Artificial intelligence-based method for predicting dose distribution in boron neutron capture therapy and its application and device

Abstract

The present invention provides a method for predicting dose distribution of boron neutron capture therapy based on artificial intelligence, and its application and device. The method includes: obtaining CT images, characteristic prompt gamma ray yield distribution samples and dose distribution samples of patients; using characteristic prompt gamma ray yield distribution samples, dose distribution samples and CT images to train and test artificial intelligence network models; in the process of boron neutron capture therapy, using radiation detection devices to detect characteristic prompt gamma rays, and reconstructing characteristic prompt gamma ray yield distribution in the target area; and predicting dose distribution based on the trained artificial intelligence network model, the reconstructed characteristic prompt gamma ray yield distribution and CT images. The present invention has the characteristics of fast, accurate, simple and easy to implement, and can effectively solve the problem of inaccurate dose prediction during BNCT treatment, and effectively ensure the treatment effect of BNCT.

Description

Technical Field

[0001] The present invention relates to the technical field of radiotherapy, and in particular to an artificial intelligence-based boron neutron capture therapy dose distribution prediction method, the application of the prediction method in boron neutron capture therapy, a device and a non-transient computer storage medium. Background Art

[0002] Cancer has become the number one killer threatening human life and health. Radiotherapy, as a local, non-invasive treatment method, has been clinically proven to have excellent therapeutic effects and has become one of the three major means of cancer treatment. Although photon therapy is currently the most mature radiotherapy technology, it is limited by the energy deposition method of photons and will inevitably produce dose deposition on endangered organs in the target area, increasing the risk of radiation-induced secondary cancer. Therefore, new radiotherapy technologies need to be further studied.

[0003] Boron Neutron Capture Therapy (BNCT) is a new type of tumor targeted treatment method. Its superior therapeutic effect has attracted the attention of more and more researchers. The treatment principle of BNCT is to inject tumor-targeted boron drugs into the patient's body before treatment. According to the pharmacokinetic properties of the drug, the boron drug will accumulate in the tumor tissue. When the concentration of the boron drug in the tumor tissue reaches a therapeutic concentration, the patient's target area is irradiated externally with a thermal / epi-thermal neutron beam. Neutrons will interact with the concentrated neutrons in the tumor during their travel through human tissues. 10 The B nuclide undergoes a boron neutron capture reaction, emitting two heavily charged ions, α and 7 Li, thereby achieving the goal of killing tumor tissue while reducing radiation damage to normal tissues around the tumor.

[0004] The total dose generated during BNCT treatment mainly includes: ① boron dose (radiation dose generated by neutron and boron capture reaction, accounting for about 60%-80% of the total BNCT dose), ② neutron dose (radiation dose of secondary protons produced by neutrons and hydrogen / nitrogen), ③ gamma dose (irradiation dose produced by secondary gamma rays). Due to individual differences among patients and the dynamic distribution of drugs, the real-time dose during BNCT treatment also changes dynamically. At present, the evaluation of treatment dose in clinical practice still uses the distribution of boron drugs in the patient's body obtained before treatment, combined with the treatment planning system to calculate the deposited dose in the patient's target area during the entire treatment process, and the actual treatment effect is also determined by estimation. This evaluation method is bound to be affected by the changes in the distribution of boron drug concentration in the patient's target area during the treatment process. In addition, due to the large individual differences among patients, the calculation time required to calculate dose deposition is long and the efficiency is low, which restricts the clinical development of BNCT technology.

[0005] The contents of the background technology section are merely the technologies known to the inventors and do not necessarily represent the prior art in the field. Summary of the invention

[0006] In order to solve the problem that the dose distribution cannot be accurately predicted in the existing boron neutron capture therapy process, the present application proposes a boron neutron capture therapy dose distribution prediction method based on artificial intelligence.

[0007] Specifically, the method includes:

[0008] Acquire a patient's CT image, characteristic prompt gamma ray yield distribution samples, and dose distribution samples;

[0009] Using the characteristic prompt gamma ray yield distribution samples, the dose distribution samples and the CT images to train and test the artificial intelligence network model;

[0010] During the BNCT process, the characteristic prompt gamma rays are detected by the radiation detection device, and the characteristic prompt gamma ray yield distribution in the target area is reconstructed; and

[0011] The dose distribution is predicted based on the trained artificial intelligence network model, the reconstructed characteristic prompt gamma-ray yield distribution and the CT image.

