Joint prediction method and device for radiotherapy dose distribution and dose volume histogram

By constructing a joint prediction model including shared encoder and decoder, the problems of efficiency and accuracy in radiotherapy dose distribution and dose volume histogram prediction are solved, and more efficient and more accurate joint prediction effects are achieved.

CN114236589BActive Publication Date: 2024-12-03UNITED IMAGING RES INST OF INNOVATIVE MEDICAL EQUIP
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202111558680.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-12-17
Publication Date
2024-12-03
Estimated Expiration
2041-12-17

AI Technical Summary

Technical Problem

The prior art cannot take into account efficiency and accuracy in the prediction of radiotherapy dose distribution and dose volume histogram, resulting in errors in the prediction results.

Method used

The initial joint prediction model is constructed using shared encoder, radiotherapy dose distribution decoder and dose volume histogram decoder. Sample characteristics are obtained through encoding and decoding processing, the number of model parameters is reduced and the structure is simplified, and the histogram of radiotherapy dose distribution and dose volume is realized. joint prediction.

Benefits of technology

The combined prediction efficiency and accuracy of radiotherapy dose distribution and dose volume histogram are improved, and the generalization ability of the model is enhanced, and the prediction accuracy is higher than that of a single-task deep learning model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114236589B_ABST
    Figure CN114236589B_ABST
Patent Text Reader

Abstract

The present invention provides a method and device for joint prediction of radiotherapy dose distribution and dose volume histogram, the method comprising: constructing an initial joint prediction model, the initial joint prediction model comprising a shared encoder, a radiotherapy dose distribution decoder and a dose volume histogram decoder; obtaining a training set, and encoding the training samples in the training set based on the shared encoder to obtain sample features of the training samples; decoding the sample features based on the radiotherapy dose distribution decoder and the dose volume histogram decoder respectively, and correspondingly obtaining a predicted radiotherapy dose distribution and a predicted dose volume histogram; training the initial joint prediction model according to the training samples, the predicted radiotherapy dose distribution and the predicted dose volume histogram to obtain a target joint prediction model; and performing joint prediction on the image to be predicted based on the target joint prediction model. The present invention takes into account both the prediction efficiency and the prediction accuracy of the radiotherapy dose distribution and the dose volume histogram.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of medical radiotherapy, and in particular to a combined prediction method and device for radiotherapy dose distribution and dose volume histogram. Background Art

[0002] Radiotherapy, also known as radiotherapy, is one of the main means of cancer treatment. It has the advantages of a wide range of applications, few side effects, and low invasiveness and painlessness. It can effectively improve the cure rate of patients and prolong their lifespan. In clinical applications, professionals are required to formulate a radiotherapy plan before radiotherapy is implemented. That is, according to the patient's target area and organs at risk outlined by the doctor, combined with the prescribed dose, the angle, intensity and other information of each radiation line are optimized to obtain a treatment plan that meets the patient's needs. In order to improve the quality of radiotherapy, it is usually necessary to predict the patient's dose information before designing the radiotherapy plan, that is, predicting the patient's radiation dose to be received based on the patient's target area and organs at risk, which is called dose prediction. In order to improve the accuracy of dose prediction, it is usually necessary to predict the three-dimensional dose (Dose) distribution and dose volume histogram (Dose Volume Histogram, DVH).

[0003] There are two methods for predicting three-dimensional dose distribution and dose volume histogram in the prior art. One is to predict the three-dimensional dose distribution of radiotherapy point by point based on deep learning, that is, to predict the dose value corresponding to each voxel in the dose field and extract DVH from the dose value corresponding to each voxel. The other is to predict the three-dimensional dose distribution of radiotherapy and DVH based on deep learning respectively.

[0004] The existing technology has the following problems: the first prediction method cannot fully guarantee the accuracy of DVH extracted from the three-dimensional dose distribution, because the three-dimensional dose distribution of radiotherapy focuses more on the predicted value of each voxel point in the entire dose field, resulting in low accuracy of the predicted DVH. The second prediction method requires the establishment of two prediction models, which have a complex model structure and too many model parameters, resulting in a long prediction time. Summary of the invention

[0005] In view of this, it is necessary to provide a joint prediction method and device for radiotherapy dose distribution and dose volume histogram to solve the technical problem in the prior art that it is impossible to take into account both the prediction efficiency and prediction accuracy of radiotherapy dose distribution and dose volume histogram.

