Brachytherapy automatic planning method, device and system, and storage medium

Abstract

The invention discloses an automatic planning method, device and system for brachytherapy, and a storage medium. Acquiring a radiation plan and a structure file; preprocessing the radiation plan and the structure file; inputting the preprocessed radiation plan and the structure file into a visual language large model, and obtaining residence time and residence position information required by the plan through multiple rounds of iteration; performing dose calculation on the residence time and the position generated by each round of iteration according to TG43, comparing with a radiation plan requirement, feeding back information and indication after comparison to the visual language large model, performing further optimization iteration, generating new residence time and position until the radiation plan requirement is met, and obtaining a basic plan; the basic plan is optimized and upgraded, and a better plan meeting the individual requirements of the patient is obtained. By adopting the technical scheme, the plan making time of a physicist is shortened, the clinical efficiency is improved, and the plan quality is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of brachytherapy, and particularly relates to an automatic planning method and device, system, and storage medium for brachytherapy. Background Art

[0002] High-dose rate brachytherapy (HDR-BT) is a specialized radiotherapy technique used to treat various cancers. It delivers high-dose radiation directly to the tumor site, minimizing radiation exposure to surrounding healthy tissues. HDR-BT is commonly used as a primary treatment method or in combination with other therapies and has become the standard treatment for cancers requiring high-dose radiation.

[0003] Current HDR-BT methods rely on interactive ("manual") treatment planning, which has limitations such as long time consumption, large workload, lack of consistency due to variability between and within planners, plan quality depending on the planner's experience and time investment, and limited plan reproducibility. In addition, the treatment planning for HDR-BT faces great time pressure because the patient must remain fixed on the treatment couch and wait for the completion of the treatment plan before receiving treatment. The long waiting time can make the patient's legs numb, and involuntary movements increase the risk of the applicator shifting or falling out of the patient's body. This time urgency forces medical physicists to complete the treatment plan within a limited time, thus increasing the risk of planning errors.

[0004] Automated HDR-BT treatment planning is an important solution to address the above challenges. In recent years, many studies have been dedicated to automating the planning process. Significant progress has been made in automatically identifying applicators and automatically segmenting normal tissue organs (OARs) and high-risk clinical target volumes (HRCTVs), improving efficiency and accuracy. In the process of plan generation and evaluation, some studies have explored using deep learning neural networks to predict dose-volume histograms (DVHs) and dose distribution maps. However, although these predictions provide valuable methods, they cannot be directly applied to clinical practice because the treatment machine requires detailed dwell time and position information to accurately execute the plan, and these studies only generate intermediate products of the plan and do not directly generate dwell time and position, so automatic planning cannot be completed. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide an automatic planning method and device, system, and storage medium for brachytherapy.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] An automatic planning method for brachytherapy, comprising:

[0008] Step S1: Obtain the radiotherapy plan and the structure file;

[0009] Step S2: Preprocess the radiotherapy plan and the structure file;

[0010] Step S3: Input the preprocessed radiotherapy plan and structure file into the vision-language large model, and through multiple rounds of iteration, obtain the dwell time and dwell position information required by the plan;

[0011] Step S4: Calculate the dose according to TG43 for the dwell time and position generated in each round of iteration, compare it with the requirements of the radiotherapy plan, and feed back the compared information and instructions to the vision-language large model for further optimization and iteration to generate new dwell time and position until the requirements of the radiotherapy plan are met; obtain the basic plan;

[0012] Step S5: Optimize and upgrade the basic plan to obtain a better plan that meets the personalized needs of the patient.

[0013] Preferably, step S2 includes:

[0014] Step S21: Extract the dose information according to the radiotherapy plan;

[0015] Step S22: Extract the trajectory of the radiation source in the applicator, the contours of the organs at risk and the high-risk clinical target volume according to the structure file;

[0016] Step S23: Obtain the initial dwell position and initial dwell time of each channel of the applicator according to the trajectory of the radiation source in the applicator.

[0017] Preferably, in step S3, the plan requirements, the trajectory of the radiation source in the applicator, the initial dwell position, the initial dwell time, the organs at risk and the contour of the target volume to be treated are input into the vision-language large model, and through multiple rounds of iteration, the dwell time and dwell position information required by the plan are obtained.

