Adaptive planning method based on orthogonal DR and related equipment

Abstract

The invention discloses an orthogonal DR-based adaptive planning method and related equipment, and relates to the technical field of radiotherapy, and the method comprises the steps: carrying out the automatic segmentation of a planned CT image of a patient, and obtaining a segmentation result; the segmentation result comprises a target region of the patient or a delineation set of an organ at risk; after a patient is positioned, DR images of the patient in two orthogonal directions are collected; projecting the planned CT image of the patient in the space of the positioning coordinate system to obtain a corresponding orthogonal DRR image; performing registration on the DR images in the two orthogonal directions and the corresponding DRR images to obtain 2D deformation grids in the two orthogonal directions; reconstructing a 3D deformation field through the 2D deformation grid; acting the 3D deformation field on the sketching set to obtain a deformation component of the corresponding target area or the endangered organ of the patient; determining whether to recollect the CT image of the patient according to the deformation component; the CT scanning frequency can be reduced, and meanwhile real-time evaluation of tumor changes is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of radiotherapy, and in particular to an adaptive planning method based on orthogonal DR and related equipment. Background Art

[0002] In the field of radiotherapy, ensuring the accuracy of treatment is crucial, especially when facing a complex and changeable disease like tumors. Traditional radiotherapy plans are often based on the location and shape of the tumor when the patient is first diagnosed. However, during the radiotherapy process, the tumor may shrink, move or change shape due to treatment response, which may cause the original treatment plan to fail or need to be adjusted. For this reason, Adaptive Planning came into being. It can adjust the treatment plan in real time according to the actual changes of the patient's tumor during the treatment process.

[0003] Adaptive planning refers to the real-time adjustment of the treatment plan according to changes in the position and shape of the patient's tumor during radiotherapy to ensure the accuracy of treatment; adaptive planning allows real-time adjustment of radiation dose and beam direction based on the latest imaging data (such as CT images).

[0004] However, the implementation of adaptive planning faces a series of challenges. First, in order to monitor tumor changes, traditional practices require regular CT scans to ensure that there are no significant changes in the tumor and surrounding tissues, which increases the patient's radiation exposure and takes up a certain amount of time. Second, traditional adaptive planning is highly dependent on 3D imaging data and planning CT to evaluate changes in the target area, which not only prolongs the treatment cycle, but also increases the burden on hospital equipment and patient treatment costs. Therefore, it is necessary to further improve traditional adaptive planning. Summary of the invention

[0005] Based on the above problems, the purpose of the present invention is to provide an adaptive planning method based on orthogonal DR and related equipment, which can achieve real-time evaluation of tumor changes while reducing the number of CT scans by combining orthogonal 2D imaging and deformation field reconstruction technology.

[0006] The purpose of the present invention is achieved by the following technical solutions:

[0007] In a first aspect, the present invention proposes an adaptive planning method based on orthogonal DR, the method comprising:

[0008] Automatically segmenting the planned CT image of the patient to obtain a segmentation result; the segmentation result includes a delineation set of the patient's target area or organs at risk;

[0009] After positioning the patient, DR images of the patient in two orthogonal directions are acquired; and the planned CT image of the patient is projected in the space of the positioning coordinate system to obtain corresponding orthogonal DRR images;

[0010] The DR images in two orthogonal directions are respectively registered with the corresponding DRR images to obtain 2D deformable meshes in two orthogonal directions;

[0011] Reconstructing the 3D deformation field through the 2D deformation grid;

[0012] The 3D deformation field is applied to the delineation set to obtain a deformation component of the patient's corresponding target area or organ at risk; and whether to re-acquire the patient's CT image is determined according to the deformation component.

[0013] Preferably, the patient's planned CT image is automatically segmented by a 3DUNet automatic segmentation network.

[0014] Preferably, the automatic segmentation further comprises obtaining a loss function of a 3DUNet automatic segmentation network, wherein the loss function is determined by a Dice loss and a cross information entropy loss between the true label and the predicted segmentation;

[0015] The loss function of the 3DUNet automatic segmentation network is:

[0016]

[0017] Among them, L Dice () is the Dice loss function; L CE () is the cross information entropy loss function; L loss () is the total loss function; e is the true label; e ′ is the predicted segmentation; t represents the voxel index; N is the total number of voxels; a g is the weight of the g-th segmentation category; w is the balance coefficient, which is used to balance the weights of the Dice loss function and the cross information entropy loss function; g is the g-th segmentation category; dn is the total number of segmentation categories.

[0018] Preferably, after positioning the patient, DR images of the patient in two orthogonal directions are acquired; and the planned CT image of the patient is projected in the space of the positioning coordinate system to obtain corresponding orthogonal DRR images; including:

[0019] The DR images of the patient are acquired through two orthogonal DR acquisition devices;

[0020] In the positioning coordinate system space, the patient's planned CT image is projected along the directions of the two DR acquisition devices to generate corresponding orthogonal DRR images.

[0021] Preferably, the registering of the DR images in two orthogonal directions with the corresponding DRR images to obtain 2D deformable meshes in two orthogonal directions comprises:

[0022] For each direction, the DR image and the corresponding DRR image are input into the VoxelMorph network as a pair to obtain the output velocity field;

[0023] The velocity field is converted into a differential deformation field using a differentiable squared and scaled integration layer;

[0024] For each direction, deformation information is extracted from the differential deformation field;

[0025] A 2D deformable mesh is generated according to the deformation information.

