Method and device for determining radiotherapy plan, storage medium and processor

Abstract

The present application discloses a method and device, storage medium and processor for determining a radiotherapy plan. The method includes: photographing a medical image 1 for a target patient for reference planning, and outlining a region of interest on the medical image 1 to obtain an outlining result 1; generating a radiotherapy plan based on the outlining result 1 to obtain a reference plan; reviewing the reference plan, and if the reference plan passes the review, photographing a medical image 2 for the target patient, registering the medical image 2 with the medical image 1 to obtain a registration result; in the outlining platform, outlining the region of interest on the medical image 2 based on the registration result to obtain an outlining result 2; determining the target radiotherapy plan based on the outlining result 2 and the reference plan. Through the present application, the problem that designing an adaptive radiotherapy plan is too time-consuming in the related art is solved.

Description

Technical Field

[0001] The present application relates to the field of medical technology, and in particular to a method and device for determining a radiotherapy plan, a storage medium, and a processor. Background Art

[0002] Radiotherapy is one of the three major tumor treatment methods in clinical practice. Modern radiotherapy has a complete treatment process, which mainly includes the delineation of critical organs and target areas, the formulation of radiotherapy plans based on dosimetry principles, and the implementation of plans. Among them, the delineation of target areas and critical organs has a crucial impact on the accuracy and efficacy of radiotherapy, and directly determines the accuracy of radiotherapy plans. At present, the delineation of target areas and critical organs is usually completed manually by doctors and physicists, which greatly wastes the time and energy of doctors and physicists. More importantly, in the current radiotherapy process, the delineation work is usually completed only before the first treatment. Each subsequent treatment is based on the first treatment plan as a template, combined with the patient's daily images for rigid registration, evaluation and implementation. Since the organs in the patient's body undergo a small amount of displacement and volume changes every day, this method cannot be adaptive in terms of dose implementation.

[0003] To address this problem, Unity MR-linac has been introduced in the existing technology. It uses magnetic resonance imaging (MR) as a guide. Before each treatment, the ROI of the first treatment plan will be flexibly registered with the ROI of the day's image. This requires first delineating the lesions in the complete target area and the ROI of the critical organ on the day's image to achieve accurate adaptive radiotherapy. The current process of MR-linac is as follows: Figure 1 As shown, the delineation, registration, and evaluation process in designing adaptive plans is too time-consuming to be applied on a large scale in clinical practice.

[0004] With regard to the problem that designing adaptive radiotherapy plans in related technologies is too time-consuming, no effective solution has been proposed so far. Summary of the invention

[0005] The main purpose of the present application is to provide a method and device for determining a radiotherapy plan, a storage medium and a processor to solve the problem that designing an adaptive radiotherapy plan in the related art is too time-consuming.

[0006] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a method for determining a radiotherapy plan is provided. The method includes: taking a medical image 1 for a reference plan of a target patient, and outlining a region of interest on the medical image 1 to obtain an outlining result 1; generating a radiotherapy plan based on the outlining result 1 to obtain a reference plan; reviewing the reference plan, and if the reference plan passes the review, taking a medical image 2 of the target patient, registering the medical image 2 with the medical image 1 to obtain a registration result; in the outlining platform, outlining a region of interest on the medical image 2 based on the registration result to obtain an outlining result 2; determining the target radiotherapy plan based on the outlining result 2 and the reference plan.

[0007] Furthermore, after determining the target radiotherapy plan based on the second outlining result and the reference plan, the method also includes: reviewing the target radiotherapy plan; after the target radiotherapy plan passes the review, obtaining the second outlining result through the outlining platform, and sending the second outlining result to the model training platform, performing learning and training on the model training platform based on the second outlining result to generate a target model; and sending the target model to the outlining platform.

[0008] Furthermore, after sending the target model to the delineation platform, the method also includes: before the target patient undergoes the Nth radiotherapy, obtaining the medical image of the target patient on that day, wherein N is a natural number greater than 1; in the delineation platform, using the target model to delineate the region of interest of the medical image on that day to obtain a target delineation result; based on the target delineation result, generating a target radiotherapy plan for the target patient on that day.

[0009] Furthermore, after generating the target radiotherapy plan for the target patient on that day based on the target outlining result, the method also includes: sending the second outlining result and the target outlining result to the model training platform, performing learning and training on the model training platform based on the second outlining result and the target outlining result to update the target model; and sending the updated target model to the outlining platform.

[0010] Furthermore, in the model training platform, learning and training are performed based on the second outlining result and the target outlining result to update the target model, including: obtaining a training data set, wherein the training data set is composed of the second outlining result and the target outlining result; constructing a student model and a teacher model, and using pre-trained weights to initialize the student model and the teacher model, and entering the following training cycle: I, using the student model and the teacher model to extract features from the images in the training data set respectively to obtain feature one and feature two; II, calculating the feature one and the outlining result in the training data set to obtain approximately Constraint one; III, respectively calculate the description operator of the feature one and the feature two to obtain description one and description two, wherein the description operator is at least one of the following: volume, center of mass, distance from each pixel to the center of mass, contour length, calculate the description one and the description two to obtain constraint two; IV, use the constraint one and the constraint two to update the weight of the student model; V, use the updated weight of the student model to update the teacher model, wherein the updating method is the exponential sliding average method; repeat steps I to V until the training is completed, and the target model is generated by the teacher model.

[0011] Furthermore, the method for obtaining the pre-training weights is at least one of the following: supervised learning, semi-supervised learning, and unsupervised learning.

