Functional guide radiotherapy plan generation method and device, electronic equipment and program product

Abstract

The application is suitable for the medical technical field, and provides a functional guidance radiotherapy plan generation method and device, electronic equipment and program product, comprising: acquiring anatomical data and functional data corresponding to a target user, wherein the anatomical data at least reflects anatomical structures of a lung tumor target area and an organ at risk of the target user, and the functional data is used for reflecting a distribution condition of lung function of the target user; determining a target dose distribution according to the anatomical data and the functional data; and determining a functional guidance radiotherapy plan of the target user based on the target dose distribution. The application can reduce the probability of radiation pneumonitis and improve the radiotherapy effect.

Description

Technical Field

[0001] This application belongs to the field of medical technology, and in particular relates to methods, devices, electronic devices and program products for generating function-guided radiotherapy plans. Background Technology

[0002] Lung cancer is one of the most common malignant tumors, and radiation therapy (RT) is a common treatment for tumors, with over 77% of lung cancer patients requiring RT during their treatment. Due to factors such as the penetrability of radiation, the sensitivity of organs and tissues, and the cumulative effects of multiple treatments, while RT kills lung tumor cells or inhibits their growth, it may also cause damage to other lung tissues and even surrounding organs, leading to a series of adverse reactions such as pulmonary fibrosis or radiation pneumonitis.

[0003] In existing technologies, traditional radiotherapy plans for the lungs typically rely solely on anatomical images such as CT scans of the patient to ensure that the tumor receives a sufficient radiation dose while minimizing damage to normal tissues. However, anatomical images can only provide information about the anatomical structure and cannot reflect the functional heterogeneity of the lungs. This makes it impossible to distinguish the functional levels of normal tissues when developing a radiotherapy plan, resulting in high-functioning areas receiving unnecessary high-dose radiation and affecting the treatment outcome. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and program product for generating function-guided radiotherapy plans, which can reduce the probability of radiation pneumonitis and improve the effectiveness of radiotherapy.

[0005] In a first aspect, embodiments of this application provide a method for generating a function-guided radiotherapy plan, including:

[0006] Obtain anatomical and functional data corresponding to the target user. The anatomical data at least reflects the anatomical structure of the target user's lung tumor target area and organs at risk, and the functional data is used to reflect the distribution of the target user's lung function.

[0007] The target dose distribution is determined based on the anatomical data and the functional data;

[0008] The function-guided radiotherapy plan for the target user is determined based on the target dose distribution.

[0009] Secondly, embodiments of this application provide a function-guided radiotherapy planning device, comprising:

[0010] The data acquisition module is used to acquire anatomical data and functional data corresponding to the target user. The anatomical data at least reflects the anatomical structure of the target user's lung tumor target area and organs at risk, and the functional data is used to reflect the distribution of the target user's lung function.

[0011] A dose distribution prediction module is used to determine a target dose distribution based on the anatomical data and the functional data;

[0012] The automatic radiotherapy planning module is used to determine the functionally guided radiotherapy plan for the target user based on the target dose distribution.

[0013] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the function-guided radiotherapy planning generation method described in the first aspect.

[0014] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the function-guided radiotherapy plan generation method described in the first aspect.

[0015] Fifthly, embodiments of this application provide a computer program product that, when run on an electronic device, causes the electronic device to perform the function-guided radiotherapy plan generation method described in the first aspect.

[0016] The beneficial effects of the embodiments in this application compared with the prior art are:

[0017] In this embodiment, the target dose distribution is determined based on both anatomical and functional data. Compared to radiotherapy plans that rely solely on anatomical images, this approach integrates information on the heterogeneous distribution of lung function provided by pulmonary function images. This allows for the protection of high-functioning areas as much as possible when determining the target dose distribution, reducing lung function damage caused by radiotherapy and improving radiotherapy efficacy. Furthermore, the automatic determination of the target dose distribution and generation of function-guided radiotherapy plans based on anatomical and functional data eliminates the need for manual analysis and planning, effectively improving the efficiency, consistency, and accuracy of function-guided radiotherapy planning. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0019] Figure 1 This is a flowchart illustrating a function-guided radiotherapy plan generation method according to an embodiment of this application;

[0020] Figure 2 This is a structural diagram of a prediction model based on function-guided dose distribution provided in an embodiment of this application;

[0021] Figure 3 This is a structural diagram of another prediction model based on function-guided dose distribution provided in an embodiment of this application;

[0022] Figure 4 This is a schematic diagram of the structure of the function-guided radiotherapy planning generation device provided in the embodiments of this application;

[0023] Figure 5 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0024] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0025] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0026] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0027] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0028] References to "one embodiment" or "some embodiments" in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized.

[0029] Example 1:

[0030] Treatment planning is a crucial step in the implementation of radiotherapy. Its core is developing a plan that guides the radiotherapy machine to deliver the intended dose distribution. Conventional radiotherapy planning aims to ensure sufficient and uniform dose to the tumor target area while minimizing the dose to normal tissues. Therefore, current radiotherapy planning primarily relies on anatomical images to delineate the tumor target area and normal tissues. Parameters in the radiotherapy plan are then manually and repeatedly adjusted to ensure the tumor target area receives a sufficient and uniform dose. This process is time-consuming and labor-intensive, and is influenced by experience and subjective factors, potentially leading to inconsistencies and low accuracy in the resulting plans.