[0012] According to some embodiments of the present application, the characteristic prompt gamma ray yield distribution includes 478keV prompt gamma ray yield distribution and 2.224MeV prompt gamma ray yield distribution, and the dose distribution includes total biological dose distribution, boron dose distribution, neutron dose distribution and gamma dose distribution.

[0013] According to some embodiments of the present application, acquiring a characteristic prompt gamma ray yield distribution sample, a dose distribution sample, and a CT image of a patient includes:

[0014] acquiring the CT image; and

[0015] The Monte Carlo method is used to simulate the particle transport process during boron neutron capture therapy to obtain characteristic prompt gamma-ray yield distribution samples and dose distribution samples.

[0016] According to some embodiments of the present application, using the Monte Carlo method to simulate the particle transport process during the boron neutron capture therapy, obtaining characteristic prompt gamma ray yield distribution samples and dose distribution samples includes:

[0017] The Monte Carlo method is used to model the boron neutron capture therapy process based on the CT image, and the energy deposition and particle transport processes under different treatment schemes are simulated to obtain characteristic prompt gamma-ray yield distribution maps and three-dimensional dose distribution maps in the target area under different tumor shapes, different tumor types, different tumor locations and different boron concentration distributions.

[0018] According to some embodiments of the present application, using the characteristic prompt gamma ray yield distribution sample, the dose distribution sample and the CT image to train and test the artificial intelligence network model includes:

[0019] The artificial intelligence network model is trained and tested using the boron dose distribution and the 478keV prompt gamma ray yield distribution;

[0020] Using the neutron dose distribution, the 2.224 MeV prompt gamma ray yield distribution and the CT image, training and testing the artificial intelligence network model; and

[0021] The CT image and the photon dose distribution are used to train and test the artificial intelligence network model.

[0022] According to some embodiments of the present application, during the boron neutron capture therapy, detecting characteristic prompt gamma rays using a radiation detection device and reconstructing the characteristic prompt gamma ray yield distribution in the target area includes:

[0023] During the BNCT process, the radiation detection device is used to detect the 478keV prompt gamma ray and the 2.224MeV prompt gamma ray to obtain the 478keV prompt gamma ray information and the 2.224MeV prompt gamma ray information; and

[0024] The 478keV prompt gamma ray information and the 2.224MeV prompt gamma ray information are reconstructed through an image reconstruction algorithm and the CT image to obtain the 478keV prompt gamma ray yield distribution and the 2.224MeV prompt gamma ray yield distribution in the target area.

[0025] According to some embodiments of the present application, based on the trained artificial intelligence network model, the reconstructed characteristic prompt gamma ray yield distribution and the CT image, predicting the dose distribution includes:

[0026] Inputting the reconstructed 478keV prompt gamma-ray yield distribution into the trained artificial intelligence network model to obtain the boron dose distribution;

[0027] Inputting the reconstructed 2.224 MeV prompt gamma ray yield distribution and the CT image into the trained artificial intelligence network model respectively to obtain a first neutron dose distribution and a second neutron dose distribution;

[0028] Inputting the CT image into the trained artificial intelligence network model to obtain gamma dose distribution;

[0029] The total biological dose distribution is calculated from the boron dose distribution, the first neutron dose distribution, the second neutron dose distribution, and the gamma dose distribution.

[0030] The present application further provides an application of the above prediction method in boron neutron capture therapy, including:

[0031] Obtaining the dose deposition results calculated by the treatment planning system before treatment; and

[0032] The dose distribution and the dose deposition result are compared to analyze the difference between the treatment effect and the expected effect.

[0033] Furthermore, the present application also provides a device, comprising:

[0034] Processor; and

[0035] The memory stores computer instructions, and when the computer instructions are executed by the processor, the processor is enabled to execute the above prediction method.

[0036] In addition, the present application also provides a non-transitory computer storage medium storing a computer program, which, when executed by one or more processors, enables the processors to execute the above prediction method.