[0026] Preferably, reconstructing the 3D deformation field through the 2D deformation grid comprises:

[0027] The node displacements in the 3D deformation field are calculated by using the reversible relationship between the components of the 2D deformed mesh on the orthogonal plane and the 3D vector field through back projection from 2D to 3D and geometric transformation.

[0028] Based on the node displacements, a complete 3D deformation field is constructed.

[0029] Preferably, applying the 3D deformation field to the delineation set to obtain a deformation component of the target area or organ at risk corresponding to the patient; and determining whether to re-acquire the CT image of the patient according to the deformation component, comprises:

[0030] If the deformation component exceeds its deformation threshold, reacquiring the patient's CT image;

[0031] If the deformation component is less than or equal to its deformation threshold, the patient's planned CT and delineation set are corrected according to the deformation component, and the treatment plan is regenerated using the original optimized parameters.

[0032] In a second aspect, the present invention provides an adaptive planning device based on orthogonal DR, the device comprising:

[0033] A segmentation and delineation module is used to automatically segment the planned CT image of the patient to obtain a segmentation result; the segmentation result includes a delineation set of the patient's target area or organs at risk;

[0034] An image acquisition module is used to acquire DR images of the patient in two orthogonal directions after positioning the patient; and to project the planned CT image of the patient in the space of the positioning coordinate system to obtain corresponding orthogonal DRR images;

[0035] A 2D deformation acquisition module is used to register the DR images in two orthogonal directions with the corresponding DRR images to obtain 2D deformation meshes in two orthogonal directions;

[0036] A 3D deformation field acquisition module, used to reconstruct the 3D deformation field through the 2D deformation grid;

[0037] The determination module is used to apply the 3D deformation field to the delineation set to obtain the deformation component of the patient's corresponding target area or organ at risk; and determine whether to re-acquire the patient's CT image according to the deformation component.

[0038] In a third aspect, the present invention provides a treatment system, the system comprising:

[0039] Radiotherapy devices, used to deliver radiation therapy to patients according to treatment plans;

[0040] The orthogonal DR-based adaptive planning device of the present invention is used to obtain an adaptive treatment plan for a patient.

[0041] In a fourth aspect, the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the aforementioned method of the present invention or the functions of the apparatus of the present invention when executing the computer program.

[0042] In a fifth aspect, the present invention provides a computer-readable storage medium, wherein the storage medium stores computer instructions. When a computer reads the computer instructions, the computer executes the steps of the aforementioned method of the present invention.

[0043] Compared with the prior art, the beneficial effects of the present invention include at least: using the orthogonal 2D images collected during each treatment process, quickly constructing an orthogonal 2D deformable grid through deformation registration technology, and reconstructing a 3D deformation field from it; by evaluating the displacement of each voxel and the change of the threshold in the corresponding area of ​​the target area and the contour of the organ at risk, it can automatically determine whether it is necessary to re-collect CT data or make adaptive plan adjustments, which significantly reduces the number of CT scans, and only performs necessary CT image collection when the tumor change is greater than the set threshold, thereby greatly reducing the patient's radiation dose; by reconstructing the 3D deformation field through the orthogonal 2D deformable grid, the excessive reliance of traditional adaptive plans on 3D image data is avoided, further shortening the treatment cycle and reducing the treatment cost; under the premise of ensuring the real-time and accuracy of the adaptive treatment process, no additional detection or acquisition equipment is introduced, making the treatment process more efficient and convenient. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a schematic diagram of an adaptive planning method for orthogonal DR according to an embodiment of the present invention;

[0045] Figure 2 is a flow chart of an adaptive planning method for orthogonal DR according to an embodiment of the present invention;

[0046] Figure 3 is a schematic diagram of a VoxelMorph end-to-end unsupervised architecture according to an embodiment of the present invention;

[0047] Figure 4 is a schematic diagram of an orthogonal imaging system according to an embodiment of the present invention;

[0048] Figure 5 is a schematic diagram of a surface in a 3D deformation field back-projected by a deformed mesh component in an orthogonal plane according to an embodiment of the present invention;

[0049] Figure 6 is a schematic diagram of a 2D registration grid after orthogonal DR and DRR registration according to an embodiment of the present invention;

[0050] Figure 7 It is a schematic diagram of an adaptive planning method device for orthogonal DR according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided to make the present invention more comprehensive and complete and to fully convey the concepts of example embodiments to those skilled in the art. The same reference numerals in the figures represent the same or similar structures, and thus their repeated description will be omitted.

[0052] The words expressing positions and directions described in the present invention are all explained with reference to the accompanying drawings as examples, but they can be changed as needed, and all such changes are included in the protection scope of the present invention.

[0053] Reference Figure 1-2 :The present invention provides an adaptive planning method based on orthogonal DR, the method comprising:

[0054] Automatically segmenting the planned CT image of the patient to obtain a segmentation result; the segmentation result includes a delineation set of the patient's target area or organs at risk;

[0055] After positioning the patient, DR images of the patient in two orthogonal directions are acquired; and the planned CT image of the patient is projected in the space of the positioning coordinate system to obtain corresponding orthogonal DRR images;

[0056] The DR images in two orthogonal directions are respectively registered with the corresponding DRR images to obtain 2D deformable meshes in two orthogonal directions;

[0057] Reconstructing the 3D deformation field through the 2D deformation grid;

[0058] The 3D deformation field is applied to the delineation set to obtain a deformation component of the patient's corresponding target area or organ at risk; and whether to re-acquire the patient's CT image is determined according to the deformation component.