[0031] In view of this, embodiments of this application provide a function-guided radiotherapy plan generation method, which acquires anatomical data and functional data corresponding to a target user. The anatomical data at least reflects the anatomical structure of the target user's lung tumor target area and organs at risk, and the functional data reflects the distribution of the target user's lung function. Then, a target dose distribution is determined based on the anatomical data and the functional data; and a function-guided radiotherapy plan for the target user is generated based on the target dose distribution.

[0032] The function-guided radiotherapy planning method provided in this application incorporates lung function limitations. By combining anatomical data reflecting the lung tumor target area and organs at risk, with functional data reflecting the actual distribution of lung function, the method can redistribute the dose from high-functioning lung regions to low-functioning lung regions while meeting traditional dose distribution requirements, thus obtaining a function-guided target dose distribution. Furthermore, a function-guided radiotherapy plan is generated based on this target dose distribution. This ensures that the tumor target area receives a sufficient and uniform dose while better protecting the lung function of the target user, effectively reducing radiation-induced lung injury and improving radiotherapy efficacy. Moreover, it is unaffected by human factors, effectively guaranteeing the consistency and accuracy of the function-guided radiotherapy plan.

[0033] Figure 1 A flowchart illustrating a function-guided radiotherapy plan generation method according to an embodiment of this application is shown below in detail:

[0034] S101, Obtain the anatomical data and functional data corresponding to the target user. The anatomical data at least reflects the anatomical structure of the target user's lung tumor target area and organs at risk, and the functional data is used to reflect the target user's lung function.

[0035] It should be understood that the target users are usually patients who are about to undergo radiotherapy.

[0036] It should be understood that the anatomical data at least reflects the tumor target volume (also known as the planning target volume, PTV) of the target user's organ to be treated, as well as the anatomical structure of organs at risk. In this embodiment, the lung is used as an example of the organ to be treated; therefore, the anatomical data at least reflects the lung tumor target volume and the anatomical structure of organs at risk. Organs at risk typically refer to normal tissues or organs that may be involved within the radiation field, i.e., organs or tissues surrounding the tumor target volume that need to be protected.

[0037] In other embodiments, the anatomical data may also reflect the anatomical structure of other organs or tissues that are associated with the organ to be treated.

[0038] It should be understood that anatomical structures can include the outline, size, and location of organs or tissues. Optionally, anatomical data can be represented as one or more of the following: images, text, and 3D models, depending on the specific application requirements.

[0039] It should be understood that functional data at least reflects the functional distribution of the target user's organs to be treated. In the embodiments of this application, the functional data is used to reflect the distribution of the target user's lung function, such as the distribution of ventilation function. It should be understood that functional data can be represented as one or more of the following: images, text, and graphic-text data, and can be specifically set according to actual application requirements.

[0040] As an example, functional data can be represented in the form of images, i.e., functional data can be functional images, including but not limited to SPECT (Single-Photon Emission Computed Tomography) ventilation images, SPECT perfusion images, MRI (Magnetic Resonance Imaging) ventilation images, MRI perfusion images, PET (Positron Emission Tomography) ventilation images, and PET perfusion images.

[0041] In some embodiments, anatomical data may include CT images, target planning maps, and organ at risk contour maps, while functional data may include pulmonary function images (i.e., images that directly reflect lung function) and lung function contour maps (i.e., structured images obtained based on pulmonary function images).

[0042] Optionally, after acquiring CT images and pulmonary function images, the pulmonary function images can be preprocessed by registration and resampling to align the position and resolution of the preprocessed pulmonary function images with those of the CT images, so that pulmonary function information can be provided more intuitively and accurately in the future.

[0043] It should be noted that this application uses lung radiotherapy as an example. In other embodiments, when the above-mentioned radiotherapy planning method is applied to radiotherapy scenarios for other organs, the anatomical data is used to reflect at least the anatomical structure of the tumor target area corresponding to the organ to be treated and its organs at risk (usually the organ where the tumor is located). The functional data can be used to reflect the functional distribution of the organ to be protected. The organ to be protected can be the organ to be treated or an organ associated with the organ to be treated, such as an organ adjacent to the organ to be treated or an organ at risk. For example, when the organ to be protected is the liver, the functional data can be used to reflect the spatial heterogeneity distribution information of liver function, etc.

[0044] S102, Determine the target dose distribution based on the above anatomical data and the above functional data.

[0045] Dose distribution usually refers to the spatial distribution of radiation energy within a patient's body structures (also known as tissues, such as the lungs), and can generally be described as the radiation dose per unit volume.

[0046] Since radiation energy may damage the lung tissue of patients, easily leading to radiation-induced lung injury and affecting the function of the lungs, in order to reduce the probability of radiation-induced lung injury, the required target dose distribution can be obtained by comprehensively analyzing anatomical and functional data when determining the dose distribution for the target patient.

[0047] It should be understood that when determining the target dose distribution by combining information from two different modalities—anatomical data that at least reflects the anatomical structure of the tumor target area and functional data that reflects lung function—automatic analysis can be performed by combining the treatment goals of treating tumor tissue and prioritizing the protection of high-functioning lung tissue. That is, when determining the target dose distribution, the actual status of the patient's current lung function should be fully considered, and high-functioning areas of the lungs should be avoided as much as possible, so that radiation energy is concentrated in the tumor tissue area (i.e., the tumor target area) and low-functioning areas. This can minimize the damage to the patient's lung function during subsequent radiotherapy based on the target dose distribution, and thus reduce the probability of radiation-induced lung injury.

