Model acquisition method, device and dose distribution generation method and device

Abstract

The present disclosure provides a model acquisition method including: obtaining training samples, wherein the training samples include image slices and a first dose distribution and a second dose distribution corresponding to the image slices, wherein the accuracy of the second dose distribution is higher than the accuracy of the first dose distribution; training a prediction model based on the image slices, the second dose distribution and the first dose distribution to obtain a target prediction model, wherein the target prediction model is used to obtain a second target dose distribution based on a target image slice and a first target dose distribution of the target image slice, wherein the accuracy of the second target dose distribution is higher than the accuracy of the first target dose distribution; wherein the training samples are obtained based on a single carbon ion pencil beam. The present disclosure also provides a dose distribution generation method for predicting a target Monte Carlo dose distribution at a treatment plan level through a target prediction model.

Description

Technical Field

[0001] The present disclosure relates to the fields of computer technology and medical technology, and more specifically, to a model acquisition method and device, and a dose distribution generation method and device. Background Art

[0002] The accuracy of dose distribution in heavy ion and proton therapy directly affects the treatment effect and side effects. The structure of human tissue is complex and irregular. The Monte Carlo method can handle complex geometric shapes and tissue interfaces. It simulates the transmission and interaction of particles in these irregular structures through random sampling, which involves complex physical processes, including ionization, inelastic scattering, elastic scattering and nuclear reactions. The interaction between these processes when passing through human tissue is very complex. The Monte Carlo method can simulate these microscopic processes in detail, thereby achieving high-precision dose calculation.

[0003] However, most traditional Monte Carlo methods simulate the transport process of particles in the human body, which requires a lot of computing resources, from several hours for CPU calculations to several minutes for GPU calculations. In addition, the Monte Carlo method relies on the accuracy of complex physical model parameters and interaction cross sections. The accuracy of these parameters directly affects the accuracy and reliability of the calculation results, and consumes a lot of computing resources. Summary of the invention

[0004] In view of this, the present disclosure provides a model acquisition method, apparatus, device, medium, program product, and dose distribution generation method, apparatus, device, medium, program product.

[0005] One aspect of the present disclosure provides a model acquisition method, comprising: acquiring training samples, wherein the training samples include image slices and first and second dose distributions corresponding to the image slices, wherein the accuracy of the second dose distribution is higher than the accuracy of the first dose distribution; training a prediction model based on the image slices, the first dose distribution, and the second dose distribution to obtain a target prediction model, wherein the target prediction model is used to obtain a target second dose distribution based on the target image slices; wherein the accuracy of the second target dose distribution is higher than the accuracy of the first target dose distribution; wherein the training samples are acquired based on a single carbon ion pencil beam.

[0006] According to an embodiment of the present disclosure, a prediction model is trained according to image slices, a second dose distribution and a first dose distribution, including: training the prediction model according to a loss function until the difference between the second dose distribution prediction value generated by the prediction model according to the image slices and the first dose distribution and the second dose distribution is less than a preset threshold, thereby obtaining a target prediction model; wherein the loss function is determined according to the number of image slices, the number of voxels in the image slices and the weight factor.

[0007] According to an embodiment of the present disclosure, the weight factor is determined based on a first quantity, where the first quantity is the proportion of voxels in the second dose distribution whose voxel dose values ​​are greater than a preset dose threshold.

[0008] According to an embodiment of the present disclosure, the preset dose threshold is determined according to the maximum voxel dose value in the second dose distribution.

[0009] According to an embodiment of the present disclosure, when the dose value of the target voxel in the second dose distribution is less than a preset dose threshold, when calculating the value of the corresponding loss function of the target voxel, the weight factor is the ratio of the first number to the number of voxels in the image slice.

[0010] According to an embodiment of the present disclosure, when the dose value of the voxel in the second dose distribution is greater than or equal to a preset dose threshold, when calculating the value of the corresponding loss function of the target voxel, the weight factor is 1 minus the ratio of the first number and the number of voxels in the image slice.

[0011] According to an embodiment of the present disclosure, the target prediction model is a convolutional neural network model for Monte Carlo simulation.

[0012] Another aspect of the present disclosure provides a method for generating a dose distribution, including: determining the beam position of each of a plurality of pencil beams in a treatment plan; acquiring image slices of each pencil beam position according to the pencil beam position; predicting a second dose distribution corresponding to a single beam according to the image slices and a target prediction model, wherein the target prediction model is a target prediction model obtained by any embodiment of the aforementioned model obtaining method; and superimposing the second dose distributions corresponding to each pencil beam to obtain a target Monte Carlo dose distribution.

