Dose distribution prediction method, device, equipment, storage medium and product

Abstract

The present application relates to a dose distribution prediction method, apparatus, computer device, storage medium, and computer program product. The method comprises: acquiring a medical image of a subject; processing the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image; and performing dose calculation on the flux map using a second network branch in the first neural network model to determine a target dose distribution for each point on the medical image. This method can improve the accuracy of the predicted dose distribution while ensuring that the obtained dose distribution prediction results conform to the laws of physics.

Description

Technical Field

[0001] The present application relates to the technical field of medical radiotherapy, and in particular to a dose distribution prediction method, apparatus, device, storage medium and product. Background Art

[0002] Radiotherapy is a method of treating tumors in patients using radiation. Before using radiotherapy, the dose distribution corresponding to the radiotherapy plan is usually predicted. The predicted results can then be used to formulate a corresponding radiotherapy plan, so that the dose to the patient can be controlled to an optimal range.

[0003] In related technologies, when predicting dose distribution, it is usually necessary to first obtain relevant delineation information of the lesion area on the patient's image, and then use deep learning technology to process this relevant delineation information to obtain the dose distribution of each point on the image.

[0004] However, the above technology has the problem that the dose distribution of each point on the obtained image is not accurate enough. Summary of the Invention

[0005] Based on this, it is necessary to provide a dose distribution prediction method, device, computer equipment, computer-readable storage medium and computer program product that can improve the accuracy of dose distribution at each point on the image in order to address the above technical problems.

[0006] In a first aspect, the present application provides a method for predicting dose distribution. The method comprises:

[0007] obtaining medical images of the subject;

[0008] Processing the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image; the flux map is used to characterize the corresponding relationship between the intensity and shape of the beam in various directions;

[0009] The second network branch in the first neural network model is used to perform dose calculation on the flux map to determine the target dose distribution of each point on the medical image.

[0010] In one embodiment, before determining the target dose distribution for each point on the medical image, the method further includes:

[0011] Using a preset second neural network model to predict the dose distribution of each point on the medical image, and determining a first dose distribution of each point on the medical image;

[0012] Accordingly, the second network branch in the first neural network model is used to perform dose calculation on the flux map to determine the target dose distribution at each point on the medical image, including:

[0013] The second network branch in the first neural network model is used to perform dose calculation on the flux map to determine a second dose distribution at each point on the medical image;

[0014] A target dose distribution for each point on the medical image is determined according to the first dose distribution for each point on the medical image and the second dose distribution for each point on the medical image.

[0015] In one embodiment, determining the target dose distribution for each point on the medical image based on the first dose distribution for each point on the medical image and the second dose distribution for each point on the medical image includes:

[0016] Mathematical operations are performed on the first dose distribution and the corresponding second dose distribution of each point on the medical image to obtain the target dose distribution of each point on the medical image.

[0017] In one embodiment, the training method of the first neural network model includes:

[0018] Acquire multiple training images; each training image includes a gold standard dose distribution for each point on the training image;

[0019] Each training image is used as the input of the initial first neural network model, and the predicted dose distribution of each point on each training image is used as the output of the initial first neural network model. The initial first neural network model is trained to determine the first neural network model.

[0020] In one embodiment, the training method of the first neural network model includes:

[0021] Acquire multiple training images; each training image includes a gold standard dose distribution for each point on the training image;

[0022] Each training image is used as the input of the initial first neural network model, and the initial flux map corresponding to each training image and the predicted dose distribution of each point on each training image are used as the output of the initial first neural network model. The initial first neural network model is trained to determine the first neural network model.

[0023] In one embodiment, the above-mentioned method uses each training image as the input of the initial first neural network model, uses the initial flux map corresponding to each training image and the predicted dose distribution of each point on each training image as the output of the initial first neural network model, and trains the initial first neural network model to determine the first neural network model, including:

[0024] Calculate the first loss between the predicted dose distribution at each point on each training image and the gold standard dose distribution at the corresponding point;

[0025] Calculate the second loss based on the initial flux map corresponding to each training image;

[0026] The initial first neural network model is trained according to the first loss and the second loss to determine the first neural network model.

[0027] In a second aspect, the present application also provides a dose distribution prediction device. The device comprises:

[0028] an acquisition module, for acquiring medical images of an object;

[0029] a processing module, configured to process the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image; the flux map is used to characterize the correspondence between the intensity and shape of the beam in various directions;

[0030] The dose prediction module is used to use the second network branch in the above-mentioned first neural network model to perform dose calculation on the flux map to determine the target dose distribution of each point on the medical image.

