Radiation irradiation system and control method thereof

Abstract

An embodiment of the present invention provides a radiation irradiation system, comprising a beam irradiation device configured to generate a therapeutic beam and irradiate the therapeutic beam onto an irradiated body to form an irradiated area; a treatment planning module configured to perform dose simulation calculations and generate a treatment plan based on parameters of the therapeutic beam and medical imaging data of the irradiated area, wherein the treatment planning module utilizes variance reduction to simulate particles; and a control module configured to control the irradiation of the beam irradiation device according to the treatment plan. Other embodiments of the present invention also provide a corresponding method for controlling a radiation irradiation system.

Description

Technical Field

[0001] One aspect of the present invention relates to a radiation irradiation system; another aspect of the present invention relates to a control method for a radiation irradiation system. Background Art

[0002] With the advancement of atomic science, radiation therapy, such as cobalt-60, linear accelerators, and electron beams, has become a primary approach to cancer treatment. However, conventional photon or electron therapy is limited by the physical properties of radiation. While killing tumor cells, it can also damage a significant amount of normal tissue in the beam's path. Furthermore, due to the varying sensitivity of tumor cells to radiation, conventional radiation therapy is often ineffective in treating more radioresistant malignancies, such as glioblastoma multiforme and melanoma.

[0003] To reduce radiation damage to normal tissues surrounding the tumor, the concept of targeted therapy from chemotherapy has been applied to radiotherapy. Furthermore, for highly radioresistant tumor cells, radiation sources with high relative biological effectiveness (RBE), such as proton therapy, heavy particle therapy, and neutron capture therapy, are currently being actively developed. Neutron capture therapy, such as boron neutron capture therapy (BNCT), combines these two concepts. By leveraging the specific accumulation of boron-containing drugs in tumor cells and combining them with precise beam control, BNCT offers a superior cancer treatment option compared to traditional radiation.

[0004] Boron neutron capture therapy is the use of boron ( 10 B) The drug has a high capture cross section for thermal neutrons, which is produced by 10B(n,α)7Li neutron capture and nuclear fission reaction. 4 He and 7 Li has two heavily charged particles, and their combined range is approximately equivalent to the size of a cell. Therefore, the radiation damage caused to the organism can be limited to the cellular level. When boron-containing drugs selectively accumulate in tumor cells and are combined with an appropriate neutron radiation source, the purpose of locally killing tumor cells can be achieved without causing too much damage to normal tissues.

[0005] In order to ensure that the radiation particles kill as many cancer cells as possible while minimizing damage to normal cells, a CT or PET scan is usually performed before treating the patient. The tissue material information of the human body is obtained based on the scan results. A computational model is established based on the material information and the radiation source to simulate the transport process of the radiation particles in the human body. Ultimately, the dose distribution of the radiation particles in the human body is obtained, and the plan with the optimal dose distribution for the patient is then selected as the treatment plan for the patient.

[0006] The dose calculation module in current radiotherapy planning systems primarily uses the Monte Carlo method to simulate radiation particles. Traditional radiotherapy requires simulating the motion of photons and electrons, while radiation therapy requires simulating the motion of neutrons and photons. While the Monte Carlo method is currently the most accurate method for dose calculation, it is time-consuming and memory-intensive.

[0007] Currently, radiotherapy calculations are mostly based on general-purpose Monte Carlo programs such as MCNP and Geant4. MCNP was originally developed for reactor design calculations, while Geant4 was originally designed for high-energy physics calculations. These programs were not designed with the physical context of radiotherapy in mind and, therefore, were not specifically optimized for this application. Radiotherapy plans typically must be completed within a specified timeframe. For example, radiation therapy requires the planning system to produce a treatment plan within one hour. Dose calculations account for a significant portion of the planning process, necessitating optimized dose calculation methods to reduce this time. Summary of the Invention

[0008] To overcome the deficiencies of the prior art, the present invention provides, in a first aspect, a radiation irradiation system, comprising: a beam irradiation device, the beam irradiation device being configured to generate a therapeutic beam and irradiate the therapeutic beam onto an irradiated body to form an irradiated portion; a treatment planning module, the treatment planning module being configured to perform dose simulation calculations and generate a treatment plan based on parameters of the therapeutic beam and medical imaging data of the irradiated portion, wherein the treatment planning module uses variance reduction to simulate particles; and a control module, the control module being configured to control the irradiation of the beam irradiation device according to the treatment plan.

[0009] Furthermore, variance reduction includes latent capture, window game and bet splitting; the radiation irradiation system is a neutron capture therapy system, the beam irradiation device includes a neutron generator, a beam shaper and a treatment table, the neutron generator includes an accelerator and a target material, the accelerator accelerates charged particles to generate charged particle beams and interacts with the target material to generate neutron beams, the beam shaper can adjust the neutron beams generated by the neutron generator to a preset beam quality, and the neutron beams generated by the neutron generator are irradiated to the irradiated object on the treatment table through the beam shaper.

[0010] A second aspect of the present invention provides a control method for a radiation irradiation system, wherein the beam irradiation system includes: a beam irradiation device, the beam irradiation device is used to generate a therapeutic beam and irradiate the therapeutic beam to an irradiated body to form an irradiated part; a treatment planning module, the treatment planning module is used to perform dose simulation calculations and generate a treatment plan based on the parameters of the therapeutic beam and medical imaging data of the irradiated part; and a control module, the control module is used to control the irradiation of the beam irradiation device according to the treatment plan; the control method of the radiation irradiation system includes the treatment planning module using variance reduction to simulate particles.

[0011] Furthermore, variance reduction includes implicit capture, weight window game and bet splitting; using variance reduction to simulate particles includes: S1: obtaining a source particle; S2: judging whether the particle collides in the gate element, if so, executing S3 and S4 in sequence, if not, executing S5 and S6 in sequence; S3: implicit capture processing; S4: judging whether the weight is lower than the weight window, if so, executing S7, if not, returning to S2; S5: bet splitting processing; S6: judging whether to perform bet processing, if yes, executing S7, if not, returning to S2; S7: judging whether the bet is dead, if yes, executing S8, if not, executing S9 and returning to S2; S8: judging whether the particle has been processed, if yes, the process ends, if not, returning to S1; S9: dividing the weight by the probability of bet death.