[0006] In one aspect, the present invention provides a method for jointly predicting radiotherapy dose distribution and dose volume histogram, comprising:

[0007] Constructing an initial joint prediction model, wherein the initial joint prediction model includes a shared encoder, a radiotherapy dose distribution decoder, and a dose volume histogram decoder;

[0008] Acquire a training set, and encode training samples in the training set based on the shared encoder to obtain sample features of the training samples;

[0009] Decoding the sample features based on the radiotherapy dose distribution decoder and the dose volume histogram decoder respectively, and obtaining a predicted radiotherapy dose distribution and a predicted dose volume histogram correspondingly;

[0010] Training the initial joint prediction model according to the training samples, the predicted radiotherapy dose distribution, and the predicted dose volume histogram to obtain a target joint prediction model;

[0011] Based on the target joint prediction model, the radiotherapy dose distribution and dose volume histogram are jointly predicted for the image to be predicted.

[0012] In some possible implementations, the training samples include historical images, historical radiotherapy dose distributions of the historical images, and regions of interest of the historical images; the regions of interest include organs at risk and planned target areas of the historical images.

[0013] In some possible implementations, encoding the training samples in the training set based on the shared encoder to obtain sample features of the training samples includes:

[0014] Based on the shared encoder, convolution processing and down-sampling processing are performed on the training samples to obtain sample features of the training samples.

[0015] In some possible implementations, decoding the sample features based on the radiotherapy dose distribution decoder to obtain a predicted radiotherapy dose distribution includes:

[0016] Based on the radiotherapy dose distribution decoder, upsampling, convolution and skip connection are performed on the sample features to obtain the predicted radiotherapy dose distribution;

[0017] The step of decoding the sample features based on the dose volume histogram decoder to obtain a predicted dose volume histogram includes:

[0018] Based on the dose volume histogram decoder, the sample features are flattened and fully connected to obtain a predicted dose volume histogram.

[0019] In some possible implementations, training the initial joint prediction model according to the training sample, the predicted radiotherapy dose distribution, and the predicted dose volume histogram to obtain a target joint prediction model includes:

[0020] Setting a loss function, and determining a loss value of the loss function according to the training sample, the predicted radiotherapy dose distribution, and the predicted dose volume histogram;

[0021] Adjusting the model parameters of the initial joint prediction model according to the loss value to obtain a transitional joint prediction model;

[0022] A validation set is obtained, and whether the transition joint prediction model has completed training is verified through the validation set and the loss function. When the transition joint prediction model has completed training, the transition joint prediction model that has completed training is used as the target joint prediction model.

[0023] In some possible implementations, determining the loss value of the loss function according to the training sample, the predicted radiotherapy dose distribution, and the predicted dose volume histogram includes:

[0024] Determining a first loss value between the predicted radiotherapy dose distribution and the historical radiotherapy dose distribution in the training sample;

[0025] Determining a historical dose volume histogram according to the historical radiotherapy dose distribution in the training sample and the region of interest of the historical image;

[0026] determining a second loss value between the predicted dose-volume histogram and the historical dose-volume histogram;

[0027] An estimated dose volume histogram is determined according to the predicted radiotherapy dose distribution, and a third loss value of the estimated dose volume histogram and the predicted dose volume histogram is determined; the sum of the first loss value, the second loss value and the third loss value is the loss value of the loss function.

[0028] In some possible implementations, before performing the joint prediction of the radiotherapy dose distribution and the dose volume histogram on the image to be predicted based on the target joint prediction model, the method further includes:

[0029] Obtaining a test set, and judging whether the prediction performance of the target joint prediction model is reliable by using the test set;

[0030] Then the joint prediction of the radiotherapy dose and the dose volume histogram of the image to be predicted based on the target joint prediction model includes:

[0031] When the prediction performance of the target joint prediction model is reliable, the radiotherapy dose and the dose volume histogram are jointly predicted for the image to be predicted based on the target joint prediction model.