[0037] The prediction method of the present application uses the patient's CT images, characteristic prompt gamma ray yield distribution samples and dose distribution samples to train a deep learning network model. For the trained network model structure, it can achieve accurate evaluation of the radiation dose received by patients of different ages, genders and cases during treatment. The present invention has the characteristics of being fast, accurate, simple and easy to implement, and can effectively solve the problem of inaccurate dose prediction during BNCT treatment, effectively ensuring the treatment effect of BNCT.

[0038] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 It is a schematic diagram of a method for predicting dose distribution of boron neutron capture therapy based on artificial intelligence in one embodiment of the present invention.

[0040] Figure 2 FIG. 4 is a schematic diagram of simulating the particle transport process during boron neutron capture therapy using the Monte Carlo method in one embodiment of the present invention.

[0041] Figure 3 It is a framework diagram of an artificial intelligence network model in one embodiment of the present invention.

[0042] Figure 4 This is a diagram showing the evaluation results of a boron dose distribution prediction image at different tissue depths in one embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following is a more detailed description of the specific embodiments of the present invention in conjunction with the accompanying drawings and examples, so that the scheme of the present invention and its advantages in various aspects can be better understood. However, the specific embodiments and examples described below are only for the purpose of illustration, rather than for limiting the present invention.

[0044] It is particularly important to point out that all similar substitutions and modifications are obvious to those skilled in the art and are considered to be included in the present invention. The methods and applications of the present invention have been described through preferred embodiments, and relevant personnel can obviously modify or appropriately change and combine the methods and applications described herein without departing from the content, spirit and scope of the present invention to implement and apply the technology of the present invention.

[0045] Figure 1 The present invention provides a method for predicting dose distribution of boron neutron capture therapy based on artificial intelligence, comprising steps S100 to S400.

[0046] S100: Acquire a CT image of a patient, a characteristic prompt gamma ray yield distribution sample, and a dose distribution sample.

[0047] Optionally, the characteristic prompt gamma ray yield distribution includes 478keV prompt gamma ray yield distribution and 2.224MeV prompt gamma ray yield distribution, and the dose distribution includes total biological dose distribution, boron dose distribution, neutron dose distribution and gamma dose distribution.

[0048] Optionally, a Monte Carlo method is used to simulate the particle transport process during the boron neutron capture therapy to obtain characteristic prompt gamma ray yield distribution samples and dose distribution samples.

[0049] Specifically, the Monte Carlo method was used to model the boron neutron capture therapy process based on CT images, and the energy deposition and particle transport processes under different treatment schemes were simulated to obtain the characteristic prompt gamma-ray yield distribution map and three-dimensional dose distribution map in the target area under different tumor shapes, different tumor types, different tumor locations and different boron concentration distributions.

[0050] The Monte Carlo method, also known as the random sampling method or the statistical experimental method, is a branch of computational mathematics that can realistically simulate the physical process of actual particle transport and solve practical problems. Optionally, a Monte Carlo program is used to construct a model of the patient based on the CT image and the material composition of different tissues, and to determine their response to different particles. The Monte Carlo program is a general software package based on the Monte Carlo method for calculating neutron, photon, electron or multi-particle transport problems in three-dimensional complex geometric structures. The Monte Carlo program can use one or more of MCNP, GEANT4, TOPAS, GATE, FLUKA, SuperMC or Phits. The Monte Carlo program is used to simulate the energy deposition and particle transport process under different treatment plans, and obtain characteristic prompt gamma ray yield distribution maps and three-dimensional dose distribution maps in the target area under different tumor shapes, different tumor types, different tumor locations and different boron concentration distributions.

[0051] Figure 2 The illustrated embodiment uses GEANT4 to simulate the characteristic prompt gamma ray yield distribution and different component dose distribution information under BNCT treatment of different head gliomas to obtain a sufficient number of training samples.

[0052] like Figure 2 As shown in the figure, a head tumor is set in the phantom, and a corresponding voxelized human model is designed in the GEANT4 program, and the boron concentration in the tumor is set to 80ppm, the boron concentration in the skin is set to 25ppm, and the boron concentration in other tissues is set to 18ppm. An epithermal neutron beam is designed to treat the case, and the transport process of each particle is tracked, and the yield of characteristic prompt gamma rays of different energies in each voxel and the energy deposition information of different particles are recorded.

[0053] According to the calculation method of energy deposition and dose, the information of different dose components is obtained according to formulas (1) to (3), and the total biological dose distribution is obtained according to formula (4).