[0012] Furthermore, the bet splitting includes: S1: calculating the grid space importance of each grid and recording them one by one; S2: obtaining particles; S3: performing weight window inspection and operation on particles; S4: calculating the grid space importance of particles before and after crossing the grid boundary. n and I n+1 ; S5: Comparison I n Is it greater than I n+1 If not, execute S6 and return to S3. If yes, execute S7. S6: split particles and reduce particle weights. S7: determine whether to kill particles. If so, execute S9. If not, execute S8 and return to S3. S8: increase particle weights. S9: determine whether the particle simulation is complete. If not, return to S2. If the simulation is complete, end.

[0013] Furthermore, implicit capture includes: S1: acquiring particles; S2: determining whether the particles collide in the cell, if so, executing S3, if not, returning to S1; S3: multiplying the weight by the probability of scattering; S4: determining whether the particle weight is less than the minimum weight, if so, executing S5, if not, returning to S2; S5: determining whether the particle is killed, if so, executing S7; if not, executing S6 and returning to S2; S6: dividing the weight by the probability of death; S7: determining whether the particle is simulated, if determined not to be simulated, returning to S1; if determined to be simulated to be completed, ending.

[0014] Furthermore, the weight window game includes: S1: simulating particle movement; S2: judging whether the particle weight is within the weight window range, if so, returning to S1, if not, executing S3; S3: judging whether the weight is greater than the weight window, if so, executing S4 and returning to S1, if not, executing S5; S4: splitting particles and reducing weight; S5: judging whether the bet is dead, if so, ending, if not, executing S6 and returning to S1; S6: increasing the particle weight.

[0015] In one embodiment, the grid space importance is obtained by solving the adjoint transport equation, which has the following form:

[0016]

[0017] where φ * is the accompanying flux, S * is the accompanying source, v is the speed of the particle, Ω is the direction of the particle, Σ t is the reaction cross section of the collision between particles and matter, Σ s is the scattering cross section, r is the position of the particle, E is the energy of the particle, and t is time.

[0018] More specifically, the particle is from the grid space importance I n The importance of mesh movement to mesh space is I n+1 Grid, if I n+1 >I n , let m = I n+1 / I n , the particle splits into m particles, and the weight of each particle is reduced to 1 / m of the original; if I n+1 n , particles make bets, let P = I n+1 / I n , sample a random number x from 0 to 1. If x is less than P, the particle survives and the weight is multiplied by 1 / P. Otherwise, the particle is killed and the simulation of the particle is terminated.

[0019] In other embodiments, the upper limit of particle weight is 10 and the lower limit is 0.25. When the particle weight w is greater than 10, let the integer part of w be w1 and the fractional part be w2. A random number x is sampled between 0 and 1. When x is less than w2, let w = w1 + 1; when x is greater than w2, let w = w1. Then, a particle is split into w particles and the simulation is repeated until the weight of each particle drops to 1.

[0020] The radiation irradiation system and control method thereof described in the embodiments of the present invention use variance reduction to simulate particles, which can reduce calculation time while ensuring calculation accuracy. BRIEF DESCRIPTION OF THE DRAWINGS​

[0021] Figure 1 Schematic diagram of the modules of the boron neutron capture therapy system according to an embodiment of the present invention.

[0022] Figure 2 Schematic diagram of the structure of a boron neutron capture therapy system according to an embodiment of the present invention.

[0023] Figure 3 This is a flowchart of process parallelism and thread parallelism calculations of the CPU in an embodiment of the present invention.

[0024] Figure 4 This is a flowchart of GPU accelerated simulation calculation in an embodiment of the present invention.

[0025] Figure 5 This is a flow chart of simulating particles using the variance reduction technique in an embodiment of the present invention.

[0026] Figure 6 1 is a flow chart of bet splitting in an embodiment of the present invention.

[0027] Figure 7 4 is a flowchart of implicit capture in an embodiment of the present invention.

[0028] Figure 8 It is a flow chart of the right window game in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The embodiments of the present invention are further described in detail below with reference to the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0030] Neutron capture therapy has been increasingly used as an effective means of treating cancer in recent years. Boron neutron capture therapy is the most common. Neutrons for boron neutron capture therapy can be supplied by nuclear reactors or accelerators. Boron neutron capture therapy (BNCT) is a method of using boron-containing ( 10 B) The drug has a high capture cross section for thermal neutrons. 10 B(n,α) 7 Li neutron capture and nuclear fission reaction production 4 He and 7 The average energy of the two heavily charged particles is about 2.33 MeV, which has the characteristics of high linear energy transfer (LET) and short range. The linear energy and range of α particles are 150 KeV / μm and 8 μm respectively. 7Li heavy-charged particles have a range of 175 keV / μm and 5 μm, respectively. The combined range of these two particles is roughly equivalent to the size of a cell, thus limiting radiation damage to the organism to the cellular level. When boron-containing drugs selectively accumulate in tumor cells, combined with an appropriate neutron source, they can precisely kill tumor cells without causing significant damage to normal tissue.

[0031] Whether the neutron source for boron neutron capture therapy comes from a nuclear reactor or a nuclear reaction between charged particles and a target, the resulting radiation field is a mixed one—that is, the beam contains neutrons and photons ranging from low to high energies. For boron neutron capture therapy of deep-seated tumors, the greater the amount of radiation other than epithermal neutrons, the greater the proportion of nonselective dose deposition in normal tissue. Therefore, the amount of radiation that causes unnecessary dose deposition should be minimized. To better understand neutron dose distribution in the human body, in addition to the air beam quality factor, embodiments of the present invention use a human head tissue prosthesis for dose distribution calculations, and use the prosthesis beam quality factor as a reference for neutron beam design.