[0032] On the other hand, the present invention also provides a combined prediction device for radiotherapy dose distribution and dose volume histogram, comprising:

[0033] A prediction model construction unit, used to construct an initial joint prediction model, wherein the initial joint prediction model includes a shared encoder, a radiotherapy dose distribution decoder, and a dose volume histogram decoder;

[0034] An encoding processing unit, used to obtain a training set, and perform encoding processing on training samples in the training set based on the shared encoder to obtain sample features of the training samples;

[0035] A decoding prediction unit, used for decoding the sample features based on the radiotherapy dose distribution decoder and the dose volume histogram decoder, respectively, to obtain a predicted radiotherapy dose distribution and a predicted dose volume histogram;

[0036] A model training unit, configured to train the initial joint prediction model according to the training samples, the predicted radiotherapy dose distribution, and the predicted dose volume histogram to obtain a target joint prediction model;

[0037] The joint prediction unit is used to perform joint prediction of the radiotherapy dose distribution and the dose volume histogram on the image to be predicted based on the target joint prediction model.

[0038] On the other hand, the present invention also provides an electronic device, including a memory and a processor, wherein:

[0039] The memory is used to store programs;

[0040] The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the method for jointly predicting radiotherapy dose distribution and dose volume histogram described in any of the above implementations.

[0041] On the other hand, the present invention also provides a computer-readable storage medium for storing computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the method for jointly predicting radiotherapy dose distribution and dose volume histogram described in any of the above-mentioned implementation methods.

[0042] The beneficial effect of adopting the above embodiment is that the joint prediction method of radiotherapy dose distribution and dose volume histogram provided by the present invention constructs an initial joint prediction model including a shared encoder, a radiotherapy dose distribution decoder and a dose volume histogram decoder, and encodes the acquired training samples based on the shared encoder to obtain the sample characteristics of the training samples. The shared encoder is used to encode the training samples, so that the initial joint prediction model shares the encoder, which can reduce the number of model parameters of the initial joint prediction model and simplify the model structure. In addition, the present invention obtains the predicted radiotherapy dose distribution and the predicted dose volume histogram based on the radiotherapy dose distribution decoder and the dose volume histogram decoder, respectively, and trains the initial joint prediction model according to the training samples, the predicted radiotherapy dose distribution and the predicted dose volume histogram to obtain the target joint prediction model to perform joint prediction on the image to be predicted. The target joint prediction model has both the radiotherapy dose distribution prediction capability of the radiotherapy dose distribution decoder and the dose volume histogram prediction capability of the dose volume histogram decoder, which can improve the prediction efficiency of the joint prediction of the radiotherapy dose distribution and the dose volume histogram. Furthermore, the target joint prediction model has stronger generalization ability than the existing single-task deep learning prediction model, thereby improving the accuracy of the target joint prediction model in jointly predicting the radiotherapy dose distribution and dose volume histogram of the predicted image. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0044] Figure 1 A schematic flow chart of an embodiment of a method for jointly predicting radiotherapy dose distribution and dose volume histogram provided by the present invention;

[0045] Figure 2 A schematic diagram of the structure of an embodiment of the initial joint prediction model provided by the present invention;

[0046] Figure 3 For the present invention Figure 1 A schematic flow chart of an embodiment of S104;

[0047] Figure 4 For the present invention Figure 3 A schematic flow chart of an embodiment of S301;

[0048] Figure 5 A schematic structural diagram of an embodiment of a device for jointly predicting radiotherapy dose distribution and dose volume histogram provided by the present invention;

[0049] Figure 6 A schematic structural diagram of an embodiment of an electronic device provided by the present invention. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0051] In the description of the embodiments of the present invention, unless otherwise specified, "multiple" means two or more than two. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" may mean: A exists alone, A and B exist at the same time, and B exists alone.

[0052] The descriptions of "first", "second", etc. involved in the embodiments of the present invention are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the technical features defined as "first" or "second" may explicitly or implicitly include at least one of the features.

[0053] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0054] The present invention provides a method and device for jointly predicting radiotherapy dose distribution and dose volume histogram, which are respectively described below.