[0012] Furthermore, if the method for obtaining pre-training weights is unsupervised learning, obtaining the weights of the student model includes: obtaining data of medical images of multiple different modalities as a training set; before the start of iteration, randomly extracting a sample from the training set, and randomly enhancing the sample twice to obtain two enhanced samples A and B; inputting samples A and B into the encoding model to obtain feature codes X and Y; inputting feature codes X and Y into the projection model to obtain projected feature codes XX and YY, and calculating the Euclidean distance or cosine similarity of XX and YY as a loss function; and using the loss function to update the student model to obtain the weight of the student model.

[0013] Furthermore, before determining the target radiotherapy plan based on the second outlining result and the reference plan, the method also includes: if a modification instruction for the outlining result is detected; in response to the modification instruction, modify the second outlining result to obtain a modified outlining result two; determining the target radiotherapy plan based on the second outlining result and the reference plan includes: determining the target radiotherapy plan based on the modified second outlining result and the reference plan.

[0014] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a device for determining a radiotherapy plan is provided. The device comprises: a first delineation unit, which is used to take a medical image 1 for a target patient for reference planning, and delineate a region of interest on the medical image 1 to obtain a delineation result 1; a first construction unit, which is used to generate a radiotherapy plan based on the delineation result 1 to obtain a reference plan; a registration unit, which is used to review the reference plan, and if the reference plan passes the review, take a medical image 2 for the target patient, and register the medical image 2 with the medical image 1 to obtain a registration result; a second delineation unit, which is used to delineate a region of interest on the medical image 2 based on the registration result in the delineation result 1 in the delineation platform to obtain a delineation result 2; a determination unit, which is used to determine the target radiotherapy plan based on the delineation result 2 and the reference plan.

[0015] Furthermore, the device also includes: a review unit, used to review the target radiotherapy plan after the target radiotherapy plan is determined based on the second outlining result and the reference plan; a first acquisition unit, used to obtain the second outlining result through the outlining platform after the target radiotherapy plan passes the review, and send the second outlining result to the model training platform, and perform learning and training based on the second outlining result in the model training platform to generate a target model; a first sending unit, used to send the target model to the outlining platform.

[0016] Furthermore, the device also includes: a second acquisition unit, used to acquire the medical image of the target patient on the day after sending the target model to the delineation platform and before the target patient undergoes the Nth radiotherapy, wherein N is a natural number greater than 1; a third delineation unit, used to use the target model to delineate the region of interest of the medical image of the day on the delineation platform to obtain a target delineation result; and a second construction unit, used to generate a target radiotherapy plan for the target patient on the day based on the target delineation result.

[0017] Furthermore, the device also includes: a training unit, which is used to send the second outlining result and the target outlining result to the model training platform after generating the target radiotherapy plan for the target patient on that day based on the target outlining result, and perform learning and training based on the second outlining result and the target outlining result in the model training platform to update the target model; a second sending unit, which is used to send the updated target model to the outlining platform.

[0018] Furthermore, the training unit includes: a first acquisition subunit, used to obtain a training data set, wherein the training data set consists of the delineation result two and the target delineation result; a second acquisition subunit, used to construct a subunit, used to construct a student model and a teacher model, and use the pre-trained weights to initialize the student model and the teacher model, and enter the following training cycle: an extraction subunit, used to use the student model and the teacher model to extract features from the images in the training data set respectively, to obtain feature one and feature two; a first calculation subunit, used to calculate the feature one and the delineation result in the training data set to obtain constraint condition one; a second calculation subunit, used to calculate the feature one and the feature two respectively Calculate the description operator to obtain description one and description two, wherein the description operator is at least one of the following: volume, center of mass, distance from each pixel to the center of mass, contour length, calculate the description one and the description two to obtain constraint condition two; a first updating subunit is used to update the weight of the student model using the constraint condition one and the constraint condition two; a second updating subunit is used to update the teacher model using the weight of the updated student model, wherein the updating method is an exponential moving average method; repeat the extraction subunit, the first calculation subunit, the second calculation subunit, the first updating subunit and the second updating subunit until the training is completed, and the target model is generated by the teacher model.

[0019] Further, the method for obtaining the pre-training weights is at least one of the following: supervised learning, semi-supervised learning, and unsupervised learning. Further, if the method for obtaining the pre-training weights is unsupervised learning, the construction subunit includes: an acquisition module for obtaining data of medical images of various different modalities as a training set; an extraction module for randomly extracting a sample from the training set before the iteration starts, and performing two random enhancements on the sample to obtain two enhanced samples A and B; an input module for inputting samples A and B into the encoding model to obtain feature codes X and Y; a calculation module for inputting feature codes X and Y into the projection model to obtain projected feature codes XX and YY, and calculating the Euclidean distance or cosine similarity of XX and YY as a loss function; an update module for updating the student model using the loss function to obtain the weight of the student model.

[0020] Furthermore, the device also includes: a detection unit, which is used to modify the outline result two to obtain a modified outline result two in response to a modification instruction for the outline result before determining the target radiotherapy plan based on the outline result two and the reference plan; the determination unit is also used to determine the target radiotherapy plan based on the modified outline result two and the reference plan.

[0021] In order to achieve the above-mentioned purpose, according to another aspect of the present application, a computer-readable storage medium is further provided, wherein the storage medium stores a program, wherein when the program is executed, the device where the storage medium is located is controlled to execute any one of the above-mentioned methods for determining a radiotherapy plan.

[0022] In order to achieve the above objective, according to another aspect of the present application, a processor is further provided, wherein the processor is used to run a program, wherein when the program is run, any one of the above-mentioned methods for determining a radiotherapy plan is executed.