[0013] Another aspect of the present disclosure provides a model acquisition device, including: an acquisition module, used to acquire training samples, wherein the training samples include image slices and a second dose distribution and a first dose distribution corresponding to the image slices, wherein the accuracy of the second dose distribution is higher than the accuracy of the first dose distribution; a training module, used to train a prediction model based on the image slices, the first dose distribution and the second dose distribution to obtain a target prediction model, wherein the target prediction model is used to obtain a second target dose distribution based on the target image slice and the first target dose distribution of the target image slice, wherein the accuracy of the second target dose distribution is higher than the accuracy of the first target dose distribution; wherein the training samples are acquired based on a single carbon ion pencil beam.

[0014] Another aspect of the present disclosure provides a dose distribution generating device, including: a determination module, used to determine the beam position of each of a plurality of pencil beams in a treatment plan; an acquisition module, used to acquire image slices of each pencil beam position according to the pencil beam position; a prediction module, used to predict a second dose distribution corresponding to a single beam according to the image slices and a target prediction model, wherein the target prediction model is a target prediction model obtained by any embodiment of the aforementioned model acquisition method; and a superposition module, used to superimpose the second dose distributions corresponding to each pencil beam to obtain a target Monte Carlo dose distribution.

[0015] Another aspect of the present disclosure provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the model acquisition method as described above.

[0016] Another aspect of the present disclosure provides a computer-readable storage medium storing computer-executable instructions, which are used to implement the model obtaining method as described above when executed.

[0017] Another aspect of the present disclosure provides a computer program product, which includes computer executable instructions, and when the instructions are executed, are used to implement the model obtaining method as described above.

[0018] Another aspect of the present disclosure provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the dose distribution generation method as described above.

[0019] Another aspect of the present disclosure provides a computer-readable storage medium storing computer-executable instructions, which are used to implement the dose distribution generating method as described above when executed.

[0020] Another aspect of the present disclosure provides a computer program product, which includes computer executable instructions, and when the instructions are executed, are used to implement the dose distribution generation method as described above.

[0021] According to the embodiments of the present disclosure, the Monte Carlo data at the single beam level is trained, and then the treatment plan level reasoning is performed based on the generated model. The computing resources consumed by the training process are greatly reduced, and the model obtained by training the model training method provided by the present disclosure is used to predict the second dose distribution. Compared with the traditional Monte Carlo calculation, there is no need to set a large number of empirical parameters, simulate particle tracking and complex physical reactions, and the trained network model can achieve an acceleration ratio of more than 100 times while meeting clinical needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The above and other objects, features and advantages of the present disclosure will become more apparent through the following description of the embodiments of the present disclosure with reference to the accompanying drawings, in which:

[0023] Figure 1 A prediction model training block diagram according to an embodiment of the present disclosure is schematically shown;

[0024] Figure 2 The following schematically shows an architecture diagram of a prediction model according to an embodiment of the present disclosure;

[0025] Figure 3 A flowchart of a model acquisition method according to an embodiment of the present disclosure is schematically shown;

[0026] Figure 4 A flow chart of a method for generating a dose distribution according to an embodiment of the present disclosure is schematically shown;

[0027] Figure 5 A block diagram of a model obtaining device according to an embodiment of the present disclosure is schematically shown;

[0028] Figure 6 A block diagram schematically shows a dose distribution generating device according to an embodiment of the present disclosure; and

[0029] Figure 7 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0030] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0031] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0032] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification and should not be interpreted in an idealized or overly rigid manner.

[0033] When using expressions such as "at least one of A, B, and C, etc.", they should generally be interpreted according to the meaning of the expression commonly understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).

[0034] In the embodiments of the present disclosure, the collection, updating, analysis, processing, use, transmission, provision, disclosure, storage, etc. of the data involved (for example, including but not limited to user personal information) are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security, and national security.

[0035] In the embodiments of the present disclosure, the user's authorization or consent is obtained before obtaining or collecting the user's personal information.

[0036] An embodiment of the present disclosure provides a model acquisition method, comprising: acquiring training samples, wherein the training samples include image slices and a second dose distribution and a first dose distribution corresponding to the image slices; training a prediction model based on the image slices, the second dose distribution and the first dose distribution to obtain a target prediction model, wherein the target prediction model is used to obtain a target second dose distribution based on the target image slices; wherein the second dose distribution and the first dose distribution are obtained based on the image slices; wherein the image slices are generated based on a carbon ion pencil beam.