[0059] The working principle of the above technical solution is as follows: first, the patient's planned CT image is automatically segmented using image segmentation technology to accurately distinguish the patient's different tissue structures, especially the target area and organs at risk; after the segmentation is completed, a delineation set containing the target area and organs at risk is automatically generated; after the patient is accurately positioned, DR (digital X-ray) imaging technology is used to collect the patient's DR image in two orthogonal directions (such as front and back and left and right); the patient's planned CT image is projected in the space of the positioning coordinate system to generate a digitally reconstructed image (DRR) image corresponding to the DR image; these DRR images are simulated X-ray images, which are used for registration with the actual DR image. Using image registration technology, the DR images in two orthogonal directions are registered with the corresponding DRR images respectively; the purpose of registration is to find the best alignment between the DR image and the DRR image, so as to determine the actual deformation of the patient; after the registration is completed, 2D deformation grids in two orthogonal directions are generated according to the registration results; the deformation grids reflect the deformation of the patient in different directions; the 2D deformation grids in the two orthogonal directions are fused and reconstructed to generate a 3D deformation field; the 3D deformation field comprehensively reflects the deformation of the patient in three-dimensional space; the generated 3D deformation field is applied to the previously generated delineation set to obtain the deformation component of the patient's corresponding target area or organ at risk; based on the obtained deformation component, the patient's deformation is analyzed in detail; if the degree of deformation exceeds the preset threshold or reaches the standard that requires re-evaluation, the next step is required; if the deformation analysis results indicate that the patient's CT image needs to be re-collected to update the treatment plan, the CT image is re-collected to ensure the accuracy and effectiveness of the treatment plan.

[0060] The effects of the above technical solution are: by automatically segmenting the planned CT images, the patient's target area and the outline set of organs at risk can be accurately obtained, providing accurate basic data for subsequent deformation analysis and treatment plan adjustment; by using the registration of orthogonal DR images and DRR images, the patient's deformation during treatment can be accurately captured, thereby generating an accurate 2D deformation grid and 3D deformation field. Applying the 3D deformation field to the outline set can accurately calculate the deformation components of the target area and organs at risk, providing a reliable basis for adjusting the treatment plan. By analyzing the deformation components, it can be determined whether the deformation degree of the patient has reached the standard that requires re-collection of CT images, thereby avoiding unnecessary CT image collection, reducing the patient's radiation exposure and medical costs. An ordinary CT scan requires 5-10mSv, while an orthogonal X-ray only requires 0.1mSv. If a plan requires 30 fractions, reducing one CT scan is equivalent to reducing the dose to 50-100th of the original dose, which greatly reduces the damage of the radiation dose to the patient. At the same time, it also reduces the time delay caused by frequent collection of CT images and improves the treatment efficiency. Originally, ART (Adaptive Radiation Therapy) must use 3D data to evaluate the deformation field, that is, only from the adaptive plan itself, 3D data is required to calculate the deformation field, and the dose of the original plan is deformed by the deformation field. The present invention avoids the necessity of 3D data by reconstructing the 3D deformation field on the orthogonal 2D deformation grid, while shortening the treatment cycle and reducing the treatment cost. Under the premise of ensuring the real-time and accuracy of the adaptive treatment process, no additional detection or acquisition equipment is introduced, making the treatment process more efficient and convenient.

[0061] In some embodiments, the planned CT image of the patient is automatically segmented by a 3DUNet automatic segmentation network; and the loss function of the 3DUNet automatic segmentation network is obtained by using the Dice loss and the cross information entropy loss between the true label and the predicted segmentation; the loss function of the 3DUNet automatic segmentation network is:

[0062]

[0063] Among them, L Dice () is the Dice loss function (Dice loss function); L CE () is the cross information entropy loss function; L loss () is the total loss function; e is the true label; e ′ is the predicted segmentation; t represents the voxel index; N is the total number of voxels; a g is the weight of the g-th segmentation category; w is the balance coefficient, which is used to balance the weights of the Dice loss function and the cross information entropy loss function; g is the g-th segmentation category; dn is the total number of segmentation categories.

[0064] The working principle and effect of the above technical solution are:

[0065] 3DUNet is a three-dimensional image segmentation network based on deep learning. Its design is inspired by the traditional U-Net architecture, but it is expanded and optimized for three-dimensional image data. The network consists of an encoder and a decoder. The encoder is responsible for extracting image features, while the decoder is responsible for accurate segmentation based on these features. The 3DUNet network framework used in this embodiment has excellent automatic adaptation capabilities and can adaptively adjust its network structure, preprocessing method, and training parameters according to the different characteristics of the input three-dimensional CT image data (such as resolution, size, dimension, etc.); this means that for organs at risk (OAR) or target areas in different parts of the body, there is no need to redesign the network structure or adjust the hyperparameters. The network can automatically adjust the core structures such as the depth of U-Net, the size of the convolution kernel, and the pooling layer based on the unique properties of the data set (such as the size, dimension, and annotation details of the image).

[0066] By analyzing the statistical characteristics of the input data, 3DUNet can intelligently identify and apply the optimal network configuration; during the training phase, the network will learn how to extract key features from the input CT images and accurately segment the target area and organs at risk based on these features.