[0023] Through this application, the following steps are adopted: take a medical image 1 for reference planning of the target patient, and outline the region of interest of the medical image 1 to obtain an outline result 1; generate a radiotherapy plan based on the outline result 1 to obtain a reference plan; review the reference plan, and if the reference plan passes the review, take a medical image 2 of the target patient, and align the medical image 2 with the medical image 1 to obtain an alignment result, wherein the medical image 2 and the medical image 1 are medical images of different modalities; in the outline platform, outline the region of interest of the medical image 2 based on the alignment result to obtain an outline result 2; determine the target radiotherapy plan based on the outline result 2 and the reference plan, and solve the problem that the design of the adaptive radiotherapy plan in the related technology is too time-consuming. The entire process is automated, avoiding manual outlining, reducing the workload of doctors and physicists, and reducing the time for designing adaptive radiotherapy plans, thereby greatly shortening the time required for the adaptive radiotherapy process. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0025] Figure 1 It is a schematic diagram of the process of MR-linac in the prior art;

[0026] Figure 2 is a flow chart of a method for determining a radiotherapy plan according to an embodiment of the present application;

[0027] Figure 3 is a schematic diagram of an optional method for determining a radiotherapy plan according to an embodiment of the present application Figure 1 ;

[0028] Figure 4 is a schematic diagram of an optional method for determining a radiotherapy plan according to an embodiment of the present application Figure 2 ;

[0029] Figure 5It is a schematic diagram of a device for determining a radiotherapy plan according to an embodiment of the present application. DETAILED DESCRIPTION

[0030] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0031] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0032] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0033] The present invention is described below in conjunction with preferred implementation steps. Figure 2 is a flow chart of a method for determining a radiotherapy plan according to an embodiment of the present application, such as Figure 2 As shown, the method comprises the following steps:

[0034] Step S101 , taking a medical image 1 of a target patient for reference planning, and outlining a region of interest on the medical image 1 to obtain an outlining result 1 .

[0035] For example, the first medical image used for reference planning may be a CT image, and the above-mentioned region of interest may be a critical organ, a lesion and / or a target area. The region of interest is delineated on the CT image taken of the target patient to obtain a delineation result 1. It should be noted that in the present application, the first medical image is not limited to a CT image, and may also be other medical images that can be used as positioning images for the target patient, which is not limited in the present application.

[0036] Step S102, generating a radiotherapy plan according to the delineation result to obtain a reference plan.

[0037] Step S103, reviewing the reference plan, and if the reference plan passes the review, taking a second medical image of the target patient, registering the second medical image with the first medical image, and obtaining a registration result.

[0038] The above-mentioned review can be the approval of the reference plan after being reviewed by a doctor, and the second medical image can be an MR image. Of course, the second medical image is not limited to an MR image and is not limited in the present application.

[0039] Step S104: In the delineation platform, the first delineation result delineates the region of interest of the second medical image based on the registration result to obtain the second delineation result.

[0040] Step S105, determining the target radiotherapy plan according to the second delineation result and the reference plan.

[0041] The above steps S101 to S105 automate the entire process, avoiding manual outlining, reducing the workload of doctors and physicists, and reducing the time for designing adaptive radiotherapy plans, thereby significantly shortening the time required for the adaptive radiotherapy process.

[0042] Optionally, in the method for determining a radiotherapy plan provided in an embodiment of the present application, after determining the target radiotherapy plan based on the outlining result two and the reference plan, the method further includes: reviewing the target radiotherapy plan; after the target radiotherapy plan passes the review, obtaining the outlining result two through the outlining platform, and sending the outlining result two to the model training platform, performing learning and training based on the outlining result two in the model training platform to generate a target model; and sending the target model to the outlining platform.

[0043] Through the above scheme, after obtaining the target radiotherapy plan, it can be reviewed and approved by the doctor. After the target radiotherapy plan is reviewed, the delineation result 2 is obtained through the delineation platform. The model training platform is trained based on the delineation result 2 to generate a target model, and the target model is sent to the delineation platform. Subsequently, the target model is used on the delineation platform to delineate the medical images that need to be delineated, thereby improving the delineation efficiency of medical images.

[0044] Optionally, in the method for determining the radiotherapy plan provided in an embodiment of the present application, after sending the target model to the delineation platform, the method further includes: before the target patient undergoes the Nth radiotherapy, obtaining the medical image of the target patient on that day, where N is a natural number greater than 1; in the delineation platform, using the target model to delineate the region of interest in the medical image of that day to obtain a target delineation result; based on the target delineation result, generating a target radiotherapy plan for the target patient on that day.

[0045] Before the target patient undergoes the Nth radiotherapy, medical images of the target patient will be taken on that day. For example, MR images of the target patient will be taken on that day. In the delineation platform, a pre-generated target model is used to delineate the region of interest in the MR image to obtain a target delineation result. Based on the target delineation result, a target radiotherapy plan for the target patient on that day is generated. Therefore, the technical effect of quickly determining the target radiotherapy plan for the target patient on that day based on the target patient's medical images on that day is achieved.

[0046] Optionally, in the method for determining the radiotherapy plan provided in an embodiment of the present application, after generating the target radiotherapy plan for the target patient on that day based on the target delineation result, the method further includes: sending the second delineation result and the target delineation result to a model training platform, performing learning and training on the model training platform based on the second delineation result and the target delineation result to update the target model; and sending the updated target model to the delineation platform.