[0031] In a third aspect, the present application further provides a computer device. The computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the following steps are performed:

[0032] obtaining medical images of the subject;

[0033] Processing the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image; the flux map is used to characterize the corresponding relationship between the intensity and shape of the beam in various directions;

[0034] The second network branch in the first neural network model is used to perform dose calculation on the flux map to determine the target dose distribution of each point on the medical image.

[0035] In a fourth aspect, the present application further provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the following steps:

[0036] obtaining medical images of the subject;

[0037] Processing the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image; the flux map is used to characterize the corresponding relationship between the intensity and shape of the beam in various directions;

[0038] The second network branch in the first neural network model is used to perform dose calculation on the flux map to determine the target dose distribution of each point on the medical image.

[0039] In a fifth aspect, the present application further provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the following steps:

[0040] obtaining medical images of the subject;

[0041] Processing the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image; the flux map is used to characterize the corresponding relationship between the intensity and shape of the beam in various directions;

[0042] The second network branch in the first neural network model is used to perform dose calculation on the flux map to determine the target dose distribution of each point on the medical image.

[0043] The above-mentioned dose distribution prediction method, apparatus, computer equipment, storage medium and computer program product, by using the first network branch in the preset first neural network model to process the acquired medical image of the object, determine the flux map corresponding to the medical image, and use the second network branch in the first neural network model to perform dose calculation on the flux map, and determine the target dose distribution of each point on the medical image, wherein the flux map is used to characterize the correspondence between the intensity and shape of the beam in each direction. In this method, since the flux map that characterizes the correspondence between the intensity and shape of the beam in each direction is applied to the dose prediction process of the medical image, that is, the beam information is integrated into the dose prediction process, and the beam information is related to the physical laws of the dose prediction results, then the dose prediction results will be more consistent with the actual physical laws, and on this basis, the dose distribution prediction is performed through the two network branches of the neural network, thereby improving the accuracy of the predicted dose distribution. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a diagram of the internal structure of a computer device in one embodiment;

[0045] Figure 2 is a schematic flow chart of a dose distribution prediction method in one embodiment;

[0046] Figure 2a is a schematic structural diagram of a first neural network model in one embodiment;

[0047] Figure 3 is a schematic flow chart of a dose distribution prediction method in another embodiment;

[0048] Figure 3a A schematic diagram of the structure of a first neural network model and a second neural network model in another embodiment;

[0049] Figure 4 is a schematic flow chart of a dose distribution prediction method in another embodiment;

[0050] Figure 5 is a schematic flow chart of a dose distribution prediction method in another embodiment;

[0051] Figure 6 FIG. 4 is a structural block diagram of a dose distribution prediction device in one embodiment. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0053] The dose distribution prediction method provided in the embodiment of the present application can be applied to Figure 1 The computer device shown can be a terminal or a server. Taking a terminal as an example, the computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner. The wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a dose distribution prediction method is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.

[0054] Those skilled in the art will understand that Figure 1 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0055] In one embodiment, Figure 2 As shown, a dose distribution prediction method is provided, which is applied to Figure 1 Taking the computer device in the example as an example, the method may include the following steps:

[0056] S202: Obtain medical images of the subject.

[0057] The object may be a human body, an animal body, etc. The medical image of the object may be an X-ray image, a CT image, etc., or other medical images. In addition, the medical image of the object may be an image of a single body part or multiple body parts.

[0058] Specifically, one or more body parts of a subject may be scanned in advance using a scanning device, or X-rays may be emitted onto one or more body parts of the subject to obtain scan data or projection data of the subject, and then the scan data or projection data may be reconstructed to obtain a medical image of the subject. Alternatively, the medical image of the subject may be stored in advance on a server or cloud and then directly retrieved from the server or cloud when needed.

[0059] S204: Process the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image.

[0060] In this step, the first neural network model can be a U-net model, a V-net model, a CNN model, etc. Figure 2a As shown, the first neural network model includes two network branches, namely a first network branch and a second network branch, wherein the first network branch and the second network branch are connected, and the output of the first network branch can be used as the input of the second network branch.