[0067] Dice loss is used to measure the similarity between the predicted segmentation result and the true label; the higher the Dice coefficient, the closer the predicted result is to the true situation; the Dice loss function is the negative value of the Dice coefficient, so the optimization goal is to minimize the loss.

[0068] The cross entropy loss is used to evaluate the difference between the probability distribution of the predicted segmentation result and the true label; the smaller the cross entropy loss, the more accurate the prediction result.

[0069] The Dice loss and cross information entropy loss are combined, and the weight coefficient and balance coefficient are introduced to balance the weights between different segmentation categories and the two loss functions; the total loss function is the key to optimizing the network model. By minimizing this loss, the accuracy and robustness of the network segmentation can be improved.

[0070] The specific segmentation process is:

[0071] The patient's planned CT image is used as input data, and the image contains the patient's three-dimensional anatomical structure information;

[0072] The 3DUNet network extracts features from the input CT images and gradually abstracts the key information in the image through layer-by-layer convolution and pooling operations.

[0073] In the decoding stage, the network will make segmentation predictions based on the extracted features and generate predicted segmentation results;

[0074] The predicted segmentation results are compared with the true labels, the Dice loss and cross information entropy loss are calculated, and the network is optimized based on the total loss function. Through multiple iterative training, the accuracy of segmentation can be gradually improved. The total loss function can flexibly adapt to multi-region segmentation tasks by introducing segmentation region weights and balance coefficients. g It allows different levels of emphasis to be given to different segmentation categories, which is particularly important for processing OARs or target areas with different importance. For example, the segmentation of the head and neck includes multiple OARs, each OAR is a category, and each OAR may include multiple regions of interest (ROIs). The balance coefficient w is used to adjust the relative importance between the Dice loss and the cross information entropy loss to ensure that the two can work together during the training process to jointly optimize the segmentation performance.

[0075] Finally, the 3DUNet network outputs accurate segmentation results, including the patient's target area and the outline set of organs at risk; these segmentation results will serve as the basic data for subsequent adaptive planning methods.

[0076] In some embodiments, after positioning the patient, acquiring DR images of the patient in two orthogonal directions; and projecting the planned CT image of the patient in the space of the positioning coordinate system to obtain corresponding orthogonal DRR images; including:

[0077] The DR images of the patient are acquired through two orthogonal DR acquisition devices;

[0078] In the positioning coordinate system space, the patient's planned CT image is projected along the directions of the two DR acquisition devices to generate corresponding orthogonal DRR images.

[0079] The working principle of the above technical solution is:

[0080] According to the specific situation of the patient and the requirements of the radiotherapy plan, a suitable position is selected to fix the patient, such as supine or prone position; after the patient's position is fixed, the DR acquisition device is used to respectively acquire DR images of the patient in two orthogonal directions; wherein the DR acquisition device can be an imaging flat panel, which is installed on the top of the treatment room, with the isocenter of the treatment room as the origin; the DR image of the patient is acquired through two orthogonal imaging flat panels; according to the direction of the acquired DR image, the projection direction of the planned CT image in the positioning coordinate system is determined; the planned CT image is projected by a ray projection algorithm (or other similar algorithms) to simulate the image formed after the X-ray passes through the patient's body, which is similar to "flattening" (converting) a three-dimensional image into a two-dimensional image, but retains key structural information; the planned CT image is projected along two orthogonal directions to generate corresponding orthogonal DRR images.

[0081] The effects of the above technical solution are: by acquiring DR images in two orthogonal directions, the image information of the patient at different perspectives can be obtained; it provides a data basis for subsequent image registration, helps to complete the construction from 2D deformation grid to 3D deformation field, and improves the accuracy of registration; the planned CT image is projected into the positioning coordinate system space to generate an orthogonal DRR image, making the registration between DR images and DRR images more intuitive and accurate.

[0082] In some embodiments, the registering of the DR images in two orthogonal directions with the corresponding DRR images to obtain 2D deformable meshes in two orthogonal directions comprises:

[0083] For each direction, the DR image and the corresponding DRR image are input as a pair into the VoxelMorph network (deformable medical image registration learning network) to obtain the output velocity field;

[0084] The velocity field is converted into a differential deformation field using a differentiable squared and scaled integration layer;

[0085] For each direction, deformation information is extracted from the differential deformation field;

[0086] A 2D deformable mesh is generated according to the deformation information.

[0087] The working principle of the above technical solution is:

[0088] As attached Figure 3 Shown: Attached Figure 3 Schematic diagram of VoxelMorph end-to-end unsupervised architecture; VoxelMorph is a framework that combines traditional methods and learning-based methods. It is based on a probabilistic generative model and derives an unsupervised learning-based reasoning algorithm. Figure 3The first part of the input image to the network def ψ (m,f), the output represents the mean value of the velocity field μ z|x:y and variance Σ z|x:y The velocity field z is sampled and converted into a differential deformation field φ using a differentiable squared and scaled integration layer z Finally, perform spatial transformation on m to obtain the transformed

[0089] In this embodiment, the VoxelMorph network is used to register the DR images in two orthogonal directions with the corresponding DRR images to obtain 2D deformable meshes in two orthogonal directions. Specifically:

[0090] For each orthogonal direction, the DR image and the corresponding DRR image are input as a pair into the VoxelMorph network; the VoxelMorph network adopts the U-Net architecture and introduces a differential homeomorphism integration layer and a spatial transformation layer; the network processes the input image pair as a parameterized function that outputs a velocity field; the velocity field represents the speed and direction that each pixel or voxel in the image needs to move during the registration process; in the VoxelMorph architecture, the first output is the approximate posterior probability parameters of the mean and variance of the velocity field, which are used to sample the velocity field; the sampled velocity field is then transformed by a differentiable squared and scaled integration layer; this layer integrates the velocity field into a differential deformation field, that is, the actual displacement of each pixel or voxel during the registration process; the differential deformation field represents the position to which each pixel point in the DR image needs to move, and the differential deformation field is then used in the spatial transformation layer, which transforms the input DR image so that it is spatially aligned with the DRR image. Deformation information is extracted from the differential deformation field, where the deformation information describes the deformation process from the DR image to the DRR image.

[0091] For each orthogonal direction, the deformation information extracted from the differential deformation field is used to generate a 2D deformable mesh.

[0092] The grid represents the displacement of each pixel in the DR image when it is registered to the DRR image. Finally, for each orthogonal direction, a 2D deformable grid is obtained, which represents the accurate registration result of the DR image to the DRR image.

[0093] The effects of the above technical solution are:

[0094] The VoxelMorph network combines traditional methods with learning-based methods, and uses the powerful learning ability of convolutional neural networks (CNNs) to automatically extract image features and perform accurate registration. By outputting the velocity field and converting it into a differential deformation field, fine displacement control of each pixel or voxel in the image is achieved, thereby improving the registration accuracy; using DR images and DRR images for registration can reduce dependence on traditional CT scans, thereby reducing the cost of purchasing and maintaining medical equipment. At the same time, since the DR image acquisition process is simple and fast, it also helps to reduce the cost of patient examinations.

[0095] In some embodiments, reconstructing the 3D deformation field by using the 2D deformation grid comprises:

[0096] The node displacements in the 3D deformation field are calculated by using the reversible relationship between the components of the 2D deformed mesh on the orthogonal plane and the 3D vector field through back projection from 2D to 3D and geometric transformation.

[0097] Based on the node displacements, a complete 3D deformation field is constructed.

[0098] The working principle of the above technical solution is:

[0099] The two planes are orthogonal in the 3D patient coordinate system; since the two planes are orthogonal, different views of the object in three-dimensional space can be captured, and through these two views, the three-dimensional shape and deformation of the object can be reconstructed.

[0100] See attached Figure 4 , two X-ray sources in an orthogonal imaging system (S A ,S B ) and two X-ray detectors (A, B), located on two orthogonal planes; the two 2D projection coordinate systems are (x A ,y A ) and (x B, y B ), the 3D patient coordinate system is (x, y, z); because (x A ,y A ) and (x B ,y B ) are orthogonal in the 3D patient coordinate system, and the 3D deformable mesh can be reversely reconstructed using the orthogonal 2D deformable mesh; for example, each voxel in the vector field represents a displacement component, and there is a reversible relationship between the component on the orthogonal plane and the direct projection of the 3D vector field.

[0101] See attached Figure 5 , attached Figure 5The nodes of the mid-projection and mid-grid midline (deformed mesh) define a plane H, which is the three-dimensional patient coordinate system; the three-dimensional displacement of the nodes on plane H is calculated through two-dimensional to three-dimensional back projection and geometric transformation.

[0102] First, the DR images in two orthogonal directions are registered with the corresponding DRR images through the VoxelMorph network to obtain the 2D deformable meshes in two orthogonal directions. Figure 6 .

[0103] For each 2D deformable mesh, it is regarded as a displacement field on the corresponding orthogonal plane, and each node in the displacement field represents a displacement vector; the displacement vector is projected from the 2D plane to the 3D space using the 2D to 3D back-projection technique; during the back-projection process, the projection points need to be geometrically transformed to calculate their actual positions in the 3D space, and the geometric transformation includes operations such as rotation, translation, and scaling; through the geometric transformation, the displacement vector of each projection point in the 3D space is calculated; the displacement vector constitutes the node displacement in the 3D deformation field. Based on the calculated node displacements, a complete 3D deformation field can be constructed.

[0104] The effects of the above technical solution are: by utilizing the reversible relationship between the 2D deformation grid and the 3D vector field on two orthogonal planes, the node displacement in the 3D deformation field is accurately calculated, thereby constructing a complete 3D deformation field to ensure the accuracy of subsequent analysis and operations; traditional deformation analysis is usually limited to the two-dimensional plane, while this method elevates deformation analysis to the three-dimensional level by introducing an orthogonal imaging system and a three-dimensional reconstruction method, so that doctors and researchers can understand the three-dimensional shape and deformation of objects (such as tumors, endangered organs, etc.) more comprehensively and intuitively, providing richer information for clinical diagnosis and treatment; through two-dimensional to three-dimensional back projection and geometric transformation, the node displacement in the 3D deformation field can be efficiently calculated, avoiding complex mathematical models and calculation processes, which not only improves the calculation efficiency but also reduces the calculation cost.

[0105] In some embodiments, applying the 3D deformation field to the delineation set to obtain a deformation component of a target area or an organ at risk corresponding to the patient; and determining whether to re-acquire a CT image of the patient according to the deformation component, comprises:

[0106] If the deformation component exceeds its deformation threshold, a prompt is given to re-acquire the patient's CT image;

[0107] If the deformation component is less than or equal to its deformation threshold, the patient's planned CT and delineation set are corrected according to the deformation component, and the treatment plan is regenerated using the original optimized parameters.