[0054]

[0055] Dose neutron =Energy proton / m (2),

[0056] Dose gamma =Energy gamma / m (3),

[0057] Dose REB =RBE boron ·Dose boron +RBE neutron ·Dose neutron +RBEgamma ·Dose gamma (4).

[0058] Among them, Dose boron Boron dose, Dose neutron is the neutron dose, Dose gamma is the gamma dose, Energy α is the energy deposition of α particles in the voxel, for 7 The energy deposition of Li particles in a voxel, m is the mass of the voxel, Energy proton Energy is the energy deposition of protons in the voxel. gamma Dose is the energy deposition of photons in the voxel. REB is the total biological dose in the voxel, RBE boron is the relative biological effect value of boron dose, which is 3.8 for tumor voxels, 1.4 for normal tissue voxels, and 2.5 for skin voxels. neutron is the relative biological effect value of neutron dose. For tumor voxels, normal tissues and skin tissues, the value is 3.2. gamma is the relative biological effect value of the gamma dose. For tumor voxels, normal tissues and skin tissues, the value is 1.

[0059] At the same time, the yield distribution of 478keV prompt gamma rays and the yield distribution of 2.224MeV prompt gamma rays were recorded.

[0060] The present invention adopts the Monte Carlo simulation method to provide samples for the training of artificial intelligence models, which expands the application scope of the artificial intelligence method and improves the applicability of the artificial intelligence method.

[0061] S200: Use characteristic prompt gamma-ray yield distribution samples, dose distribution samples and CT images to train and test the artificial intelligence network model.

[0062] Optionally include:

[0063] The artificial intelligence network model is trained and tested using the boron dose distribution obtained in step S100 and the 478keV prompt gamma-ray yield distribution, so that the boron dose distribution generated by the artificial intelligence network model is close to the boron dose distribution obtained in step S100.

[0064] The neutron dose distribution obtained in step S100, the 2.224MeV prompt gamma-ray yield distribution and the CT images were used to train and test the artificial intelligence network model, so that the neutron dose distribution generated by the artificial intelligence network model was close to the neutron dose distribution obtained in step S100.

[0065] And the CT image and photon dose distribution obtained by step S100 are used to train and test the artificial intelligence network model, so that the photon dose distribution generated by the artificial intelligence network model is close to the photon dose distribution obtained by step S100. The framework of the artificial intelligence algorithm that can be used in the present invention is one or more of DeepLearnToolbox, Caffe, CNTK, TensorFlow, Theano, Torch, Keras, Lasagne, DSSTNE, MXNet, DeepLearning4J, ConvNetJs, Chainer, Scikit-Learn, and Matlab, and the artificial intelligence network model uses one or more of a deep belief network, a deep neural network, a convolutional neural network, a stacked autoencoder, or a convolutional autoencoder.

[0066] Figure 3 The embodiment shown uses the Pix2pix artificial intelligence network. After obtaining sample pairs under different tumor shapes and boron concentration distributions through simulation calculation, the samples are brought into the artificial intelligence network framework Pix2pix for training.

[0067] Pix2pix is ​​an artificial intelligence network based on the GAN network, which can mainly realize image-to-image prediction. For example, by feeding the generator ( Figure 3 The generator generates a predicted dose distribution image based on the image of the 478keV characteristic prompt gamma ray yield distribution, and the image generated by the generator and the real boron dose distribution image are input into the discriminator ( Figure 3 D), the discriminator will evaluate the similarity of the two images. When the two images are very different, the discriminator will determine that the image generated by the generator is a "fake" image and score the quality of the generated image. The generator will adjust the weight parameters of its own algorithm based on the evaluation results of the discriminator to generate more similar images. As the number of training times increases, the dose distribution map generated by the generator becomes more and more similar to the real dose distribution map, and the discriminator gradually cannot distinguish the authenticity of the image. At this time, the boron dose distribution generated by the generator is basically consistent with the real boron dose distribution.