[0037] Figure 1 A prediction model training block diagram according to an embodiment of the present disclosure is schematically shown.

[0038] like Figure 1 As shown, the prediction model training block diagram includes a training sample data preparation module 110 , a training data generation module 120 and a training module 130 .

[0039] The training sample data preparation module 110 is used to obtain initial training data, for example, CT image data 111 that complies with the DICOM (Digital Imaging and Communications in Medicine) standard is obtained through existing medical imaging equipment and information systems, and the image data 111 is analyzed and processed to obtain multiple image slices 112. Image slices are cross-sectional images that capture internal structures, such as CT image slices. Image slices can be randomly selected from a specific number of patient CT images and include a CT image slice corresponding to the location of various human tissues. The resolution of the slice can be 1 mm, 1 mm and 3 mm, and the size is 32 × 32 × 96. It should be noted that the specific parameters of the above slices are only examples, and those skilled in the art can make corresponding adjustments according to actual conditions, and the present disclosure does not limit this.

[0040] It should be noted that the image slices and image slices mentioned in the present disclosure all represent a cross section of a certain thickness in the body, which has information about the internal structure of the region. Continuous slices can be used to reconstruct a complete three-dimensional image. That is, the image slices and image slices mentioned in the present disclosure are not two-dimensional pictures, but three-dimensional image data with a certain thickness.

[0041] The training data generation module 120 is used to generate a second dose distribution 123 and a first dose distribution 122 corresponding to each image slice according to multiple image slices 112 and pencil beam 121. It should be noted that the second dose distribution 123 and the first dose distribution 122 here are the slices with specific thickness mentioned above.

[0042] In the embodiment of the present disclosure, the pencil beam is a carbon ion pencil beam. Specific parameters of the pencil beam, such as energy, energy dispersion, and angular dispersion, can be determined by those skilled in the art according to actual conditions.

[0043] The training module 130 is used to train the prediction model 131 according to the image slice 112, the second dose distribution 123 and the first dose distribution 122, wherein the input of the model is the image slice and the first dose distribution corresponding to the image slice, and the output of the model is the predicted value 132 of the second dose distribution. The loss calculation 133 is performed on the second dose distribution 123 in the training sample by the predicted value, and the parameters of the model are adjusted to obtain the target prediction model.

[0044] Figure 2 The schematic diagram shows an architecture diagram of a prediction model according to an embodiment of the present disclosure.

[0045] like Figure 2As shown, the prediction model in the embodiment of the present disclosure is a PUnet network structure, a combination of U-net and FPN, such as Figure 2 As shown in , the PUnet network structure of the disclosed embodiment includes two parts, one is a U-shaped encoder-decoder architecture similar to the U-net backbone, which gradually extracts features through multi-layer convolution and pooling. The feature map extracted in each layer will be used for subsequent multi-scale feature fusion. The encoder may include multiple feature extraction modules, each module consisting of a convolution module, instance normalization and an activation function. Each module in the encoder downsamples the input feature map through the convolution module to capture the contextual information of the feature map and reduce the spatial dimension of the feature. After the convolution operation, the feature map output by the convolution module is normalized and adjusted through instance normalization, and then nonlinear features are introduced and regularized through the activation function to improve the expression ability of the network. In the encoder stage, the input of each feature extraction module is the output of the previous layer.

[0046] The decoder can include multiple feature extraction modules, each of which is composed of a convolution module, instance normalization, activation function and feature fusion module. Each module in the decoder performs transposed convolution on the input feature map through the convolution module. After instance normalization and activation function, it is spliced ​​with the feature map generated by the corresponding level in the decoder, that is, skip connection, which gradually restores the spatial dimension of the feature map while retaining more context information.

[0047] According to the embodiment of the present disclosure, during the upsampling process of the decoder, since the feature map is upsampled multiple times through the feature extraction module, a pyramid structure (FPN structure) is formed. By adding direct supervision signals at multiple levels of the neural network, rather than only supervising at the terminal output layer of the network, the learning objectives of each level are clearly pointed to the final task goal, which helps to solve the gradient disappearance problem and improve the learning efficiency during network training and the prediction accuracy of the final layer. It can be understood that since the downsampling process of the encoder is also similar to a pyramid structure, based on the same reason, those skilled in the art can adaptively apply the design of the deep supervision pyramid to the encoder stage, and this disclosure will not elaborate on this.

[0048] Figure 3 The flowchart of the model acquisition method according to the embodiment of the present disclosure is schematically shown.