[0048] S103, Based on the above target dose distribution, determine the function-guided radiotherapy plan for the above target user.

[0049] It should be understood that a radiotherapy plan can be used to guide a radiotherapy machine to perform radiotherapy. For example, a radiotherapy plan can be executable machine parameters of the radiotherapy machine, such as the position, trajectory, velocity, dose rate (which can also be expressed as beam intensity), and subfield weights of the multileaf collimator (MLC).

[0050] In this embodiment, since anatomical data can reflect the anatomical structure of the tumor target area and organs at risk of the target user, that is, the distribution of the tumor target area and surrounding organs to be protected, and functional data can reflect the actual distribution of the lung function of the target user, the distribution of the tumor target area, organs at risk, and lung function in the user's lungs can be fully understood based on these data. When determining the target dose distribution, tissues with good lung function can be avoided as much as possible, that is, high-function areas can be avoided, and the dose distribution can be concentrated on the tumor area of ​​interest and low-function areas. This can minimize the radiation dose received by high-function tissues and reduce the damage to lung function, resulting in a target dose distribution that is beneficial to lung function protection. Furthermore, a function-guided radiotherapy plan corresponding to the target user can be automatically generated based on the target dose distribution. This can effectively improve the efficiency and accuracy of function-guided radiotherapy plan formulation while reducing the probability of radiation-induced lung injury in the target user and improving the radiotherapy effect.

[0051] In some embodiments, step S102 includes:

[0052] The aforementioned anatomical data and functional data are used as inputs to a trained prediction model to obtain the aforementioned target dose distribution output by the prediction model. The prediction model is used to analyze the dose distribution that matches the lung function of the aforementioned target user based on the aforementioned anatomical data and functional data to obtain the aforementioned target dose distribution.

[0053] It should be understood that the prediction model can be a model built based on network structures such as convolutional neural networks, U-Net, generative adversarial networks (GAN), graph neural networks (GNN), or Transformer. The embodiments of this application do not impose specific restrictions on the network structure of the prediction model.

[0054] As an example, the trained predictive model can be a model fine-tuned from a general large model using constructed training samples (typically including user-specific functional data, anatomical data, and actual functionally guided dose distribution). Since the large model is a deep learning model with massive parameters (usually billions or even hundreds of billions), fine-tuning this general large model using training samples allows it to learn the relationship between different patients' anatomical structures, lung function, and functionally guided dose distribution. Therefore, the fine-tuned large model can better predict functionally guided dose distributions that match lung function based on anatomical and functional data, improving the accuracy of the predicted target dose distribution.

[0055] In this embodiment, since the model can learn the complex relationship between anatomical and functional data of different modalities, the trained prediction model can more accurately predict the dose distribution that matches the actual lung function of the target user based on the anatomical and functional data, thereby improving the accuracy of the obtained target dose distribution and significantly improving the efficiency of determining the target dose distribution and reducing human error.

[0056] In some embodiments, the prediction model includes an anatomical coding network, a functional coding network, a fusion network, and a decoding network. The method of using the anatomical data and the functional data as input to a trained prediction model to obtain the target dose distribution output by the prediction model includes:

[0057] The anatomical data is encoded using the aforementioned anatomical coding network to obtain anatomical features, and the functional data is encoded using the aforementioned functional coding network to obtain functional features.

[0058] The aforementioned anatomical features and functional features are fused together using the fusion network described above to obtain fused features;

[0059] The target dose distribution is obtained by predicting the dose distribution that matches the lung function based on the above-mentioned fusion features using the above-mentioned decoding network.

[0060] Specifically, considering that anatomical and functional data reflect different modal information, to improve feature accuracy, different encoders can be used to encode the anatomical and functional data separately (also known as feature extraction processing) to obtain the anatomical features corresponding to the anatomical data and the functional features corresponding to the functional data. Since anatomical and functional encoding networks can focus on extracting features from their corresponding modal data, they can better extract and retain the unique information of each modality, which helps to provide richer feature representations for subsequent feature fusion and prediction of target dose distribution. Furthermore, the prediction model can adapt well to changes in different modal data, helping to improve the generalization ability of the prediction model on data from different patients.

[0061] Optionally, the fusion network can fuse anatomical and functional features based on attention mechanisms (such as channel attention or spatial attention). Since attention mechanisms enable the fusion network to focus more on features helpful for dose prediction during feature fusion, attention-based fusion networks can improve the accuracy of feature representation, resulting in more accurate fused features. This helps the subsequent guided decoding network to more precisely adjust the function-guided dose distribution, thereby improving the accuracy of target dose distribution prediction.

[0062] In some embodiments, the fusion network can fuse anatomical and functional features based on channel attention and spatial attention. Since channel attention can focus on the channel dimension of a feature, highlighting the channels relevant to the dose prediction task, and spatial attention can focus on the spatial dimension of a feature, highlighting the region that contributes the most to the dose prediction task, fusing anatomical and functional features based on channel attention and spatial attention can enhance feature representation from different dimensions and extract complementary feature information from anatomical and functional features of different modalities for fusion, further enhancing the fused features and enabling the prediction model to better learn and understand the complex relationship between anatomical and functional features.

[0063] Figure 2 A structural diagram of the prediction model in some embodiments is shown.

