Radiotherapy plan dose evaluation method and system based on intelligent prediction

Abstract

The invention discloses a radiation treatment plan dose evaluation method and system based on intelligent prediction, and the method comprises the steps: obtaining a treatment scheme formulated by a radiation treatment plan system, and carrying out the calculation to obtain corresponding plan dose distribution data Dose-A; utilizing the trained neural network model S-NET to predict the actual beam dose rate of each sub-field; utilizing the trained neural network model M-NET to predict the actual position of each leaf of the grating in each sub-field; based on the predicted actual beam dose rate of each sub-field and the actual position of each blade of the grating in each sub-field, performing dose calculation to obtain predicted actual dose distribution data Dose-B; calculating first deviation data DELTA-A between the planned dose distribution data Dose-A and the predicted actual dose distribution data Dose-B; calculating second deviation data DELTA-P between the predicted dose distribution data Dose-B and a prescription dose; and generating a dose evaluation report based on the first deviation data DELTA-A and the second deviation data DELTA-P.

Description

Technical Field

[0001] This invention relates to the field of radiotherapy technology, and more specifically to a method and system for evaluating radiotherapy planned dose based on intelligent prediction. Background Technology

[0002] Radiation therapy is one of the main methods of cancer treatment. Its goal is to deliver a precise radiation dose sufficient to kill cancer cells to the tumor target area while maximizing the protection of surrounding normal tissues and organs at risk. To achieve this goal, a detailed and optimized treatment plan must be developed before treatment using a professional radiation therapy planning system. Based on this plan, the three-dimensional distribution data of the radiation dose in the patient's body under ideal conditions, i.e., the planned dose distribution, is calculated. This data is the core basis for doctors to assess the rationality and safety of the treatment plan.

[0003] However, in clinical practice, radiotherapy equipment such as medical linear accelerators inevitably experience execution errors when implementing treatment plans. These errors mainly include two aspects: first, deviations in the beam dose rate, meaning there is a difference between the actual dose rate output by the accelerator and the ideal dose rate set in the treatment plan; second, deviations in the movement position of the multi-leaf grating blades, meaning the actual position reached by the blades during dynamic treatment is not entirely consistent with the target position required by the plan instructions. These execution deviations stem from various factors such as the mechanical precision of the equipment, the response of the control system, and the complex operating environment.

[0004] Deviations in these machine parameters directly lead to a difference between the actual dose distribution absorbed by the patient and the ideal dose distribution calculated by the treatment planning system (i.e., the planned dose distribution). This dose distribution deviation is crucial for accurately predicting and evaluating the final effectiveness of the treatment plan. If the deviation is too large, it may result in insufficient dose to the tumor target area, affecting the treatment effect, or excessive irradiation of normal tissues, causing complications.

[0005] Currently, the industry widely employs rigorous quality assurance processes to ensure treatment accuracy. Conventional methods include pre-irradiation verification using phantoms; that is, before actually treating patients, the treatment plan is executed in a humanoid model, and the dose distribution is measured using equipment such as film, ionization chambers, or two-dimensional detector arrays, then compared with the planned dose distribution. In addition, some technical solutions aim to directly improve the accuracy of machine execution by improving equipment hardware or control algorithms.

[0006] However, existing technical solutions have significant limitations. For example, while conventional phantom validation methods can detect systematic errors, the process is cumbersome and time-consuming, and cannot directly and individually predict actual dose deviations under specific patient anatomy. This method is a post-hoc validation and cannot provide prospective warnings of potential risks before treatment is administered.

[0007] Therefore, there is an urgent need in this field for a method and system that can intelligently predict changes in the final dose distribution caused by machine execution errors, after the treatment plan is formulated but before treatment execution. This method can assist clinicians in more comprehensively and proactively evaluating the robustness and safety of the treatment plan during actual execution, thereby making timely optimization decisions and ensuring the overall accuracy and reliability of radiotherapy.