[0049] like Figure 3 As shown, the method includes operations S310 to S320.

[0050] In operation S310 , a training sample is acquired, wherein the training sample includes an image slice, a first dose distribution corresponding to the image slice, and a second dose distribution corresponding to the image slice.

[0051] According to an embodiment of the present disclosure, training samples may be image data that complies with the DICOM (Digital Imaging and Communications in Medicine) standard acquired through existing medical imaging equipment and information systems, and multiple image slices may be obtained through the image data. The image slices may be, for example, CT image slices.

[0052] After obtaining the CT image slices, high-quality gold annotation data can be generated according to the existing Monte Carlo calculation method and the preset pencil beam parameters, for example, through GATE software. For example, the first dose distribution and the second dose distribution corresponding to each CT image slice are generated through the preset pencil beam parameters and the CT image slices.

[0053] The first dose distribution refers to a dose distribution with lower distribution accuracy in a specific area relative to the image slice, such as a low particle dose distribution, which is often used for preliminary estimation and prediction of particle dose distribution and can be obtained by conventional or existing simple methods. The second dose distribution refers to a dose distribution with higher distribution accuracy in a specific area relative to the image slice, which is used to accurately determine the specific and accurate dose distribution of the pencil beam. This distribution is usually concentrated in the tumor center or target area to ensure effective treatment of tumor tissue. The low particle dose area may appear as a more blurred or uneven dose distribution with greater noise. The high particle dose area appears as a more uniform and concentrated dose distribution, and the dose distribution is clear and stable.

[0054] According to the embodiments of the present disclosure, the pencil beam may be a preset carbon ion pencil beam. The specific parameters of the carbon ion pencil beam, such as energy, energy dispersion, angular dispersion, etc., can be selected by those skilled in the art to conform to the actual carbon ion treatment scenario according to actual conditions.

[0055] In operation S320, a prediction model is trained based on the image slice, the first dose distribution, and the second dose distribution to obtain a target prediction model, where the target prediction model is used to obtain a second target dose distribution based on the target image slice and the first target dose distribution corresponding to the target image slice.

[0056] According to the embodiment of the present disclosure, since the noise in the first dose distribution is relatively high and there is interference noise, by training the prediction model with CT images and the first dose distribution, the prediction model's resistance to noise can be increased, thereby improving the accuracy of the prediction of the second dose distribution.

[0057] According to an embodiment of the present disclosure, operation S320 may specifically include: training the prediction model according to the loss function until the difference between the second dose distribution prediction value generated by the prediction model according to the image slice and the first dose distribution and the second dose distribution is less than a preset threshold, thereby obtaining a target prediction model. The loss function is determined according to the number of image slices, the number of voxels in the image slice, and the weight factor.

[0058] The following is a method for calculating a loss function in the prior art, specifically a root mean square logarithmic error (RMSLE) loss function formula:

[0059]

[0060] Where n is the number of samples, is the predicted value of the model, y i is the true value. However, in the embodiments of the present disclosure, the slice or feature map is three-dimensional data with thickness, wherein each voxel in the three-dimensional data has an uneven dose value. By using a loss function similar to the above-mentioned traditional loss function, the model cannot be more focused on processing the portion with a larger dose value in the slice, and the calculation accuracy of the model for the dose value is low.

[0061] In view of this, in an embodiment of the present disclosure, during the training process, different weights are assigned to different voxels corresponding to different dose values, so that the prediction model can perform differentiated processing on the voxels corresponding to different dose values ​​during the training process, thereby enhancing the accuracy of the model.

[0062] According to an embodiment of the present disclosure, the weight factor is determined based on a first number, where the first number is the number of voxels in the second dose distribution whose dose values ​​corresponding to the voxels are greater than a preset dose threshold.

[0063] Specifically, the weight factor can be determined according to the first number and the total number of voxels in the second dose distribution. In the second dose distribution, the voxels corresponding to the dose values ​​greater than the preset threshold have a larger corresponding weight factor when calculating the value of the loss function. The voxels corresponding to the dose values ​​less than the preset threshold have a smaller corresponding weight factor when calculating the value of the loss function. This allows the model to pay more attention to the part with larger dose values ​​during the training process, thereby improving the prediction accuracy of the dose values ​​corresponding to the high-dose voxels.

[0064] According to an embodiment of the present disclosure, the preset dose threshold is determined based on the maximum value of the voxel dose value in the second dose distribution. Specifically, the maximum value of the dose value corresponding to the voxel in a single second dose distribution can be determined. Based on the maximum value, the dose threshold is determined. For example, the dose threshold is determined based on the maximum value and a preset selection ratio.