[0064] refer to Figure 2 The prediction model can include n fusion networks (n is greater than 1, for example) Figure 2 The prediction model includes three fusion networks. The anatomy coding network and the function coding network can each contain m convolutional layers (m is greater than 1, and m is usually greater than or equal to n, such as n). Figure 2 The prediction model includes three convolutional layers, and different fusion networks are used to fuse the anatomical and functional features extracted from different convolutional layers.

[0065] Figure 3 Structural diagrams of prediction models in other embodiments are shown.

[0066] refer to Figure 3 The prediction model can include n fusion networks (n is greater than 1, for example) Figure 3 The prediction model includes 5 fusion networks, and the anatomy coding network and function coding network can each include m convolutional layers (m is greater than 1, e.g., ...). Figure 3 The prediction model consists of 5 convolutional layers and m-1 downsampling layers. Different fusion networks are used to fuse anatomical and functional features extracted from different convolutional layers at different scales. That is, through the above settings, features at different scales obtained from each convolutional layer of the anatomical coding network and the functional coding network, as well as fused features at different scales obtained from different fusion networks, can all be used as inputs to the decoding network, enabling the decoding network to make the final dose distribution prediction based on features at different scales.

[0067] Since different convolutional layers in anatomical coding networks and functional coding networks can extract features at different scales, fusing the anatomical and functional features extracted by different convolutional layers through different fusion networks can better capture multi-scale information, further enhance feature representation, and help improve the overall performance of the prediction model.

[0068] In this embodiment, different encoding networks are used to encode data from different modalities, enabling the encoding networks to focus on extracting feature information from a single modality, thus improving the accuracy of the obtained features. Furthermore, the anatomical and functional features from different modalities are fused, effectively utilizing the complementarity between anatomical and functional features to obtain multimodal fused features. Based on these fused features, dose distributions that match the anatomical structure distribution and lung function distribution can be predicted more accurately, significantly enhancing the prediction accuracy and robustness of the prediction model. Simultaneously, it effectively improves the clinical usability and accuracy of the predicted dose distribution.

[0069] In some embodiments, the above prediction model is trained through the following steps:

[0070] Construct training samples, each of which includes anatomical and functional data corresponding to a user.

[0071] Using the training samples mentioned above as input to the prediction model to be trained, the predicted dose distribution output by the prediction model to be trained is obtained.

[0072] The loss value is determined based on the predicted dose distribution corresponding to the training samples and the actual functional guidance plan dose distribution.

[0073] The parameters of the prediction model to be trained are updated based on the above loss value to obtain the trained prediction model.

[0074] It should be understood that the actual functional guided plan dose distribution corresponding to the training sample usually refers to the label value of the training sample. This actual functional guided plan dose distribution can be manually formulated by experienced doctors or automatically generated by intelligent algorithms such as large models. The specific settings can be made according to the actual application requirements.

[0075] It should be understood that when constructing training samples, a certain number (e.g., 1000) of training samples are typically built to ensure the accuracy of the trained prediction model. Optionally, after constructing the training samples, a training sample set, a validation sample set, and a test sample set can be determined based on the constructed training samples. The training sample set is used to train the prediction model to be trained; the validation sample set can be used to evaluate the performance of the prediction model during training, helping to select hyperparameters and avoid overfitting; the test sample set can be used to evaluate the performance (e.g., generalization ability) of the finally trained model. By reasonably dividing the training sample set, validation sample set, and test sample set, the prediction model can be effectively evaluated and optimized, ensuring the predictive performance of the prediction model on new data.

[0076] Optionally, when updating the parameters of the prediction model to be trained based on the loss value, the parameters of the prediction model can be updated using one or more optimization algorithms such as gradient descent, momentum, or adaptive learning rate. The specific settings can be made according to the actual application scenario.

[0077] In some embodiments, after updating the parameters of the prediction model to be trained according to the aforementioned loss value, if the updated prediction model does not meet the set training requirements, the step of using training samples as input to the prediction model to be trained and updating the parameters of the prediction model to be trained according to the aforementioned loss value can be repeatedly executed based on the updated prediction model until the latest updated prediction model meets the training requirements, thus obtaining a trained prediction model. Training requirements may include one or more of the following: iteration count requirements (e.g., the number of iterations is greater than a threshold, such as 100) and performance requirements (e.g., the prediction error is less than a set error threshold, such as 0.1).

[0078] In this embodiment, by training the prediction model, the prediction model learns the complex relationship between anatomical data, functional data and dose distribution, thereby learning to predict the impact of the anatomical structure of the tumor target area and the distribution of lung function on the actual functional guidance plan dose distribution. When predicting the functional guidance dose distribution, the model can fully consider the correlation between lung function and the dose distribution, ensuring the prediction accuracy of the prediction model.

[0079] In some embodiments, determining the loss value based on the predicted dose distribution and the actual functionally guided dose distribution corresponding to the training samples includes:

[0080] The loss value corresponding to the training sample is determined based on the predicted dose distribution, the actual functional guidance plan dose distribution, and the set functional weights of the target voxels. The functional weights are determined based on the functional data in the training sample.

[0081] A voxel, also known as a volume pixel, is the smallest unit in three-dimensional space and is commonly used to describe density, signal intensity, or other physical quantities in medical images. In radiotherapy, voxels are often used to discretize dose distribution; that is, dose distribution is usually calculated in units of voxels, and the dose value of each voxel typically represents the radiation dose received at that location.