[0055] Figure 1 A schematic flow chart of an embodiment of a method for jointly predicting radiotherapy dose distribution and dose volume histogram provided by the present invention is shown in FIG. Figure 1 As shown in FIG. 1 , the joint prediction method of radiotherapy dose distribution and dose volume histogram includes:

[0056] S101, constructing an initial joint prediction model, where the initial joint prediction model includes a shared encoder, a radiotherapy dose distribution decoder, and a dose volume histogram decoder;

[0057] S102, obtaining a training set, and encoding training samples in the training set based on a shared encoder to obtain sample features of the training samples;

[0058] S103, decoding the sample features based on the radiotherapy dose distribution decoder and the dose volume histogram decoder, respectively, to obtain a predicted radiotherapy dose distribution and a predicted dose volume histogram;

[0059] S104, training an initial joint prediction model according to the training samples, the predicted radiotherapy dose distribution, and the predicted dose volume histogram to obtain a target joint prediction model;

[0060] S105 . Perform a joint prediction of the radiotherapy dose distribution and the dose volume histogram on the image to be predicted based on the target joint prediction model.

[0061] Compared with the prior art, the joint prediction method of radiotherapy dose distribution and dose volume histogram provided by the embodiment of the present invention constructs an initial joint prediction model including a shared encoder, a radiotherapy dose distribution decoder and a dose volume histogram decoder, and encodes the acquired training samples based on the shared encoder to obtain the sample features of the training samples. The shared encoder is used to encode the training samples, so that the initial joint prediction model shares the encoder, which can reduce the number of model parameters of the initial joint prediction model and simplify the model structure. In addition, the embodiment of the present invention obtains the predicted radiotherapy dose distribution and the predicted dose volume histogram based on the radiotherapy dose distribution decoder and the dose volume histogram decoder, respectively, and trains the initial joint prediction model according to the training samples, the predicted radiotherapy dose distribution and the predicted dose volume histogram to obtain the target joint prediction model to perform joint prediction on the image to be predicted. The target joint prediction model has both the radiotherapy dose distribution prediction capability of the radiotherapy dose distribution decoder and the dose volume histogram prediction capability of the dose volume histogram decoder, which can improve the prediction efficiency of the joint prediction of the radiotherapy dose distribution and the dose volume histogram. Furthermore, the target joint prediction model has stronger generalization ability than the existing single-task deep learning prediction model, thereby improving the accuracy of the target joint prediction model in jointly predicting the radiotherapy dose distribution and dose volume histogram of the predicted image.

[0062] It should be understood that the training set in step S102 can be obtained from a data set constructed based on the radiotherapy plans already collected in advance in the hospital.

[0063] In some embodiments of the present invention, the training samples in the training set include historical images, historical radiotherapy dose distributions of historical images, and regions of interest of historical images; the regions of interest include organs-at-risk (OAR) and planning target volumes (PTV) of historical images.

[0064] In some specific embodiments, the historical image may be a CT image.

[0065] In some embodiments of the present invention, step S102 includes:

[0066] Based on the shared encoder, the training samples are convolved and downsampled to obtain sample features.

[0067] In a specific embodiment of the present invention, Figure 2 As shown, the initial joint prediction model 20 includes a shared encoder 21 , a radiotherapy dose distribution decoder 22 and a dose volume histogram decoder 23 which are jump-connected to the shared encoder 21 .

[0068] The shared encoder 21 performs convolution processing and three-time downsampling processing on the training samples in sequence to obtain sample features.

[0069] The embodiment of the present invention obtains sample features from training samples by sharing the encoder 21, without the need to manually extract sample features, thereby further improving the efficiency and accuracy of radiotherapy dose distribution and dose volume histogram prediction.

[0070] In some embodiments of the present invention, the decoding process of the sample features based on the radiotherapy dose distribution decoder 22 in step S103 to obtain the predicted radiotherapy dose distribution includes:

[0071] Based on the radiotherapy dose distribution decoder 22, upsampling, convolution and skip connection are performed on the sample features to obtain a predicted radiotherapy dose distribution;

[0072] In step S103, the sample features are decoded based on the dose volume histogram decoder 23 to obtain a predicted dose volume histogram, including:

[0073] Based on the dose volume histogram decoder 23, the sample features are flattened and fully connected to decode the sample features to obtain a predicted dose volume histogram.

[0074] Since the dose-volume histogram is obtained based on multiple sampling points in the radiotherapy dose distribution, the embodiment of the present invention can flatten the sample features based on the dose-volume histogram decoder 23, and can form the dose volumes of multiple sampling points into a one-dimensional vector, thereby realizing end-to-end prediction of the dose-volume histogram and further improving the efficiency of dose-volume histogram prediction.

