose verification result diagram generated by the system of the present application. DETAILED DESCRIPTION

[0053] In order to make the purpose, technical solutions and advantages of the present application more clear, the embodiments of the present application will be further described in detail below.

[0054] Embodiment one:

[0055] The embodiment provides a VMAT-based Lattice radiotherapy bead target intelligent generation and optimization method, which comprises the following steps:

[0056] Step 1: Obtain patient DICOM CT images and Dicom RT Structure files, and extract normal tissue OAR structure and target PTV structure.

[0057] The DICOM CT image can provide information such as pixel spacing and slice thickness; the Dicom RT Structure file is used to store or add the contour information of the bead target; the normal tissue OAR structure is specifically a mask dictionary of organs at risk (OARs), which is used to avoid the overlap of the bead target and the organs at risk; and the target PTV structure is specifically a three-dimensional mask of the tumor target (ROI).

[0058] Step 2: Generate a target bead with a diameter of 1.5 cm in the target PTV structure.

[0059] First, the target PTV is eroded by using an erosion function, so that the distance constraint between the edge of the bead target and the edge of the target PTV is met, and the distance constraint in the embodiment is 0.3 cm by default.

[0060] Then, a population is initialized by using a genetic algorithm, 20 individuals (population_size=20) are generated, each individual contains a set of layer spacing (5-7 cm) and layer spacing (3-5 cm) parameters, and initial bead target candidate points are obtained.

[0061] Step 3: Check whether the candidate points meet the distance constraint by using a K-D tree, and generate a bead target that meets the constraint condition; the distance constraint includes: layer spacing between bead target centers, layer spacing between bead target centers, and spacing between bead edge and target PTV edge.

[0062] Step 4: Calculate the fitness of the current target bead according to the number of beads, target coverage, distance variance and peak valley dose ratio, and the calculation method is: fitness function=bead number*0.3+target coverage*0.2+distance variance*0.3+PVDR*0.2.

[0063] Step 5: A new population is generated by selection, crossover and mutation, steps 3 and 4 are repeated until the preset number of iterations is completed.

[0064] Wherein, the selection operation is: using roulette selection to select the individual with higher fitness; the crossover operation is: taking the average of the distance between two parents with a probability of 0.8 to generate offspring; the mutation operation is: randomly disturbing the distance within a range of ±0.5cm with a probability of 0.2, and limiting in the effective range (5-7cm and 3-5cm); the iteration is stopped after running 20 times.

[0065] Step 6: selecting the small ball target region distribution structure corresponding to the highest fitness value as the final optimization result, and outputting the result including: the generated three-dimensional mask of each small ball target region, the small ball statistical information (center coordinates, diameter, volume, etc.), the statistics of the overlap between the small ball and the OAR, and the DICOM RT Structure file containing the small ball target region contour information, which is compatible with the treatment planning system (TPS).

[0066] Embodiment two:

[0067] The embodiment provides a Lattice radiotherapy small ball target region intelligent generation and optimization system based on VMAT, which comprises a data import module, a parameter setting module, a small ball target region generation and optimization module, a K-D tree optimization module, a genetic algorithm optimization module, a small ball generation module and a result output module.

[0068] The data import module is used for importing patient DICOM CT images and Dicom RT Structure files, and extracting normal tissue OAR structures and target region PTV structures.

[0069] The parameter setting module is used for setting small ball target region distance constraint parameters and small ball target region optimization mode parameters; wherein the small ball target region distance constraint parameters mainly include the small ball target region center interval in the layer, the small ball target region center interval between layers and the interval between the small ball edge and the target region PTV edge, and the small ball target region optimization mode parameter (optimize) is used for determining whether to enable the genetic algorithm optimization module.

[0070] The small ball target region generation and optimization module selects whether to adopt the genetic algorithm to optimize the small ball target region according to the optimize parameter; when the genetic algorithm is adopted, the genetic algorithm optimization module is called to optimize the small ball target region distribution structure, otherwise the K-D tree optimization module is adopted for optimization.

[0071] The K-D tree optimization module verifies the distance constraint of the candidate small ball and the existing small ball based on the three-dimensional space data structure in real time, and the specific optimization process is as follows:

[0072] The target region mask is subjected to erosion treatment to ensure that the distance between the small ball and the target region boundary is at least 0.3cm, and the area of each layer is calculated, the layer with the maximum area is selected as the initial center layer and the geometric center is obtained;

[0073] In larger layers (area > 5000 pixels), candidate points are randomly sampled using the eroded boundary mask, with points far away from existing balls being given priority.