[0018] The present invention also provides a brachytherapy automatic planning device, including:

[0019] The first processing module is used to obtain the radiotherapy plan and the structure file;

[0020] The second processing module is used to preprocess the radiotherapy plan and the structure file;

[0021] The third processing module is used to input the preprocessed radiotherapy plan and structure file into the vision-language large model, and through multiple rounds of iteration, obtain the dwell time and dwell position information required by the plan;

[0022] The fourth processing module is used to calculate the dwell time and position generated in each round of iteration according to TG43, compare them with the requirements of the radiotherapy plan, and feedback the compared information and instructions to the vision-language large model for further optimization and iteration to generate new dwell time and position until the requirements of the radiotherapy plan are met; and obtain the basic plan;

[0023] The fifth processing module is used to optimize and upgrade the basic plan to obtain a better plan that meets the personalized needs of the patient.

[0024] Preferably, the second processing module includes:

[0025] The first processing unit is used to extract dose information according to the radiotherapy plan;

[0026] The second processing unit is used to extract the trajectory of the radiation source in the applicator, the contours of the organs at risk and the high-risk clinical target volume according to the structure file;

[0027] The third processing unit is used to obtain the initial dwell position and initial dwell time of each channel of the applicator according to the trajectory of the radiation source in the applicator.

[0028] Preferably, the third processing module inputs the plan requirements, the trajectory of the radiation source in the applicator, the initial dwell position, the initial dwell time, the contours of the organs at risk and the high-risk clinical target volume into the vision-language large model, and through multiple rounds of iteration, obtains the dwell time and dwell position information required by the plan.

[0029] The present invention also provides a brachytherapy automatic planning system, including: a memory and a processor, wherein a computer program is stored on the memory and run by the processor, and the computer program executes the brachytherapy automatic planning method when run by the processor.

[0030] The present invention also provides a storage medium, on which a computer program is stored, and the computer program executes the brachytherapy automatic planning method when running.

[0031] The present invention utilizes a computer vision-language large model to comprehensively analyze the processed patient's image, target position, and applicator position, and directly generates the dwell position and dwell time required by the plan, completing the automatic planning of brachytherapy; reducing the time for physicists to make plans, improving clinical efficiency, and enhancing the plan quality. Description of the Drawings

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on the provided drawings.

[0033] Figure 1 Flowchart of the automatic planning method for brachytherapy in an embodiment of the present invention;

[0034] Figure 2 Flowchart of the optimization iteration algorithm for automatic planning. Detailed implementation manners

[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0036] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementation manners.

[0037] Embodiment 1:

[0038] As Figure 1 shown, an embodiment of the present invention provides an automatic planning method for brachytherapy, including:

[0039] Step S1, obtain the radiation plan (RP.dcm) and the structure file (RS.dcm) of the patient's image data;

[0040] Step S2, preprocess the radiation plan and the structure file;

[0041] Step S3, input the preprocessed radiation plan and structure file into the vision-language large model, and through multiple rounds of iteration, obtain the dwell time and dwell position information required for the plan;

[0042] Step S4, perform dose calculation on the dwell time and position generated in each round of iteration according to TG43, compare with the requirements of the radiation plan, and feedback the compared information and instructions to the vision-language large model for further optimization iteration to generate new dwell time and position until the requirements of the radiation plan are met to obtain the basic plan;

[0043] Step S5, optimize and upgrade the basic plan to obtain a better plan that meets the personalized needs of the patient.

[0044] As an implementation manner of an embodiment of the present invention, step S2 includes:

[0045] Step S21: Extract dose information according to the radiotherapy plan;

[0046] Step S22: Extract the trajectory of the radiation source in the applicator, the contours of the organs at risk (OARs) and the high-risk clinical target volume (HRCTV), and their specific coordinate positions in the space coordinate system according to the structure file. All OARs and HRCTV are plotted on images of different levels according to their z-axis coordinate relationships, and different organs are represented by different colors;

[0047] Step S23: Obtain the initial dwell positions and initial dwell times of each channel of the applicator according to the trajectory of the radiation source in the applicator.

[0048] As an implementation manner of an embodiment of the present invention, in step S3, the plan requirements, the trajectory of the radiation source in the applicator, the initial dwell positions, the initial dwell times, the contours of the organs at risk and the high-risk clinical target volume are input into the vision-language large model, and through multiple rounds of iteration, the dwell times and dwell position information required by the plan are obtained; wherein, the interval distance between the dwell positions is any distance, preferably, the interval distance is 1 millimeter. Further, the vision-language large model randomly assigns the initial dwell times to each dwell position; in subsequent iterations, these dwell times are adjusted according to the feedback of the previous iteration.