[0075] In some embodiments of the present invention, Figure 3 As shown, step S104 includes:

[0076] S301, setting a loss function, and determining a loss value of the loss function according to training samples, predicted radiotherapy dose distribution, and predicted dose volume histogram;

[0077] S302, adjusting the model parameters of the initial joint prediction model according to the loss value to obtain a transitional joint prediction model;

[0078] S303, obtaining a verification set, and verifying whether the transition joint prediction model has completed training through the verification set and the loss function. When the transition joint prediction model has completed training, the trained transition joint prediction model is used as the target joint prediction model.

[0079] The validation set in step S303 may also be obtained from a data set constructed based on the hospital's existing radiotherapy plans collected in advance.

[0080] It should be understood that the model parameters in step S302 may be the convolution kernel size in the convolution process, the downsampling parameters in the downsampling process, the upsampling parameters in the upsampling process, etc., which will not be described one by one here.

[0081] In some embodiments of the present invention, the loss function in step S301 is the sum of three sub-loss functions. The three sub-loss functions are: a first sub-loss function between the predicted radiotherapy dose distribution and the historical radiotherapy dose distribution in the training sample, a second sub-loss function between the predicted dose volume histogram and the historical dose volume histogram, and a third sub-loss function between the estimated dose volume histogram determined according to the predicted radiotherapy dose distribution and the predicted dose volume histogram.

[0082] The embodiment of the present invention adds a third sub-loss function on the basis of the first sub-loss function and the second sub-loss function to jointly constitute the loss function of the initial joint prediction model. This takes into account the consistency loss of the initial joint prediction model and can further improve the prediction accuracy of the trained target joint prediction model for predicting radiotherapy dose distribution and dose volume histogram.

[0083] In a specific embodiment of the present invention, Figure 4 As shown, step S301 includes:

[0084] S401, determining a first loss value between the predicted radiotherapy dose distribution and the historical radiotherapy dose distribution in the training sample;

[0085] S402, determining a historical dose volume histogram according to the historical radiotherapy dose distribution in the training sample and the region of interest of the historical image;

[0086] S403, determining a second loss value between the predicted dose-volume histogram and the historical dose-volume histogram;

[0087] S404, determining an estimated dose volume histogram according to the predicted radiotherapy dose distribution, and determining a third loss value of the estimated dose volume histogram and the predicted dose volume histogram; the sum of the first loss value, the second loss value and the third loss value is the loss value of the loss function.

[0088] In some embodiments of the present invention, the first loss value in step S401 is the mean square error between the predicted radiotherapy dose distribution and the historical radiotherapy dose distribution. The second loss value in step S403 is the mean absolute error between the predicted dose volume histogram and the historical dose volume histogram. The third loss value in step S404 is the mean absolute error between the estimated dose volume histogram and the predicted dose volume histogram.

[0089] Specifically, the mean square error between the predicted radiotherapy dose distribution and the historical radiotherapy dose distribution is:

[0090]

[0091] Where, MSE is the mean square error between the predicted radiotherapy dose distribution and the historical radiotherapy dose distribution; n is the number of voxels in the predicted radiotherapy dose distribution or the historical radiotherapy dose distribution; To predict the radiotherapy dose of each voxel in the radiotherapy dose distribution; i is the radiotherapy dose of each voxel in the historical radiotherapy dose distribution.

[0092] The mean absolute error between the predicted dose-volume histogram and the historical dose-volume histogram is:

[0093]

[0094] Wherein, MAE is the mean absolute error between the predicted dose volume histogram and the historical dose volume histogram; m is the number of sampling points in the predicted dose volume histogram or the historical dose volume histogram; is the dose volume of each sampling point in the predicted dose volume histogram; xi is the dose volume of each sampling point in the historical dose volume histogram; || is the absolute value function.

[0095] The calculation formula of the mean absolute error between the estimated dose volume histogram and the predicted dose volume histogram is the same as the calculation formula of the mean absolute error between the predicted dose volume histogram and the historical dose volume histogram, which will not be described in detail here.

[0096] In some embodiments of the present invention, the verification sample in step S303 is the same as the training sample in step S102.