[0061] The first network branch can also be a U-net model, a V-net model, a CNN model, etc. The first network branch is mainly used to output a flux map corresponding to the above-mentioned medical image, and the above-mentioned flux map is used to characterize the correspondence between the intensity and shape of the beam in various directions. The flux map refers to the correspondence between the intensity and shape of the beam in various directions when the beam is emitted to the object through the collimator of the radiation equipment, where the shape refers to the shape of the beam to the object. The beam information (including the intensity, direction, shape, etc. of the beam) is added to the dose distribution prediction process through the flux map. The beam information is related to the physical laws of the dose prediction results, so that the subsequent dose distribution results can be more consistent with the physical laws.

[0062] Specifically, after obtaining the medical image of the object, the medical image of the object can be input into the first neural network model, and the medical image can be processed by up- and down-sampling, skip-layer connection, convolution, pooling, etc. through the first network branch of the first neural network model to finally obtain the flux map corresponding to the medical image.

[0063] S206 , performing dose calculation on the flux map using the second network branch in the first neural network model to determine the target dose distribution of each point on the medical image.

[0064] In this step, the first network branch was mentioned above. Here, the second network branch will be described. The second network branch can also be a U-net model, a V-net model, a CNN model, etc., and is primarily used to calculate dose distribution. A known mathematical dose calculation method can be used as the second network branch. This known mathematical dose method can be a method for calculating dose using a known algorithm, such as a pencil beam algorithm, a convolution algorithm, or a Monte Carlo algorithm.

[0065] Specifically, after obtaining the flux map output by the first network branch, the flux map can be input into the second network branch for dose distribution calculation to obtain the dose distribution results for each point on the flux map. Since the flux map corresponds to the medical image, the dose distribution results for each point on the medical image are obtained, and the target dose distribution for each point is determined based on the dose distribution results for each point on the medical image (for example, the dose distribution results for each point can be directly used as the target dose distribution for each point, or the dose distribution results for each point can be further processed to obtain the target dose distribution for each point). It should be noted that the size of the flux map is generally the same as the size of the medical image.

[0066] In addition, it should be noted that the use of flux maps for dose distribution calculation can make the calculated dose distribution results more consistent with the laws of physics, and at the same time, the dose distribution results are more accurate on this basis.

[0067] Taking human organs and tumors as an example, when using existing dose distribution prediction methods, organ regions and tumor regions are generally outlined. When performing dose distribution prediction, the prediction results for the outlined regions (including organ regions and tumor regions) are relatively accurate, but the prediction results for non-outlined regions (including non-organ regions and non-tumor regions) are less accurate. In the embodiments of the present application, beam information is integrated into the dose distribution prediction. While ensuring the accuracy of the prediction results for the outlined regions (including organ regions and tumor regions), the accuracy of the prediction results for the non-outlined regions (including non-organ regions and non-tumor regions) can also be improved, thereby providing a more accurate data basis for subsequent further analysis of the prediction results for the outlined and non-outlined regions.

[0068] In the above-mentioned dose distribution prediction method, the acquired medical image of the object is processed by using the first network branch in the preset first neural network model to determine the flux map corresponding to the medical image, and the second network branch in the first neural network model is used to perform dose calculation on the flux map to determine the target dose distribution of each point on the medical image, wherein the flux map is used to characterize the correspondence between the intensity and shape of the beam in each direction. In this method, since the flux map that characterizes the correspondence between the intensity and shape of the beam in each direction is applied to the dose prediction process of the medical image, that is, the beam information is integrated into the dose prediction process, and the beam information is related to the physical laws of the dose prediction results, the dose prediction results will be more consistent with the actual physical laws, and on this basis, the dose distribution prediction is performed through the two network branches of the neural network, thereby improving the accuracy of the predicted dose distribution.

[0069] In the above embodiment, it is mentioned that the various network branches of the first neural network model can be used to predict the dose distribution. Based on this, the second neural network model is introduced to simultaneously predict the dose distribution. The process is described in detail below. In one embodiment, another dose distribution prediction method is provided. Based on the above embodiment, before determining the target dose distribution of each point on the medical image in S206, the above method can also include the following steps:

[0070] A preset second neural network model is used to predict the dose distribution of each point on the medical image to determine the first dose distribution of each point on the medical image.

[0071] In this step, the second neural network model can have the same structure as the first neural network model, for example, a U-net model, a V-net model, a CNN model, etc. The second neural network model can have only one network branch for calculating the dose distribution of each point on the medical image.