[0049] As an implementation manner of an embodiment of the present invention, in step S4, the dwell positions and times are then exported to a dose calculation program to automatically calculate the radiation dose distribution in the patient's body and generate a dose-volume histogram (DVH), dose-volume indices (DVIs), and a dose distribution map. In brachytherapy planning, the dose calculation is based on the TG-43 consensus guidelines. By calculating the dose distribution of the radiation source in the patient's body, the accuracy and safety of the treatment plan are ensured.

[0050] Further, the selected dose parameters can be completed according to user-defined settings. In the embodiment of the present invention, the most commonly used D2cc is adopted for the organs at risk, indicating that the volume in the organ at risk is 2 cm 3(i.e. 2cc) is the highest irradiated dose, and the most commonly used D90% is used for HRCTV, which means the minimum dose received by 90% of the volume of the target area. The number and type of organs at risk can also be customized to ensure that they meet the requirements of the brachytherapy plan (radiotherapy plans can also be customized by the user). If the plan does not meet these standards, the system will generate corresponding feedback and instructions to help the visual language model optimize the plan. These feedbacks and the dose distribution map and dose volume histogram of the previous iteration are returned to the visual language model again. The model then identifies the areas that need adjustment and generates a new dwell time for each dwell position to solve problems such as excessive or insufficient dose parameters for normal tissues, organs and targets.

[0051] Furthermore, all images and textual communication processes and interactive memories are continuously saved throughout the entire process, allowing subsequent iterations to reference previous optimization history and re-optimize dwell time based on this information. Once the plan meets the dose parameters for all target volumes and organs at risk, the final dwell time and dwell position are exported to the treatment machine for planned implementation.

[0052] Furthermore, in step S4, in order to obtain a more reasonable and optimized dose distribution, the target area can be guaranteed to receive a sufficient dose and the organs at risk can be protected from the dose. Figure 2 As shown, an embodiment of the present invention proposes an iterative automatic plan optimization algorithm that employs a round-by-round optimization strategy. By gradually relaxing hard constraints, the algorithm improves optimization feasibility and ensures the quality of the final solution. In the first round of optimization, the initial hard constraints are used for optimization, and a determination is made as to whether the optimization objective is met. If the objective is met, the algorithm proceeds to the next round of optimization, further adjusting the constraints to find a more optimal solution. If not, the algorithm proceeds to the second round of optimization, adjusting the hard constraints to 90% of the original constraints and reoptimizing. If still unsuccessful, the algorithm proceeds to the third round of optimization, adjusting the constraints to 80% of the original constraints and reoptimizing. After each round of optimization, if the optimization succeeds, the algorithm proceeds to the next round of optimization to continuously improve the solution. If the optimization fails, the optimization process terminates, and the optimal solution from the current round is selected as the final output. If all rounds fail to meet the objective, the optimal solution is found in the final round of optimization and output. This method ensures that the optimization process both satisfies the hard constraints as much as possible and allows for appropriate strategy adjustments when necessary to achieve the optimal solution. At the same time, if the user prefers to prioritize protecting a certain critical organ of the patient, taking the bladder as an example, during the optimization process, the optimization target of the bladder is adjusted to 90%, 80%, 70% of the original constraint in succession, and so on, until the optimization cannot be completed and stops.

[0053] As an implementation manner of the embodiment of the present invention, to meet the usage requirements of users, the system can customize plans according to the needs of doctors. When further optimizing a basic plan that already meets the conditions, a plan can be formulated according to the patient's condition. For example, if a certain critical organ (taking the bladder as an example here) of the patient needs special protection, while meeting the requirements of the basic plan, the optimization weight can be tilted towards the bladder to ensure that the bladder receives a minimal dose. Step S4 is to generate a basic plan (for example, the dose received by all critical organs is less than 4 Gy), and the basic plan can meet the basic requirements of the EMBRACE II guideline for plan making and patient protection. Step S5 is to further tighten the optimization conditions on the basis of S4 to generate an optimal plan; among them, there are two options in step S5: 1. The dose protection of all organs is increased by 20%. 2. If a certain organ (such as the bladder) needs to be specially protected according to the patient's personal situation, the protection of the bladder alone is increased to 40%.

[0054] The specific process of the brachytherapy automatic planning method in the embodiment of the present invention includes:

[0055] 1. Data preprocessing

[0056] Identify the position of the applicator, and generate corresponding dwell positions according to the feasible trajectories of the sources in the applicator.

[0057] Read DICOM image data, and extract the position information of the target area, organs at risk (OAR) and the applicator.