[0082] Optionally, the target voxel can be a voxel determined based on the tumor target area; for example, the target voxel can be the voxel corresponding to the tumor target area and the organs at risk. Alternatively, the target voxel can be a voxel determined based on the lung function distribution reflected by functional data; for example, the target voxel can be the voxel corresponding to the high-function region. Or, the target voxel can also be a voxel determined based on the tumor target area and functional distribution; the specific settings can be configured according to actual application requirements.

[0083] When calculating the loss value for the training samples, the prediction error between the predicted and actual dose values ​​of the target voxels can be calculated first based on the predicted dose distribution and the actual function-guided plan dose distribution. Specifically, for the prediction error value corresponding to a target voxel, it can be weighted according to the functional weight of the target voxel, and the weighted prediction error value can be used as the loss value for that target voxel. For non-target voxels (if any), the calculated prediction error value can be directly used as the loss value for that voxel. Then, the loss value for the training samples is calculated based on the loss values ​​for each voxel.

[0084] Since the functional weight of the target voxel is determined based on the distribution of lung function, the above processing can amplify or reduce the dose difference value of the target voxel in combination with the actual lung function corresponding to the target voxel. This can improve the prediction accuracy of high-function areas, thereby better protecting high-function tissues while achieving better treatment of tumor tissues, improving the effectiveness and safety of radiotherapy, and thus improving the radiotherapy effect.

[0085] Optionally, the set loss function may include an error loss function and a dose-volume-based loss function.

[0086] As an example, error loss functions can include the Mean Absolute Error (MAE) loss function, and dose volume-based loss functions can include the Dose Volume Histogram (DVH) loss function. The MAE loss function is typically used to measure the difference between the predicted and actual dose values, yielding an error loss value. This voxel-level error loss ensures that the predicted dose value for each target voxel is as close as possible to the planned actual dose value. The DVH loss function is typically used to measure the difference between the predicted and actual dose volume histograms, yielding a dose volume histogram error loss value. This histogram loss function ensures that the key dose volume parameters of the predicted dose distribution are close to the actual dose volume parameters, optimizing the overall dose distribution characteristics and effectively improving the accuracy of the predicted dose distribution.

[0087] As an example, the MAE loss function can be expressed in the following form:

[0088]

[0089] Among them, L MAE n represents the error loss value. total n represents the total number of all voxels. lungThis indicates the number of target voxels (in this embodiment, the target voxels are voxels corresponding to the lung structure, also known as lung voxels). This represents the predicted dose value of the i-th voxel. f represents the actual dose value of the i-th voxel. j This represents the functional weight of the j-th lung voxel. This represents the predicted dose value of the j-th lung voxel. This represents the actual dose value of the j-th lung voxel. || indicates taking the absolute value, and ∑ represents the accumulation operator.

[0090] It should be understood that, in some other embodiments, the target voxel may also include voxels corresponding to organs and tissues other than the lungs.

[0091] As an example, the dose-volume histogram loss function can calculate the dose-volume parameter loss value based on the dose-volume histogram. Alternatively, the dose-volume histogram can be approximated by a series of moments of different orders, and the dose-volume loss value can be obtained based on the error between the predicted moments and the actual moments.

[0092] As an example, the dose-volume histogram loss function can be expressed in the following form:

[0093]

[0094] L DVH The dose-volume histogram loss value is represented by p, where p represents the order, s represents the s-th structure (i.e., organ, such as lung, heart, and tumor target area) in the set of tissues and organs involved in the radiotherapy planning, and |||2 represents the L2 norm (also known as the second norm). This represents the prediction moment for the s-th organ. Let represent the actual moment of the s-th organ.

[0095] It should be understood that the structure of the organ and tissue set S involved in the radiotherapy planning design can be manually set or automatically generated based on empirical data, and the embodiments of this application do not impose specific limitations on this.

[0096] Alternatively, a moment can be represented in the following form:

[0097]

[0098] in, Let n represent the moment of order p of the s-th structure. s D represents the number of voxels in the s-th structure. i denoted by , where represents the dose value of the i-th voxel in the s-th structure, and p represents the order.

[0099] Optionally, for parallel organs to be protected (including the esophagus, heart, and lungs, etc.), p = (1,2) can be set to approximate the average dose; for tandem organs to be protected (such as the spinal cord), p = (5,10) can be set to approximate the maximum dose; and for representing the planned target area, p = (2,4,6) can be set.

[0100] In some embodiments, the dose-volume-based loss function may further include a dose-function histogram loss function. Optionally, the dose-function histogram loss function may be expressed in the following form:

[0101]

[0102] L DFH This represents the dose-function histogram loss value. This represents the predictive moments based on lung function (equivalent to a predictive dose-function histogram). This represents the actual moments based on lung function (equivalent to the actual dose-function histogram).

[0103] Alternatively, the moments based on lung function (i.e., the dose-function histogram) can be represented in the following form:

[0104]

[0105] The moments (i.e., dose-function histograms) based on lung function, n lung f represents the number of lung voxels. i This represents the functional weight of the i-th lung voxel.

[0106] in, A predictive dose-function histogram representing lung structure. A dose-function histogram representing the actual dose to lung structure.