[0097] Furthermore, in step S303, the transition joint prediction model is verified by verifying the sample and the loss function to see whether the training is completed, specifically:

[0098] After the transition joint prediction model is trained with a preset number of training samples, the validation sample is used to evaluate the loss value of the loss function of the current transition joint prediction model. When the loss value of the current transition joint prediction model does not decrease compared to the loss value of the previous transition joint prediction model, the current transition joint prediction model is trained.

[0099] In order to further ensure the accuracy of the target joint prediction model in predicting the radiotherapy dose distribution and dose volume histogram of the image to be predicted, in some embodiments of the present invention, before step S105, the following is further included:

[0100] Obtain a test set and use it to determine whether the prediction performance of the target joint prediction model is reliable.

[0101] The test set includes multiple test samples, and the test samples are the same as the training samples in step S102.

[0102] Then the prediction performance of the target joint prediction model can be judged by the test set as follows:

[0103] Input the test samples in the test set into the target joint prediction model to obtain the predicted radiotherapy dose distribution and the predicted dose volume histogram;

[0104] determining a loss value based on a predicted radiotherapy dose distribution, a predicted dose volume histogram, a historical radiotherapy dose distribution, a historical dose volume histogram, and a loss function;

[0105] When the loss value is less than the threshold loss value, the prediction performance of the target joint prediction model is reliable.

[0106] It should be understood that the threshold loss value can be adjusted according to actual conditions, which will not be elaborated here.

[0107] After the reliability of the prediction performance of the target joint prediction model is judged by the test set, step S105 is specifically as follows:

[0108] When the prediction performance of the target joint prediction model is reliable, the radiotherapy dose and dose volume histogram are jointly predicted for the image to be predicted based on the target joint prediction model.

[0109] The embodiment of the present invention tests the prediction performance of the target joint prediction model through a test set, which can further ensure the prediction accuracy of the target joint prediction model for the joint prediction of the radiotherapy dose distribution and the dose volume histogram of the image to be predicted.

[0110] In order to better implement the combined prediction method of radiotherapy dose distribution and dose volume histogram in the embodiment of the present invention, based on the combined prediction method of radiotherapy dose distribution and dose volume histogram, correspondingly, Figure 5As shown, the embodiment of the present invention further provides a combined prediction device for radiotherapy dose distribution and dose volume histogram, and the combined prediction device 500 for radiotherapy dose distribution and dose volume histogram includes:

[0111] A prediction model building unit 501 is used to build an initial joint prediction model, the initial joint prediction model includes a shared encoder, a radiotherapy dose distribution decoder and a dose volume histogram decoder;

[0112] The encoding processing unit 502 is used to obtain a training set, and perform encoding processing on the training samples in the training set based on the shared encoder to obtain sample features of the training samples;

[0113] A decoding prediction unit 503 is used to decode the sample features based on the radiotherapy dose distribution decoder and the dose volume histogram decoder, respectively, to obtain a predicted radiotherapy dose distribution and a predicted dose volume histogram;

[0114] A model training unit 504 is used to train an initial joint prediction model according to the training samples, the predicted radiotherapy dose distribution and the predicted dose volume histogram to obtain a target joint prediction model;

[0115] The joint prediction unit 505 is used to perform joint prediction of the radiotherapy dose distribution and the dose volume histogram on the image to be predicted based on the target joint prediction model.

[0116] The device 500 for jointly predicting radiotherapy dose distribution and dose-volume histogram provided in the above embodiment can implement the technical solution described in the above embodiment of the method for jointly predicting radiotherapy dose distribution and dose-volume histogram. The specific implementation principles of the above modules or units can refer to the corresponding contents in the above embodiment of the method for jointly predicting radiotherapy dose distribution and dose-volume histogram, which will not be repeated here.

[0117] like Figure 6 As shown, the present invention also provides an electronic device 600. The electronic device 600 includes a processor 601, a memory 602 and a display 603. Figure 6 Only some components of the electronic device 600 are shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0118] In some embodiments, the processor 601 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run program codes or process data stored in the memory 602, such as the joint prediction method of radiotherapy dose distribution and dose volume histogram in the present invention.

[0119] In some embodiments, the processor 601 may be a single server or a server group. The server group may be centralized or distributed. In some embodiments, the processor 601 may be local or remote. In some embodiments, the processor 601 may be implemented in a cloud platform. In one embodiment, the cloud platform may include a private cloud, a public cloud, a hybrid cloud, a community cloud, a distributed cloud, an internal cloud, a multi-cloud, etc., or any combination thereof.