[0032] The International Atomic Energy Agency (IAEA) has established five recommended air beam quality factors for neutron sources used in clinical boron neutron capture therapy. These five recommendations can be used to compare the advantages and disadvantages of different neutron sources and serve as a reference for selecting neutron production methods and designing beam shapers. The five recommendations are as follows:

[0033] Epithermal neutron flux>1 x 10 9 n / cm 2 s

[0034] Fast neutron contamination <2 x 10 -13 Gy-cm 2 / n

[0035] Photon contamination<2 x 10 -13 Gy-cm 2 / n

[0036] Thermal to epithermal neutron flux ratio<0.05

[0037] Epithermal neutron current to flux ratio>0.7

[0038] Note: The energy range of epithermal neutrons is between 0.5eV and 40keV, the energy range of thermal neutrons is less than 0.5eV, and the energy range of fast neutrons is greater than 40keV.

[0039] Reference Figure 1 As shown, the radiation irradiation system in this embodiment is a boron neutron capture therapy system 100, which includes a neutron beam irradiation device 10, a treatment planning module 20, and a control module 30. The neutron beam irradiation device 10 is used to generate a therapeutic neutron beam N and irradiate the therapeutic neutron beam N onto an irradiated object 200 to form an irradiated area. The treatment planning module 20 performs dose simulation calculations based on the parameters of the therapeutic neutron beam N generated by the neutron beam irradiation device 10 and the medical imaging data of the irradiated area, and generates a treatment plan. The treatment plan determines the position of the irradiated area relative to the neutron beam irradiation device 10 during the irradiation treatment and the corresponding irradiation time. After the irradiated object 200 is positioned according to the position determined by the treatment plan, treatment can begin. The control module 30 retrieves the treatment plan corresponding to the current irradiated object 200 from the treatment planning module 20 and controls the irradiation of the neutron beam irradiation device 10 according to the treatment plan. The control module 30 can also receive other data information, such as data from the neutron beam irradiation device 10 and data from the irradiated object 200.

[0040] Reference Figure 2 As shown, in this embodiment, the neutron beam irradiation device 10 includes a neutron generator 11, a beam shaper 12, a collimator 13 and a treatment table 14. The neutron generator 11 includes an accelerator 111 and a target material T. The accelerator 111 accelerates charged particles (such as protons, deuterons, etc.) to generate a charged particle line P such as a proton line. The charged particle line P irradiates the target material T and interacts with the target material T to generate a neutron line (neutron beam) N. The target material T is preferably a metal target material. The appropriate nuclear reaction is selected based on the required neutron yield and energy, the energy and current of the accelerated charged particles that can be provided, the physical and chemical properties of the metal target material, and other characteristics. Commonly discussed nuclear reactions include 7 Li(p,n) 7 Be and 9 Be(p,n) 9B, both reactions are endothermic reactions. The energy thresholds of the two nuclear reactions are 1.881MeV and 2.055MeV, respectively. Since the ideal neutron source for boron neutron capture therapy is epithermal neutrons at the keV energy level, theoretically, if protons with energies just slightly above the threshold are used to bombard a lithium metal target, relatively low-energy neutrons can be produced, which can be used clinically without much slowing down. However, the cross-sections of lithium metal (Li) and beryllium metal (Be) targets with protons at the threshold energy are not high. To generate a sufficiently large neutron flux, higher-energy protons are usually selected to trigger the nuclear reactions. The ideal target should have a high neutron yield, a neutron energy distribution close to the epithermal neutron energy region (described in detail below), no excessive strong penetrating radiation, be safe, cheap, easy to operate, and resistant to high temperatures. However, in reality, it is impossible to find a nuclear reaction that meets all these requirements. In the embodiments of the present invention, a target made of lithium metal is used. However, as is well known to those skilled in the art, the target material T can also be made of metal materials other than lithium and beryllium, such as tantalum (Ta) or tungsten (W); the target material T can be in the shape of a disk, or other solid shapes, or a liquid (liquid metal). The accelerator 111 can be a linear accelerator, a cyclotron, a synchrotron, or a synchrocyclotron, and the neutron generating device 11 can also be a nuclear reactor instead of an accelerator and a target. Regardless of whether the neutron source of boron neutron capture therapy comes from a nuclear reactor or a nuclear reaction between charged particles in the accelerator and the target material, what is actually generated is a mixed radiation field, that is, the beam contains neutrons and photons ranging from low energy to high energy. For boron neutron capture therapy of deep tumors, in addition to epithermal neutrons, the more radiation content, the greater the proportion of non-selective dose deposition to normal tissues, so these radiations that cause unnecessary doses should be reduced as much as possible. In addition, for the normal tissues of the irradiated body, excessive amounts of various radiations should be avoided, which also causes unnecessary dose deposition.

[0041] The neutron beam N generated by the neutron generator 11 sequentially passes through the beam shaper 12 and collimator 13 and irradiates the irradiated object 200 on the treatment table 14. The beam shaper 12 adjusts the beam quality of the neutron beam N generated by the neutron generator 11, while the collimator 13 focuses the neutron beam N, ensuring high targeting during treatment. It is understood that the present invention can also be implemented without a collimator; the beam, after exiting the beam shaper 12, directly irradiates the irradiated object 200 on the treatment table 14.