[0072] Before using the second neural network model, the second neural network model may also be trained. The training process includes: obtaining multiple training images, each of which includes a gold standard dose distribution for each point on the training image, inputting each training image into the initial second neural network model, obtaining a predicted dose distribution for each point on each training image, calculating the loss between the predicted dose distribution for each point on each training image and the gold standard dose distribution for the corresponding point, and training the initial second neural network model using the loss to obtain a trained second neural network model. The gold standard dose distribution for each point on the training image can be the dose distribution result for each point obtained by performing dose calculation on the existing dose distribution plan on the training image using the known mathematical dose calculation method as described above, and using the dose distribution result for each point as its corresponding gold standard dose distribution.

[0073] After the second neural network model is trained, the trained second neural network model can be used to calculate the dose distribution of each point on the medical image to obtain the dose distribution of each point on the medical image, which is recorded as the first dose distribution.

[0074] In this embodiment, before determining the target dose distribution of each point on the medical image, a second neural network model can also be used to calculate the first dose distribution of each point on the medical image. In this way, the results obtained by combining the first neural network model can be used to determine the target dose distribution, thereby improving the accuracy of the determined target dose distribution.

[0075] In the above embodiment, it is mentioned that each network branch of the first neural network model can be used to predict the dose distribution, and the second neural network model is introduced to predict the dose distribution at the same time. The following is a detailed description of how to obtain the target dose distribution through these two models. In one embodiment, another dose distribution prediction method is provided, such as Figure 3 As shown, based on the above embodiment, the above S206 may include the following steps:

[0076] S302: Perform dose calculation on the flux map using the second network branch in the first neural network model to determine a second dose distribution for each point on the medical image.

[0077] In this step, see Figure 3a As shown, as mentioned in S206 above, a known mathematical dose calculation method can be used as the second network branch. After obtaining the flux map output by the first network branch, the flux map can be input into the second network branch for dose distribution calculation to obtain the dose distribution results of each point on the flux map. Since the flux map corresponds to the medical image, the dose distribution results of each point on the medical image are obtained, which are recorded as the second dose distribution of each point, namely dose1 in the figure.

[0078] Continue to see Figure 3a As shown, after obtaining the medical image (i.e., input in the figure), the medical image is input into the first neural network model, and at the same time, the medical image can be input into the second neural network model for dose distribution calculation to obtain the dose distribution of each point on the medical image, which is recorded as the first dose distribution of each point on the medical image, i.e., dose2 in the figure.

[0079] S304 , determining a target dose distribution for each point on the medical image according to the first dose distribution for each point on the medical image and the second dose distribution for each point on the medical image.

[0080] In this step, after obtaining the first dose distribution and the second dose distribution for each point on the medical image, the target dose distribution for each point (dose in the figure) can be obtained by combining these two dose distributions. Optionally, mathematical operations can be performed on the first dose distribution and the corresponding second dose distribution for each point on the medical image to obtain the target dose distribution for each point on the medical image.

[0081] That is, mathematical operations can be performed on the first dose distribution and the second dose distribution at each point to obtain the target dose distribution at each point. The mathematical operations here are mainly averaging operations, which can be weighted summation followed by averaging, or direct summation followed by averaging.

[0082] In this embodiment, based on the dose distribution of each point on the medical image obtained by the second network branch of the first neural network model, the dose distribution of each point obtained by the second neural network model can be combined to ultimately obtain the target dose distribution for each point. Here, the dose distribution results of the two network models are combined to obtain the final dose distribution result, which can further improve the accuracy of the predicted dose distribution result. Furthermore, the final dose distribution result is obtained by performing mathematical operations on the dose distribution results of the two network models. This calculation process is relatively simple and fast, thereby improving the efficiency of the dose distribution prediction process.

[0083] In the above embodiment, it is mentioned that the dose distribution prediction can be performed using each network branch of the first neural network model. As for the training method of the first neural network model, two possible implementation methods are given below. The following embodiment first describes a possible implementation method. Figure 4 As shown, based on the above embodiment, the training method of the first neural network model may include the following steps:

[0084] S402 , acquiring a plurality of training images; each training image includes a gold standard dose distribution of each point on the training image.

[0085] The multiple training images here may be multiple medical images of the same subject at different periods / times, or multiple medical images of different subjects at the same period / time, or multiple medical images of different subjects at different periods / times.