[0074] Generate a small ball mask with the candidate point as the center to check whether it is completely within the erosion target area and does not overlap with the risk organ;

[0075] Use the KD tree algorithm to verify whether the candidate points and the existing balls meet the distance constraints, and retain the candidate points that meet the constraints; at the same time, provide an alternative distance calculation method (such as Euclidean distance) to ensure that the algorithm continues to run when the KD tree fails;

[0076] After generating a valid ball, add candidate points in the nearby layer. If the number of candidate points is insufficient, add them from the high area layer.

[0077] Output DICOM RT Structure file containing the ball outline.

[0078] The optimization process of the genetic algorithm optimization module is the same as the optimization process of the first embodiment.

[0079] The ball generation module is used to generate a ball target area. After obtaining the optimal distribution parameters using the KD tree optimization module or the genetic algorithm optimization module, the ball target area is generated within the target area PTV structure according to the parameters.

[0080] The output module outputs the following results: a 3D mask of each spherical target area, spherical statistical information (center coordinates, diameter, volume, etc.), statistics of spherical and OAR overlap, and a DICOM RTStructure file containing spherical target area contour information, which is compatible with the treatment planning system (TPS).

[0081] like Figure 2 The interface of the intelligent generation and optimization system for the spatially segmented radiotherapy spherical target area of ​​the present invention is shown, and the corresponding operation process is as follows:

[0082] (1) Import the patient's positioning CT data in the format of Dicom CT images.

[0083] (2) Import the Dicom RTStructure file that contains the normal tissue OAR structure and the target area PTV structure that has been outlined by the radiotherapist.

[0084] (3) The program reads all OARs in the Dicom RT Structure file and selects the exempted areas.

[0085] (4) The physicist selects the target area for the ball target to be placed, which is usually the PTV.

[0086] (5) Set the spatial relationship limit area of the small ball target area, including the layer small ball target area center interval, the range is set to any number between 5-7cm; the layer small ball target area center interval, the range is set to any number between 3-5cm; the small ball edge and the target area PTV edge interval distance is 3mm by default.

[0087] (6) Select the generation mode of the generated Dicom RT Structure file containing the small ball target area structure, including Standalone and Append two modes:

[0088] Standalone mode: independently generate small ball structure file;

[0089] Append mode: generate small ball structure file on the imported Dicom RT Structure file structure of the radiotherapy doctor's delineation.

[0090] (7) Select the path and name of the Dicom RT Structure file to be generated, with the default.dcm suffix.

[0091] (8) Select the generated small ball target area, then run the program, and jump out some statistical information of the target area structure (including length, width, volume, etc.) and the number of predicted small ball target areas.

[0092] (9) After generating the target area, give a completion prompt dialog box.

[0093] (10) Provide a preview visualization tool to show the spatial relationship of the small ball target area and the target area, which is convenient for physicists to verify.

[0094] (11) The statistical information of the generated target small ball can be selected, including the involved CT layer number and the three-dimensional coordinates of the small ball center.

[0095] Figure 3 The target small ball schematic diagram generated by the system of the application shows the spatial distribution of the small ball under different vision and the positional relationship with the target area PTV.

[0096] Figure 4 The visualization interface diagram of the small ball target area generated by the system of the application.

[0097] Figure 5 It is the dose distribution and DVH (dose volume histogram) display of the small ball target area calculated by the radiotherapy planning system (Varian Eclipse 15.6), which shows that the calculation results of the application meet the clinical requirements.

[0098] Figure 6-8The peak-valley dose ratio (PVDR) result chart generated by the system of the present application shows that the PVDR result of the small ball target region generated by the treatment planning system of the present application is significantly improved to 3-4 compared with the prior art (generally 2.5).

[0099] The plans are verified by using a Varian accelerator Portal Dosimetry dose verification system, and the passing rate of the final Gamma analysis evaluated by 2mm / 2% is 100%, which is much higher than the standard requirement of 95%, indicating that the dose accuracy and precision completely meet the clinical requirements.

[0100] Figure 9 The Portal Dosimetry dose verification result chart after the treatment plan of the small ball target region generated by the system of the present application shows that the dose verification result of the treatment plan generated by the small ball target region generated by the system of the present application is significantly improved to more than 99% in Gamma value compared with the prior art, which completely meets the requirements of clinical quality control.

[0101] Part of the steps in the embodiments of the present application can be realized by software, and the corresponding software program can be stored in a readable storage medium, such as an optical disc or a hard disk.

