Dose distribution acquisition method

Abstract

The application relates to the technical field of medical devices, and discloses a dose distribution acquisition method, wherein the method comprises the following steps: constructing a voxel grid based on tissue information, the voxel grid comprising a plurality of voxels, each voxel having a voxel cross section, the voxel cross section representing the probability of a nuclear reaction between the voxel and a particle, the particle representing a radiation particle with a motion direction in the voxel grid; determining the motion distance of the particle based on the voxel main cross section, the voxel main cross section being the largest voxel cross section in the voxel grid; predicting the target voxel that can be reached by the particle after the particle moves based on the motion distance and the motion direction; updating the particle weight of the particle based on the difference between the voxel cross section of the target voxel and the voxel main cross section; and obtaining the tissue dose distribution based on the updated particle weight. The method has the beneficial effect of improving the simulation efficiency of particle motion.

Description

Technical Field

[0001] This application relates to the field of medical device technology, and in particular to a method for obtaining dose distribution. Background Technology

[0002] Binary radiation techniques with strong targeting and high energy transfer density can include, but are not limited to, boron neutron capture technique (BNCT). These techniques can precisely eliminate tumor cells by introducing radiation particles (such as neutrons) into tumor tissue, while maximizing the protection of healthy tissue surrounding the tumor.

[0003] Currently, implementing these technologies requires determining the number of radiation particles incident on tumor tissue through simulation. However, the simulation efficiency in some technologies is low and cannot meet practical needs. Summary of the Invention

[0004] This application provides a dose distribution acquisition method, a computer device, and a readable storage medium, which solves the technical problem of low simulation efficiency and achieves the technical effect of improving simulation efficiency.

[0005] To achieve the above objectives, the main technical solutions adopted in this application include: In a first aspect, embodiments of this application provide a method for obtaining dose distribution, including: Based on organizational information, a voxel grid is constructed, which includes multiple voxels. Each voxel has a voxel cross section, which represents the probability of the voxel reacting with a particle. The particle represents a radiating particle with a direction of motion in the voxel grid. The movement distance of the particle is determined based on the principal voxel cross section, where the principal voxel cross section is the largest voxel cross section in the voxel grid. Based on the distance and direction of motion, predict the target voxel that the particle can reach after its motion; Based on the difference between the voxel cross section of the target voxel and the main cross section of the voxel, the particle weights of the particles are updated, and the tissue dose distribution is obtained based on the updated particle weights.

[0006] Secondly, embodiments of this application provide a dose distribution acquisition device, comprising: A mesh construction module is used to construct a voxel mesh based on organizational information. The voxel mesh includes multiple voxels, each of which has a voxel cross section. The voxel cross section represents the probability of the voxel undergoing a nuclear reaction with a particle, and the particle represents a radiating particle with a direction of motion in the voxel mesh. The distance determination module is used to determine the movement distance of the particle based on the voxel principal cross section, wherein the voxel principal cross section is the largest voxel cross section in the voxel grid; A motion prediction module is used to predict the target voxel that the particle can reach after its motion, based on the motion distance and the motion direction. The weight update module is used to update the particle weights of the particles based on the difference between the voxel cross section of the target voxel and the main cross section of the voxel, and to obtain the tissue dose distribution based on the updated particle weights.

[0007] Thirdly, embodiments of this application provide a graphics processor including multiple computing units, which can run in parallel to execute the dose distribution acquisition method as described in any of the preceding claims.

[0008] Fourthly, embodiments of this application provide a computer device, including: The memory and the graphics processor as described above are communicatively connected, the memory stores computer instructions, and the graphics processor executes the dose distribution acquisition method as described above by executing the computer instructions.

[0009] Fifthly, embodiments of this application provide a computer-readable storage medium storing computer instructions that cause a computer to perform the dose distribution acquisition method as described in any of the preceding claims.

[0010] In some embodiments of this application, the particle's movement distance is determined based on the largest voxel cross-section among multiple voxel cross-sections. This allows the particle to traverse multiple voxels simultaneously during a single movement, eliminating the need for separate probability calculations of nuclear reactions when the particle passes through each voxel. This significantly reduces redundant calculations and improves simulation efficiency when simulating particle movement. Based on the movement distance and direction, the target voxel that the particle can reach after its movement is predicted. Furthermore, the particle weight is updated based on the difference between the target voxel's voxel cross-section and the main voxel cross-section. This corrects the probability deviation of nuclear reactions between the particle and the atomic nuclei in the target voxel, thereby improving the accuracy of tissue dose distribution. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0012] Figure 1 A schematic flowchart illustrating a dose distribution acquisition method provided for some embodiments of this application; Figure 2 Schematic diagram of coordinate system creation provided for some embodiments of this application; Figure 3 Simulation flow of a graphics processor provided for some embodiments of this application; Figure 4 A schematic diagram of a dose distribution acquisition device provided in the first embodiment of this application; Figure 5 A schematic diagram of a graphics processor module provided for some embodiments of this application; Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] The following section uses boron neutron capture technology as an example to introduce some core principles of binary radioactive technologies.

[0015] In boron neutron capture technology, when tumor tissue is present in a biological organism, boron neutrons can be captured... 10 B-targeted boron agents are injected into biological tissues. 10 B specifically accumulates in tumor cells and healthy cells. 10 The B concentration is low. Subsequently, a hyperthermal neutron beam with an energy range of 1 eV to 10 keV can be incident on the tumor tissue. The neutrons can move within the tumor tissue and are probabilistically destroyed during their movement. 10 B captures, thus triggering 10 B(n,ɑ) 7 Li nuclear reaction. The nuclear reaction can release a large number of α particles with a range of only 4–10 μm and 7 Li particles. This range is comparable to the diameter of tumor cells, allowing the energy carried by these two particles to be deposited within the tumor cells, achieving precise elimination of tumor cells and maximizing the protection of healthy tissue surrounding the tumor.