[0086] The gold standard dose distribution here is the same as the gold standard dose distribution of the second neural network model mentioned above. It can be the dose distribution results of each point obtained by calculating the training image using the known mathematical dose calculation method mentioned above, and the dose distribution result of each point is used as its corresponding gold standard dose distribution.

[0087] S404: Using each training image as input of the initial first neural network model, using the predicted dose distribution of each point on each training image as output of the initial first neural network model, training the initial first neural network model, and determining the first neural network model.

[0088] Specifically, each training image can be input into the initial first neural network model to obtain the predicted dose distribution of each point on each training image, calculate the loss between the predicted dose distribution of each point on each training image and the gold standard dose distribution of the corresponding point, and use the loss to train the initial first neural network model to obtain a trained first neural network model.

[0089] In this embodiment, an initial first neural network model is trained using multiple training images including gold standard dose distributions for each point and predicted dose distributions for each point in each training image to obtain a trained first neural network model. Training the first neural network model using multiple training images can improve the accuracy of the trained first neural network model, resulting in more accurate dose distribution predictions obtained using the trained first neural network model.

[0090] Regarding the training method of the first neural network model, based on the above embodiment, the following embodiment describes another possible implementation method, such as Figure 5 As shown, the training method of the first neural network model may include the following steps:

[0091] S502 , obtaining a plurality of training images; each training image includes a gold standard dose distribution of each point on the training image.

[0092] The explanation of this step can be found in the explanation of S402 above, which will not be repeated here.

[0093] S504: Use each training image as the input of the initial first neural network model, use the initial flux map corresponding to each training image and the predicted dose distribution of each point on each training image as the output of the initial first neural network model, train the initial first neural network model, and determine the first neural network model.

[0094] In this step, after obtaining each training image, each training image can be input into the first network branch and the second network branch of the initial first neural network model for processing to obtain the predicted dose distribution of each point on each training image. Afterwards, optionally, the first loss between the predicted dose distribution of each point on each training image and the gold standard dose distribution of the corresponding point can be calculated, and the second loss can be calculated based on the initial flux map corresponding to each training image, and the initial first neural network model can be trained based on the first loss and the second loss to determine the trained first neural network model.

[0095] Among them, the initial flux map can be inputting each training image into the first network branch of the initial first neural network model, the flux map output by the first network branch, and further inputting the initial flux map output by the first network branch of the initial first neural network model into the second network branch of the initial first neural network model to obtain the predicted dose distribution of each point on the training image.

[0096] Furthermore, when calculating the first loss, the variance, standard deviation, error, etc. between the predicted dose distribution at each point and the gold standard dose distribution at the corresponding point can be calculated to obtain the first loss for each point on each training image. Furthermore, the second loss can be calculated using the initial flux map corresponding to each training image by calculating the variance, standard deviation, error, gradient, or reciprocal of each initial flux map, and using the calculated variance, standard deviation, error, gradient, or reciprocal as the second loss.

[0097] After obtaining the first loss of each point on each training image and the second loss corresponding to each training image, the first total loss corresponding to each training image can be obtained by averaging or weighted summing the first losses of each point on each training image, and summing or weighted summing the first total loss and the second loss of each training image to obtain the total image loss, and using the total image loss corresponding to each training image to train the initial first neural network model to obtain a trained first neural network model. Here, the first neural network model is trained, which means that the first network branch and the second network branch in the first neural network model are both trained.

[0098] In this embodiment, a first neural network model is trained using multiple training images including gold standard dose distributions for each point, predicted dose distributions for each point on each training image, and initial flux maps corresponding to each training image. Combining the predicted dose distributions for each point with the flux map allows the trained first neural network model to be more consistent with the physical laws governing dose prediction results, and on this basis, the dose distribution prediction results obtained using the trained first neural network model are more accurate. Furthermore, training the first neural network model using two losses (including the loss of the predicted dose distribution for each point and the loss of the initial flux map) can increase the speed of the first neural network model training process and further improve the efficiency of dose distribution prediction.

[0099] In order to explain the technical solution of the present application in more detail, a more detailed embodiment is given below to illustrate the technical solution of the present application. Based on the above embodiment, the method may include the following steps:

[0100] This method is mainly divided into two parts: model training stage and model usage stage. The two parts are explained below.