[0058] Generate two-dimensional images of each layer, and through image fusion, ensure that the relative positions of the target area, the applicator, and the dwell positions are consistent on all sections.

[0059] 2. Data input

[0060] Input all key data into the model to make it have complete treatment information and support optimization iteration.

[0061] Input the three-dimensional images related to the target area, organs at risk, and dwell positions, the dwell positions, the initial dwell time, the plan requirements, and the dose distribution map generated in the previous optimization process into the large model.

[0062] Save the input information through the memory module, and at the same time control the amount of stored information, only save the information within 20 iterations to avoid information explosion.

[0063] If the plan does not meet the requirements after being evaluated by the evaluation module, feedback information related to the plan will be automatically generated as data input. This feedback information is a specific instruction in written form, such as "The current D90 dose of the target area is xxx, the planned requirement is xxx, the current D2cc dose of the bladder is xxx, the planned requirement is xxx; the current dose distribution of the patient is not conformal (taking the pear-shaped distribution of cervical cancer as an example), please adjust the dwell time ratio", etc. Along with other data, it will be input into the model again. At the same time, dose distribution maps at different levels of the generated plan will also be transmitted into the model for analysis.

[0064] 3. Data Output

[0065] The model evaluates based on the input information (including feedback on the previous plan, dose distribution maps at different levels in the previous plan, relative distribution maps of target areas, organs at risk, and dwell positions). For areas with too high or too low doses, adjustments are made respectively. The adjustment strategy is to comprehensively consider the dose coverage of the target area and the protection of organs at risk. The dose needs to meet the plan requirements, and at the same time, the dose distribution of the patient should be a pear-shaped distribution.

[0066] After evaluation, the model generates a new and optimized dwell time distribution based on its understanding of the task. This dwell time will be used for dose calculation together with the dose distribution. The dose calculation uses a dose calculation model based on the TG43 formula to obtain the new doses of the target area and organs at risk (dose indicators are D2cc, D90%, etc.), dose volume histograms, and dose distribution maps.

[0067] 5. Plan Evaluation

[0068] The plan evaluation module evaluates the generated new plan (taking cervical cancer as an example). The evaluation content includes: ① evaluating the dose distribution map of the generated plan to check if it conforms to the pear-shaped dose distribution; ② evaluating whether the key dose indicators (D2cc and D90%) meet the plan requirements; ③ whether the dwell time distribution of the implant needles meets the requirements; ④ whether the dose distribution is smoothly transitioned without causing obvious dose cold spots and hot spots.

[0069] 6. Plan Optimization

[0070] After generating a basic plan that meets the basic requirements, the plan is further optimized and iterated. The optimization methods include two types: 1. Further tighten the indicators of all organs at risk and optimize all organs synchronously; 2. According to the doctor's plan-making preferences, select the organs at risk that need to be protected first. In the optimization, give the highest protection weight to the organs at risk that need to be protected first to obtain a more suitable plan protection than the basic plan.

[0071] 7. Plan Transmission

[0072] After the final plan selection and evaluation, the system automatically transmits the residence position and residence time to the radiotherapy instrument for patient treatment. If the evaluation is unqualified, the plan will be continuously optimized and a new treatment plan will be regenerated.

[0073] Embodiment 2:

[0074] The embodiment of the present invention also provides a brachytherapy automatic planning device, including:

[0075] The first processing module is used to obtain the radiotherapy plan and the structure file;

[0076] The second processing module is used to preprocess the radiotherapy plan and the structure file;

[0077] The third processing module is used to input the preprocessed radiotherapy plan and structure file into the vision-language large model, and through multiple rounds of iteration, obtain the residence time and residence position information required by the plan;

[0078] The fourth processing module calculates the dose according to TG43 for the residence time and position generated in each round of iteration, compares it with the requirements of the radiotherapy plan, and feeds back the compared information to the vision-language large model for further optimization and iteration to generate new residence time and position until the requirements of the radiotherapy plan are met, obtaining the basic plan;

[0079] The fifth processing module is used to optimize and upgrade the basic plan to obtain a better plan that meets the personalized needs of the patient.