[0016] In practical applications, before incident a neutron beam onto tumor tissue, a voxel grid can be constructed based on slice images of the tumor tissue. The voxel grid is used to characterize the three-dimensional image obtained after dividing the tumor tissue into multiple sub-tissues. The voxel grid consists of multiple voxels, each representing one of the sub-tissues of the tumor tissue.

[0017] After creating the voxel grid, multiple particles can be initialized within it using simulation methods such as Monte Carlo (MC). These particles are used to characterize the ultrathermal neutron beam injected into tumor tissue. Each particle can have a randomly initialized energy and direction of motion. Based on the particle's energy and direction of motion, the particle's path within the voxel grid can be predicted. During the particle's motion along this path, it can be probabilistically... 10 B captures, thus triggering 10 B(n,ɑ) 7 Li nuclear reactions. This allows for the simulation of the actual movement and nuclear reaction processes of neutrons within tumor tissue. Based on the simulation results, the dose distribution within the voxel grid can be evaluated. The dose distribution refers to the energy distribution deposited after a particle undergoes a nuclear reaction with the nuclei in a voxel. Based on the dose distribution, it can be determined whether the number of neutrons incident on the tumor tissue is appropriate.

[0018] For example, in the first simulation, a first number of particles can be initialized in the voxel grid to represent the first number of neutrons. Based on the first dose distribution obtained from the first simulation, it can be determined whether the first dose distribution can meet the elimination requirements of tumor tissue cells. If it can, the simulation can be stopped, and a first number of neutron beams can be injected into the tumor tissue. If not, a second simulation can be performed. In the second simulation, a second number of particles can be initialized in the voxel grid to represent the second number of neutrons. Based on the second dose distribution obtained from the second simulation, it can be further determined whether the second dose distribution can meet the elimination requirements of tumor tissue cells. If it can, the simulation can be stopped, and a second number of neutron beams can be injected into the tumor tissue. If not, a third simulation can be performed. And so on.

[0019] Currently, in some simulation techniques, the distance a particle travels within a voxel grid is determined by the type of matter within each voxel and the density of atomic nuclei for each type of matter. Simply put, a voxel can contain multiple substances, such as hydrogen, oxygen, etc. 10B. The nuclei of each substance can potentially react with the particle through nuclear reactions. In the particle's path, if the nucleus density of most voxels is high, or if the probability of nuclear reactions between the nuclei of at least some substances within a voxel and the particle is high, then the particle's path can be shorter. Conversely, if the nucleus density of most voxels is low, and the probability of nuclear reactions between the nuclei of at least some substances within a voxel and the particle is low, then the particle's path can be longer.

[0020] Because different voxels contain different types of matter and atomic nuclei densities, the probability of a nuclear reaction between the particle and the nuclei within a voxel must be recalculated each time a particle passes through it, based on the type of matter and atomic nuclei density of that voxel. This calculated probability is then used as a constraint, and a random sampling method is used to determine whether the particle will react with the nuclei within the voxel. For example, when particle A moves from its initial position to voxel B, the first probability (e.g., 60%) of a nuclear reaction between particle A and the nuclei within voxel B can be determined based on the type of matter and atomic nuclei density of voxel B. Using this first probability as a constraint, a random sampling method is used to determine whether a nuclear reaction occurs between particle A and the nuclei within voxel B. If particle A does not react with the nuclei within voxel B, and particle A's energy is not below the energy threshold and particle A is not located at the boundary of the voxel grid, then particle A can continue moving. When particle A continues moving to voxel C, the second probability (e.g., 75%) of a nuclear reaction between particle A and the nuclei within voxel C can be determined based on the type of matter and atomic nuclei density of voxel C. Using the second probability as a constraint, a sampling method can be used to randomly determine whether particle A undergoes a nuclear reaction with the nuclei within voxel C. If particle A undergoes a nuclear reaction with the nuclei within voxel C, then particle A will perish when it reaches voxel C, and the distance traveled by particle A is the distance between its initial position and voxel C. If particle A does not undergo a nuclear reaction with the nuclei within voxel C, and particle A's energy is not lower than the energy threshold, and particle A is not located at the boundary of the voxel grid, then particle A can continue to move, and so on.

[0021] This simulation method, which performs probability calculations at each voxel separately, involves a large amount of redundant computation, significantly reducing simulation efficiency. Furthermore, in these simulation techniques, when a particle moves to a certain position, it is necessary to traverse all voxels in the tumor tissue based on the coordinates of that position to determine the particle's current voxel location and perform probability calculations based on the material type and atomic density of the corresponding voxel. This voxel-traversal approach also reduces simulation efficiency.

[0022] In view of this, this application provides a dose distribution acquisition method that can solve the above problems. The dose distribution acquisition method is applied to a graphics processing unit (GPU) or an electronic device including a graphics processing unit. The electronic device may include, but is not limited to, servers, tablet computers, desktop computers, laptop computers, etc.

[0023] Based on the above description, and in conjunction with the references Figure 1 This is a flowchart illustrating a dose distribution acquisition method provided in some embodiments of this application. Figure 1 In this study, the method for obtaining dose distribution includes the following steps: Step S101: Based on the organization information, construct a voxel grid. The voxel grid includes multiple voxels, each voxel has a voxel cross section, the voxel cross section represents the probability of a nuclear reaction between the voxel and the particle, and the particle represents a radiating particle with a motion direction in the voxel grid.