[0102] The above description is only the preferred embodiments of the present application, and is not intended to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A VMAT-based Lattice radiotherapy spherical target volume intelligent generation and optimization method, characterized by: The method comprises: Step 1: Obtain the patient's DICOM CT images and Dicom RT Structure files, and extract the normal tissue OAR structure and target area PTV structure; Step 2: In the target area PTV structure, a population is initialized using a genetic algorithm. Each individual contains a set of intra-layer spacing and inter-layer spacing parameters to obtain the initial small ball target area candidate points. Step 3: Use the KD tree to check whether the candidate points meet the distance constraints and generate a small ball target area that meets the constraints; Step 4: Calculate the fitness of the current target area based on the number of balls, target area coverage, distance variance, and peak-to-valley dose ratio; Step 5: Generate a new population through selection, crossover and mutation, and repeat steps 3 and 4 until the preset number of iterations is completed; Step 6: Select the ball target area distribution structure corresponding to the highest fitness value as the final optimization result.

2. The method for intelligent generation and optimization of Lattice radiotherapy spherical target volume based on VMAT according to claim 1, characterized in that: The distance constraints include: the center spacing of the ball target area within a layer, the center spacing of the ball target area between layers, and the spacing between the edge of the ball and the edge of the target area PTV.

3. The method for intelligent generation and optimization of Lattice radiotherapy spherical target volume based on VMAT according to claim 1, characterized in that: The method further includes: utilizing an erosion function to erode the target region PTV, so as to ensure that a distance constraint is satisfied between an edge of the ball target region and an edge of the target region PTV.

4. A VMAT-based Lattice radiotherapy spherical target volume intelligent generation and optimization system, characterized by: The system comprises: Data import module, used to import patient DICOM CT images and Dicom RT Structure files, extract normal tissue OAR structure and target area PTV structure; A parameter setting module is used to set the small ball target area distance constraint parameters and the small ball target area optimization mode parameters; A ball target area generation and optimization module selects whether to use a genetic algorithm to optimize the ball target area according to the ball target area optimization mode parameters. When the genetic algorithm is used, the genetic algorithm optimization module is called to optimize the ball target area distribution structure. Otherwise, the KD tree optimization module is used for optimization. The KD tree optimization module verifies the spacing constraints between candidate balls and existing balls in real time based on the three-dimensional spatial data structure; A genetic algorithm optimization module, which optimizes the distribution structure of the ball target area using the method described in any one of claims 1 to 3; Ball generation module: generates a compliant ball target area within the target area PTV structure according to the optimized parameters; Result output module: Generates DICOM RT Structure file, including ball outline, center coordinates and statistical information.

5. The VMAT-based Lattice radiotherapy spherical target volume intelligent generation and optimization system according to claim 4, characterized in that: The process of the KD tree optimization module includes: Erosion is performed on the target area mask, the area of ​​each layer is calculated, the layer with the largest area is selected as the initial center layer and the geometric center is obtained; Generate candidate points based on random sampling of the eroded mask; Generate a small ball mask with the candidate point as the center to check whether it is completely within the erosion target area and does not overlap with the risk organ; Use the KD tree algorithm to verify whether the candidate points and the existing balls meet the distance constraints, and retain the candidate points that meet the constraints; After generating a valid ball, add candidate points in the nearby layer. If the number of candidate points is insufficient, add them from the high area layer. Output DICOM RT Structure file containing the ball outline.

6. The VMAT-based Lattice radiotherapy spherical target volume intelligent generation and optimization system according to claim 4, characterized in that: The system also includes a preview visualization module.

7. The VMAT-based Lattice radiotherapy spherical target volume intelligent generation and optimization system according to claim 4, characterized in that: The system further comprises a statistical information generating module for generating statistical information of the target area sphere, including the number of CT layers involved and the three-dimensional coordinates of the center of the sphere.

8. The VMAT-based Lattice radiotherapy spherical target volume intelligent generation and optimization system according to claim 4, characterized in that: In the ball target area distance constraint parameters, the center spacing range of the ball target area within a layer is 5-7 cm, the center spacing range of the ball target area between layers is 3-5 cm, and the default distance between the ball edge and the target area PTV edge is 3 mm.

9. A device for intelligent generation and optimization of Lattice radiotherapy spherical target volume based on VMAT, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is configured to implement the VMAT-based Lattice radiotherapy spherical target area intelligent generation and optimization method as described in any one of claims 1 to 3 when executing the computer program.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by the processor, the VMAT-based Lattice radiotherapy spherical target area intelligent generation and optimization method as described in any one of claims 1 to 3 is implemented.

Images