[0108] The working principle of the above technical solution is: first, the acquired 3D deformation field is fused with the patient's outline set; by applying the 3D deformation field to each structure in the outline set, the new position and shape of these structures after deformation can be calculated; during the fusion process, the deformation components of each key structure (such as the target area and organs at risk) are extracted; the deformation components describe the displacement, rotation and deformation changes of these structures in 3D space; the extracted deformation components are evaluated to determine whether the deformation exceeds a preset threshold or standard; the threshold or standard can be set according to clinical experience and treatment requirements.

[0109] Based on the evaluation results of the deformation component, decide whether it is necessary to re-acquire the patient's CT image: if the deformation component is small and the impact on the treatment plan is within an acceptable range, there is no need to re-acquire the CT image; if the deformation component is large and has a significant impact on the treatment plan, or the deformation exceeds the safe range, it is necessary to re-acquire the CT image in order to update the treatment plan and ensure the accuracy and safety of the treatment.

[0110] If the CT image is re-acquired, the outline set is updated and the adaptive treatment plan is regenerated based on the re-acquired CT image; if it is determined that the CT image does not need to be re-acquired, the outline set is deformed and corrected according to the current deformation field, the treatment plan is optimized, and the treatment is performed according to the optimized treatment plan.

[0111] In some embodiments, the deformation component includes the position change, area change and perimeter change of the patient's target area or organ at risk, and if the deformation component exceeds its deformation threshold, it prompts to re-acquire the patient's CT image; including:

[0112] The weighted average of the position change, area change and perimeter change is performed to obtain the weighted deformation;

[0113] If any deformation component exceeds its preset threshold, it will prompt to re-acquire the patient's CT image;

[0114] If each deformation component is within its preset threshold range, but the weighted deformation amount exceeds its preset threshold, it is prompted to re-acquire the patient's CT image.

[0115] If any deformation component exceeds the preset threshold, or the weighted deformation exceeds its preset threshold, a prompt is given to re-acquire the patient's CT image; this decision logic helps to promptly detect significant deformations that may occur in the patient's body, thereby avoiding potential risks during radiotherapy and ensuring the safety of the patient.

[0116] Through automated deformation component monitoring and decision logic, manual intervention in radiotherapy can be reduced and the efficiency of the radiotherapy process can be improved; through precise deformation component monitoring and decision logic, unnecessary CT image acquisition can be avoided, which helps reduce patients' radiation exposure and reduce potential health risks.

[0117] In some embodiments, the method further comprises:

[0118] Through the patient's historical data, multiple deformation components of the same target area or organ at risk, changes in deformation components, and treatment records based on time series are obtained. Through machine learning, the deformation component during the next treatment is predicted; if the deformation component exceeds its deformation threshold, it is prompted to re-acquire the patient's CT image.

[0119] The working principle and effect of the above technical solution are as follows: extract multiple deformation components of the same target area or organ at risk from the patient's historical data, including position change, area change and perimeter change; at the same time, record the corresponding treatment record for each deformation component measurement, including key information such as radiotherapy dose and radiotherapy time. Preprocess the collected data, including data cleaning, format conversion and standardization, to ensure data quality and consistency.

[0120] By using time series analysis, the deformation components of the same target area or organ at risk are modeled in time series. Through time series analysis, the deformation component value at a certain time point in the future (i.e., the next treatment) can be predicted.

[0121] Based on the preprocessed historical data and time series analysis results, a machine learning model is trained. The goal of the model is to predict the deformation component of the next treatment based on the historical deformation component and treatment records. The machine learning model can be a support vector machine, a random forest, or a neural network; after training, the model is used to predict the deformation component of the next treatment. The predicted deformation component value is compared with the preset deformation threshold. If the predicted deformation component exceeds the deformation threshold, it is considered that the target area or the organ at risk has undergone significant deformation. When the predicted deformation component exceeds the deformation threshold, it is prompted to re-collect the patient's CT image. During the actual radiotherapy process, the prediction accuracy of the machine learning model is evaluated based on the re-collected CT images and the new deformation component measurement results. If it is found that the prediction accuracy of the model needs to be improved, the model can be further optimized and adjusted to improve its prediction ability.

[0122] At the same time, new data can also be added to the historical data set to enrich the training samples and improve the generalization ability of the model.

[0123] In summary, by integrating patients' historical data, time series analysis, machine learning predictions, and the relationship between deformation components and the decision to re-acquire CT images, accurate prediction and timely response to the deformation components of the target area or organs at risk are achieved, avoiding unnecessary CT image acquisition, helping to reduce patients' radiation exposure and potential health risks, while improving the accuracy and safety of radiotherapy plans.