[0024] Specifically, tissue information can characterize the parameters of tumor tissue. Based on these parameters, information such as the structure and composition of tumor tissue can be determined.

[0025] In this embodiment, the tissue information specifically refers to multiple slice images of the tumor tissue received by the graphics processor, and these multiple slice images are obtained by continuously slicing the tumor tissue. By stacking these multiple slice images, three-dimensional information of the tumor tissue can be obtained. This three-dimensional information may include, but is not limited to, the types of substances and tissue density at various locations within the tumor tissue. 10 B concentration, etc.

[0026] Based on three-dimensional information, a coordinate system can be created, and a voxel mesh can be constructed based on this coordinate system. For ease of understanding, please refer to the relevant references. Figure 2 Using the target location (e.g., the lower left corner) of one of the target slice images (e.g., slice image 1) as the origin O of the coordinate system, two mutually perpendicular coordinate axes X and Y can be created in the target slice image, and a coordinate axis Z perpendicular to the X and Y axes can be created in the stacking direction of the slice images. Based on the coordinate system formed by the X, Y, and Z axes, a three-dimensional image of the tumor tissue can be constructed. Dividing the three-dimensional image into multiple voxels yields a voxel mesh.

[0027] Specifically, when dividing a 3D image into multiple voxels, the side lengths Δx, Δy, and Δz of a single voxel in the X, Y, and Z coordinate axes can be predefined. The 3D image of the tumor tissue is then divided into multiple voxels according to these Δx, Δy, and Δz. Each voxel has side lengths Δx, Δy, and Δz in the X, Y, and Z coordinate axes.

[0028] In this embodiment, △x = △y = △z, meaning the three-dimensional image of the tumor tissue is divided into multiple cubic sub-units. It is understood that in practical applications, △x, △y, and △z can be set according to actual needs; that is, △x, △y, and △z may not be equal. This application does not impose any limitations on this.

[0029] The voxel cross section refers to the sum of probabilities of nuclear reactions between the atomic nuclei and particles of all substances within a voxel. Specifically, the voxel cross section of each voxel can be calculated based on expression (1).

[0030] (1) in, This represents the voxel cross section of the m-th voxel. This represents the atomic nucleus density of the first substance in the m-th voxel. This represents the probability that a single atomic nucleus of the first substance in the m-th vacuole will undergo a nuclear reaction with a particle. This represents the atomic nucleus density of the nth substance in the mth voxel. This represents the probability that a single atomic nucleus of the nth substance in the m-th voxel will undergo a nuclear reaction with a particle. The values ​​of m and n are integers greater than 0.

[0031] It is understandable that, since different voxels may contain different types of matter and have different nucleus densities, each voxel can have its own corresponding voxel cross section, and at least some voxels may have different voxel cross sections. Among the multiple voxel cross sections of the voxel mesh, the largest voxel cross section can be taken as the principal voxel cross section.

[0032] In this embodiment, based on the divided voxels and the parameter information of each voxel (such as the tissue mass density of the voxel, ... 10 (B concentration, voxel cross-section, etc.), a voxel matrix corresponding to each voxel can be pre-constructed. The voxel matrix can store the parameter information of the corresponding voxel. Specifically, the voxel matrix can be as shown in expression (2).

[0033] (2) in, The voxel matrix representing the m-th voxel. This represents the tissue mass density of the m-th voxel. Represents the m-th voxel 10 B concentration, This represents the voxel cross section of the m-th voxel. This represents the difference between the voxel cross section of the m-th voxel and the principal voxel cross section.

[0034] Each voxel can be associated with a corresponding voxel matrix. Based on this association, the parameter information of each voxel can be quickly retrieved. This eliminates the need to traverse all voxels in the voxel mesh, thereby improving simulation efficiency.

[0035] After completing the above operations, multiple particles can be initialized in the voxel grid. These particles are used to simulate radiation particles (such as neutrons) incident on tumor tissue in real-world applications. Specifically, when initializing particles, the initial energy, initial position, and initial direction of motion of each particle in the voxel grid can be randomly initialized based on the energy, incident position, and incident direction of the radiation particles in the real-world application. For example, if the energy of the radiation particles to be incident on the tumor tissue in a real-world application is between 1 eV and 10 keV, then the initial energy of each particle in the voxel grid can be randomly initialized within the range of 1 eV to 10 keV. Another example is that if radiation particles need to be incident on the tumor tissue from three directions F1, F2, and F3 in a real-world application, then the direction of motion of each particle can be randomly initialized based on directions F1, F2, and F3. In other words, for any particle, one of directions F1, F2, and F3 can be used as the initial direction of motion of the particle. Yet another example is that if radiation particles need to be incident from region A of the tumor tissue in a real-world application, then multiple particles can be randomly initialized in the region corresponding to region A in the voxel grid. It should be noted that when a particle moves within a voxel grid, its location can change. At the same time, when a particle collides with a voxel, its direction of motion and energy can be altered.

[0036] Step S102: Determine the particle's movement distance based on the voxel principal cross section, where the voxel principal cross section is the largest voxel cross section in the voxel grid.

[0037] Specifically, the principal voxel section can be used to determine the average distance traveled by multiple particles. Within an allowable range of distance fluctuations, the average distance can be randomly modified to obtain the distance traveled by each individual particle.