[0101] 1. Model training stage:

[0102] The first training method of the neural network model is:

[0103] Acquire multiple training images; each training image includes a gold standard dose distribution for each point on the training image; use each training image as input to an initial first neural network model, use the predicted dose distribution for each point on each training image as output of the initial first neural network model, train the initial first neural network model, and determine the first neural network model.

[0104] The second training method of the first neural network model:

[0105] Acquire multiple training images; each training image includes a gold standard dose distribution for each point on the training image; calculate a first loss between the predicted dose distribution for each point on each training image and the gold standard dose distribution for the corresponding point; calculate a second loss based on an initial flux map corresponding to each training image; train an initial first neural network model based on the first loss and the second loss to determine the first neural network model.

[0106] 2. Model usage stage:

[0107] S1, obtain medical images of the subject.

[0108] S2. Process the medical image using the first network branch in the first neural network model to determine a flux map corresponding to the medical image; the flux map is used to characterize the correspondence between the intensity and shape of the beam in various directions.

[0109] S3, using the second network branch in the first neural network model to perform dose calculation on the flux map to determine a second dose distribution for each point on the medical image.

[0110] S4, using a preset second neural network model to predict the dose distribution of each point on the medical image, and determining a first dose distribution of each point on the medical image.

[0111] S5 , performing a summation operation on the first dose distribution and the corresponding second dose distribution of each point on the medical image to obtain a target dose distribution of each point on the medical image.

[0112] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0113] Based on the same inventive concept, embodiments of the present application also provide a dose distribution prediction device for implementing the aforementioned dose distribution prediction method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more dose distribution prediction device embodiments provided below can be found in the aforementioned limitations of the dose distribution prediction method and will not be further elaborated here.

[0114] In one embodiment, Figure 6 As shown, a dose distribution prediction device is provided, comprising: an acquisition module 10, a processing module 11 and a dose prediction module 12, wherein:

[0115] an acquisition module 10, for acquiring medical images of an object;

[0116] A processing module 11 is configured to process the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image; the flux map is configured to represent the corresponding relationship between the intensity and shape of the beam in various directions;

[0117] The dose prediction module 12 is used to perform dose calculation on the flux map using the second network branch in the first neural network model to determine the target dose distribution of each point on the medical image.

[0118] In one embodiment, another dose distribution prediction device is provided. Before the dose prediction module 12 determines the target dose distribution of each point on the medical image, the device may further include:

[0119] The first dose prediction unit is used to predict the dose distribution of each point on the medical image using a preset second neural network model to determine a first dose distribution of each point on the medical image.

[0120] Accordingly, the dose prediction module 12 may include:

[0121] a second dose prediction unit, configured to perform dose calculation on the flux map using the second network branch in the first neural network model to determine a second dose distribution for each point on the medical image;

[0122] The target dose prediction unit is used to determine the target dose distribution of each point on the medical image according to the first dose distribution of each point on the medical image and the second dose distribution of each point on the medical image.

[0123] Optionally, the target dose prediction unit is specifically configured to perform mathematical operations on the first dose distribution and the corresponding second dose distribution of each point on the medical image to obtain the target dose distribution of each point on the medical image.

[0124] In one embodiment, another dose distribution prediction device is provided. Based on the above embodiment, the device may further include:

[0125] A first training image acquisition module is used to acquire a plurality of training images; each training image includes a gold standard dose distribution of each point on the training image;

[0126] The first training module is used to use each training image as the input of the initial first neural network model, use the predicted dose distribution of each point on each training image as the output of the initial first neural network model, train the initial first neural network model, and determine the first neural network model.

[0127] In one embodiment, another dose distribution prediction device is provided. Based on the above embodiment, the device may further include:

[0128] A second training image acquisition module is used to acquire multiple training images; each training image includes a gold standard dose distribution of each point on the training image;

[0129] The second training module is used to use each training image as the input of the initial first neural network model, and use the initial flux map corresponding to each training image and the predicted dose distribution of each point on each training image as the output of the initial first neural network model, train the initial first neural network model, and determine the first neural network model.

[0130] Optionally, the second training module may include:

[0131] a first loss calculation unit, configured to calculate a first loss between the predicted dose distribution at each point on each training image and the gold standard dose distribution at the corresponding point;

[0132] a second loss calculation unit, configured to calculate a second loss according to an initial flux map corresponding to each training image;

[0133] The training unit is used to train the initial first neural network model according to the first loss and the second loss to determine the first neural network model.