[0120] In some embodiments, the memory 602 may be an internal storage unit of the electronic device 600, such as a hard disk or memory of the electronic device 600. In other embodiments, the memory 602 may also be an external storage device of the electronic device 600, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device 600.

[0121] Furthermore, the memory 602 may include both an internal storage unit of the electronic device 600 and an external storage device. The memory 602 is used to store application software installed in the electronic device 600 and various data.

[0122] In some embodiments, the display 603 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 603 is used to display information on the electronic device 600 and to display a visual user interface. The components 601-603 of the electronic device 600 communicate with each other via a system bus.

[0123] In one embodiment, when the processor 601 executes the joint prediction program of the radiotherapy dose distribution and the dose volume histogram in the memory 602, the following steps may be implemented:

[0124] An initial joint prediction model is constructed, wherein the initial joint prediction model includes a shared encoder, a radiotherapy dose distribution decoder, and a dose volume histogram decoder;

[0125] A training set is obtained, and training samples in the training set are encoded based on a shared encoder to obtain sample features of the training samples;

[0126] Decoding the sample features based on a radiotherapy dose distribution decoder to obtain a predicted radiotherapy dose distribution, and decoding the sample features based on a dose volume histogram decoder to obtain a predicted dose volume histogram;

[0127] An initial joint prediction model is trained according to the training samples, the predicted radiotherapy dose distribution and the predicted dose volume histogram to obtain a target joint prediction model;

[0128] The radiotherapy dose distribution and dose volume histogram are jointly predicted for the image to be predicted based on the target joint prediction model.

[0129] It should be understood that: when the processor 601 executes the joint prediction program of radiotherapy dose distribution and dose volume histogram in the memory 602, in addition to the above functions, other functions can also be realized. For details, please refer to the description of the corresponding method embodiment above.

[0130] Furthermore, the embodiment of the present invention does not specifically limit the type of the electronic device 600 mentioned, and the electronic device 600 may be a portable electronic device such as a mobile phone, a tablet computer, a personal digital assistant (PDA), a wearable device, a laptop computer, etc. Exemplary embodiments of portable electronic devices include but are not limited to portable electronic devices equipped with IOS, Android, Microsoft or other operating systems. The above-mentioned portable electronic device may also be other portable electronic devices, such as a laptop computer with a touch-sensitive surface (e.g., a touch panel). It should also be understood that in some other embodiments of the present invention, the electronic device 600 may not be a portable electronic device, but a desktop computer with a touch-sensitive surface (e.g., a touch panel).

[0131] Accordingly, an embodiment of the present application also provides a computer-readable storage medium, which is used to store computer-readable programs or instructions. When the program or instructions are executed by a processor, it can implement the steps or functions in the joint prediction method of radiotherapy dose distribution and dose volume histogram provided in the above-mentioned method embodiments.

[0132] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the computer program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0133] The above is a detailed introduction to the combined prediction and device of radiotherapy dose distribution and dose volume histogram provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A joint prediction method for radiotherapy dose distribution and dose volume histogram, characterized in that: include: Constructing an initial joint prediction model, wherein the initial joint prediction model includes a shared encoder, a radiotherapy dose distribution decoder, and a dose volume histogram decoder; Acquire a training set, and encode training samples in the training set based on the shared encoder to obtain sample features of the training samples; Decoding the sample features based on the radiotherapy dose distribution decoder and the dose volume histogram decoder respectively, and obtaining a predicted radiotherapy dose distribution and a predicted dose volume histogram correspondingly; Training the initial joint prediction model according to the training samples, the predicted radiotherapy dose distribution, and the predicted dose volume histogram to obtain a target joint prediction model; Based on the target joint prediction model, the radiotherapy dose distribution and dose volume histogram are jointly predicted for the image to be predicted.

2. The method for joint prediction of radiotherapy dose distribution and dose volume histogram according to claim 1, characterized in that: The training samples include historical images, historical radiotherapy dose distributions of the historical images, and regions of interest of the historical images; the regions of interest include organs at risk and planned target areas of the historical images.