[0008] In view of this, this invention patent is hereby proposed. Summary of the Invention

[0009] To address the aforementioned problems, this invention provides a method and system for evaluating radiotherapy planned dose based on intelligent prediction, specifically employing the following technical solution: A method for evaluating radiotherapy planning dose based on intelligent prediction, comprising: Obtain the treatment plan formulated by the radiotherapy planning system and calculate the corresponding planned dose distribution data Dose-A; The actual beam dose rate of each subfield is predicted using the trained neural network model S-NET. Using the trained neural network model M-NET, the actual positions of each blade of the grating in each subfield are predicted; Based on the predicted actual beam dose rate of each subfield and the actual position of each grating blade in each subfield, dose calculation is performed to obtain the predicted actual dose distribution data Dose-B. Calculate the first deviation data DELTA-A between the planned dose distribution data Dose-A and the predicted actual dose distribution data Dose-B; Calculate the second deviation data DELTA-P between the predicted dose distribution data Dose-B and the prescribed dose; A dose evaluation report is generated based on the first deviation data DELTA-A and the second deviation data DELTA-P.

[0010] As an optional embodiment of the present invention, a radiotherapy planning method based on intelligent prediction is provided. In the dose evaluation method, the neural network model S-NET is trained as follows: The planned dose rate and actual dose rate recorded in the historical operation log of the medical accelerator are used as training data. According to this, the neural network is trained; The input parameters of the neural network model S-NET include the planned dose rate Dp and the gantry angle Ang; The output of the neural network model S-NET is the predicted actual dose rate Dq. When training the neural network model S-NET, the cost function used is the mean squared error (MSE), which is calculated as follows: Where n is the number of shooting fields, For the first The actual dose rate of each subfield For the first Predicted dose rate for individual fields.

[0011] As an optional embodiment of the present invention, a radiotherapy planning method based on intelligent prediction is provided. In the dose evaluation method, the neural network model S-NET is trained as follows: The neural network model M-NET is trained as follows: The planned and actual positions of each blade of the grating, recorded in the historical operation log of the medical accelerator, are used as... The neural network is trained using training data; The input parameters of the neural network model M-NET include: the planned positions of each blade of the grating. Frame angle Ang and blade speed ; The output of the neural network model M-NET is the predicted actual position of each blade of the grating. When training the neural network model M-NET, the cost function used is the mean squared error (MSE), which is calculated as follows:

[0012] Where n is the number of radiation fields and m is the number of grating blades. For the first Ge Ziye Di The position of each leaf Actual value For the first Ge Ziye Di Predicted position values ​​for each leaf.

[0013] As an optional embodiment of the present invention, a radiotherapy planning method based on intelligent prediction is provided. In the dose evaluation method, the dose calculation based on the predicted actual beam dose rate of each subfield and the actual position of each grating blade in each subfield to obtain the predicted actual dose distribution data Dose-B includes: The predicted actual beam dose rate for each subfield { } and the actual positions of each blade of the grating in each subfield { The data is input as a parameter to the dose calculation engine, which performs the calculation and outputs the predicted actual dose distribution data Dose-B. The dose calculation engine employs one of the following dose calculation algorithms: a convolution-based pen-beam model algorithm, a Monte Carlo algorithm, or a linear Boltzmann equation solving algorithm. The geometric model on which the dose calculation is based is a three-dimensional anatomical structure reconstructed from the patient's CT image data. The predicted actual dose distribution data Dose-B is a three-dimensional dose distribution matrix, which is used to characterize the spatial distribution of radiation energy deposited in the patient's body under the predicted actual machine operating parameters. When calculating the Dose-B, the dose calculation engine combines one or more parameters from the field direction, field shape, and wedge angle.

[0014] As an optional embodiment of the present invention, a radiotherapy planning method based on intelligent prediction is provided. In the dose evaluation method, calculating the first deviation data DELTA-A between the planned dose distribution data Dose-A and the predicted actual dose distribution data Dose-B includes: A quantitative comparison is performed on Dose-A and Dose-B to generate the first deviation data DELTA-A, which is used to quantify the difference between the two. The first deviation data DELTA-A includes a global dose deviation index, which is one or more of the following: Mean, standard deviation, maximum, or median dose difference of voxels throughout the body or in a specific region of interest; The percentage difference in volume at key dose levels in the dose-volume histogram; And / or, the first deviation data DELTA-A includes a spatial consistency assessment index, which is obtained through three-dimensional gamma analysis; the three-dimensional gamma analysis simultaneously considers dose difference criteria and distance difference criteria, and outputs gamma pass rate; And / or, the first deviation data DELTA-A includes deviation indicators based on clinical objectives, including: For the tumor target area, calculate the difference between Dose-A and Dose-B in the volume percentage of the tumor that received the prescribed dose or more. For organs at risk, calculate the difference between Dose-A and Dose-B in terms of the percentage of volume that received a specific tolerable dose.