[0065] According to an embodiment of the present disclosure, when the dose value of the target voxel in the second dose distribution is less than a preset dose threshold, when calculating the value of the loss function, the corresponding weight factor of the target voxel is the ratio of the first number to the number of voxels in the image slice.

[0066] According to an embodiment of the present disclosure, when the dose value of the target voxel in the second dose distribution is greater than or equal to a preset dose threshold, when calculating the value of the loss function, the corresponding weight factor of the target voxel is 1 minus the ratio of the first number and the number of voxels in the image slice.

[0067] Based on the above examples, the specific method for determining the weight factor and the corresponding loss function are given below.

[0068] For example, the loss function could be:

[0069]

[0070] Wherein, n is the number of samples, that is, the number of CT image slices or second dose distribution slices in the sample. In the embodiment of the present disclosure, the CT image slices and the second dose distribution image correspond one to one. m is the number of voxels in each slice, w ij represents the weight factor corresponding to the jth voxel in the i-th slice, represents the prediction model's prediction of the dose value corresponding to the jth voxel in the i-th slice, Represents the actual value of the dose value corresponding to the jth voxel in the ith slice.

[0071] For example, the dose threshold is 1% of the maximum dose value corresponding to a voxel in a single second dose distribution.

[0072] Then correspondingly,

[0073]

[0074] in,

[0075]

[0076] in, represents the maximum dose value in each voxel in the i-th slice, K i represents the number of voxels in the i-th slice whose dose value is greater than 1% of the maximum dose, W i It can be regarded as the proportion of the foreground in the training process, that is, the proportion of voxels with larger dose values ​​to the total voxels.

[0077] In this case, the trained model will focus more on the prediction of the foreground, making the model's prediction more accurate.

[0078] Figure 4The flowchart of the dose distribution generating method according to the embodiment of the present disclosure is schematically shown.

[0079] According to an embodiment of the present disclosure, a dose distribution generating method is also provided, including operations S410 to S440.

[0080] In operation S410, beam positions of respective ones of a plurality of pencil beams in a treatment plan are determined.

[0081] In operation S420 , image slices at each pencil beam position are acquired according to the pencil beam position.

[0082] In operation S430, a second dose distribution corresponding to a single beam is predicted based on the image slice, the first dose distribution corresponding to the image slice, and the target prediction model. The accuracy of the second target dose distribution is higher than the accuracy of the first target dose distribution. The first dose distribution refers to a dose distribution with a lower distribution accuracy in a specific area relative to the image slice, such as a low particle dose distribution, which is often used for preliminary estimation and prediction of particle dose distribution and can be obtained by conventional or existing simple methods. The second dose distribution refers to a dose distribution with a higher distribution accuracy in a specific area relative to the image slice, which is used to accurately determine the specific and accurate dose distribution of the pencil beam.

[0083] In operation S440, the second dose distributions corresponding to the pencil beams are superimposed to obtain a target Monte Carlo dose distribution.

[0084] According to an embodiment of the present disclosure, the target prediction model is the target prediction model obtained in the aforementioned embodiment.

[0085] According to the embodiment of the present disclosure, the Monte Carlo data at the single beam level is trained, and then the treatment plan level reasoning is performed based on the generated model. The computing resources consumed by the training process are greatly reduced, and the model trained by the model training method provided by the present disclosure can make a more accurate prediction of the second dose distribution. Compared with the traditional Monte Carlo calculation, there is no need to set a large number of empirical parameters, simulate particle tracking and complex physical reactions, and the trained network model can achieve an acceleration ratio of more than 100 times while meeting clinical needs.

[0086] Figure 5 The block diagram of a model obtaining device according to an embodiment of the present disclosure is schematically shown.

[0087] like Figure 5 As shown, the model acquisition device 500 includes a first acquisition module 510 and a training module 520 .

[0088] The first acquisition module 510 is used to acquire training samples, wherein the training samples include image slices and first and second dose distributions corresponding to the image slices, wherein the accuracy of the second dose distribution is higher than the accuracy of the first dose distribution; wherein the training samples are acquired based on a single carbon ion pencil beam. In some instances, it can be used to perform operation S310 in the above method, which will not be described in detail here.