[0134] Each module in the aforementioned dose distribution prediction device may be implemented in whole or in part through software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a computer device memory in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0135] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0136] Acquire a medical image of the object; process the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image; the flux map is used to characterize the correspondence between the intensity and shape of the beam in various directions; and perform dose calculation on the flux map using a second network branch in the first neural network model to determine a target dose distribution for each point on the medical image.

[0137] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0138] A preset second neural network model is used to predict the dose distribution of each point on the medical image to determine the first dose distribution of each point on the medical image; the second network branch in the first neural network model is used to calculate the dose of the flux map to determine the second dose distribution of each point on the medical image; based on the first dose distribution of each point on the medical image and the second dose distribution of each point on the medical image, the target dose distribution of each point on the medical image is determined.

[0139] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0140] Mathematical operations are performed on the first dose distribution and the corresponding second dose distribution of each point on the medical image to obtain the target dose distribution of each point on the medical image.

[0141] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0142] Acquire multiple training images; each training image includes a gold standard dose distribution for each point on the training image; use each training image as input to an initial first neural network model, use the predicted dose distribution for each point on each training image as output of the initial first neural network model, train the initial first neural network model, and determine the first neural network model.

[0143] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0144] Acquire multiple training images; each training image includes a gold standard dose distribution for each point on the training image; use each training image as input to an initial first neural network model, use an initial flux map corresponding to each training image and a predicted dose distribution for each point on each training image as outputs of the initial first neural network model, train the initial first neural network model, and determine the first neural network model.

[0145] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0146] Calculate the first loss between the predicted dose distribution of each point on each training image and the gold standard dose distribution of the corresponding point; calculate the second loss based on the initial flux map corresponding to each training image; train the initial first neural network model based on the first loss and the second loss to determine the first neural network model.

[0147] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0148] Acquire a medical image of the object; process the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image; the flux map is used to characterize the correspondence between the intensity and shape of the beam in various directions; and perform dose calculation on the flux map using a second network branch in the first neural network model to determine a target dose distribution for each point on the medical image.

[0149] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0150] A preset second neural network model is used to predict the dose distribution of each point on the medical image to determine the first dose distribution of each point on the medical image; the second network branch in the first neural network model is used to calculate the dose of the flux map to determine the second dose distribution of each point on the medical image; based on the first dose distribution of each point on the medical image and the second dose distribution of each point on the medical image, the target dose distribution of each point on the medical image is determined.

[0151] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0152] Mathematical operations are performed on the first dose distribution and the corresponding second dose distribution of each point on the medical image to obtain the target dose distribution of each point on the medical image.

[0153] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0154] Acquire multiple training images; each training image includes a gold standard dose distribution for each point on the training image; use each training image as input to an initial first neural network model, use the predicted dose distribution for each point on each training image as output of the initial first neural network model, train the initial first neural network model, and determine the first neural network model.

[0155] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0156] Acquire multiple training images; each training image includes a gold standard dose distribution for each point on the training image; use each training image as input to an initial first neural network model, use an initial flux map corresponding to each training image and a predicted dose distribution for each point on each training image as outputs of the initial first neural network model, train the initial first neural network model, and determine the first neural network model.

[0157] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0158] Calculate the first loss between the predicted dose distribution of each point on each training image and the gold standard dose distribution of the corresponding point; calculate the second loss based on the initial flux map corresponding to each training image; train the initial first neural network model based on the first loss and the second loss to determine the first neural network model.

[0159] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0160] Acquire a medical image of the object; process the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image; the flux map is used to characterize the correspondence between the intensity and shape of the beam in various directions; and perform dose calculation on the flux map using a second network branch in the first neural network model to determine a target dose distribution for each point on the medical image.

[0161] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0162] A preset second neural network model is used to predict the dose distribution of each point on the medical image to determine the first dose distribution of each point on the medical image; the second network branch in the first neural network model is used to calculate the dose of the flux map to determine the second dose distribution of each point on the medical image; based on the first dose distribution of each point on the medical image and the second dose distribution of each point on the medical image, the target dose distribution of each point on the medical image is determined.

[0163] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0164] Mathematical operations are performed on the first dose distribution and the corresponding second dose distribution of each point on the medical image to obtain the target dose distribution of each point on the medical image.