[0015] As an optional embodiment of the present invention, a radiotherapy planning method based on intelligent prediction is provided. In the dose evaluation method, calculating the second deviation data DELTA-P between the predicted dose distribution data Dose-B and the prescribed dose includes: The Dose-B is compared with a predefined clinical prescription dosage protocol to generate the second deviation data DELTA-P; The second bias data DELTA-P includes bias indices for the tumor target area, which are one or more of the following: The difference between the volume percentage of the target area receiving the prescribed dose and the clinical target value; The difference between the maximum, minimum, or average dose within the target area and the prescribed dose; The difference between the dose uniformity index of the target area and the target value; And / or, the second deviation data DELTA-P includes a deviation index for at least one organ at risk, said index being any one or more of the following: The volume percentage of the organ at risk receiving a dose exceeding its tolerance level; The difference between the maximum dose to which an organ is at risk and its tolerable dose; The difference between the average dose to the organ at risk and its tolerable dose.

[0016] As an optional embodiment of the present invention, a radiotherapy planning method based on intelligent prediction is provided. Dosage evaluation methods include: The indicators in the second deviation data DELTA-P are compared with the preset clinically acceptable thresholds; When any indicator exceeds the clinically acceptable threshold, a prompt message is generated in the dose evaluation report, which includes a suggestion to optimize the treatment plan.

[0017] As an optional embodiment of the present invention, a radiotherapy planning method based on intelligent prediction is provided. In the dose evaluation method, generating a dose evaluation report based on the first deviation data DELTA-A and the second deviation data DELTA-P includes: Integrate the DELTA-A and DELTA-P to generate a structured dose evaluation report; The dose evaluation report includes one or more of the following elements: The planned dose distribution Dose-A and the predicted actual dose distribution Dose-B are displayed side-by-side on the same set of patient anatomical images; A dose difference distribution map used to illustrate the difference in dose between Dose-A and Dose-B; A dose-volume histogram comparison curve used to illustrate the relationship between Dose-A, Dose-B and prescription dose; A table containing the specific values ​​of the first deviation data DELTA-A and the second deviation data DELTA-P.

[0018] As an optional embodiment of the present invention, a radiotherapy planning method based on intelligent prediction is provided. In the dose evaluation method, the step of acquiring the treatment plan formulated by the radiotherapy planning system and calculating the corresponding planned dose distribution data Dose-A includes: The treatment plan data can be directly retrieved from the database of the radiotherapy planning system, or imported by parsing the digital treatment plan file. The calculation yields the corresponding planned dose distribution data Dose-A, including: The radiotherapy planning system performs dose calculations based on the treatment plan and outputs the planned dose distribution data Dose-A; or, the dose calculation engine within the radiotherapy planning system recalculates the planned dose distribution data Dose-A based on machine parameters obtained from the treatment plan and the patient's anatomical model. The planned dose distribution data Dose-A is a three-dimensional dose distribution matrix, which is consistent with the predicted actual dose distribution data Dose-B in terms of spatial dimension and grid resolution.

[0019] This invention also provides a radiotherapy planning dose evaluation system based on intelligent prediction, comprising: The planned dose distribution calculation module acquires the treatment plan formulated by the radiotherapy planning system and calculates the corresponding planned dose distribution data Dose-A. The beam dose rate prediction module uses a trained neural network model S-NET to predict the actual beam dose rate of each subfield. The grating blade position prediction module uses a trained neural network model M-NET to predict the actual position of each grating blade in each subfield. The dose calculation module performs dose calculation based on the predicted actual beam dose rate of each subfield and the actual position of each grating blade in each subfield to obtain the predicted actual dose distribution data Dose-B. The dose distribution data evaluation module calculates the first deviation data DELTA-A between the planned dose distribution data Dose-A and the predicted actual dose distribution data Dose-B, and calculates the second deviation data DELTA-P between the predicted dose distribution data Dose-B and the prescription dose. The dose evaluation report module is based on the first deviation data DELTA-A and the second deviation data. DELTA-P generates a dose evaluation report.