[0089] The training module 520 is used to train a prediction model according to the image slice, the first dose distribution, and the second dose distribution to obtain a target prediction model, wherein the target prediction model is used to obtain a second target dose distribution according to the target image slice and the first target dose distribution of the target image slice, wherein the accuracy of the second target dose distribution is higher than the accuracy of the first target dose distribution. In some instances, it can be used to perform operation S320 in the above method, which will not be described in detail here.

[0090] According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits, or at least part of the functions of any one of them can be implemented in one module. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be split into multiple modules for implementation. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be at least partially implemented as hardware circuits, such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), systems on chips, systems on substrates, systems on packages, application specific integrated circuits (ASICs), or can be implemented by hardware or firmware in any other reasonable way of integrating or packaging circuits, or can be implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, according to the embodiments of the present invention, one or more of the modules, submodules, units, and subunits can be at least partially implemented as computer program modules, and when the computer program modules are run, the corresponding functions can be performed.

[0091] For example, any multiple of the first acquisition module 510 and the training module 520 can be combined in one module / unit / sub-unit for implementation, or any one of the modules / units / sub-units can be split into multiple modules / units / sub-units. Alternatively, at least part of the functions of one or more of these modules / units / sub-units can be combined with at least part of the functions of other modules / units / sub-units and implemented in one module / unit / sub-unit. According to an embodiment of the present disclosure, at least one of the first acquisition module 510 and the training module 520 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware and firmware or in any appropriate combination of any of them. Alternatively, at least one of the first acquisition module 510 and the training module 520 can be at least partially implemented as a computer program module, and when the computer program module is run, the corresponding function can be performed.

[0092] Figure 6 The block diagram of a dose distribution generating device according to an embodiment of the present disclosure is schematically shown.

[0093] like Figure 6 As shown, the dose distribution generating device 600 includes a determination module 610 , a second acquisition module 620 , a prediction module 630 and a superposition module 640 .

[0094] The determination module 610 is used to determine the beam positions of the multiple pencil beams in the treatment plan. In some examples, it can be used to perform the operation S410 in the above method, which will not be described in detail here.

[0095] The second acquisition module 620 is used to acquire the image slices corresponding to each pencil beam stream according to the pencil beam stream position. In some examples, it can be used to perform the operation S420 in the above method, which will not be described in detail here.

[0096] The prediction module 630 is used to predict the second dose distribution corresponding to each pencil beam according to the image slice, the first dose distribution corresponding to the image slice and the target prediction model; in some examples, it can be used to perform operation S430 in the above method, which will not be described in detail here.

[0097] The superposition module 640 is used to superimpose the second dose distributions corresponding to the pencil beams to obtain the target Monte Carlo dose distribution. In some examples, it can be used to perform the operation S440 in the above method, which will not be described in detail here.

[0098] According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits, or at least part of the functions of any one of them can be implemented in one module. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be split into multiple modules for implementation. According to the embodiments of the present invention, any one or more of the modules, submodules, units, and subunits can be at least partially implemented as hardware circuits, such as field programmable gate arrays (FPGAs), programmable logic arrays (PLAs), systems on chips, systems on substrates, systems on packages, application specific integrated circuits (ASICs), or can be implemented by hardware or firmware in any other reasonable way of integrating or packaging circuits, or can be implemented in any one of the three implementation methods of software, hardware, and firmware, or in any appropriate combination of any of them. Alternatively, according to the embodiments of the present invention, one or more of the modules, submodules, units, and subunits can be at least partially implemented as computer program modules, and when the computer program modules are run, the corresponding functions can be performed.

[0099] For example, any multiple of the determination module 610, the second acquisition module 620, the prediction module 630, and the superposition module 640 can be combined in one module / unit / sub-unit for implementation, or any one of the modules / units / sub-units can be split into multiple modules / units / sub-units. Alternatively, at least part of the functions of one or more of these modules / units / sub-units can be combined with at least part of the functions of other modules / units / sub-units and implemented in one module / unit / sub-unit. According to an embodiment of the present disclosure, at least one of the determination module 610, the second acquisition module 620, the prediction module 630, and the superposition module 640 can be at least partially implemented as a hardware circuit, such as a field programmable gate array (FPGA), a programmable logic array (PLA), a system on a chip, a system on a substrate, a system on a package, an application specific integrated circuit (ASIC), or can be implemented by hardware or firmware such as any other reasonable way of integrating or packaging the circuit, or implemented in any one of the three implementation methods of software, hardware, and firmware or in an appropriate combination of any of them. Alternatively, at least one of the determination module 610 , the second acquisition module 620 , the prediction module 630 , and the superposition module 640 may be at least partially implemented as a computer program module, and when the computer program module is executed, a corresponding function may be performed.