[0038] Specifically, the average motion distance is inversely proportional to the principal cross-section of the voxel; that is, the larger the principal cross-section of the voxel, the greater the probability of a nuclear reaction between the particle and the atomic nucleus, and consequently, the smaller the average motion distance of the particle. Using the average motion distance as a benchmark, random sampling within an allowable range of distance fluctuation can yield the particle's motion distance. For example, assuming an average motion distance of 500 nanometers and a distance fluctuation range between -10 nanometers and 10 nanometers, random sampling between 490 nanometers and 510 nanometers can be used to obtain the motion distances of individual particles. For instance, particle 1's motion distance is 495 nanometers, and particle 2's motion distance is 501 nanometers. This allows for differences in the motion distances of different particles, consistent with the actual physical laws governing particle motion.

[0039] By determining the average motion distance based on the principal voxel cross-section and randomly modifying the average motion distance within an allowable fluctuation range, each particle can initially move at the maximum allowable distance. This means that during a single motion, a particle can simultaneously traverse multiple voxels without needing to calculate the probability of nuclear reactions for each voxel. This significantly improves simulation efficiency.

[0040] Step S103: Based on the motion distance and motion direction, predict the target voxel that the particle can reach after its motion.

[0041] Specifically, for any particle, its initial position can be used as a starting point. Based on the distance traveled, extending from the direction of the particle's motion, we can obtain the target position that the particle can reach after its motion. The voxel containing the target position is called the target voxel.

[0042] Since the initial position, direction of motion, and distance traveled by each particle may be different, the target positions that different particles can reach may be different. For any two particles, if the target positions of the two particles are significantly different, the target voxels that the two particles can reach after moving may be different; if the target positions of the two particles are relatively small, the target voxels that the two particles can reach after moving may be the same.

[0043] Understandably, the motion distance determined based on the principal cross-section of a voxel is equivalent to calculating the motion distance based on the maximum probability of a nuclear reaction occurring. Theoretically, after moving along the corresponding distance, the particle should undergo a nuclear reaction with the atomic nuclei within the target voxel, and then be destroyed in the target voxel, depositing energy. However, during the actual motion of the particle, the voxel cross-sections of each voxel it passes through may be smaller than the principal cross-section. This leads to the situation where, in reality, the particle may continue to move after reaching the target voxel, meaning it will not undergo a nuclear reaction with the atomic nuclei within the target voxel, nor will it deposit energy in the target voxel. This implies that the energy deposition that should occur within the target voxel based on theoretical calculations will not actually occur; that is, the probability of a nuclear reaction between the particle and the atomic nuclei in the target voxel calculated based on the principal cross-section is inaccurate. If the dose distribution within the voxel grid is evaluated based on the theoretically calculated energy deposition, it may cause inaccurate dose distribution. Therefore, step S104 can be executed to correct this problem.

[0044] Step S104: Based on the difference between the voxel cross section and the main voxel cross section of the target voxel, update the particle weights and obtain the tissue dose distribution based on the updated particle weights.

[0045] Tissue dose distribution refers to the dose distribution within the voxel grid.

[0046] Comparing the voxel cross-section of the target voxel with the principal voxel cross-section can be understood as a sampling comparison. Simply put, among the multiple voxels the particle passes through, a sample voxel cross-section is compared with the principal voxel cross-section. Based on the comparison result, the differences between the voxel cross-sections of each voxel along the particle's path and the principal voxel cross-section are evaluated. This reduces computational load and increases the randomness of particle motion, aligning with the actual physical laws governing particle motion.

[0047] Furthermore, based on the description in step S101, since the association between voxels and voxel matrices is established in advance, in step S104, the voxel matrix of the target voxel can be quickly found based on this association, and the difference between the voxel cross section and the main voxel cross section of the target voxel can be obtained from the found voxel matrix. Thus, during the simulation, it is not necessary to calculate the difference between the voxel cross section and the main voxel cross section of the target voxel in real time, thereby improving simulation efficiency.

[0048] Particle weight refers to the importance of a particle in determining tissue dose distribution. During particle initialization, the particle weight of each particle can be randomly assigned. Each particle can have its own corresponding particle weight, and at least some particles can have different particle weights. For any given particle, a larger particle weight indicates that the energy deposited after the particle's nuclear reaction with the voxel will be given significant consideration in determining the tissue dose distribution. Conversely, if the particle weight is small, the reference value of the particle will be reduced in determining the tissue dose distribution; that is, the energy deposited after the particle's nuclear reaction with the voxel can be almost ignored.

[0049] Based on the above description, in this embodiment, updating the particle weight mainly involves reducing the particle weight. Specifically, when the difference between the voxel cross-section and the principal voxel cross-section of the target voxel is large, the particle weight can be reduced. When the difference between the voxel cross-section and the principal voxel cross-section is small, the particle weight can be kept unchanged. In this way, the probability deviation of nuclear reactions between particles and atomic nuclei in the target voxel can be corrected, ensuring the accuracy of tissue dose distribution.

[0050] In summary, in the technical solutions of some embodiments of this application, determining the particle's movement distance based on the largest voxel cross-section among multiple voxel cross-sections allows the particle to simultaneously traverse multiple voxels during a single movement, eliminating the need for probability calculations of nuclear reactions each time the particle passes through each voxel. This significantly reduces redundant calculations when simulating particle movement, thereby improving simulation efficiency. Based on the movement distance and direction, predicting the target voxel the particle can reach after its movement, and updating the particle weights based on the difference between the target voxel's voxel cross-section and the main voxel cross-section, can correct the probability deviation of nuclear reactions between the particle and the atomic nuclei in the target voxel, thereby improving the accuracy of tissue dose distribution.