[0042] The beam shaper 12 further includes a reflector 121, a retarder 122, a thermal neutron absorber 123, a radiation shield 124, and a beam outlet 125. Since the neutrons generated by the neutron generator 11 have a wide energy spectrum, in addition to epithermal neutrons that meet treatment needs, it is necessary to minimize the content of other types of neutrons and photons to avoid harm to the operator or the irradiated object. Therefore, the neutrons emitted from the neutron generator 10 need to pass through the retarder 22 to adjust the fast neutron energy (>40keV) to the epithermal neutron energy range (0.5eV-40keV) and minimize the thermal neutrons (<0.5eV). The retarder 22 is made of a material with a large cross section for fast neutrons and a small cross section for epithermal neutrons. As a preferred embodiment, the retarder 122 is made of D2O, AlF3, Fluental TM , CaF2, Li2CO3, MgF2 and Al2O3; the reflector 121 surrounds the retarder 122 and reflects the neutrons that pass through the retarder 122 and diffuse to the surroundings back to the neutron beam N to improve the utilization rate of the neutrons. It is made of a material with strong neutron reflection ability. As a preferred embodiment, the reflector 121 is made of at least one of Pb or Ni; there is a thermal neutron absorber 123 at the rear of the retarder 122, which is made of a material with a large cross-section for interacting with thermal neutrons. As a preferred embodiment, the thermal neutron absorber 123 is made of Li-6. 3 is used to absorb thermal neutrons that pass through the retarder 122, thereby reducing the thermal neutron content in the neutron beam N and preventing excessive dose to superficial normal tissue during treatment. It is understood that the thermal neutron absorber can also be integrated with the retarder, and the retarder material contains Li-6. The radiation shield 124 is used to shield neutrons and photons that leak out of the beam exit 125. The material of the radiation shield 124 includes at least one of a photon shielding material and a neutron shielding material. In a preferred embodiment, the material of the radiation shield 124 includes lead (Pb) as a photon shielding material and polyethylene (PE) as a neutron shielding material. The collimator 13 is positioned behind the beam exit 125. The epithermal neutron beam exiting the collimator 13 irradiates the irradiated object 200, and after passing through superficial normal tissue, it is retarded into thermal neutrons that reach the tumor cells M. It is understood that the beam shaping body 20 may have other configurations, as long as it can produce the epithermal neutron beam required for treatment. For ease of description, when the collimator 13 is provided, the exit of the collimator 13 may also be referred to as the beam exit 125 below. In this embodiment, a radiation shielding device 15 is also provided between the irradiated object 200 and the beam exit 125 to shield the normal tissues of the irradiated object from the radiation emitted by the beam exit 125. However, it is understood that the radiation shielding device 15 may also be omitted.

[0043] After the irradiated body 200 takes or is injected with a boron-containing (B-10) drug, the boron-containing drug selectively accumulates in the tumor cells M. Then, the boron-containing (B-10) drug has a high capture cross section for thermal neutrons. 10 B(n,α) 7 Li neutron capture and nuclear fission reaction production 4 He and 7 The average energy of the two charged particles is about 2.33 MeV, with high linear energy transfer (LET) and short range. The linear energy transfer and range of alpha particles are 150 keV / μm and 8 μm respectively. 7 The Li heavy-charged particles are 175keV / μm and 5μm, and the total range of the two particles is approximately equivalent to the size of a cell. Therefore, the radiation damage caused to the organism can be limited to the cellular level, achieving the purpose of locally killing tumor cells without causing too much damage to normal tissues.

[0044] The boron neutron capture therapy system 100 is housed entirely within a concrete building. Specifically, it includes an irradiation chamber 101 and a charged particle beam generation chamber 102. An irradiated object 200 on a treatment table 14 undergoes neutron beam N irradiation therapy in the irradiation chamber 101. The charged particle beam generation chamber 102 at least partially houses an accelerator 111. The beam shaper 12 is at least partially housed within a partition 103 separating the irradiation chamber 101 and the charged particle beam generation chamber 102. It is understood that the partition 103 can completely separate the irradiation chamber 101 from the charged particle beam generation chamber 102, or it can partially separate the irradiation chamber 101 and the charged particle beam generation chamber 102, with the irradiation chamber 101 and the charged particle beam generation chamber 102 communicating with each other. There can be one or more targets T, and the charged particle beam P can selectively interact with one or more targets T, or simultaneously interact with multiple targets T, to generate one or more therapeutic neutron beams N. Depending on the number of targets T, there can be one or more beam shapers 12, collimators 13, and treatment tables 14. Multiple treatment tables can be located within the same irradiation chamber, or each treatment table can be housed in a separate irradiation chamber. The irradiation chamber 101 and charged particle beam generation chamber 102 are enclosed by concrete walls W (including partitions 103). The concrete structure shields neutrons and other radiation that may leak during operation of the boron neutron capture therapy system 100.

[0045] In order to kill cancer cells to the greatest extent possible while reducing the damage of radiation to normal tissues, the accuracy of the dose distribution of epithermal neutrons and photons is particularly important in the setting of the treatment plan module 20. In the application scenario of the radiation irradiation system, the dose calculation needs to read the photon and neutron cross-section database. Among them, the neutron cross-section database is very large, which causes the software for formulating the treatment plan of the radiation irradiation system to occupy a large storage space when installed, and when the software is used for dose calculation simulation, the memory space occupied is large and the calculation time is long. The following embodiments of the present invention provide a series of optimization methods for the treatment plan module, which reduce the running memory and the dose calculation time at the same time to meet the needs of quickly formulating radiotherapy plans. Each optimization method is described below in conjunction with the accompanying drawings.

[0046] Embodiment 1: Optimizing the database in the treatment planning module 20 .

[0047] The human body contains over 60 elements, over 20 of which are essential and crucial for maintaining normal physiological functions. Carbon, hydrogen, oxygen, nitrogen, phosphorus, chlorine, sodium, magnesium, potassium, and calcium are the most abundant elements in the human body. Carbon, hydrogen, oxygen, and nitrogen are the primary elements that make up the body's organic matter, accounting for 96% of the body's total weight. The remaining elements, which account for more than 0.01% of the body's total weight, are calcium, potassium, phosphorus, sulfur, chlorine, magnesium, and sodium, accounting for 1.5%, 0.35%, 1%, 0.25%, 0.15%, 0.05%, and 0.15% of the body's total weight, respectively. Therefore, the content of carbon, hydrogen, oxygen, nitrogen, calcium, potassium, phosphorus, sulfur, chlorine, magnesium, and sodium accounts for 99.45% of the body's total weight. Monte Carlo simulations and calculations store data for over 100 elements. Each Monte Carlo simulation and calculation requires a significant time-consuming process, as each element must be simulated and calculated for each element. In addition, because it stores calculations and simulation systems corresponding to a large number of elements, the Monte Carlo database is extremely large. However, in the application scenario of radiation irradiation systems, when using Monte Carlo to simulate photons, neutrons and other particles in the human body, a relatively accurate calculation result can be obtained by simulating only the elements that account for a large proportion of the total human weight (greater than 0.01%). The simulation calculation of some trace elements will take more time, but the difference between the calculation result and the result obtained by simulating only the elements that account for 1 / 10,000 of the total human weight or more is within 0.001%.