[0124] An embodiment of the present invention provides an adaptive planning device based on orthogonal DR, the device comprising:

[0125] A segmentation and delineation module is used to automatically segment the planned CT image of the patient to obtain a segmentation result; the segmentation result includes a delineation set of the patient's target area or organs at risk;

[0126] An image acquisition module is used to acquire DR images of the patient in two orthogonal directions after positioning the patient; and to project the planned CT image of the patient in the space of the positioning coordinate system to obtain corresponding orthogonal DRR images;

[0127] A 2D deformation acquisition module is used to register the DR images in two orthogonal directions with the corresponding DRR images to obtain 2D deformation meshes in two orthogonal directions;

[0128] A 3D deformation field acquisition module, used to reconstruct the 3D deformation field through the 2D deformation grid;

[0129] The determination module is used to apply the 3D deformation field to the delineation set to obtain the deformation component of the patient's corresponding target area or organ at risk; and determine whether to re-acquire the patient's CT image according to the deformation component.

[0130] In some embodiments, the patient's planning CT image is automatically segmented by a 3DUNet automatic segmentation network.

[0131] In some embodiments, the automatic segmentation further comprises obtaining a loss function of a 3DUNet automatic segmentation network, wherein the loss function of the 3DUNet automatic segmentation network is obtained by using a Dice loss and a cross information entropy loss between the true label and the predicted segmentation;

[0132] The loss function of the 3DUNet automatic segmentation network is:

[0133]

[0134] Among them, L Dice () is the Dice loss function; L CE () is the cross information entropy loss function; L loss () is the total loss function; e is the true label; e ′is the predicted segmentation; t represents the voxel index; N is the total number of voxels; a g is the weight of the g-th segmentation category; w is the balance coefficient, which is used to balance the weights of the Dice loss function and the cross information entropy loss function; g is the g-th segmentation category; dn is the total number of segmentation categories.

[0135] In some embodiments, the image acquisition module includes:

[0136] A DR image acquisition unit, used to acquire DR images of the patient by using two orthogonal DR acquisition devices with the isocenter of the treatment room as the origin;

[0137] The DDR image generation unit is used to project the planned CT image of the patient along the directions of two DR acquisition devices in the positioning coordinate system space to generate corresponding orthogonal DRR images.

[0138] In some embodiments, the 2D deformation acquisition module includes:

[0139] A first output unit is used for inputting the DR image and the corresponding DRR image into a VoxelMorph network as a pair for each direction to obtain an output velocity field;

[0140] A differential deformation field unit, for converting a velocity field into a differential deformation field using a differentiable square and scaling integration layer;

[0141] The deformation mesh acquisition unit is used to extract deformation information from the differential deformation field for each direction; and generate a 2D deformation mesh according to the deformation information.

[0142] In some embodiments, the 3D deformation field acquisition module includes:

[0143] A node displacement acquisition unit is used to calculate the node displacement in the 3D deformation field by using the reversible relationship between the components of the 2D deformed grid on the orthogonal plane and the 3D vector field through two-dimensional to three-dimensional back projection and geometric transformation;

[0144] The 3D deformation field reconstruction unit is used to construct a complete 3D deformation field according to the node displacement.

[0145] In some embodiments, the determination module includes:

[0146] A first determination unit, configured to reacquire a CT image of the patient if the deformation component exceeds its deformation threshold;

[0147] The second determination unit is used to correct the patient's planned CT and delineation set according to the deformation component if the deformation component is less than or equal to its deformation threshold, and regenerate the treatment plan through the original optimized parameters.

[0148] In some embodiments, the deformation component includes a position change, an area change, and a circumference change of a patient's target area or an organ at risk; the first determination unit includes:

[0149] A weighted deformation acquisition unit is used to perform weighted averaging of the position change, area change and perimeter change to obtain a weighted deformation;

[0150] The first determination subunit is used to prompt to re-acquire the CT image of the patient if any deformation component exceeds its preset threshold value;

[0151] The second determination subunit is used to prompt to re-acquire the patient's CT image if all deformation components are within their preset threshold ranges but the weighted deformation amount exceeds its preset threshold.

[0152] In some embodiments, the device also includes a prediction module for obtaining multiple deformation components of the same target area or organ at risk, changes in deformation components, and treatment records based on a time series through the patient's historical data, and predicting the deformation component during the next treatment through machine learning; if the deformation component exceeds its deformation threshold, a prompt is given to re-acquire the patient's CT image.

[0153] The working principle and effect of the above technical solution are the same as those in the embodiment of the method of the present invention, and will not be described in detail here.

[0154] An embodiment of the present invention provides a treatment system, the system comprising:

[0155] Radiotherapy devices, used to deliver radiation therapy to patients according to treatment plans;

[0156] The orthogonal DR-based adaptive planning device described in any one of the embodiments of the present invention is used to obtain an adaptive treatment plan for a patient.

[0157] An embodiment of the present invention further provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method in the foregoing embodiment of the present invention or the functions of the device in the embodiment of the present invention are implemented.

[0158] An embodiment of the present invention further provides a computer-readable storage medium, which is used to store a computer program. When the computer program is executed, the steps of the method in the embodiment of the present invention are implemented. The specific implementation method is consistent with the implementation method and the technical effect achieved recorded in the above method embodiment, and some contents will not be repeated here.

[0159] In the present invention, a readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device, or device. A program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination of the above. More specific examples (a non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0160] Computer-readable storage media may include data signals propagated in baseband or as part of a carrier wave, wherein readable program codes are carried. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The readable storage medium may also be any readable medium, which may send, propagate, or transmit a program for use by an instruction execution system, an apparatus, or a device or for use in combination with it. The program code contained on the readable storage medium may be transmitted with any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above. The program code for performing the operation of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and also conventional procedural programming languages ​​such as C language or similar programming languages. The program code may be executed entirely on a user computing device, partially on an associated device, as an independent software package, partially on a user computing device, partially on a remote computing device, or entirely on a remote computing device or server. Where a remote computing device is involved, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0161] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, substitute and modify the above embodiments within the scope of the invention without departing from the principles and purpose of the present invention. All such changes should fall within the scope of protection of the claims of the present invention.