[0051] In some embodiments, based on Figure 2 In the coordinate system shown, each voxel can have multiple position coordinates. Different position coordinates can represent different positions of the voxel. Additionally, the initial position of the particle can also be represented using initial position coordinates, and the position the particle can reach during its motion can also be represented using corresponding position coordinates. Step S103, which predicts the target voxel that the particle can reach after its motion based on the motion distance and direction, may include: Obtain the first position coordinates of the particle in the voxel grid. The first position coordinates are the position coordinates of the particle before it moves. Using the first position coordinate as the starting point and the movement distance and direction as constraints, predict the second position coordinate that the particle can reach after its movement; The voxel at the second position coordinate is taken as the target voxel.

[0052] In the above embodiments, the position of the particle during its motion is represented by position coordinates, which can accurately locate the particle position, ensure the accuracy of the particle motion path, and thus improve the simulation accuracy.

[0053] In some embodiments, each voxel has a voxel index. The voxel index can be unique, allowing for the identification of a unique voxel. The above-described method of using the voxel containing the second position coordinates as the target voxel, based on the voxel index, includes: Based on the mapping relationship between position coordinates and voxel indices, the target voxel index corresponding to the second position coordinates is determined; The voxel identified by the target voxel index is used as the target voxel.

[0054] Specifically, in the mapping relationship, multiple position coordinates within the same voxel can be associated with the voxel index of the corresponding voxel. For example, a mapping relationship table similar to Table 1 can be created, associating each position coordinate with the voxel index of its corresponding voxel. In this way, by querying the mapping relationship table, the target voxel index corresponding to the second position coordinate can be quickly located.

[0055] Table 1 Mapping Relationship Table In the above embodiments, after creating the mapping relationship between position coordinates and voxel indices, the target voxel corresponding to the second position coordinates can be quickly found based on this mapping relationship. Compared with some simulation techniques that find the voxel to which the voxel has moved by traversing voxels, the above embodiments can reduce traversal calculations and thus improve simulation efficiency.

[0056] In some embodiments, the mapping relationship is created based on the following method: Obtain the coordinate system information of the voxel mesh. The coordinate system information includes the origin of the coordinate system and the direction of the coordinate axes. The position coordinates include the sub-coordinates located in each coordinate axis direction. A mapping relationship is created based on the distance between each sub-coordinate and the origin of the coordinate system, and the side length of a single voxel in each coordinate axis direction.

[0057] Specifically, based on Figure 2 The coordinate system shown can include sub-coordinates located in the X, Y, and Z axes for each position. Based on the sub-coordinates, a mapping relationship can be created according to expressions (3)-(5).

[0058] (3) (4) (5) In expressions (3)-(5), x, y, and z represent position coordinates (x, y, z), where x is the sub-coordinate of the position coordinate on the X-axis, y is the sub-coordinate of the position coordinate on the Y-axis, and z is the sub-coordinate of the position coordinate on the Z-axis. , , Represents the coordinates of the origin of the coordinate system ( , , ). This represents the side length of a single voxel on the X-axis. This represents the side length of a single voxel on the Y-axis. This represents the side length of a single voxel on the Z-axis. (i,j,k) is the voxel index corresponding to the position coordinates (x,y,z).

[0059] In the above embodiments, a mapping relationship is created based on the distance between each sub-coordinate and the origin of the coordinate system, and the side length of a single voxel in each coordinate axis direction, which can improve the accuracy of the mapping relationship.

[0060] Furthermore, based on the voxel indices (i,j,k) determined by expressions (3)-(5), the voxel matrix in the above expression (2) can be changed to expression (6).

[0061] (6) In this way, each voxel matrix can be identified by a voxel index, thereby improving the efficiency of voxel matrix lookup.

[0062] In some embodiments, the method of this application further includes: In the direction of motion, determine the distance between the first position coordinates and the boundary coordinates of the voxel mesh; If the interval distance is less than the movement distance, then the movement distance is updated to the interval distance.

[0063] Specifically, the boundary coordinates of the voxel grid refer to the coordinates corresponding to the edge positions of the voxel grid. In practice, the first position coordinates can be used as the starting point, and the distance between the first position coordinates and the boundary coordinates of the voxel grid can be determined along the particle's movement direction. If the distance is less than the movement distance, it indicates that the particle will move into an invalid region outside the voxel grid. Since the invalid region does not contain voxels and voxel parameters, it may cause calculation errors during the simulation process. Therefore, by updating the movement distance to the distance, the particle movement can be confined within the voxel grid, avoiding calculation errors during the simulation process. It should be noted that although the particle movement is confined within the voxel grid, in step S104, a probability deviation correction method can be used to determine whether the particle undergoes a nuclear reaction with the atomic nuclei within the voxel at the boundary of the voxel grid. This can correct the energy deposition at the voxel grid boundary and ensure the accuracy of tissue dose distribution.

[0064] In the above embodiments, when the interval between the first position coordinate and the boundary coordinate of the voxel grid is less than the movement distance, the movement distance is updated to the interval distance, which can limit the particle movement within the voxel grid and avoid calculation errors during the simulation process.

[0065] In some embodiments, determining the particle's motion distance based on the voxel principal section in step S102 includes: Generate the first random number; Based on the first random number and the principal voxel section, a random mapping is performed to obtain the motion distance. In the random mapping, the principal voxel section is used to determine the average motion distance of the particle, and the first random number is used to randomly generate the motion distance of the particle based on the average motion distance.