[0048] Furthermore, some elements, such as Fe, although present in the human body by weight, do not react with neutrons and photons, or their reaction with neutrons and photons is weak. Therefore, the impact of these elements on the simulation results can be ignored. On the other hand, some elements, such as boron, present in the human body by weight, but in the application scenario of radiation irradiation systems, large amounts of boron are injected into the human body, and their reaction with neutrons is relatively strong, which has a significant impact on the final calculation results.

[0049] By comprehensively considering the weight proportion of elements in the human body and the reaction intensity with neutrons and photons, elements (selected from one or more of H, He, Li, Be, B, C, N, O, F, Na, Mg, Al, Si, P, S, Cl, Ar, K and Ca) that have an impact on the simulation calculation results of neutrons and photons in the application scenario of the radiation irradiation system are screened out. During the simulation process, only the screened elements are simulated, thereby reducing the simulation calculation time while ensuring the calculation accuracy, thereby improving the operation efficiency of Monte Carlo. In addition, the database capacity corresponding to these screened elements is only 2% to 3% of the original data path capacity, which greatly reduces the disk memory requirement and reduces the manufacturing cost.

[0050] On the other hand, within the neutron cross-section database corresponding to each element, the Monte Carlo system stores cross-section data for multiple temperatures to choose from, such as 0K, 1200K, 2500K, 250K, 294K, 600K, and 900K. However, in the application scenario of the radiation irradiation system, only the human body model needs to be simulated, and the normal human body temperature is between 310K and 315K. According to the operating principle of Monte Carlo, after the temperature is entered, the system automatically matches the temperature closest to the input temperature among the multiple stored temperatures and uses this temperature as the simulation parameter for the simulation calculation. Therefore, when the human body temperature is entered, the system automatically matches the temperature of 294K stored in the database and performs the simulation calculation using 294K as the parameter. In addition, the temperature used for Doppler effect calculations is 0K. That is to say, in the application scenario of the radiation irradiation system, only 294K and 0K temperatures stored in the database will be used, and the remaining temperatures are redundant for the application of the radiation irradiation system. Therefore, by removing the databases corresponding to temperatures other than 294K and 0K, without affecting the advance of the simulation calculation structure, the size of the database can be reduced to about one-third of the original size, thereby reducing the operating cost of the database.

[0051] Table 1 shows the results of neutron dose simulation calculations performed by the Monte Carlo system on the same human body model at temperatures of 294K, 310K, and 330K.

[0052] Table 1: Results of neutron dose simulations performed by the Monte Carlo system on the same human body model at temperatures of 294K, 310K, and 330K

[0053] temperature 294K 310K 330K Neutron dose 146.63eV / g 146.085eV / g 145.596eV / g

[0054] From the data in Table 1, it can be seen that the influence of small temperature fluctuations on the final calculation results of neutron dose can be ignored. In the application of radiation irradiation system, only retaining the database corresponding to 294K stored in the system can meet the use requirements and the error of the final calculation result is within an acceptable range.

[0055] Embodiment 2: Cutting off the process during particle simulation to reduce the particle simulation time.

[0056] In the application scenario of a radiation irradiation system, dose calculation includes neutron dose calculation and photon dose calculation. As photons move through the body, their energy and half-absorption thickness gradually decrease, while the absorption cross section increases rapidly as the energy and half-absorption thickness decrease. When the half-absorption thickness of a photon is less than the size of a cell, the photon is likely absorbed within the cell. In this case, the photon is assumed to deposit all its energy within the cell, and no further simulation is performed. This treatment is consistent with the dose distribution calculated using continuous photon simulation, and the error is within 0.1%.

[0057] The size of human cells is 2-200 microns. Usually, in the application of radiation irradiation system, the minimum size of the model grid is 0.8mm. When the half-absorption thickness of the photon is less than one-quarter of the minimum grid size, the simulation of the photon can be stopped.

[0058] Specifically, the half-absorption thickness t of photons is calculated using formula (1-1):

[0059]

[0060] Where μ is the linear attenuation factor of the photon, which is determined by the material the photon passes through and the photon energy.

[0061] Taking human bones as an example, the following calculations show the half-absorption thickness of bones for different photon energy values.

[0062] Table 2: Half-absorption thickness in bone for different photon energies

[0063] Photon energy (KeV) Semi-absorption thickness (mm) 0.1 0.0006 1 0.001 4 0.029 6 0.089 8 0.105 10 0.137 12 0.365 14 0.471 16 0.586 18 0.867 20 0.970

[0064] As can be seen from Table 2, within the bone, when the photon energy is less than 10 KeV, the half-absorption thickness of the photon is less than 0.2 mm. In other words, when the photon energy is less than 10 KeV, the photon is very likely to be absorbed within the grid in which it is currently located. Therefore, the difference between the dose obtained by stopping the photon simulation at this time and the dose obtained by further simulating the photon is within an acceptable range.

[0065] Because the half-absorption thickness of photons in cell tissue depends on the material they pass through and the photon energy, a corresponding cutoff energy can be set for each material. When the photon energy falls below this cutoff energy, the photon simulation is terminated. The following simulation calculations in a human body model show the corresponding photon doses for different photon cutoff energies.

[0066] Table 3: Photon energies calculated by simulations with different photon cutoff energies in the human body model

[0067]

[0068]

[0069] As can be seen from Table 2, when the photon energy is less than or equal to 10 KeV, the simulation calculation of the photon is stopped, and the error between the final calculated photon dose and the dose distribution calculated by continuing the simulation of the photon is within 0.1%.

[0070] In summary, during the photon simulation calculation process, when the photon energy is less than the preset cutoff energy, the half-absorption thickness of the photon is less than the size of a cell. At this time, stopping the simulation of the photon can reduce the calculation time of the photon dose without affecting the dose distribution calculation results.

[0071] In other embodiments, the first preset value and the second preset value can be set according to the actual simulation accuracy requirements. When the half-absorption thickness of the photon is less than or equal to the first preset value, or when the photon energy is less than or equal to the second preset value, the simulation of the photon is stopped. The first preset value can be greater than or less than the size of a cell, and the second preset value can be 12KeV, 16KeV, etc.