[0100] Figure 7 A block diagram of an electronic device suitable for implementing the method described above according to an embodiment of the present disclosure is schematically shown. Figure 7 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0101] like Figure 7 As shown, the electronic device 700 according to an embodiment of the present disclosure includes a processor 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage part 708 to a random access memory (RAM) 703. The processor 701 may include, for example, a general-purpose microprocessor (such as a CPU), an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (for example, an application-specific integrated circuit (ASIC)), etc. The processor 701 may also include an onboard memory for caching purposes. The processor 701 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0102] In RAM 703, various programs and data required for the operation of electronic device 700 are stored. Processor 701, ROM 702 and RAM 703 are connected to each other via bus 704. Processor 701 performs various operations of the method flow according to the embodiment of the present disclosure by executing the program in ROM 702 and / or RAM 703. It should be noted that the program can also be stored in one or more memories other than ROM 702 and RAM 703. Processor 701 can also perform various operations of the method flow according to the embodiment of the present disclosure by executing the program stored in the one or more memories.

[0103] According to an embodiment of the present disclosure, the electronic device 700 may further include an input / output (I / O) interface 707, and the input / output (I / O) interface 705 is also connected to the bus 704. The electronic device 700 may further include one or more of the following components connected to the input / output (I / O) interface 705: an input portion 706 including a keyboard, a mouse, etc.; an output portion 707 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 708 including a hard disk, etc.; and a communication portion 709 including a network interface card such as a LAN card, a modem, etc. The communication portion 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the input / output (I / O) interface 705 as needed. A removable medium 711, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 710 as needed, so that a computer program read therefrom is installed into the storage portion 708 as needed.

[0104] According to an embodiment of the present disclosure, the method flow according to an embodiment of the present disclosure can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable storage medium, and the computer program contains a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 709, and / or installed from the removable medium 711. When the computer program is executed by the processor 701, the above-mentioned functions defined in the system of the embodiment of the present disclosure are executed. According to an embodiment of the present disclosure, the system, equipment, device, module, unit, etc. described above can be implemented by a computer program module.

[0105] The present disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or may exist independently without being assembled into the device / apparatus / system. The above computer-readable storage medium carries one or more programs, and when the above one or more programs are executed, the method according to the embodiment of the present disclosure is implemented.

[0106] According to an embodiment of the present disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium. For example, it may include, but is not limited to: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0107] For example, according to an embodiment of the present disclosure, the computer-readable storage medium may include the ROM 702 and / or the RAM 703 described above and / or one or more memories other than the ROM 702 and the RAM 703 .

[0108] An embodiment of the present disclosure also includes a computer program product, which includes a computer program, which contains program code for executing the method provided by the embodiment of the present disclosure. When the computer program product runs on an electronic device, the program code is used to enable the electronic device to implement the model acquisition method or dose distribution generation method provided by the embodiment of the present disclosure.

[0109] When the computer program is executed by the processor 701, the above functions defined in the system / device of the embodiment of the present disclosure are executed. According to the embodiment of the present disclosure, the system, device, module, unit, etc. described above can be implemented by a computer program module.

[0110] In one embodiment, the computer program may rely on tangible storage media such as optical storage devices, magnetic storage devices, etc. In another embodiment, the computer program may also be transmitted and distributed in the form of signals on a network medium, and downloaded and installed through the communication part 709, and / or installed from the removable medium 711. The program code contained in the computer program may be transmitted using any appropriate network medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0111] According to an embodiment of the present disclosure, the program code for executing the computer program provided by the embodiment of the present disclosure can be written in any combination of one or more programming languages. Specifically, these computing programs can be implemented using high-level process and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, Java, C++, python, "C" language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0112] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions. It can be understood by those skilled in the art that the features recorded in the various embodiments of the present disclosure can be combined and / or combined in a variety of ways, even if such a combination or combination is not explicitly recorded in the present disclosure. In particular, without departing from the spirit and teaching of the present disclosure, the features described in the various embodiments of the present disclosure may be combined and / or combined in a variety of ways. All of these combinations and / or combinations fall within the scope of the present disclosure.

[0113] The embodiments of the present disclosure are described above. However, these embodiments are only for illustrative purposes and are not intended to limit the scope of the present disclosure. Although the embodiments are described above, this does not mean that the measures in the various embodiments cannot be used in combination to advantage. Without departing from the scope of the present disclosure, those skilled in the art may make a variety of substitutions and modifications, which should all fall within the scope of the present disclosure.