[0066] Specifically, the relationship between the first random number, the voxel principal section, and the motion distance can be expressed as in expression (7).

[0067] (7) in, The first random number, The principal cross section of the voxel.

[0068] In the above embodiments, the average motion distance of particles is constrained by the principal cross-section of a voxel, and the motion distance of each particle is randomly generated based on a first random number. On the one hand, this ensures that the motion distances of different particles are within the same dimension. For example, when the average motion distance is 10 nanometers, the motion distances of all particles will fluctuate around 10 nanometers, and there will be no situation where the motion distance of a particle becomes 1000 nanometers. On the other hand, this ensures the randomness of particle motion and guarantees that the simulation process conforms to actual physical laws.

[0069] In some embodiments, updating the particle weights in step S104 based on the difference between the target voxel cross-section and the principal voxel cross-section may include: Determine the section ratio between the voxel section and the principal voxel section of the target voxel; If the cross-section ratio is less than the second random number, then update the particle weights.

[0070] Specifically, the cross-sectional ratio can be expressed as shown in expression (8).

[0071] (8) in, Principal cross section of voxel The voxel cross section of the target voxel.

[0072] Specifically, let's assume the second random number is... If the cross-sectional ratio is greater than or equal to the second random number (i.e. This can indicate that after a particle moves to the target voxel, it will undergo a nuclear reaction with the atomic nuclei in the target voxel. At this point, the particle weight can be reduced. If the cross-sectional ratio is less than the second random number (i.e., ... This can indicate that after a particle moves to the target voxel, it will not undergo a nuclear reaction with the atomic nuclei in the target voxel. In this case, the particle weight can remain unchanged.

[0073] In the above embodiments, determining whether to update the particle weight by generating a second random number can ensure the randomness of particle motion, which conforms to the actual physical laws of particle motion.

[0074] In some embodiments, updating the particle weights if the cross-sectional ratio is less than the second random number may include: The update magnitude of particle weights is determined based on the cross-section ratio, where the update magnitude is inversely proportional to the cross-section ratio. Update the particle weights according to the update magnitude.

[0075] Specifically, the relationship between the cross section ratio and the particle weight can be expressed as in expression (9).

[0076] (9) in, For the updated particle weights, The particle weights before the update.

[0077] In the above embodiments, by limiting the update range of particle weights by the cross-section ratio, the simulated tissue dose distribution can be kept as consistent as possible with the real tissue dose distribution, thereby improving the simulation accuracy.

[0078] In some embodiments, after a particle moves to a target voxel, the particle's position coordinates can be updated based on expression (10).

[0079] (10) in, Represents the position coordinates of the particle after its motion. This represents the particle's position coordinates before its motion. Indicates the distance traveled.

[0080] In some embodiments, if the particle does not undergo a nuclear reaction with the nucleus of the target voxel, the particle can theoretically use the target voxel as a starting point to continue moving. However, during the initial movement, the particle will consume energy after colliding with the voxel. Therefore, after the particle moves to the target voxel, even if the particle does not undergo a nuclear reaction with the nucleus of the target voxel (i.e., the particle does not perish), whether the particle can continue moving depends on the remaining energy of the particle. Of course, it is also necessary to determine whether the particle is already at the boundary of the voxel grid; if so, the particle needs to stop moving. In summary, the method of this application also includes: If the particle does not meet the following conditions, then using the target voxel as the starting point, predict the next voxel that the particle can reach after continuing its motion: The particle undergoes a nuclear reaction with the target voxel; The energy of the particles is below the energy threshold; The target voxel where the particle is located is the boundary voxel of the voxel grid.

[0081] Specifically, if the particle does not undergo a nuclear reaction with the target voxel, it means the particle has not perished. If the particle's energy is not lower than the energy threshold, it means the particle still possesses kinetic energy to continue moving. If the particle is located in a non-voxel grid boundary voxel of the target voxel, it means the particle is located within the effective region of the voxel grid. When the particle simultaneously meets the above conditions, the target voxel can be used as the starting point to predict the next voxel the particle can reach after continuing its movement. If the particle does not meet any of the above conditions, its movement ceases. The prediction of the next voxel the particle can reach after continuing its movement can be performed according to the methods in steps S102 to S103, which will not be elaborated here.

[0082] In the above embodiments, if the particle does not meet the above conditions for stopping motion after it reaches the target voxel, the next voxel that the particle can reach is predicted. This conforms to the actual physical laws of particle motion and can ensure the accuracy of the simulation.

[0083] In some embodiments, while the particles continue to move, the particle weights can be updated based on the difference between the voxel cross-section and the principal voxel cross-section of the next voxel. Specifically, if the particle does not undergo a nuclear reaction with the nuclei within the next voxel, the particle weights can be further reduced based on the modified particle weights in step S104. This further reduces the importance of the particles in determining tissue dose distribution, ensuring the accuracy of the simulation results.

[0084] In some embodiments, when executing the dose distribution acquisition method, the graphics processor of this application can run at least two threads, wherein each thread corresponds to at least one particle, and for any given thread, the thread is used to perform the following operations: Obtain the first position coordinates and direction of motion of the corresponding particle; Based on the principal voxel section, determine the motion distance of the corresponding particle; Based on the first position coordinates, direction of motion, and distance of motion of the corresponding particle, predict the second position coordinates that the corresponding particle can reach after its motion and the target voxel where the second position coordinates are located; The particle weights of the corresponding particles are updated based on the difference between the voxel cross section and the principal voxel cross section of the target voxel.