[0020] Compared with the prior art, the beneficial effects of the present invention are as follows: 1. It enables forward-looking intelligent prediction of dose distribution deviation.

[0021] This invention proposes a radiotherapy planning dose evaluation method based on intelligent prediction. Through neural network models (S-NET and M-NET), it can accurately predict the beam dose rate and grating blade position deviations that may occur during the actual operation of a medical accelerator. This allows for the pre-calculation of a dose distribution (Dose-B) that more closely matches the actual situation before the patient receives irradiation, transforming the traditional "post-treatment verification" into "pre-treatment prediction," greatly improving the proactiveness and safety of the treatment process.

[0022] 2. It has improved the comprehensiveness and depth of the evaluation of treatment plans.

[0023] This invention proposes a method for evaluating radiotherapy treatment plans based on intelligent prediction. By calculating the deviation between the planned dose and the predicted dose (DELTA-A) and the deviation between the predicted dose and the prescribed dose (DELTA-P), this invention provides a dual evaluation perspective. The former reveals the "robustness" of the plan in actual execution, while the latter directly assesses the "clinical effectiveness" of the predicted treatment outcome. This multi-dimensional evaluation provides a more comprehensive and profound reflection of the quality and potential risks of the treatment plan than a single-dimensional comparison.

[0024] 3. The clinical decision-making process has been optimized, and work efficiency has been improved.

[0025] This invention proposes a radiotherapy plan dose evaluation method based on intelligent prediction. The automatically generated dose evaluation report integrates key deviation data and provides clear overall conclusions (such as "pass" or "optimization recommended"). This provides physicians with intuitive and reliable decision support, reducing their time and burden from manually analyzing complex data. Simultaneously, the system can provide specific optimization suggestions for non-compliance items, directly guiding the re-optimization of the treatment plan, forming an efficient "evaluation-feedback-optimization" closed loop and shortening the treatment preparation cycle.

[0026] 4. It has promoted the precision and individualization of radiotherapy quality assurance.

[0027] This invention proposes a radiotherapy treatment plan dose evaluation method based on intelligent prediction, which eliminates the over-reliance on universal phantoms and periodic QA, and enables personalized pre-treatment risk prediction for each specific patient and treatment plan. This shifts the focus of quality assurance from "equipment status" to "the safety and effectiveness of this treatment," representing the future direction of precision radiotherapy. Attached Figure Description

[0028] Figure 1 A flowchart of a radiotherapy planning dose evaluation method based on intelligent prediction according to an embodiment of the present invention; Figure 2 The network structure of the S-NET neural network model for beam dose rate prediction in this embodiment of the invention; Figure 3 The network structure of the M-NET neural network model for predicting the position of grating blades in this embodiment of the invention; Figure 4 This invention provides a system architecture diagram of a radiotherapy planning dose evaluation system based on intelligent prediction. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0030] Therefore, the following detailed description of embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely illustrates some embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0031] It should be noted that, unless otherwise specified, the embodiments and features and technical solutions in the present invention can be combined with each other.

[0032] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0033] In the description of this invention, it should be noted that the terms "upper," "lower," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use, or the orientation or positional relationship commonly understood by those skilled in the art. These terms are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention. In addition, the terms "first," "second," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0034] See Figure 1 As shown, this embodiment proposes a method for evaluating radiotherapy planning dose based on intelligent prediction, including: Obtain the treatment plan formulated by the radiotherapy planning system and calculate the corresponding planned dose distribution data Dose-A; The actual beam dose rate of each subfield is predicted using the trained neural network model S-NET. Using the trained neural network model M-NET, the actual positions of each blade of the grating in each subfield are predicted; Based on the predicted actual beam dose rate of each subfield and the actual position of each grating blade in each subfield, dose calculation is performed to obtain the predicted actual dose distribution data Dose-B. Calculate the first deviation data DELTA-A between the planned dose distribution data Dose-A and the predicted actual dose distribution data Dose-B; Calculate the second deviation data DELTA-P between the predicted dose distribution data Dose-B and the prescribed dose; A dose evaluation report is generated based on the first deviation data DELTA-A and the second deviation data DELTA-P.