[0068] For neutron dose distribution, by inputting 2.224MeV prompt gamma ray yield distribution and CT images into the generator, the generator generates predicted dose distribution images based on the 2.224MeV prompt gamma ray yield distribution and CT images, and inputs the images generated by the generator and the real neutron dose distribution images into the discriminator, which will evaluate their similarity respectively. Similarly, when the image difference is large, the discriminator will determine that the image generated by the generator is a "fake" image and score the quality of the generated image. The generator will adjust the weight parameters of its own algorithm based on the evaluation results of the discriminator to generate more similar images. As the number of training times increases, the dose distribution map generated by the generator becomes more and more similar to the real dose distribution map, and the discriminator gradually cannot distinguish the authenticity of the image. At this time, the boron dose distribution generated by the generator is basically consistent with the real boron dose distribution.

[0069] Similarly, for photon dose distribution, by inputting CT images into the generator, the generator generates a predicted dose distribution image based on the CT images, and inputs the image generated by the generator and the real photon dose distribution image into the discriminator, and the discriminator will evaluate their similarity respectively. Similarly, when the image difference is large, the discriminator will determine that the image generated by the generator is a "fake" image and score the quality of the generated image. The generator will adjust the weight parameters of its own algorithm based on the evaluation results of the discriminator to generate more similar images. As the number of training times increases, the dose distribution map generated by the generator becomes more and more similar to the real dose distribution map, and the discriminator gradually cannot distinguish the authenticity of the image. At this time, the boron dose distribution generated by the generator is basically consistent with the real boron dose distribution.

[0070] For the images generated by the generator, it is also necessary to combine relevant evaluation parameters to make a more accurate and objective evaluation of the treatment of the generated images. Figure 4 The following is a description of the generated image evaluation method:

[0071] Taking the simulated boron dose distribution results at different tissue depths as the standard, the similarity between the boron dose distribution map predicted by the artificial intelligence network and the standard image is analyzed, focusing on the two parameters of structural similarity (SSIM) and peak signal-to-noise ratio (PSNR). SSIM is an indicator to measure the similarity between two images, and its value ranges from 0 to 1. When the two images are exactly the same, the value of SSIM is equal to 1. The calculation of SSIM is shown in formula (5), where μ x and μ y Represent the average values ​​of input images x and y, σ x and σ y Represent the standard deviation of the two images, σ xy Represents the covariance of the two images, c1 and c2 are certain constants.

[0072]

[0073] In addition to evaluating the similarity between the generated dose distribution image and the real dose distribution image obtained by simulation, it is also necessary to accurately evaluate the quality of the generated image. The peak signal-to-noise ratio (PSNR) is an objective standard for evaluating images. The larger the value, the smaller the image distortion. The definition is shown in formula (6), where MSE represents the mean square error between the current image and the reference image, and the definition is shown in formula (7):

[0074]

[0075]

[0076] The images predicted by the Pix2pix network are analyzed based on these two image evaluation parameters. The results are as follows: Figure 4 It can be seen that at different depth levels, the SSIM of the dose distribution images predicted by the network and the dose distribution images obtained by the simulation method are basically greater than 0.96, and the SSIM value of the shallow image is close to 0.99, indicating that the dose distribution image predicted by the network is basically consistent with the real dose distribution image. In addition, it can be seen from the PSNR parameter that the quality of the dose distribution image predicted by the network is good. Therefore, it is feasible to use this method in predicting dose distribution from characteristic prompt gamma-ray yield distribution, with high accuracy, less calculation time, and can achieve parameter conditions that are not in the training samples, broadening the application of this method.

[0077] S300: During the BNCT process, the characteristic prompt gamma rays are detected using a radiation detection device, and the characteristic prompt gamma ray yield distribution in the target area is reconstructed.

[0078] Optionally include:

[0079] During the BNCT process, the radiation detection device is used to detect the 478keV prompt gamma ray and the 2.224MeV prompt gamma ray to obtain the 478keV prompt gamma ray information and the 2.224MeV prompt gamma ray information; and

[0080] The information of 478keV prompt gamma rays and 2.224MeV prompt gamma rays is reconstructed through image reconstruction algorithm and CT images to obtain the yield distribution of 478keV prompt gamma rays and 2.224MeV prompt gamma rays in the target area.

[0081] The above-mentioned image reconstruction algorithms include: one or more of SBP, ART, SART, MLEM, List Mode-MLEM or OSEM.