[0085] That is, multiple threads in the graphics processor can execute at least some steps of the dose distribution acquisition method in parallel, thereby improving simulation efficiency.

[0086] For details, please refer to the following: Figure 3 This is a simulation process for a graphics processor provided in some embodiments of this application. Figure 3 The graphics processor can execute the following steps in sequence: 1) Data loading. Specifically, the graphics processor can load slice images of tumor tissue and create a voxel mesh based on the slice images.

[0087] 2) Create the association between voxel indexes and voxel matrices.

[0088] 3) Run multiple threads for parallel computation. In this process, each thread can calculate and predict the motion of its corresponding particle based on the pre-created relationships between voxel grids, voxel indices, and voxel matrices, and modify the particle weights accordingly. This improves simulation efficiency.

[0089] Corresponding to the dose distribution acquisition method, this application also provides a dose distribution acquisition device. (See also...) Figure 4 This is a schematic diagram of a dose distribution acquisition device provided in the first embodiment of this application. The dose distribution acquisition device includes: Mesh building module 401 is used to build a voxel mesh based on organization information. The voxel mesh includes multiple voxels, each voxel has a voxel cross section, the voxel cross section represents the probability of a nuclear reaction between a voxel and a particle, and the particle represents a radiating particle with a motion direction in the voxel mesh. The distance determination module 402 is used to determine the movement distance of the particles based on the voxel principal cross section, where the voxel principal cross section is the largest voxel cross section in the voxel grid. The motion prediction module 403 is used to predict the target voxel that the particle can reach after its motion, based on the motion distance and motion direction. The weight update module 404 is used to update the particle weights based on the difference between the voxel cross section and the main voxel cross section of the target voxel, and to obtain the tissue dose distribution based on the updated particle weights.

[0090] In some embodiments, each voxel has multiple position coordinates; the motion prediction module 403 is used for: Obtain the first position coordinates of the particle in the voxel grid. The first position coordinates are the position coordinates of the particle before it moves. Using the first position coordinate as the starting point and the movement distance and direction as constraints, predict the second position coordinate that the particle can reach after its movement; The voxel at the second position coordinate is taken as the target voxel.

[0091] In some embodiments, each voxel has a voxel index; the motion prediction module 403 is used for: Based on the mapping relationship between position coordinates and voxel indices, the target voxel index corresponding to the second position coordinates is determined; The voxel identified by the target voxel index is used as the target voxel.

[0092] In some embodiments, the motion prediction module 403 is used to create a mapping relationship based on the following method: Obtain the coordinate system information of the voxel mesh. The coordinate system information includes the origin of the coordinate system and the direction of the coordinate axes. The position coordinates include the sub-coordinates located in each coordinate axis direction. A mapping relationship is created based on the distance between each sub-coordinate and the origin of the coordinate system, and the side length of a single voxel in each coordinate axis direction.

[0093] In some embodiments, the voxel mesh includes multiple particles; the method further includes: The distance determination module 402 runs at least two threads, each thread corresponding to at least one particle. For any given thread, the thread performs the following operation: determining the motion distance of the corresponding particle based on the voxel principal section; The motion prediction module 403 runs at least two threads, each thread corresponding to at least one particle. For any given thread, the thread performs the following operations: obtains the first position coordinates and direction of motion of the corresponding particle, and based on the first position coordinates, direction of motion, and distance of motion of the corresponding particle, predicts the second position coordinates that the corresponding particle can reach after its motion and the target voxel where the second position coordinates are located. The weight update module 404 runs at least two threads, each thread corresponding to at least one particle. For any given thread, the thread performs the following operation: based on the difference between the voxel cross section and the main voxel cross section of the target voxel, the particle weight of the corresponding particle is updated.

[0094] In some embodiments, the distance determination module 402 is further configured to: In the direction of motion, determine the distance between the first position coordinates and the boundary coordinates of the voxel mesh; If the interval distance is less than the movement distance, then the movement distance is updated to the interval distance.

[0095] In some embodiments, the distance determination module 402 is specifically used for: Generate the first random number; Based on the first random number and the principal voxel section, a random mapping is performed to obtain the motion distance. In the random mapping, the principal voxel section is used to determine the average motion distance of the particle, and the first random number is used to randomly generate the motion distance of the particle based on the average motion distance.

[0096] In some embodiments, the weight update module 404 is specifically used for: Determine the section ratio between the voxel section and the principal voxel section of the target voxel; If the cross-section ratio is less than the second random number, then update the particle weights.

[0097] In some embodiments, the weight update module 404 is specifically used for: The update magnitude of particle weights is determined based on the cross-section ratio, where the update magnitude is inversely proportional to the cross-section ratio. Update the particle weights according to the update magnitude.

[0098] In some embodiments, after the particle moves to the target voxel, the motion prediction module 403 is further configured to: If the particle does not meet the following conditions, then using the target voxel as the starting point, predict the next voxel that the particle can reach after continuing its motion: The particle undergoes a nuclear reaction with the target voxel; The energy of the particles is below the energy threshold; The target voxel where the particle is located is the boundary voxel of the voxel grid.

[0099] In some embodiments, while the particles continue to move, the weight update module 404 is further configured to: The particle weights are updated based on the difference between the voxel cross section and the principal voxel cross section of the next voxel.

[0100] In this embodiment, the dose distribution acquisition device is presented in the form of a functional unit. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0101] See also Figure 5 This is a schematic diagram of a graphics processor module provided in some embodiments of this application. Figure 5 In a graphics processing unit (GPU), multiple computing units are included, which can run in parallel for parallel execution. Figure 1 The method for obtaining dose distribution in [the context of the text]. Specifically, multiple computational units can be used in parallel. Figure 1 Steps S102-S104 in the process.