[0107] Optionally, the loss value corresponding to the training sample obtained in the final calculation can be expressed in the following form:

[0108] L total =L MAE +w DVH (L DVH +L DFH )

[0109] Among them, L total L represents the total loss value corresponding to the training samples. MAE L represents the mean absolute error loss value. DVH L represents the dose-volume histogram loss value. DFH The dose-function histogram loss value, w DVHThis represents the relative weights of the dose-function histogram and the dose-volume histogram. Optionally, w DVH It can take the value 0.01.

[0110] In some embodiments, step S103 includes:

[0111] The function-guided dose simulation algorithm generates the function-guided radiotherapy plan based on the target dose distribution. The function-guided dose simulation algorithm is used to convert the target dose distribution into a radiotherapy plan that conforms to a set optimization function. The optimization function includes at least a target term based on lung function.

[0112] It should be understood that, to ensure the accuracy of the generated function-guided radiotherapy plan, the aforementioned optimization function (also known as the objective function) can be used to guide the generation of a function-guided radiotherapy plan that meets the target dose distribution, clinical needs, and the physical limitations of the radiotherapy machine. The generation process of the function-guided radiotherapy plan is achieved by minimizing the difference between the calculated dose distribution and the target dose distribution as the optimization objective. That is, the set optimization function can be determined based on this optimization problem.

[0113] In some embodiments, the objective and constraints of this optimization problem (i.e., the optimization function mentioned above) are expressed by the following equation:

[0114]

[0115] stx≥0

[0116]

[0117] Where x represents the flux diagram, and the physical limitations of the radiotherapy machine require x to be greater than or equal to 0; This represents the optimization objective of minimizing the difference between the planned dose distribution and the predicted target dose distribution by optimizing the flux map.

[0118] w PTV The weight of the tumor target region, n PTV Indicates the number of voxels in the tumor target area. s n represents the weight of the s-th structure in the set of organs and tissues S. s This represents the number of voxels of the s-th structure in the organ and tissue set S.

[0119] The calculated dose value of the k-th voxel in the tumor target region (determined based on the planned dose distribution) represents the total dose. The predicted dose value for the k-th voxel in the tumor target region (determined based on the target dose distribution). w s The weight of the s-th structure in the organ and tissue set S is represented by n. sThis represents the number of voxels in the s-th structure within the organ / tissue set S. lung w represents the number of target voxels (in this embodiment, the target voxels are voxels corresponding to the lung structure). F The weight of the structure corresponding to the target voxel (which can be used to reflect the clinical priority of the structure).

[0120] This represents the calculated dose value of the i-th voxel. f represents the predicted dose value of the i-th voxel. i represents the functional weight of the i-th voxel. D represents the dose matrix, which is typically used to describe the dose contribution of each beamlet to each voxel.

[0121] It should be understood that the functional weight of non-target voxels (i.e., voxels other than the target voxels) can be set to 1. In the embodiments of this application, the target voxels include voxels corresponding to lung structures, so the functional weight of lung voxels can be determined based on the lung function distribution reflected by functional data, and the functional weight of non-lung voxels can be regarded as 1.

[0122] Optionally, the weights of the planned target area (i.e., the tumor target area), lungs, spinal cord, esophagus, heart, and normal tissues are 800, 800, 600, 400, 400, and 200, respectively.

[0123] In this embodiment, considering the need to improve actual treatment efficacy, after determining the target dose distribution, the determined target dose distribution can be used as an optimization guide to solve the optimization function corresponding to the above optimization problem, thereby generating a function-guided radiotherapy plan that conforms to the target dose distribution, clinical needs, and the physical limitations of the radiotherapy machine. Since the optimization function includes at least a target term based on lung function—that is, the guided dose based on the actual lung function of the target user shifts from high-function regions to low-function regions—the function-guided dose simulation algorithm can prioritize the protection of high-function regions when converting the target dose distribution into a function-guided radiotherapy plan. This process can automatically generate a function-guided radiotherapy plan based on the actual distribution of lung function, improving the efficiency, consistency, and accuracy of function-guided radiotherapy plan development, which is beneficial for improving the effectiveness of subsequent radiotherapy.

[0124] In some embodiments, the above-mentioned generation of the radiotherapy plan based on the target dose distribution using the function-guided dose simulation algorithm includes:

[0125] Based on the target dose distribution described above, the optimization function is solved to obtain the target flux map;

[0126] The target flux map is converted into machine parameters of the radiotherapy machine based on the multileaf grating sequence optimization algorithm and the direct subfield optimization algorithm, thus obtaining the above-mentioned function-guided radiotherapy plan.

[0127] Multileaf grating sequence optimization algorithms are typically used to convert flux maps into specific leaf positions and motion sequences of multileaf gratings. The leaf positions and motion sequences of the multileaf grating determine the shape and intensity distribution of the beam, thereby determining the actual dose distribution.

[0128] Optionally, the multi-leaf grating sequence optimization algorithm includes, but is not limited to, sequence optimization algorithms based on static segmentation or sequence optimization algorithms based on dynamic segmentation.

[0129] Direct subfield optimization algorithms are typically used to directly optimize the shape and intensity of a beam to achieve an ideal dose distribution. Optionally, direct subfield optimization algorithms include, but are not limited to, optimization algorithms such as Interior Point OPTimize (IPTOT) or genetic algorithms.