[0080] As an implementation manner of the embodiment of the present invention, the second processing module includes:

[0081] The first processing unit is used to extract dose information according to the radiotherapy plan;

[0082] The second processing unit is used to extract the trajectory of the radiation source in the applicator, the initial residence position, the initial residence time, the contours of the organs at risk and the high-risk clinical target volume according to the structure file;

[0083] Step S23: According to the trajectory of the radiation source in the applicator, obtain the initial residence position and initial residence time of each channel of the applicator;

[0084] As an implementation manner of the embodiment of the present invention, the third processing module inputs the plan requirements, the trajectory of the radiation source in the applicator, the contours of the organs at risk and the high-risk clinical target volume into the vision-language large model, and through multiple rounds of iteration, obtains the residence time and residence position information required by the plan.

[0085] Embodiment 3:

[0086] An embodiment of the present invention further provides a brachytherapy automatic planning system, including: a memory and a processor, where a computer program run by the processor is stored on the memory, and the computer program executes a brachytherapy automatic planning method when run by the processor.

[0087] Embodiment 4:

[0088] An embodiment of the present invention further provides a storage medium, where a computer program is stored on the storage medium, and the computer program executes a brachytherapy automatic planning method when running.

[0089] The above embodiments are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. An automatic planning method for brachytherapy, characterized in that, Including: Step S1: Obtain the radiotherapy plan and the structure file; Step S2: Preprocess the radiotherapy plan and the structure file; Step S3: Input the preprocessed radiotherapy plan and structure file into the vision-language large model, and through multiple rounds of iteration, obtain the dwell time and dwell position information required by the plan; Step S4: Calculate the dose according to TG43 for the dwell time and position generated in each round of iteration, compare it with the requirements of the radiotherapy plan, and feedback the compared information and instructions to the vision-language large model for further optimization and iteration to generate new dwell time and position until the requirements of the radiotherapy plan are met; Obtain the basic plan; Step S5: Optimize and upgrade the basic plan to obtain a better plan that meets the personalized needs of the patient.

2. The automatic brachytherapy treatment planning method according to claim 1, wherein Step S2 includes: Step S21: Extract dose information according to the radiotherapy plan; Step S22: Extract the trajectory of the radiation source in the applicator, the contours of the organs at risk and the high-risk clinical target volume according to the structure file; Step S23: Obtain the initial dwell position and initial dwell time of each channel of the applicator according to the trajectory of the radiation source in the applicator.

3. The automatic planning method for brachytherapy according to claim 2, wherein In Step S3, input the plan requirements, the trajectory of the radiation source in the applicator, the initial dwell position, the initial dwell time, the contours of the organs at risk and the high-risk clinical target volume into the vision-language large model, and through multiple rounds of iteration, obtain the dwell time and dwell position information required by the plan.

4. An automatic planning device for brachytherapy, characterized in that, Including: The first processing module is used to obtain the radiotherapy plan and the structure file; The second processing module is used to preprocess the radiotherapy plan and the structure file; The third processing module is used to input the preprocessed radiotherapy plan and structure file into the vision-language large model, and through multiple rounds of iteration, obtain the dwell time and dwell position information required by the plan; The fourth processing module is used to calculate the dose according to TG43 for the dwell time and position generated in each round of iteration, compare it with the requirements of the radiotherapy plan, and feedback the compared information and instructions to the vision-language large model for further optimization and iteration to generate new dwell time and position until the requirements of the radiotherapy plan are met; Obtain the basic plan; The fifth processing module is used to optimize and upgrade the basic plan to obtain a better plan that meets the personalized needs of the patient.

5. The automatic brachytherapy treatment planning device according to claim 4, wherein The second processing module includes: The first processing unit is used to extract dose information according to the radiotherapy plan; The second processing unit is used to extract the trajectory of the radiation source in the applicator, the contours of the organs at risk and the high-risk clinical target volume according to the structure file; The third processing unit is used to obtain the initial dwell position and initial dwell time of each channel of the applicator according to the trajectory of the radiation source in the applicator.

6. The automatic brachytherapy treatment planning device according to claim 5, characterized in that, The third processing module inputs the plan requirements, the trajectory of the radiation source in the applicator, the initial dwell position, the initial dwell time, the contours of the organs at risk and the high-risk clinical target volume into the vision-language large model, and through multiple rounds of iteration, obtains the dwell time and dwell position information required by the plan.

7. An automatic planning system for brachytherapy, characterized in that, Including: A memory and a processor, wherein a computer program is stored on the memory and run by the processor, and the computer program, when run by the processor, executes the brachytherapy automatic planning method according to any one of claims 1-3.

8. A storage medium, characterized in that, A computer program is stored on the storage medium, and the computer program, when running, executes the brachytherapy automatic planning method according to any one of claims 1-3.

Images