Claims

1. An adaptive planning method based on orthogonal DR, characterized in that: The method comprises: Automatically segmenting the planned CT image of the patient to obtain a segmentation result; the segmentation result includes a delineation set of the patient's target area or organs at risk; After positioning the patient, DR images of the patient in two orthogonal directions are acquired; and the planned CT image of the patient is projected in the space of the positioning coordinate system to obtain corresponding orthogonal DRR images; The DR images in two orthogonal directions are respectively registered with the corresponding DRR images to obtain 2D deformable meshes in two orthogonal directions; Reconstructing the 3D deformation field through the 2D deformation grid; The 3D deformation field is applied to the delineation set to obtain a deformation component of the patient's corresponding target area or organ at risk; and whether to re-acquire the patient's CT image is determined according to the deformation component.

2. The adaptive planning method based on orthogonal DR according to claim 1, characterized in that: The patient's planned CT images are automatically segmented using the 3DUNet automatic segmentation network.

3. The adaptive planning method based on orthogonal DR according to claim 2, characterized in that: The automatic segmentation also includes obtaining a loss function of a 3DUNet automatic segmentation network, wherein the loss function is determined by a Dice loss and a cross information entropy loss between the true label and the predicted segmentation; The loss function of the 3DUNet automatic segmentation network is: Among them, L Dice is the Dice loss function; L CE is the cross information entropy loss function; L loss is the total loss function; e is the true label; e ′ is the predicted segmentation; t represents the voxel index; N is the total number of voxels; a g is the weight of the g-th segmentation category; w is the balance coefficient, which is used to balance the weights of the Dice loss function and the cross information entropy loss function; g is the g-th segmentation category; dn is the total number of segmentation categories.

4. The adaptive planning method based on orthogonal DR according to claim 1, characterized in that: After positioning the patient, acquiring DR images of the patient in two orthogonal directions; and projecting the planned CT image of the patient in the space of the positioning coordinate system to obtain a corresponding orthogonal DRR image; including: The DR images of the patient are acquired through two orthogonal DR acquisition devices; In the positioning coordinate system space, the patient's planned CT image is projected along the directions of the two DR acquisition devices to generate corresponding orthogonal DRR images.

5. The adaptive planning method based on orthogonal DR according to claim 1, characterized in that: The DR images in two orthogonal directions are respectively registered with the corresponding DRR images to obtain 2D deformable meshes in two orthogonal directions; comprising: For each direction, the DR image and the corresponding DRR image are input into the VoxelMorph network as a pair to obtain the output velocity field; The velocity field is converted into a differential deformation field using a differentiable squared and scaled integration layer; For each direction, deformation information is extracted from the differential deformation field; A 2D deformable mesh is generated according to the deformation information.

6. The adaptive planning method based on orthogonal DR according to claim 1, characterized in that: The reconstructing the 3D deformation field by using the 2D deformation grid comprises: The node displacements in the 3D deformation field are calculated by using the reversible relationship between the components of the 2D deformed mesh on the orthogonal plane and the 3D vector field through back projection from 2D to 3D and geometric transformation. Based on the node displacements, a complete 3D deformation field is constructed.

7. The adaptive planning method based on orthogonal DR according to claim 1, characterized in that: applying the 3D deformation field to the delineation set to obtain a deformation component of the patient's corresponding target area or organ at risk; Determining whether to reacquire the CT image of the patient according to the deformation component includes: If the deformation component exceeds its deformation threshold, reacquiring the patient's CT image; If the deformation component is less than or equal to its deformation threshold, the patient's planned CT and delineation set are corrected according to the deformation component, and the treatment plan is regenerated using the original optimized parameters.

8. An adaptive planning device based on orthogonal DR, characterized in that: The device comprises: A segmentation and delineation module is used to automatically segment the planned CT image of the patient to obtain a segmentation result; the segmentation result includes a delineation set of the patient's target area or organs at risk; An image acquisition module is used to acquire DR images of the patient in two orthogonal directions after positioning the patient; and to project the planned CT image of the patient in the space of the positioning coordinate system to obtain corresponding orthogonal DRR images; A 2D deformation acquisition module is used to register the DR images in two orthogonal directions with the corresponding DRR images to obtain 2D deformation meshes in two orthogonal directions; A 3D deformation field acquisition module, used to reconstruct the 3D deformation field through the 2D deformation grid; The determination module is used to apply the 3D deformation field to the delineation set to obtain the deformation component of the patient's corresponding target area or organ at risk; and determine whether to re-acquire the patient's CT image according to the deformation component.

9. A treatment system, characterized in that: The system comprises: Radiotherapy devices, used to deliver radiation therapy to patients according to treatment plans; The orthogonal DR-based adaptive planning device described in claim 8 is used to obtain an adaptive treatment plan for a patient.

10. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method according to any one of claims 1 to 7 when executing the computer program.

11. A computer-readable storage medium, characterized in that: The storage medium stores computer instructions. When a computer reads the computer instructions, the computer executes the steps of the method according to any one of claims 1 to 7.

Images