[0102] Compared to the serial operation of a traditional central processing unit (CPU), a graphics processing unit (GPU) can execute the dose distribution acquisition method of this application in parallel, thereby greatly improving simulation efficiency.

[0103] Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application, such as... Figure 6 As shown, the computer device includes: Figure 5The system includes one or more graphics processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other using different buses and can be mounted on a common motherboard or otherwise as required. The graphics processor can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of the GUI on external input / output devices (such as display devices coupled to the interface). In some alternative implementations, multiple graphics processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multi-graphics processor system). Figure 6 Take a graphics processor 10 as an example.

[0104] The graphics processor 10 may be a central graphics processor, a network graphics processor, or a combination thereof. The graphics processor 10 may further include a hardware chip. This hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GPRS), or any combination thereof.

[0105] The memory 20 stores instructions executable by at least one graphics processor 10 to cause the at least one graphics processor 10 to perform the method shown in the above embodiments.

[0106] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the graphics processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0107] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0108] This application provides a computer program product including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method of any embodiment of this application.

[0109] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for obtaining dose distribution, characterized in that, The method includes: Based on organizational information, a voxel grid is constructed, which includes multiple voxels. Each voxel has a voxel cross section, which represents the probability of the voxel reacting with a particle. The particle represents a radiating particle with a direction of motion in the voxel grid. The movement distance of the particle is determined based on the principal voxel cross section, where the principal voxel cross section is the largest voxel cross section in the voxel grid. Based on the distance and direction of motion, predict the target voxel that the particle can reach after its motion; Based on the difference between the voxel cross section of the target voxel and the main cross section of the voxel, the particle weights of the particles are updated, and the tissue dose distribution is obtained based on the updated particle weights.

2. The method according to claim 1, characterized in that, Each voxel has multiple position coordinates; the prediction of the target voxel that the particle can reach after its motion, based on the motion distance and the motion direction, includes: Obtain the first position coordinates of the particle on the voxel grid, where the first position coordinates are the position coordinates of the particle before the motion; Using the first position coordinates as the starting point, and the movement distance and the movement direction as constraints, predict the second position coordinates that the particle can reach after its movement; The voxel at the second position coordinate is taken as the target voxel.

3. The method according to claim 2, characterized in that, Each voxel has a voxel index; the step of using the voxel where the second position coordinates are located as the target voxel includes: Based on the mapping relationship between position coordinates and voxel indices, the target voxel index corresponding to the second position coordinates is determined; The voxel identified by the target voxel index is used as the target voxel.

4. The method according to claim 3, characterized in that, The mapping relationship is created based on the following method: Obtain the coordinate system information of the voxel mesh, the coordinate system information including the origin of the coordinate system and the direction of the coordinate axes, and the position coordinates including the sub-coordinates located in each of the coordinate axis directions; The mapping relationship is created based on the distance between each of the sub-coordinates and the origin of the coordinate system, and the side length of each voxel in each of the coordinate axis directions.

5. The method according to claim 2, characterized in that, The voxel grid comprises multiple particles; the method further includes: At least two threads are run, each thread corresponding to at least one particle, and for any given thread, the thread is used to perform the following operations: Obtain the first position coordinates and direction of motion of the corresponding particle; Based on the principal cross section of the voxel, the motion distance of the corresponding particle is determined; Based on the first position coordinates, direction of motion, and distance of motion of the corresponding particle, predict the second position coordinates that the corresponding particle can reach after its motion and the target voxel where the second position coordinates are located; The particle weights of the corresponding particles are updated based on the difference between the voxel cross section of the target voxel and the main cross section of the voxel.

6. The method according to claim 2, characterized in that, The method further includes: In the direction of motion, determine the distance between the first position coordinates and the boundary coordinates of the voxel mesh; If the interval distance is less than the movement distance, then the movement distance is updated to the interval distance.

7. The method according to claim 1, characterized in that, The determination of the particle's motion distance based on the voxel principal section includes: Generate the first random number; Based on the first random number and the voxel principal section, a random mapping is performed to obtain the motion distance. In the random mapping, the voxel principal section is used to determine the average motion distance of the particle, and the first random number is used to randomly generate the motion distance of the particle based on the average motion distance.

8. The method according to claim 1, characterized in that, The process of updating the particle weights based on the difference between the voxel cross-section of the target voxel and the principal voxel cross-section includes: Determine the cross-sectional ratio between the voxel cross section of the target voxel and the principal cross section of the voxel; If the cross-section ratio is less than the second random number, then the particle weights are updated.

9. The method according to claim 8, characterized in that, If the cross-sectional ratio is less than the second random number, then the particle weight is updated, including: Based on the cross-sectional ratio, the update magnitude of the particle weight is determined, wherein the update magnitude is inversely proportional to the cross-sectional ratio; The particle weights are updated according to the stated update magnitude.

10. The method according to claim 1, characterized in that, After the particle moves to the target voxel, the method further includes: If the particle does not meet the following conditions, then taking the target voxel as the starting point, predict the next voxel that the particle can reach after continuing to move, and based on the difference between the voxel cross-section of the next voxel and the main voxel cross-section, continue to update the particle weight: The particle undergoes a nuclear reaction with the target voxel; The energy of the particle is below the energy threshold; The target voxel containing the particle is the boundary voxel of the voxel grid.

Images