[0082] The prompt gamma ray distribution information of different energies detected by the radiation detection device and the CT image of the patient are input into the reconstruction algorithm, so as to reconstruct the prompt gamma ray yield distribution of different energies in the patient's target area.

[0083] S400: Predict dose distribution based on the trained artificial intelligence network model, reconstructed characteristic prompt gamma-ray yield distribution and CT images.

[0084] Optionally include:

[0085] The reconstructed 478keV prompt gamma-ray yield distribution is input into the trained artificial intelligence network model to obtain the boron dose distribution.

[0086] The reconstructed 2.224MeV prompt gamma-ray yield distribution and CT image were respectively input into the trained artificial intelligence network model to obtain the first neutron dose distribution and the second neutron dose distribution.

[0087] The CT images are input into the trained artificial intelligence network model to obtain the gamma dose distribution.

[0088] The total biological dose distribution is calculated from the boron dose distribution, the first neutron dose distribution, the second neutron dose distribution, and the gamma dose distribution.

[0089] Optionally, the first neutron dose distribution and the second neutron dose distribution may be averaged to obtain a final predicted neutron dose distribution. The total biological dose distribution may be calculated by formula (4).

[0090] Furthermore, the present application further provides an application of the above prediction method in boron neutron capture therapy, including:

[0091] Obtaining the dose deposition results calculated by the treatment planning system before treatment; and

[0092] The dose distribution and the dose deposition result are compared to analyze the difference between the treatment effect and the expected effect.

[0093] Specifically, the real-time irradiation dose in the target area predicted by the artificial intelligence network model is combined with the dose deposition results calculated by the treatment planning system before treatment to accurately evaluate the treatment effect and accurately guide the implementation of subsequent treatment plans. For example, when the actual dose received by the tumor area is lower than the plan, a further treatment plan is set based on the current treatment results. The treatment planning system can be one or more of NCTplan, SERA, JCDS, THORplan or NeubronPlan.

[0094] The present invention uses a trained network to predict the real-time dose distribution results during the treatment process, which speeds up the prediction speed, is insensitive to reconstruction artifacts, and improves the accuracy of dose prediction. The present invention is based on an artificial intelligence algorithm, and can achieve accurate prediction of boron dose distribution, neutron dose distribution, photon dose distribution and total dose distribution for different lesions in different locations under different treatment plans through training with a large number of CT images and characteristic prompt gamma-ray yield distribution samples. The present invention, based on the trained network structure, can achieve accurate evaluation of the radiation dose received by patients of different ages, genders and cases during the treatment process, thereby improving the practicality and convenience of the present invention. The present invention compares the prediction results of artificial intelligence with the results calculated by the treatment planning system, thereby improving the accuracy of the treatment effect evaluation.

[0095] The present application further provides a device, comprising: a processor; and a memory, wherein the memory stores computer instructions, and when the computer instructions are executed by the processor, the processor executes the above-mentioned artificial intelligence-based boron neutron capture therapy dose distribution prediction method.

[0096] In addition, the present application also provides a non-transitory computer storage medium storing a computer program, which, when executed by one or more processors, enables the processors to execute the above-mentioned artificial intelligence-based boron neutron capture therapy dose distribution prediction method.

[0097] It should be noted that the division of units in the embodiments of the present application is schematic and is merely a logical function division. There may be other division methods in actual implementation.

[0098] In addition, those skilled in the art should be easily aware that, in combination with the units and algorithm steps of each example described in the embodiments provided herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0099] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0100] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0101] In the several embodiments provided in the present application, it should be understood that the disclosed device can be implemented in other ways. For example, the device embodiments described above are only schematic, such as the division of the units, which is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the device or unit can be electrical or other forms.

[0102] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0103] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of a software program module.

[0104] If the integrated unit is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a memory and includes several plug-ins to enable a computer device (which can be a personal computer, server or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned memory includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0105] Those skilled in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by plugging in relevant hardware through a program, and the program may be stored in a computer-readable memory, which may include a flash drive, a read-only memory (ROM), a random access memory (RAM), or an optical disk.