[0047] In order to ensure the accuracy of the target model outlining, in the present application, after generating the target radiotherapy plan for the target patient on that day, the target model is iterated based on the outlining results, and learning and training are performed on the model training platform based on the outlining result 2 and the target outlining result, and the target model is updated, and the updated target model is sent to the outlining platform, thereby improving the accuracy of the target model outlining.

[0048] Optionally, in the method for determining the radiotherapy plan provided in the embodiment of the present application, learning and training are performed in the model training platform based on the second delineation result and the target delineation result to update the target model, including: obtaining a training data set, wherein the training data set consists of the second delineation result and the target delineation result; constructing a student model and a teacher model, and using pre-trained weights to initialize the student model and the teacher model, and entering the following training cycle: I, using the student model and the teacher model to extract features from the images in the training data set respectively to obtain feature one and feature two; II, calculating feature one and the delineation result in the training data set Calculate to obtain constraint one; III, calculate the description operator for feature one and feature two respectively, and obtain description one and description two, wherein the description operator is at least one of the following: volume, center of mass, distance from each pixel to the center of mass, contour length, calculate description one and description two, and obtain constraint two; IV, use constraint one and constraint two to update the weight of the student model; V, use the updated weight of the student model to update the teacher model, wherein the updating method is the exponential sliding average method; repeat steps I to V until the training is completed, and the target model is generated by the teacher model.

[0049] Through the above scheme, a method of obtaining a target model is proposed, which ensures the feasibility of the scheme, thereby ensuring that the target model can be used to improve the delineation efficiency in the future.

[0050] Optionally, in the method for determining a radiotherapy plan provided in an embodiment of the present application, a method for obtaining pre-trained weights is at least one of the following: supervised learning, semi-supervised learning, and unsupervised learning.

[0051] Optionally, in the method for determining the radiotherapy plan provided in the embodiment of the present application, if the method for obtaining the pre-trained weights is unsupervised learning, obtaining the weights of the student model includes: obtaining data of medical images of multiple different modalities as a training set; before the iteration starts, randomly extracting a sample from the training set, and randomly enhancing the sample twice to obtain two enhanced samples A and B; inputting samples A and B into the encoding model to obtain feature codes X and Y; inputting feature codes X and Y into the projection model to obtain projected feature codes XX and YY, and calculating the Euclidean distance or cosine similarity of XX and YY as a loss function; and using the loss function to update the student model to obtain the weight of the student model.

[0052] Through the above scheme, a method for obtaining the weight of the student model is proposed, which ensures the feasibility of the scheme, thereby ensuring that the weight of the student model can be used to update the target model in the future.

[0053] Optionally, in the method for determining a radiotherapy plan provided in an embodiment of the present application, before determining a target radiotherapy plan based on outlining result two and a reference plan, the method further includes: if a modification instruction for outlining result two is detected; in response to the modification instruction, modifying outlining result two to obtain a modified outlining result two; determining a target radiotherapy plan based on outlining result two and a reference plan includes: determining the target radiotherapy plan based on the modified outlining result two and the reference plan.

[0054] In the above scheme, before determining the target radiotherapy plan based on the outlining result 2 and the reference plan, if it is detected that the doctor or physicist has modified the outlining result 2, the target radiotherapy plan is determined based on the modified outlining result 2 and the reference plan, thereby ensuring the accuracy of the generated target radiotherapy plan.

[0055] In summary, the method for determining the radiotherapy plan provided in the embodiment of the present application is to obtain a delineation result 1 by taking a medical image 1 for reference planning of the target patient, and outlining the region of interest of the medical image 1; generating a radiotherapy plan based on the outlining result 1 to obtain a reference plan; reviewing the reference plan, and if the reference plan passes the review, taking a medical image 2 of the target patient, registering the medical image 2 with the medical image 1 to obtain a registration result, wherein the medical image 2 and the medical image 1 are medical images of different modalities; in the outlining platform, outlining the region of interest of the medical image 2 based on the registration result to obtain an outlining result 2; determining the target radiotherapy plan based on the outlining result 2 and the reference plan solves the problem that the design of the adaptive radiotherapy plan in the related technology is too time-consuming. The entire process is automated, avoiding manual outlining, reducing the workload of doctors and physicists, and reducing the time for designing the adaptive radiotherapy plan, thereby greatly shortening the time required for the adaptive radiotherapy process.

[0056] Optional, such as Figure 3 As shown, the process of constructing the outline model for the first time (corresponding to the target model mentioned above) can be as follows:

[0057] After the patient has taken the positioning CT, he or she obtains image CT1, and then takes a set of MR images on Unity (a medical image acquisition platform) to obtain image MR1;

[0058] Push CT1 to AccuContour (corresponding to the above-mentioned delineation platform) through the dicom3.0 protocol for automatic delineation to obtain Rt1, and push CT1 and Rt1 to Offline Monaco (planning system);

[0059] The plan is made using CT1 in Offline Monaco and approved by the doctor after review, generating the approved Plan1, Dose1, and Rt1. At this time, AccuContour will automatically grab the CT1, Plan1, Dose1, and Rt1 files;

[0060] Push MR1 to AccuContour and Offline Monaco. In AccuContour, MR1 and CT1 will be automatically registered (flexible registration), and RT1 will be generated as MR1 / w RT1 based on the registration result, and then pushed to OfflineMonaco.

[0061] Modify MR / w RT1 in Offline Monaco to get MR / w RT1+, use CT1, Plan1, Dose1, MR1, and MR / w RT1+ to optimize the ATS plan and get the Plan (corresponding to the target radiotherapy plan mentioned above), verify the Dose through ArtCheck, and approve it after being reviewed by the doctor. The Plan obtained this time is not used for treatment;

[0062] AccuContour automatically captures MR1 ​​and the approved MR / w RT1+, pushes them to AccuLeaning (corresponding to the above-mentioned model training platform) for model training, obtains model 0, and automatically pushes model 0 to AccuContour.