[0130] In this embodiment, the optimization function can be a flux map-based function. When generating a radiotherapy plan based on the predicted target dose distribution using a function-guided dose simulation algorithm, the optimal flux map can be obtained by first solving the optimization function corresponding to the optimization problem, thus obtaining the target flux map. After determining the optimal target flux map, machine parameters (let's call them initial machine parameters) that can be used to guide radiotherapy implementation can be determined using a multileaf grating sequence optimization algorithm. This ensures that during actual treatment, a flux map matching the calculated optimal flux map can be generated based on these machine parameters, which helps improve the radiotherapy effect. To further improve the radiotherapy effect, after determining the initial machine parameters using the multileaf grating sequence optimization algorithm, the obtained initial machine parameters can be further optimized using a direct subfield optimization algorithm to obtain the final machine parameters.

[0131] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0132] Example 2:

[0133] Corresponding to the radiotherapy plan generation method described in the above embodiments, Figure 4 A structural block diagram of the function-guided radiotherapy planning device provided in the embodiments of this application is shown. For ease of explanation, only the parts related to the embodiments of this application are shown.

[0134] Reference Figure 4 The device includes: a data acquisition module 41, a dose distribution prediction module 42, and an automatic radiotherapy plan generation module 43. Among them,

[0135] The data acquisition module 41 is used to acquire the anatomical data and functional data corresponding to the target user. The anatomical data at least reflects the anatomical structure of the target user's lung tumor target area and organs at risk, and the functional data is used to reflect the distribution of the target user's lung function.

[0136] The dose distribution prediction module 42 is used to determine the target dose distribution based on the above-mentioned anatomical data and functional data.

[0137] The radiotherapy plan automatic generation module 43 is used to determine the target user's radiotherapy plan based on the above-mentioned target dose distribution.

[0138] In this embodiment, since anatomical data can reflect the anatomical structure of the tumor target area and organs at risk of the target user, that is, the distribution of the tumor target area and surrounding organs to be protected, and functional data can reflect the actual distribution of the lung function of the target user, the distribution of the tumor target area, organs at risk, and lung function in the user's lungs can be fully understood based on these data. When determining the target dose distribution, tissues with good lung function can be avoided as much as possible, that is, high-function areas can be avoided, and the dose distribution can be concentrated on the tumor area of ​​interest and low-function areas. This can minimize the radiation dose received by high-function tissues and reduce the damage to lung function, resulting in a target dose distribution that is beneficial to lung function protection. Furthermore, a function-guided radiotherapy plan corresponding to the target user can be automatically generated based on the target dose distribution. This can effectively improve the efficiency and accuracy of function-guided radiotherapy plan formulation while reducing the probability of radiation-induced lung injury in the target user and improving the radiotherapy effect.

[0139] In some embodiments, the dose distribution prediction module 42 includes:

[0140] The prediction unit is used to take the above-mentioned anatomical data and functional data as input to the trained prediction model to obtain the above-mentioned target dose distribution output by the prediction model. The prediction model is used to analyze the dose distribution that matches the lung function of the above-mentioned target user based on the above-mentioned anatomical data and functional data to obtain the above-mentioned target dose distribution.

[0141] In some embodiments, the prediction model includes an anatomical coding network, a functional coding network, a fusion network, and a decoding network, and the dose distribution prediction module 42 further includes:

[0142] The encoding unit is used to encode the anatomical data through the aforementioned anatomical encoding network to obtain anatomical features, and to encode the aforementioned functional data through the aforementioned functional encoding network to obtain functional features.

[0143] The fusion unit is used to fuse the anatomical features and functional features through the fusion network to obtain fused features.

[0144] The decoding unit is used to predict the dose distribution that matches the lung function based on the fusion features through the decoding network, and obtain the target dose distribution.

[0145] In some embodiments, the function-guided radiotherapy planning device further includes:

[0146] The sample construction module is used to construct training samples. Each training sample includes anatomical data and functional data corresponding to a user.

[0147] The prediction module is used to take the above training samples as input to the prediction model to be trained, and obtain the predicted dose distribution output by the prediction model to be trained.

[0148] The loss calculation module is used to determine the loss value based on the predicted dose distribution corresponding to the training samples and the actual functional guidance plan dose distribution.

[0149] The update module is used to update the parameters of the prediction model to be trained according to the above loss value, so as to obtain the trained prediction model.

[0150] In some embodiments, the function-guided radiotherapy planning device further includes:

[0151] The function loss calculation module is used to determine the loss value corresponding to the training sample based on the predicted dose distribution, the actual dose distribution, and the set functional weights of the target voxels. The functional weights are determined based on the functional data in the training sample.

[0152] In some embodiments, the above-mentioned radiotherapy planning automatic generation module 43 includes:

[0153] The function-guided dose simulation unit is used to generate the function-guided radiotherapy plan based on the target dose distribution using a function-guided dose simulation algorithm. The function-guided dose simulation algorithm is used to convert the target dose distribution into a radiotherapy plan that conforms to a set optimization function. The optimization function includes at least a target term based on lung function.

[0154] In some embodiments, the above-mentioned radiotherapy planning automatic generation module 43 further includes:

[0155] The target flux map calculation unit is used to solve the above optimization function based on the above target dose distribution to obtain the target flux map.

[0156] The optimization unit is used to convert the target flux map into machine parameters of the radiotherapy machine based on the multileaf grating sequence optimization algorithm and the direct subfield optimization algorithm, so as to obtain the above-mentioned function-guided radiotherapy plan.