[0072] Example 3: Using parallel technology to accelerate the dose calculation module.

[0073] The Monte Carlo method is used in the radiation irradiation system's treatment planning module. The simulation processes for different source particles are completely independent, meaning the order in which they are simulated has no effect on the calculated results. Based on this computational characteristic, simulation tasks for different source particles can be assigned to different processes or threads. After each process or thread completes its computational task, the tasks are aggregated to produce the final result. Computers offer two types of parallelism: CPU process parallelism, thread parallelism, and GPU acceleration.

[0074] Reference Figure 3 As shown in Figure 1, the so-called CPU process parallelism and thread parallelism computational process is as follows: First, the system obtains the number of processes or threads, a value n; then, the system divides the particles to be simulated into n equal parts; next, each thread or process simulates and counts each particle individually; finally, the system tallies the counts obtained by each process or thread to obtain the final dose. Because particles are assigned to different processes or threads for simultaneous simulation and computation, the simulation and computation time is compressed to one-nth (process) or one-2nth of the time required to simulate and compute all particles using a single thread or process.

[0075] Both process and thread parallelism utilize multi-core CPUs to achieve parallel computing, while GPU acceleration utilizes GPUs for multi-processor parallel computing. The effectiveness of process and thread parallelism is limited by the number of CPU cores. A typical single-machine has four cores, so process and thread parallelism can only increase computing speed by a factor of four (for processes) or eight (for threads). However, a GPU integrates multiple processors, theoretically increasing computing speed many times over.

[0076] Reference Figure 4 As shown in the figure, the process of so-called GPU accelerated simulation calculation is as follows: first, the system transfers random numbers, cross-section data, etc. from the CPU memory to the GPU memory. Then, each processor in the GPU simulates, calculates, and counts a single particle, and adds the counting results to the global count. Next, the system determines whether there are any particles that have not been simulated. If there are no particles that have not been simulated, the counting results are transferred from the GPU memory to the CPU memory. If there are still particles that have not been simulated, it returns to the previous step and continues to simulate and count the particles that have not been simulated until all particles are simulated.

[0077] Due to the abundance of GPU processors, using GPUs to accelerate simulations can significantly increase computation speed and reduce computation time. Furthermore, the CPU uses an Intel Xeon processor with a 2.27GHz frequency, priced at approximately 6,000 yuan, while the GPU uses an NVIDIA Tesla C2050 processor, priced at approximately 9,000 yuan. Each GPU has a total of 448 processors, equivalent to 448 CPUs. However, CPU computation time is 50 to 70 times that of a GPU. Therefore, using GPUs for simulations offers significant advantages over CPUs in both cost and time.

[0078] Example 4: Using the Monte Carlo variance reduction technique to accelerate convergence.

[0079] Monte Carlo methods offer a variety of variance reduction techniques to improve computational speed. Treatment planning systems employ techniques such as latent capture, weighted window games, and gambling splitting. Gambling splitting consists of two techniques: gambling and splitting. The following explains the variance reduction techniques involved in gambling, splitting, the importance of grid space, latent capture, and weighted window games.

[0080] Betting technique: Typically, the upper limit of particle weight is set to 10 and the lower limit is set to 0.25. When the particle weight w drops below a preset value (for example, 0.25), a random number x is sampled between 0 and 1. When x is less than w, the particle survives and the particle weight returns to 1. If x is greater than or equal to w, the particle is killed and the simulation is terminated.

[0081] Splitting technique: When the particle weight w is greater than a certain value (for example, 10), let the integer part of w be w1 and the decimal part be w2. Sample a random number x between 0 and 1. When x is less than w2, let w = w1 + 1. When x is greater than w2, let w = w1. Then split the particle into w particles and simulate until the weight of each particle drops to 1.

[0082] Grid spatial importance: Particles in different regions of the model contribute differently to the dose. Spatial importance is used to characterize the contribution of particles to the dose. Before simulation, the grid spatial importance of each grid is calculated based on parameters such as tumor location and model characteristics. n The importance of mesh movement to mesh space is I n+1 Grid, if I n+1 >I n , let m = I n+1 / I n , the particle splits into m particles, and the weight of each particle is reduced to 1 / m of the original; if I n+1 n , particles make bets, let P = I​n+1 / I n , sample a random number x from 0 to 1. If x is less than P, the particle survives and the weight is multiplied by 1 / P. Otherwise, the particle is killed and the simulation of the particle is terminated.

[0083] Calculation method of grid space importance: It can be obtained by calculating the adjoint flux, which is obtained by solving the adjoint transport equation. The adjoint transport equation is as follows:

[0084]

[0085] where φ * is the accompanying flux, S * is the accompanying source, v is the speed of the particle, Ω is the direction of the particle, Σ t is the reaction cross section of the collision between particles and matter, Σ s is the scattering cross section, r is the position of the particle, E is the energy of the particle, and t is time.

[0086] Implicit capture: When particles interact with matter, they are only scattered and not absorbed. The weight of each collision particle is multiplied by P. 散射 / P 总 When the particle weight drops to a certain value (such as 0.25), the particle is processed using the gambling technique.

[0087] Weight Window Game: When the particle weight is greater than a certain value, such as 10, a splitting technique is used. When the particle weight is less than a certain value, such as 0.25, a gambling technique is used. These two values ​​are set according to software performance.

[0088] In general, when a particle moves from a grid with low spatial importance to one with high spatial importance, its weight increases, preventing the particle from being killed. When a particle moves from a grid with high spatial importance to one with low spatial importance, its weight decreases, leading to particle splitting or latent capture. Because the absorption cross-section of thermal neutrons in the human body (primarily for elements such as nitrogen and boron) is large, latent capture should be used to accelerate the convergence of dose calculations for grids farther from the neutron source. When a particle is far from the treatment area, further simulation of the particle's contribution to the dose calculation of the treatment area is of little significance, and continued calculations in this situation waste computational resources. Therefore, setting grid spatial importance can reduce the likelihood of this occurring. If a particle's weight is too small, its contribution to the count is minimal, and continued simulation wastes computational resources. If a particle's weight is too large, individual counts are overstated, risking distorted results. Therefore, a weight window should be set to keep the particle weight within an appropriate range. Therefore, a combination of latent capture, appropriate grid spatial importance settings, and weight window games is necessary to accelerate the convergence of Monte Carlo simulations.