[0063] like Figure 4 As shown, the Nth treatment of the patient (N is a natural number greater than 1) and the iterative model training process can be as follows:

[0064] Shoot MRn on Unity and push it to AccuContour. AccuContour automatically outlines MRn based on the existing model (n-1) to obtain MRn w / RTn and push it to Offline Monaco.

[0065] Modify MR / w RTn in Offline Monaco (if necessary) to get MR / w RTn+, use CT1, Plan1, Dose1, MRn, MR / w RTn+ to optimize the ATS plan, get Plan(n+1), pass ArtCheck verification, and perform approval operations;

[0066] AccuContour automatically captures MRn and the approved MR / w RTn+, pushes them to AccuLeaning for model training, obtains modeln (corresponding to the updated target model above), and automatically pushes modeln to AccuContour.

[0067] In the above scheme, the entire process is automated (no manual operation is required), and the overall ATS treatment time is shortened by more than half. Specifically, the time for automatic outlining by the outlining platform is <1min, including the overall time of modification confirmation and data transmission <=5min, and the automatic check of the day's plan by ARTcheck takes about 15 seconds. In the scheme of the present application, a QA report can also be generated. Specifically, ARTcheck automatically synchronizes the Approved plan in Online Monaco and generates a QA report, and users can view and export the QA report at any time. In addition, the scheme in this application can fully utilize Monaco's existing interface without affecting the existing process. The deployment can be completed by simply connecting the workstation to the work network, so it is convenient to use.

[0068] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0069] The embodiment of the present application also provides a method and device for determining a radiotherapy plan. It should be noted that the device for determining a radiotherapy plan in the embodiment of the present application can be used to execute the method for determining a radiotherapy plan provided in the embodiment of the present application. The device for determining a radiotherapy plan provided in the embodiment of the present application is introduced below.

[0070] Figure 5 Schematic diagram of a device for determining a radiotherapy plan according to an embodiment of the present application. Figure 5 As shown, the device includes: a first delineating unit 201, a first constructing unit 202, a registration unit 203, a second delineating unit 204 and a determining unit 205.

[0071] Specifically, the first delineation unit 201 is used to take a medical image 1 of a target patient for reference planning, and delineate a region of interest on the medical image 1 to obtain a delineation result 1.

[0072] The first constructing unit 202 is used to generate a radiotherapy plan according to the delineation result to obtain a reference plan.

[0073] The registration unit 203 is used to review the reference plan. When the reference plan passes the review, a second medical image is taken of the target patient, and the second medical image is registered with the first medical image to obtain a registration result.

[0074] The second delineation unit 204 is used to perform delineation of a region of interest on the medical image 2 based on the registration result in the delineation platform to obtain a second delineation result.

[0075] The determination unit 205 is used to determine the target radiotherapy plan according to the second delineation result and the reference plan.

[0076] The apparatus for determining a radiotherapy plan provided in an embodiment of the present application uses a first delineation unit 201 to shoot a medical image 1 for a target patient for reference planning, and delineates a region of interest on the medical image 1 to obtain a delineation result 1. The first construction unit 202 generates a radiotherapy plan based on the delineation result 1 to obtain a reference plan. The registration unit 203 reviews the reference plan, and if the reference plan passes the review, shoots a medical image 2 for the target patient, and registers the medical image 2 with the medical image 1 to obtain a registration result. In the delineation platform, the second delineation unit 204 delineates a region of interest on the medical image 2 based on the registration result to obtain a delineation result 2. The determination unit 205 determines the target radiotherapy plan based on the delineation result 2 and the reference plan, solving the problem that the design of an adaptive radiotherapy plan in the related art is too time-consuming. The entire process is automated, avoiding manual delineation, reducing the workload of doctors and physicists, reducing the time for designing an adaptive radiotherapy plan, and thus greatly shortening the time required for the adaptive radiotherapy process.

[0077] Optionally, in the device for determining the radiotherapy plan provided in the embodiment of the present application, the device also includes: a review unit, used to review the target radiotherapy plan after the target radiotherapy plan is determined based on the outline result two and the reference plan; a first acquisition unit, used to obtain the outline result two through the outline platform after the target radiotherapy plan passes the review, and send the outline result two to the model training platform, and perform learning and training based on the outline result two in the model training platform to generate a target model; a first sending unit, used to send the target model to the outline platform.

[0078] Optionally, in the device for determining the radiotherapy plan provided in the embodiment of the present application, the device also includes: a second acquisition unit, used to acquire the medical image of the target patient on that day after sending the target model to the delineation platform and before the target patient undergoes the Nth radiotherapy, wherein N is a natural number greater than 1; a third delineation unit, used to use the target model to delineate the region of interest of the medical image of that day in the delineation platform to obtain a target delineation result; and a second construction unit, used to generate a target radiotherapy plan for the target patient on that day based on the target delineation result.

[0079] Optionally, in the device for determining the radiotherapy plan provided in the embodiment of the present application, the device also includes: a training unit, which is used to send the second outline result and the target outline result to the model training platform after generating the target radiotherapy plan for the target patient on the day based on the target outline result, and perform learning and training based on the second outline result and the target outline result in the model training platform to update the target model; a second sending unit, which is used to send the updated target model to the outline platform.