[0165] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0166] Acquire multiple training images; each training image includes a gold standard dose distribution for each point on the training image; use each training image as input to an initial first neural network model, use the predicted dose distribution for each point on each training image as output of the initial first neural network model, train the initial first neural network model, and determine the first neural network model.

[0167] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0168] Acquire multiple training images; each training image includes a gold standard dose distribution for each point on the training image; use each training image as input to an initial first neural network model, use an initial flux map corresponding to each training image and a predicted dose distribution for each point on each training image as outputs of the initial first neural network model, train the initial first neural network model, and determine the first neural network model.

[0169] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented:

[0170] Calculate the first loss between the predicted dose distribution of each point on each training image and the gold standard dose distribution of the corresponding point; calculate the second loss based on the initial flux map corresponding to each training image; train the initial first neural network model based on the first loss and the second loss to determine the first neural network model.

[0171] It should be noted that the data involved in this application (including but not limited to data used for analysis, stored data, displayed data, etc.) are all information and data authorized by the user or fully authorized by all parties.

[0172] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0173] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0174] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A dose distribution prediction method, characterized in that: The method comprises: obtaining medical images of the subject; Processing the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image, the flux map being used to represent a corresponding relationship between intensity and shape of a beam in various directions when a collimator of a radiation device emits a beam toward the object; Using a preset second neural network model to predict the dose distribution of each point on the medical image, and determining a first dose distribution of each point on the medical image; performing dose calculation on the flux map using a second network branch in the first neural network model to determine a second dose distribution for each point on the medical image; A target dose distribution for each point on the medical image is determined according to the first dose distribution for each point on the medical image and the second dose distribution for each point on the medical image.

2. The method according to claim 1, characterized in that Determining the target dose distribution for each point on the medical image according to the first dose distribution for each point on the medical image and the second dose distribution for each point on the medical image includes: Mathematical operations are performed on the first dose distribution and the corresponding second dose distribution of each point on the medical image to obtain a target dose distribution of each point on the medical image.

3. The method according to claim 1 or 2, characterized in that The training method of the first neural network model includes: Acquiring a plurality of training images, each of the training images including a gold standard dose distribution for each point on the training image; Each of the training images is used as the input of the initial first neural network model, and the predicted dose distribution of each point on each of the training images is used as the output of the initial first neural network model. The initial first neural network model is trained to determine the first neural network model.

4. The method according to any one of claims 1 or 2, characterized in that The training method of the first neural network model includes: Acquiring a plurality of training images, each of the training images including a gold standard dose distribution for each point on the training image; Each of the training images is used as the input of the initial first neural network model, and the initial flux map corresponding to each of the training images and the predicted dose distribution of each point on each of the training images are used as the output of the initial first neural network model. The initial first neural network model is trained to determine the first neural network model.

5. The method according to claim 4, characterized in that The method of using each of the training images as an input of an initial first neural network model, using an initial flux map corresponding to each of the training images and a predicted dose distribution of each point on each of the training images as outputs of the initial first neural network model, training the initial first neural network model, and determining the first neural network model includes: Calculating a first loss between the predicted dose distribution of each point on each training image and the gold standard dose distribution of the corresponding point; Calculating a second loss according to the initial flux map corresponding to each of the training images; The initial first neural network model is trained according to the first loss and the second loss to determine the first neural network model.

6. The method according to claim 2, characterized in that The performing mathematical operation on the first dose distribution and the corresponding second dose distribution of each point on the medical image to obtain the target dose distribution of each point on the medical image includes: A mean operation is performed on the first dose distribution and the corresponding second dose distribution of each point on the medical image to obtain a target dose distribution of each point on the medical image.

7. A dose distribution prediction device, characterized in that: The device comprises: an acquisition module, for acquiring medical images of an object; a processing module, configured to process the medical image using a first network branch in a preset first neural network model to determine a flux map corresponding to the medical image, the flux map being used to represent a correspondence between intensity and shape of a beam in various directions when a collimator of a radiation device emits a beam toward the object; a first dose prediction unit, configured to predict the dose distribution of each point on the medical image using a preset second neural network model, and determine a first dose distribution of each point on the medical image; a second dose prediction unit, configured to perform dose calculation on the flux map using the second network branch in the first neural network model to determine a second dose distribution for each point on the medical image; The target dose prediction unit is used to determine the target dose distribution of each point on the medical image according to the first dose distribution of each point on the medical image and the second dose distribution of each point on the medical image.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images