[0089] Reference Figure 5 As shown in FIG, the process of simulating particles using the variance reduction technique is as follows: S1: obtain a source particle; S2: determine whether the particle collides in the gate element, if so, execute S3 and S4 in sequence, if not, execute S5 and S6 in sequence; S3: implicit capture processing; S4: determine whether the weight is lower than the weight window, if so, execute S7, if not, return to S2; S5: bet splitting processing; S6: determine whether to perform bet processing, if yes, execute S7, if not, return to S2; S7: determine whether the bet is dead, if yes, execute S8, if not, execute S9 and return to S2; S8: determine whether the particle has been processed, if yes, the process ends, if not, return to S1; S9: the weight is divided by the probability of bet death.

[0090] Reference Figure 6 As shown in the figure, the bet splitting process is as follows: S1: calculate the grid space importance of each grid and record them one by one; S2: obtain particles; S3: perform weight window inspection and operation on particles; S4: calculate the grid space importance of particles before and after crossing the grid boundary. n and I n+1 ; S5: Comparison I n Is it greater than I n+1 If not, execute S6 and return to S3. If yes, execute S7. S6: split particles and reduce particle weights. S7: determine whether to kill particles. If so, execute S9. If not, execute S8 and return to S3. S8: increase particle weights. S9: determine whether the particle simulation is complete. If not, return to S2. If the simulation is complete, end.

[0091] Reference Figure 7 As shown, the implicit capture process is as follows: S1: obtain particles; S2: determine whether the particles collide in the gate element, if so, execute S3, if not, return to S1; S3: multiply the weight by the probability of scattering; S4: determine whether the particle weight is less than the minimum weight, if so, execute S5, if not, return to S2; S5: determine whether the particle is killed, if so, execute S7; if not, execute S6 and return to S2; S6: divide the weight by the probability of death; S7: determine whether the particle simulation is completed, if it is not completed, return to S1; if it is determined that the simulation is completed, end.

[0092] Reference Figure 8 As shown, the weight window game process is as follows: S1: simulate particle movement; S2: determine whether the particle weight is within the weight window range, if so, return to S1, if not, execute S3; S3: determine whether the weight is greater than the weight window, if so, execute S4 and return to S1, if not, execute S5; S4: split the particle and reduce the weight; S5: determine whether the bet is dead, if so, end, if not, execute S6 and return to S1; S6: increase the particle weight.

[0093] To compare the speed-up effect of the variance reduction technique on dose calculations, the computation time and standard deviation of the results were compared between the use of implicit trapping and the absence of implicit trapping. For the same volume membrane model, with 10 million particles, the implicit trapping technique took 1584 seconds to compute, with a dose standard deviation of 11%. Without implicit trapping, the technique took 1258 seconds, with a dose standard deviation of 14.5%. To reduce the standard deviation to the same level as using implicit trapping, the number of simulated particles needed to be increased to 16 million, increasing the simulation time to 2045 seconds. This represents an additional 461 seconds of computation time when using the implicit trapping technique, meaning that the implicit trapping technique reduces the computation time by approximately 20%.

[0094] Example 5: Using non-uniform rectangular grids for simulation calculations.

[0095] Traditional body membrane calculations often use a uniform grid. Sometimes, in order to achieve detailed calculations in certain areas, the number of grids must be increased, resulting in longer calculation times and exponential growth in running memory. Models read from CT or PET often use a uniform grid, in which many connected areas are made of the same material, such as air and blood. Some of these areas do not require accurate calculation of detailed dose distribution, so a coarser grid can be used instead of a fine grid. For important locations such as tumors, a finer grid is required to achieve more precise calculations. Using a non-uniform grid to calculate dose improves calculation accuracy while not significantly increasing calculation time and running memory. This can improve calculation accuracy in important areas without significantly increasing calculation time, and reduce calculation time while meeting calculation accuracy in non-important areas.

[0096] The following simulation calculations are performed using grids of different sizes to compare the effects of different grid sizes on simulation time and calculation accuracy.

[0097] Table 3: Simulation time and calculation accuracy corresponding to different mesh sizes

[0098]

[0099]

[0100] In this experiment, the hybrid grid used is composed of 0.4 mm grids in the first 5 cm of the neutron incident direction, 0.8 mm grids in the middle 5 cm, and 1.6 mm grids in the remaining area. As shown in Table 3, the use of a non-uniform grid can reduce the total number of grid cells to 2 / 5 of the original 0.4 mm grid, the required memory for the grid is also reduced to 1 / 3 of the original, and the calculation time is reduced to 34.7% of the original. Based on the results of the calculation using the 0.4 mm grid, the neutron error of the results obtained using the hybrid grid is less than 0.1%, and the photon error is less than 0.2%.

[0101] The hybrid grid configuration is not limited to the above examples and can be customized based on the specific conditions of the irradiated object. Typically, the grid size is determined by the importance of the area. For example, the tumor area has a smaller grid size, typically 4 mm or smaller. The blood, air, and bone areas have a relatively larger grid size, typically 1.6 mm or larger. The areas containing normal muscle and other tissues have a grid size greater than 0.8 mm and less than 1.6 mm.

[0102] The Monte Carlo algorithm is used for dose calculation. Its advantages are high computational accuracy, but its disadvantages are slow convergence and long computation time. Therefore, optimizing computational efficiency is a crucial aspect of optimization. The optimization methods of Examples 1 to 5 of this application can improve computational efficiency and reduce computation time to varying degrees.

[0103] Although the above describes the illustrative specific embodiments of the present invention to facilitate understanding of the present invention by those skilled in the art, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious and are within the scope of protection required by the present invention.