[0080] Optionally, in the apparatus for determining a radiotherapy plan provided in an embodiment of the present application, the training unit includes: a first acquisition subunit, used to obtain a training data set, wherein the training data set consists of a delineation result two and a target delineation result; a second acquisition subunit, used to construct a subunit, used to construct a student model and a teacher model, and use pre-trained weights to initialize the student model and the teacher model, and enter the following training loop: an extraction subunit, used to use the student model and the teacher model to perform feature extraction on the images in the training data set, respectively, to obtain feature one and feature two; a first calculation subunit, used to calculate feature one and the delineation result in the training data set, to obtain constraint one; a second calculation subunit, used to respectively Perform description operator calculation on feature one and feature two to obtain description one and description two, wherein the description operator is at least one of the following: volume, center of mass, distance from each pixel to the center of mass, contour length, and calculate description one and description two to obtain constraint two; a first updating subunit is used to update the weight of the student model using constraint one and constraint two; a second updating subunit is used to update the teacher model using the weight of the updated student model, wherein the updating method is an exponential sliding average method; the extraction subunit, the first calculation subunit, the second calculation subunit, the first updating subunit and the second updating subunit are repeated until the training is completed, and the target model is generated by the teacher model.

[0081] Optionally, in the apparatus for determining a radiotherapy plan provided in an embodiment of the present application, a method for obtaining pre-trained weights is at least one of the following: supervised learning, semi-supervised learning, and unsupervised learning.

[0082] Optionally, in the device for determining the radiotherapy plan provided in the embodiment of the present application, if the method for obtaining the pre-trained weights is unsupervised learning, the construction subunit includes: an acquisition module, used to obtain data of medical images of multiple different modalities as a training set; an extraction module, used to randomly extract a sample from the training set before the iteration starts, and randomly enhance the sample twice to obtain two enhanced samples A and B; an input module, used to input samples A and B into the encoding model to obtain feature codes X and Y; a calculation module, used to input feature codes X and Y into the projection model to obtain projected feature codes XX and YY, and calculate the Euclidean distance or cosine similarity of XX and YY as the loss function; an update module, used to update the student model using the loss function to obtain the weight of the student model.

[0083] Optionally, in the device for determining the radiotherapy plan provided in the embodiment of the present application, the device also includes: a detection unit, which is used to, before determining the target radiotherapy plan based on the outlining result two and the reference plan, if a modification instruction for the outlining result two is detected; in response to the modification instruction, modify the outlining result two to obtain a modified outlining result two; and the determination unit is also used to determine the target radiotherapy plan based on the modified outlining result two and the reference plan.

[0084] The device for determining the radiotherapy plan includes a processor and a memory. The first delineation unit 201, the first construction unit 202, the registration unit 203, the second delineation unit 204 and the determination unit 205 are all stored in the memory as program units, and the processor executes the program units stored in the memory to realize corresponding functions.

[0085] The processor includes a kernel, which calls the corresponding program unit from the memory. One or more kernels can be set, and the time for designing an adaptive radiotherapy plan is reduced by adjusting kernel parameters.

[0086] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0087] An embodiment of the present invention provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, the method for determining a radiotherapy plan is implemented.

[0088] An embodiment of the present invention provides a processor, which is used to run a program, wherein the method for determining a radiotherapy plan is executed when the program is run.

[0089] An embodiment of the present invention provides a device, which includes a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, the following steps are implemented: a medical image 1 is taken of a target patient for reference planning, and a region of interest is outlined on the medical image 1 to obtain a first outline result; a radiotherapy plan is generated based on the first outline result to obtain a reference plan; the reference plan is reviewed, and when the reference plan passes the review, a medical image 2 is taken of the target patient, and the medical image 2 is registered with the medical image 1 to obtain a registration result; in a delineation platform, based on the first outline result, a region of interest is outlined on the medical image 2 to obtain a second outline result; and a target radiotherapy plan is determined based on the second outline result and the reference plan.

[0090] Optionally, after determining the target radiotherapy plan based on the outlining result two and the reference plan, the method also includes: reviewing the target radiotherapy plan; after the target radiotherapy plan passes the review, obtaining the outlining result two through the outlining platform, and sending the outlining result two to the model training platform, performing learning and training based on the outlining result two in the model training platform to generate a target model; and sending the target model to the outlining platform.

[0091] Optionally, after sending the target model to the delineation platform, the method further includes: before the target patient undergoes the Nth radiotherapy, obtaining the medical image of the target patient on that day, where N is a natural number greater than 1; in the delineation platform, using the target model to delineate the region of interest in the medical image of that day to obtain a target delineation result; based on the target delineation result, generating a target radiotherapy plan for the target patient on that day.

[0092] Optionally, after generating the target radiotherapy plan for the target patient on that day based on the target outlining result, the method also includes: sending the second outlining result and the target outlining result to the model training platform, performing learning and training based on the second outlining result and the target outlining result in the model training platform to update the target model; and sending the updated target model to the outlining platform.

[0093] Optionally, learning and training are performed in the model training platform based on the second outlining result and the target outlining result to update the target model, including: obtaining a training data set, wherein the training data set consists of the second outlining result and the target outlining result; constructing a student model and a teacher model, and using pre-trained weights to initialize the student model and the teacher model, and entering the following training cycle: I, using the student model and the teacher model to extract features from the images in the training data set respectively to obtain feature one and feature two; II, calculating feature one and the outlining results in the training data set to obtain constraint one; III, performing description operator calculation on feature one and feature two respectively to obtain description one and description two, wherein the description operator is at least one of the following: volume, center of mass, distance from each pixel to the center of mass, and contour length, and calculating description one and description two to obtain constraint two; IV, using constraint one and constraint two to update the weight of the student model; V, using the updated weight of the student model to update the teacher model, wherein the updating method is the exponential sliding average method; repeating steps I to V until the training is completed, and the target model is generated by the teacher model.