Claims

1. A model acquisition method, comprising: Acquiring a training sample, wherein the training sample includes an image slice, a first dose distribution corresponding to the image slice, and a second dose distribution corresponding to the image slice, wherein the accuracy of the second dose distribution is higher than the accuracy of the first dose distribution; Training a prediction model according to the image slice, the first dose distribution, and the second dose distribution to obtain a target prediction model, wherein the target prediction model is used to obtain a second target dose distribution according to a target image slice and the first target dose distribution of the target image slice, wherein the accuracy of the second target dose distribution is higher than the accuracy of the first target dose distribution; Wherein, the training samples are obtained based on a single carbon ion pencil beam; The step of training a prediction model according to the image slice, the second dose distribution and the first dose distribution comprises: According to the loss function, the prediction model is trained until the difference between the second dose distribution prediction value generated by the prediction model according to the image slice and the first dose distribution and the second dose distribution is less than a preset threshold, thereby obtaining the target prediction model; The loss function is determined according to the number of image slices, the number of voxels in the image slices, and the weight factor corresponding to each voxel.

2. The method according to claim 1, wherein: The weight factor is determined according to a first number, where the first number is the number of voxels in the second dose distribution whose dose values ​​corresponding to the voxels are greater than a preset dose threshold.

3. The method according to claim 2, wherein: The preset dose threshold is determined according to the maximum dose value corresponding to the voxels in the second dose distribution.

4. The method according to claim 2, wherein: When the dose value of the target voxel in the second dose distribution is less than a preset dose threshold, when calculating the value of the loss function, the weight factor corresponding to the target voxel is the ratio of the first number to the number of voxels in the image slice.

5. The method according to claim 2, wherein: When the dose value of the voxel in the second dose distribution is greater than or equal to a preset dose threshold, when calculating the value of the loss function, the weight factor corresponding to the target voxel is 1 minus the ratio of the first number to the number of voxels in the image slice.

6. According to the method of claim 1, the target prediction model is a convolutional neural network model for Monte Carlo simulation.

7. A method for generating a dose distribution, comprising: Determine the beam positions of multiple pencil beams in the treatment plan; According to the pencil beam positions, acquiring image slices corresponding to each pencil beam; Predicting a second target dose distribution corresponding to each pencil beam according to the image slice, the first target dose distribution corresponding to the image slice, and a target prediction model, wherein the target prediction model is the target prediction model according to any one of claims 1 to 6, wherein the accuracy of the second target dose distribution is higher than the accuracy of the first target dose distribution; The second dose distributions corresponding to the pencil beams are superimposed to obtain a target Monte Carlo dose distribution.

8. A model obtaining device, comprising: A first acquisition module is used to acquire a training sample, wherein the training sample includes an image slice and a second dose distribution and a first dose distribution corresponding to the image slice, wherein the accuracy of the second dose distribution is higher than the accuracy of the first dose distribution; a training module, used for training a prediction model according to the image slice, the first dose distribution and the second dose distribution to obtain a target prediction model, wherein the target prediction model is used for obtaining a second target dose distribution according to a target image slice and the first target dose distribution of the target image slice, wherein the accuracy of the second target dose distribution is higher than the accuracy of the first target dose distribution; Wherein, the training samples are obtained based on a single carbon ion pencil beam; The step of training a prediction model according to the image slice, the second dose distribution and the first dose distribution comprises: According to the loss function, the prediction model is trained until the difference between the second dose distribution prediction value generated by the prediction model according to the image slice and the first dose distribution and the second dose distribution is less than a preset threshold, thereby obtaining the target prediction model; The loss function is determined according to the number of image slices, the number of voxels in the image slices, and the weight factor corresponding to each voxel.

9. A dose distribution generating device, comprising: A determination module, used for determining the beam positions of each of the plurality of pencil beams in the treatment plan; A second acquisition module, used for acquiring image slices corresponding to each pencil beam stream according to the pencil beam stream position; a prediction module, configured to predict a second dose distribution corresponding to each of the pencil beams according to the image slice, the first dose distribution corresponding to the image slice, and a target prediction model, wherein the target prediction model is the target prediction model of any one of claims 1 to 6, wherein the accuracy of the second dose distribution is higher than the accuracy of the first dose distribution; and The superposition module is used to superimpose the second dose distributions corresponding to the pencil beams to obtain a target Monte Carlo dose distribution.

10. An electronic device, comprising: one or more processors; a memory for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

11. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, enables the processor to implement the method according to any one of claims 1 to 7.

12. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images