[0157] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0158] Example 3:

[0159] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 5 As shown, the electronic device 5 of this embodiment includes: at least one processor 50 ( Figure 5 The diagram shows only one processor, a memory 51, and a computer program 52 stored in the memory 51 and executable on the at least one processor 50, which, when executed, performs the steps of any of the above method embodiments.

[0160] The electronic device 5 can be a desktop computer, laptop, handheld computer, or cloud server, etc. This electronic device may include, but is not limited to, a processor 50 and a memory 51. Those skilled in the art will understand that... Figure 5 This is merely an example of electronic device 5 and does not constitute a limitation on electronic device 5. It may include more or fewer components than shown in the figure, or combine certain components, or different components. For example, it may also include input / output devices, network access devices, etc.

[0161] The processor 50 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0162] In some embodiments, the memory 51 may be an internal storage unit of the electronic device 5, such as a hard disk or memory of the electronic device 5. In other embodiments, the memory 51 may be an external storage device of the electronic device 5, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 5. Furthermore, the memory 51 may include both internal and external storage units of the electronic device 5. The memory 51 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory 51 can also be used to temporarily store data that has been output or will be output.

[0163] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0164] This application also provides a network device, which includes: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above method embodiments.

[0165] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps in the above-described method embodiments.

[0166] This application provides a computer program product that, when run on an electronic device, enables the electronic device to implement the steps described in the various method embodiments above.

[0167] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographic device / electronic device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0168] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0169] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0170] In the embodiments provided in this application, it should be understood that the disclosed apparatus / network devices and methods can be implemented in other ways. For example, the apparatus / network device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0171] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0172] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for generating a function-guided radiotherapy plan, characterized in that, include: Obtain anatomical and functional data corresponding to the target user. The anatomical data at least reflects the anatomical structure of the target user's lung tumor target area and organs at risk, and the functional data is used to reflect the distribution of the target user's lung function. The target dose distribution is determined based on the anatomical data and the functional data; The function-guided radiotherapy plan for the target user is determined based on the target dose distribution.

2. The function-guided radiotherapy planning method as described in claim 1, characterized in that, Determining the target dose distribution based on the anatomical data and the functional data includes: The anatomical data and the functional data are used as inputs to a trained prediction model to obtain the target dose distribution output by the prediction model. The prediction model is used to analyze the dose distribution that matches the lung function of the target user based on the anatomical data and the functional data to obtain the target dose distribution.

3. The function-guided radiotherapy planning method as described in claim 2, characterized in that, The prediction model includes an anatomical coding network, a functional coding network, a fusion network, and a decoding network. The step of using the anatomical data and the functional data as input to the trained prediction model to obtain the target dose distribution output by the prediction model includes: The anatomical data is encoded using the anatomical coding network to obtain anatomical features, and the functional data is encoded using the functional coding network to obtain functional features. The anatomical features and functional features are fused using the fusion network to obtain fused features; The target dose distribution is obtained by predicting a dose distribution that matches the lung function based on the fusion features using the decoding network.

4. The function-guided radiotherapy planning generation method as described in claim 2, characterized in that, The prediction model is trained through the following steps: Construct training samples, each training sample including anatomical data and functional data corresponding to a user; The training samples are used as input to the prediction model to be trained, and the predicted dose distribution output by the prediction model to be trained is obtained. The loss value is determined based on the predicted dose distribution corresponding to the training samples and the actual functional guidance plan dose distribution. The parameters of the prediction model to be trained are updated based on the loss value to obtain the trained prediction model.

5. The function-guided radiotherapy planning method as described in claim 4, characterized in that, The step of determining the loss value based on the predicted dose distribution corresponding to the training samples and the actual functional guidance plan dose distribution includes: The loss value corresponding to the training sample is determined based on the predicted dose distribution, the actual functional guidance plan dose distribution, and the set functional weights of the target voxels, wherein the functional weights are determined based on the functional data in the training sample.

6. The function-guided radiotherapy planning generation method as described in any one of claims 1 to 5, characterized in that, The determination of the function-guided radiotherapy plan for the target user based on the target dose distribution includes: The function-guided dose simulation algorithm generates the function-guided radiotherapy plan based on the target dose distribution. The function-guided dose simulation algorithm is used to convert the target dose distribution into a radiotherapy plan that conforms to a set optimization function, the optimization function including at least a target term based on lung function.

7. The function-guided radiotherapy planning generation method as described in claim 6, characterized in that, The generation of the function-guided radiotherapy plan based on the target dose distribution using the function-guided dose simulation algorithm includes: The optimization function is solved based on the target dose distribution to obtain the target flux map; The target flux map is converted into machine parameters of the radiotherapy machine based on the multileaf grating sequence optimization algorithm and the direct subfield optimization algorithm, thereby obtaining the function-guided radiotherapy plan.

8. A function-guided radiotherapy planning device, characterized in that, include: The data acquisition module is used to acquire anatomical data and functional data corresponding to the target user. The anatomical data at least reflects the anatomical structure of the target user's lung tumor target area and organs at risk, and the functional data is used to reflect the distribution of the target user's lung function. A dose distribution prediction module is used to determine a target dose distribution based on the anatomical data and the functional data; The automatic radiotherapy plan generation module is used to determine the functionally guided radiotherapy plan for the target user based on the target dose distribution.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 7.

10. A computer program product, characterized in that, When the computer program product is run on an electronic device, it causes the electronic device to perform the method as described in any one of claims 1 to 7.

Images