[0094] Optionally, the method for obtaining the pre-training weights is at least one of the following: supervised learning, semi-supervised learning, and unsupervised learning.

[0095] Optionally, if the method for obtaining pre-trained weights is unsupervised learning, obtaining the weights of the student model includes: obtaining data of medical images of multiple different modalities as a training set; before the start of iteration, randomly extracting a sample from the training set, and randomly enhancing the sample twice to obtain two enhanced samples A and B; inputting samples A and B into the encoding model to obtain feature codes X and Y; inputting feature codes X and Y into the projection model to obtain projected feature codes XX and YY, and calculating the Euclidean distance or cosine similarity of XX and YY as a loss function; and using the loss function to update the student model to obtain the weight of the student model.

[0096] Optionally, before determining the target radiotherapy plan based on the second delineation result and the reference plan, the method further includes: if a modification instruction for the second delineation result is detected; in response to the modification instruction, modifying the second delineation result to obtain a modified second delineation result; determining the target radiotherapy plan based on the second delineation result and the reference plan includes: determining the target radiotherapy plan based on the modified second delineation result and the reference plan. The device in this article can be a server, a PC, a PAD, a mobile phone, etc.

[0097] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program that is initialized with the following method steps: taking a medical image one for a target patient to make a reference plan, and outlining a region of interest on the medical image one to obtain an outlining result one; generating a radiotherapy plan based on the outlining result one to obtain a reference plan; reviewing the reference plan, and if the reference plan passes the review, taking a medical image two for the target patient, and registering the medical image two with the medical image one to obtain a registration result; in an outlining platform, outlining a region of interest on the medical image two based on the registration result to obtain an outlining result two; and determining a target radiotherapy plan based on the outlining result two and the reference plan.

[0098] Optionally, after determining the target radiotherapy plan based on the outlining result two and the reference plan, the method also includes: reviewing the target radiotherapy plan; after the target radiotherapy plan passes the review, obtaining the outlining result two through the outlining platform, and sending the outlining result two to the model training platform, performing learning and training based on the outlining result two in the model training platform to generate a target model; and sending the target model to the outlining platform.

[0099] Optionally, after sending the target model to the delineation platform, the method further includes: before the target patient undergoes the Nth radiotherapy, obtaining the medical image of the target patient on that day, where N is a natural number greater than 1; in the delineation platform, using the target model to delineate the region of interest in the medical image of that day to obtain a target delineation result; based on the target delineation result, generating a target radiotherapy plan for the target patient on that day.

[0100] Optionally, after generating the target radiotherapy plan for the target patient on that day based on the target outlining result, the method also includes: sending the second outlining result and the target outlining result to the model training platform, performing learning and training based on the second outlining result and the target outlining result in the model training platform to update the target model; and sending the updated target model to the outlining platform.

[0101] Optionally, learning and training are performed in the model training platform based on the second outlining result and the target outlining result to update the target model, including: obtaining a training data set, wherein the training data set consists of the second outlining result and the target outlining result; constructing a student model and a teacher model, and using pre-trained weights to initialize the student model and the teacher model, and entering the following training cycle: I, using the student model and the teacher model to extract features from the images in the training data set respectively to obtain feature one and feature two; II, calculating feature one and the outlining results in the training data set to obtain constraint one; III, performing description operator calculation on feature one and feature two respectively to obtain description one and description two, wherein the description operator is at least one of the following: volume, center of mass, distance from each pixel to the center of mass, and contour length, and calculating description one and description two to obtain constraint two; IV, using constraint one and constraint two to update the weight of the student model; V, using the updated weight of the student model to update the teacher model, wherein the updating method is the exponential sliding average method; repeating steps I to V until the training is completed, and the target model is generated by the teacher model.

[0102] Optionally, the method for obtaining the pre-training weights is at least one of the following: supervised learning, semi-supervised learning, and unsupervised learning.

[0103] Optionally, if the method for obtaining pre-trained weights is unsupervised learning, obtaining the weights of the student model includes: obtaining data of medical images of multiple different modalities as a training set; before the start of iteration, randomly extracting a sample from the training set, and randomly enhancing the sample twice to obtain two enhanced samples A and B; inputting samples A and B into the encoding model to obtain feature codes X and Y; inputting feature codes X and Y into the projection model to obtain projected feature codes XX and YY, and calculating the Euclidean distance or cosine similarity of XX and YY as a loss function; and using the loss function to update the student model to obtain the weight of the student model.

[0104] Optionally, before determining the target radiotherapy plan based on the outlining result two and the reference plan, the method also includes: if a modification instruction for the outlining result two is detected; in response to the modification instruction, modify the outlining result two to obtain a modified outlining result two; determining the target radiotherapy plan based on the outlining result two and the reference plan includes: determining the target radiotherapy plan based on the modified outlining result two and the reference plan.