Claims

1. A radiation irradiation system, characterized in that: include: a beam irradiation device for generating a therapeutic beam and irradiating the therapeutic beam onto an irradiated body to form an irradiated portion; a treatment planning module, configured to perform dose simulation calculations and generate a treatment plan based on parameters of the treatment beam and medical imaging data of the irradiated part, wherein the treatment planning module simulates particles using variance reduction; and a control module, configured to control the irradiation of the beam irradiation device according to the treatment plan; The variance reduction includes implicit capture, weight window game and bet splitting; the treatment plan module uses variance reduction to simulate particles, including: S1: obtaining a source particle; S2: judging whether the particle collides in the gate element, if so, executing S3 and S4 in sequence, if not, executing S5 and S6 in sequence; S3: implicit capture processing; S4: judging whether the weight is lower than the weight window, if so, executing S7, if not, returning to S2; S5: bet splitting processing; S6: judging whether to perform bet processing, if yes, executing S7, if not, returning to S2; S7: judging whether the bet is dead, if yes, executing S8, if not, executing S9 and returning to S2; S8: judging whether the particle has been processed, if yes, the process ends, if not, returning to S1; S9: the weight is divided by the probability of bet death.

2. The radiation irradiation system according to claim 1, wherein: The radiation irradiation system is a neutron capture therapy system. The beam irradiation device includes a neutron generator, a beam shaper, and a treatment table. The neutron generator includes an accelerator and a target. The accelerator accelerates charged particles to generate charged particle beams and interacts with the target to generate neutron beams. The beam shaper can adjust the neutron beams generated by the neutron generator to a preset beam quality. The neutron beams generated by the neutron generator are irradiated toward the irradiated object on the treatment table through the beam shaper.

3. A method for optimizing a treatment planning module of a radiation irradiation system, characterized by: The radiation irradiation system includes: a beam irradiation device for generating a therapeutic beam and irradiating the therapeutic beam to an irradiated body to form an irradiated part; a treatment planning module for performing dose simulation calculation and generating a treatment plan based on parameters of the therapeutic beam and medical imaging data of the irradiated part; The optimization method of the treatment planning module of the radiation irradiation system includes the treatment planning module simulating particles using variance reduction; the variance reduction includes implicit capture, weight window game and bet splitting; the simulation of particles using variance reduction includes: S1: obtaining a source particle; S2: judging whether the particle collides in the gate element, if so, executing S3 and S4 in sequence, if not, executing S5 and S6 in sequence; S3: implicit capture processing; S4: judging whether the weight is lower than the weight window, if so, executing S7, if not, returning to S2; S5: bet splitting processing; S6: judging whether to perform bet processing, if yes, executing S7, if not, returning to S2; S7: judging whether the bet is dead, if yes, executing S8, if not, executing S9 and returning to S2; S8: judging whether the particle has been processed, if yes, the process ends, if not, returning to S1; S9: dividing the weight by the probability of bet death.

4. The method for optimizing a treatment planning module of a radiation irradiation system according to claim 3, wherein: The betting splitting includes: S1: calculating the grid space importance of each grid and recording them one by one; S2: obtaining particles; S3: performing weight window checks and operations on particles; S4: calculating the grid space importance In and In+1 of particles before and after crossing the grid boundary; S5: comparing whether In is greater than In+1, if not, executing S6 and returning to S3, if so, executing S7; S6: splitting particles and reducing particle weights; S7: judging whether to bet on the particles to death, if it is judged that the particles are to be killed, executing S9, if it is judged that the particles are not to be bet on, executing S8 and returning to S3; S8: increasing the particle weight; S9: judging whether the particle simulation is completed, if it is judged that the simulation is not completed, returning to S2; if it is judged that the simulation is completed, ending.

5. The method for optimizing a treatment planning module of a radiation irradiation system according to claim 3, wherein: The implicit capture includes: S1: acquiring particles; S2: determining whether the particles collide in the cell, if so, executing S3, if not, returning to S1; S3: multiplying the weight by the probability of scattering; S4: determining whether the particle weight is less than the minimum weight, if so, executing S5, if not, returning to S2; S5: determining whether the particle is killed, if so, executing S7; if not, executing S6 and returning to S2; S6: dividing the weight by the probability of death; S7: determining whether the particle is simulated, if determined not to be simulated, returning to S1; if determined to be simulated, ending.

6. The method for optimizing a treatment planning module of a radiation irradiation system according to claim 3, wherein: The weight window game includes: S1: simulate particle movement; S2: determine whether the particle weight is within the weight window range, if so, return to S1, if not, execute S3; S3: determine whether the weight is greater than the weight window, if so, execute S4 and return to S1, if not, execute S5; S4: split the particle and reduce the weight; S5: determine whether the bet is dead, if so, end, if not, execute S6 and return to S1; S6: increase the particle weight.

7. The method for optimizing a treatment planning module of a radiation irradiation system according to claim 4, wherein: The grid space importance is obtained by solving the adjoint transport equation, which has the following form: where φ * is the accompanying flux, S * is the accompanying source, v is the speed of the particle, Ω is the direction of the particle, Σ t is the reaction cross section of the collision between particles and matter, Σ s is the scattering cross section, r is the position of the particle, E is the energy of the particle, and t is time.

8. The method for optimizing a treatment planning module of a radiation irradiation system according to claim 7, wherein: Particles move from a grid with an importance of In in grid space to a grid with an importance of In+1 in grid space. If In+1 > In, let m = In+1 / In, and the particle splits into m particles, with the weight of each particle reduced to 1 / m of the original; if In+1 < In, the particle plays a gambling trick. Let P = In+1 / In, sample a random number x between 0 and 1. If x is less than P, the particle survives and its weight is multiplied by 1 / P, otherwise the particle is "gambled to death" and the simulation of the particle is terminated.

9. The method for optimizing a treatment planning module of a radiation irradiation system according to any one of claims 3 to 8, characterized in that: The upper limit of the particle weight is 10 and the lower limit is 0.25; when the particle weight w is greater than 10, let the integer part of w be w1 and the decimal part be w2. Sample a random number x between 0 and 1. When x is less than w2, let w = w1 + 1, and when x is greater than w2, let w = w1; Then split a particle into w particles for simulation until the weight of each particle drops to 1.

Images