3. The method for joint prediction of radiotherapy dose distribution and dose volume histogram according to claim 2, characterized in that: The encoding process of the training samples in the training set based on the shared encoder to obtain sample features of the training samples includes: Based on the shared encoder, convolution processing and down-sampling processing are performed on the training samples to obtain sample features of the training samples.

4. The combined prediction method of radiotherapy dose distribution and dose volume histogram according to claim 2, characterized in that: The decoding process of the sample features based on the radiotherapy dose distribution decoder to obtain a predicted radiotherapy dose distribution includes: Based on the radiotherapy dose distribution decoder, upsampling, convolution and skip connection are performed on the sample features to obtain the predicted radiotherapy dose distribution; The step of decoding the sample features based on the dose volume histogram decoder to obtain a predicted dose volume histogram includes: Based on the dose volume histogram decoder, the sample features are flattened and fully connected to obtain a predicted dose volume histogram.

5. The method for joint prediction of radiotherapy dose distribution and dose volume histogram according to claim 2, characterized in that: The step of training the initial joint prediction model according to the training sample, the predicted radiotherapy dose distribution, and the predicted dose volume histogram to obtain a target joint prediction model includes: Setting a loss function, and determining a loss value of the loss function according to the training sample, the predicted radiotherapy dose distribution, and the predicted dose volume histogram; Adjusting the model parameters of the initial joint prediction model according to the loss value to obtain a transitional joint prediction model; A validation set is obtained, and whether the transition joint prediction model has completed training is verified through the validation set and the loss function. When the transition joint prediction model has completed training, the transition joint prediction model that has completed training is used as the target joint prediction model.

6. The method for joint prediction of radiotherapy dose distribution and dose volume histogram according to claim 5, characterized in that: Determining the loss value of the loss function according to the training sample, the predicted radiotherapy dose distribution, and the predicted dose volume histogram includes: Determining a first loss value between the predicted radiotherapy dose distribution and the historical radiotherapy dose distribution in the training sample; Determining a historical dose volume histogram according to the historical radiotherapy dose distribution in the training sample and the region of interest of the historical image; determining a second loss value between the predicted dose-volume histogram and the historical dose-volume histogram; An estimated dose volume histogram is determined according to the predicted radiotherapy dose distribution, and a third loss value of the estimated dose volume histogram and the predicted dose volume histogram is determined; the sum of the first loss value, the second loss value and the third loss value is the loss value of the loss function.

7. The combined prediction method of radiotherapy dose distribution and dose volume histogram according to claim 1, characterized in that: Before performing joint prediction of radiotherapy dose distribution and dose volume histogram on the image to be predicted based on the target joint prediction model, the method further includes: Obtaining a test set, and judging whether the prediction performance of the target joint prediction model is reliable by using the test set; Then the joint prediction of the radiotherapy dose and the dose volume histogram of the image to be predicted based on the target joint prediction model includes: When the prediction performance of the target joint prediction model is reliable, the radiotherapy dose and the dose volume histogram are jointly predicted for the image to be predicted based on the target joint prediction model.

8. A combined prediction device for radiotherapy dose distribution and dose volume histogram, characterized in that: include: A prediction model construction unit, used to construct an initial joint prediction model, wherein the initial joint prediction model includes a shared encoder, a radiotherapy dose distribution decoder, and a dose volume histogram decoder; An encoding processing unit, used to obtain a training set, and perform encoding processing on training samples in the training set based on the shared encoder to obtain sample features of the training samples; A decoding prediction unit, used for decoding the sample features based on the radiotherapy dose distribution decoder and the dose volume histogram decoder, respectively, to obtain a predicted radiotherapy dose distribution and a predicted dose volume histogram; A model training unit, configured to train the initial joint prediction model according to the training samples, the predicted radiotherapy dose distribution, and the predicted dose volume histogram to obtain a target joint prediction model; The joint prediction unit is used to perform joint prediction of the radiotherapy dose distribution and the dose volume histogram on the image to be predicted based on the target joint prediction model.

9. An electronic device, characterized in that: comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the method for jointly predicting radiotherapy dose distribution and dose volume histogram as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store computer-readable programs or instructions, which, when executed by a processor, can implement the steps in the method for jointly predicting radiotherapy dose distribution and dose volume histogram as described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • A method and a system for building a combined prediction model based on metrical structures and stresses of Chinese

    CN107464559A

  • Training method for generating face change image model and related device

    CN113240115A