[0105] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0106] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0107] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0108] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0109] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0110] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0111] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0112] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0113] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0114] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for determining a radiotherapy plan, characterized in that: include: Taking a medical image 1 for reference planning of a target patient, and outlining a region of interest on the medical image 1 to obtain an outlining result 1; generating a radiotherapy plan according to the delineation result to obtain a reference plan; Reviewing the reference plan, and if the reference plan passes the review, photographing a second medical image of the target patient, registering the second medical image with the first medical image, and obtaining a registration result; In the delineation platform, the first delineation result delineates a region of interest on the second medical image based on the registration result to obtain a second delineation result; Determining a target radiotherapy plan according to the second delineation result and the reference plan; Before the target patient undergoes the Nth radiotherapy, the medical image of the day is delineated by the target model in the delineation platform to obtain a target delineation result, wherein the target model is obtained by learning and training based on the second delineation result; Based on obtaining a training data set, wherein the training data set consists of the second delineation result and the target delineation result; Construct a student model and a teacher model, and use the pre-trained weights to initialize the student model and the teacher model, and enter the following training cycle: I, using the student model and the teacher model to extract features of the images in the training data set respectively to obtain feature one and feature two; II, calculating the feature one and the delineation results in the training data set to obtain constraint condition one; III, respectively calculating the description operator on the feature one and the feature two to obtain description one and description two, wherein the description operator is at least one of the following: volume, centroid, distance from each pixel to the centroid, and contour length, and calculating the description one and the description two to obtain constraint condition two; IV, using the constraint condition 1 and the constraint condition 2 to update the weight of the student model; V, using the updated weight of the student model to update the teacher model, wherein the updating method is an exponential moving average method; Repeat steps I to V until the training is completed and the target model is generated by the teacher model.

2. The method according to claim 1, characterized in that After determining the target radiotherapy plan according to the second delineation result and the reference plan, the method further includes: Reviewing the target radiotherapy plan; After the target radiotherapy plan passes the review, the second delineation result is obtained through the delineation platform, and the second delineation result is sent to the model training platform, and learning and training are performed on the model training platform according to the second delineation result to generate a target model; The target model is sent to the delineation platform.

3. The method according to claim 2, characterized in that After sending the target model to the delineation platform, the method further comprises: Before the target patient undergoes the Nth radiotherapy, obtaining the target patient's medical image on the same day, where N is a natural number greater than 1; In the delineation platform, the target model is used to delineate the region of interest of the medical image of the day to obtain a target delineation result; Based on the target delineation result, a target radiotherapy plan for the target patient on that day is generated.

4. The method according to claim 3, characterized in that After generating a target radiotherapy plan for the target patient on that day based on the target delineation result, the method further includes: Sending the second delineation result and the target delineation result to the model training platform, and performing learning and training on the model training platform according to the second delineation result and the target delineation result to update the target model; The updated target model is sent to the delineation platform.

5. The method according to claim 1, characterized in that The method for obtaining the pre-training weights is at least one of the following: supervised learning, semi-supervised learning, and unsupervised learning.

6. The method according to claim 5, characterized in that If the method for obtaining pre-training weights is unsupervised learning, obtaining the weights of the student model includes: Obtain medical imaging data of various modalities as training sets; Before the iteration starts, a sample in the training set is randomly selected, and the sample is randomly enhanced twice to obtain two enhanced samples A and B; Input samples A and B into the encoding model to obtain feature codes X and Y; Input feature codes X and Y into the projection model to obtain projected feature codes XX and YY, and calculate the Euclidean distance or cosine similarity of XX and YY as the loss function; The loss function is used to update the student model to obtain the weight of the student model.

7. The method according to claim 1, characterized in that Before determining the target radiotherapy plan according to the second delineation result and the reference plan, the method further includes: if a modification instruction for the delineation result is detected; in response to the modification instruction, modifying the second delineation result to obtain a modified second delineation result; Determining a target radiotherapy plan based on the second delineation result and the reference plan includes: determining a target radiotherapy plan based on the modified second delineation result and the reference plan.

8. A device for determining a radiotherapy plan, characterized in that: include: A first delineation unit is used to take a medical image 1 of a target patient for reference planning, and delineate a region of interest on the medical image 1 to obtain a delineation result 1; A first construction unit is used to generate a radiotherapy plan according to the delineation result to obtain a reference plan; a registration unit, configured to review the reference plan, and if the reference plan passes the review, photograph a second medical image of the target patient, and register the second medical image with the first medical image to obtain a registration result, wherein the second medical image and the first medical image are medical images of different modalities; A second delineation unit is used for delineating a region of interest on the medical image 2 based on the registration result in the delineation result 1 in the delineation platform to obtain a second delineation result; a determination unit, configured to determine a target radiotherapy plan according to the second delineation result and the reference plan; The device is further used to delineate the medical image of the day by using the target model in the delineation platform before the target patient undergoes the Nth radiotherapy to obtain a target delineation result, wherein the target model is obtained by learning and training based on the second delineation result; Based on obtaining a training data set, wherein the training data set consists of the second delineation result and the target delineation result; Construct a student model and a teacher model, and use the pre-trained weights to initialize the student model and the teacher model, and enter the following training cycle: I, using the student model and the teacher model to extract features of the images in the training data set respectively to obtain feature one and feature two; II, calculating the feature one and the delineation results in the training data set to obtain constraint condition one; III, respectively calculating the description operator on the feature one and the feature two to obtain description one and description two, wherein the description operator is at least one of the following: volume, centroid, distance from each pixel to the centroid, and contour length, and calculating the description one and the description two to obtain constraint condition two; IV, using the constraint condition 1 and the constraint condition 2 to update the weight of the student model; V, using the updated weight of the student model to update the teacher model, wherein the updating method is an exponential moving average method; Repeat steps I to V until the training is completed and the target model is generated by the teacher model.

9. A computer-readable storage medium, characterized in that: The storage medium stores a program, wherein the program executes the method for determining a radiotherapy plan according to any one of claims 1 to 7.

10. A processor, characterized in that: The processor is used to run a program, wherein the program, when running, executes the method for determining a radiotherapy plan according to any one of claims 1 to 7.

Images