[0106] The embodiments of the present application are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core idea of ​​the present application. At the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for predicting dose distribution of boron neutron capture therapy based on artificial intelligence, characterized in that: include: Acquire a patient's CT image, characteristic prompt gamma ray yield distribution samples, and dose distribution samples; Using the characteristic prompt gamma ray yield distribution samples, the dose distribution samples and the CT images to train and test the artificial intelligence network model; During the BNCT process, the characteristic prompt gamma rays are detected by the radiation detection device, and the characteristic prompt gamma ray yield distribution in the target area is reconstructed; and Predicting dose distribution based on the trained artificial intelligence network model, the reconstructed characteristic prompt gamma-ray yield distribution and the CT image; The dose distribution includes total biological dose distribution, boron dose distribution, neutron dose distribution and gamma dose distribution.

2. The prediction method according to claim 1, characterized in that: The characteristic prompt gamma-ray yield distribution includes 478keV prompt gamma-ray yield distribution and 2.224MeV prompt gamma-ray yield distribution.

3. The prediction method according to claim 1, characterized in that: Acquiring CT images of patients, characteristic prompt gamma ray yield distribution samples and dose distribution samples include: acquiring the CT image; and The Monte Carlo method is used to simulate the particle transport process during boron neutron capture therapy to obtain characteristic prompt gamma-ray yield distribution samples and dose distribution samples.

4. The prediction method according to claim 3, characterized in that: The Monte Carlo method is used to simulate the particle transport process during BNCT to obtain characteristic prompt gamma ray yield distribution samples and dose distribution samples, including: The Monte Carlo method is used to model the boron neutron capture therapy process based on the CT image, and the energy deposition and particle transport processes under different treatment schemes are simulated to obtain characteristic prompt gamma-ray yield distribution maps and three-dimensional dose distribution maps in the target area under different tumor shapes, different tumor types, different tumor locations and different boron concentration distributions.

5. The prediction method according to claim 2, characterized in that: Using the characteristic prompt gamma ray yield distribution sample, the dose distribution sample and the CT image to train and test the artificial intelligence network model includes: The artificial intelligence network model is trained and tested using the boron dose distribution and the 478keV prompt gamma ray yield distribution; Using the neutron dose distribution, the 2.224 MeV prompt gamma ray yield distribution and the CT image, training and testing the artificial intelligence network model; and The artificial intelligence network model is trained and tested using the CT image and the gamma dose distribution.

6. The prediction method according to claim 2, characterized in that: During the BNCT process, the characteristic prompt gamma rays are detected by the radiation detection device, and the characteristic prompt gamma ray yield distribution in the target area is reconstructed, including: During the BNCT process, the radiation detection device is used to detect the 478keV prompt gamma ray and the 2.224MeV prompt gamma ray to obtain the 478keV prompt gamma ray information and the 2.224MeV prompt gamma ray information; and The 478keV prompt gamma ray information and the 2.224MeV prompt gamma ray information are reconstructed through an image reconstruction algorithm and the CT image to obtain the 478keV prompt gamma ray yield distribution and the 2.224MeV prompt gamma ray yield distribution in the target area.

7. The prediction method according to claim 2, characterized in that: Based on the trained artificial intelligence network model, the reconstructed characteristic prompt gamma ray yield distribution and the CT image, predicting the dose distribution includes: Inputting the reconstructed 478keV prompt gamma-ray yield distribution into the trained artificial intelligence network model to obtain the boron dose distribution; Inputting the reconstructed 2.224 MeV prompt gamma ray yield distribution and the CT image into the trained artificial intelligence network model respectively to obtain a first neutron dose distribution and a second neutron dose distribution; Inputting the CT image into the trained artificial intelligence network model to obtain gamma dose distribution; The total biological dose distribution is calculated from the boron dose distribution, the first neutron dose distribution, the second neutron dose distribution, and the gamma dose distribution.

8. Application of the prediction method according to any one of claims 1 to 7 in boron neutron capture therapy, characterized in that: include: Obtain the dose deposition results calculated by the treatment planning system before treatment; as well as The dose distribution and the dose deposition result are compared to analyze the difference between the treatment effect and the expected effect.

9. A device for predicting dose distribution of boron neutron capture therapy based on artificial intelligence, comprising: processor; as well as A memory storing computer instructions, which, when executed by the processor, enables the processor to execute the prediction method described in any one of claims 1 to 7.

10. A non-transitory computer storage medium storing a computer program, which, when executed by one or more processors, enables the processors to perform the prediction method according to any one of claims 1 to 7.

Images