[0035] Therefore, this embodiment proposes a radiotherapy planning dose evaluation method based on intelligent prediction. By introducing intelligent prediction and prospective dose assessment, it brings the following significant technical effects: 1. It enables forward-looking intelligent prediction of dose distribution deviation.

[0036] Through neural network models (S-NET and M-NET), it is possible to predict with high accuracy the beam dose rate and grating blade position deviations that may occur in the actual operation of medical accelerators. This allows for the pre-calculation of a dose distribution that more closely matches the actual execution situation (Dose-B) before the patient actually receives irradiation, transforming the traditional "post-event verification" into "pre-event prediction," greatly improving the proactivity and safety of the treatment process.

[0037] 2. It has improved the comprehensiveness and depth of the evaluation of treatment plans.

[0038] By separately calculating the deviation between the planned and predicted doses (DELTA-A) and the deviation between the predicted and prescribed doses (DELTA-P), this invention provides a dual evaluation perspective. The former reveals the "robustness" of the plan in actual execution, while the latter directly assesses the "clinical effectiveness" of the predicted treatment outcome. This multi-dimensional evaluation provides a more comprehensive and profound reflection of the quality and potential risks of the treatment plan than a single-dimensional comparison.

[0039] 3. The clinical decision-making process has been optimized, and work efficiency has been improved.

[0040] The automatically generated dose evaluation report integrates key deviation data and provides clear overall conclusions (such as "pass" or "optimization recommended"). This provides physicians with intuitive and reliable decision support, reducing their time and burden from manually analyzing complex data. Simultaneously, the system can provide specific optimization suggestions for non-compliance items, directly guiding the re-optimization of the treatment plan, forming an efficient "evaluation-feedback-optimization" closed loop and shortening the treatment preparation cycle.

[0041] 4. It has promoted the precision and individualization of radiotherapy quality assurance.

[0042] This method eliminates over-reliance on universal phantoms and periodic QA, enabling personalized pre-treatment risk prediction for each specific patient and treatment plan. This shifts the focus of quality assurance from "equipment status" to "the safety and effectiveness of the treatment," representing a new direction for precision radiotherapy.

[0043] Optionally, in the intelligent prediction-based radiotherapy planning dose evaluation method of this embodiment, The neural network model S-NET was trained as follows: The planned dose rate and actual dose rate recorded in the historical operation log of the medical accelerator were used as training data. According to reports, the neural network is trained.

[0044] Specifically, the planned dose rate and actual dose rate recorded in the medical accelerator operation log for the past month or a user-specified time period are used as training data for model training.

[0045] See Figure 2 As shown, the network structure of the S-NET neural network model in this embodiment includes an input layer, a hidden layer, and an output layer.

[0046] The input parameters of the neural network model S-NET include the planned dose rate Dp and the gantry angle Ang; The output of the neural network model S-NET is the predicted actual dose rate Dq. When training the neural network model S-NET, the cost function used is the mean squared error (MSE), which is calculated as follows: Where n is the number of shooting fields, For the first The actual dose rate of each subfield For the first Predicted dose rate for individual fields.

[0047] This embodiment uses the medical accelerator's own historical operating logs (including planned and actual dose rates) as training data, enabling the S-NET model to deeply learn and capture the systematic errors, random fluctuations, and complex relationships between these errors and operating parameters (such as gantry angle) present in the actual operation of a specific device. This data-driven approach, compared to theoretical models based on purely physical assumptions, can more accurately predict the actual dose rate of a specific machine under a specific plan, providing a reliable input guarantee for subsequent dose calculations.

[0048] The model uses the gantry angle (Ang) as one of the key input parameters, fully considering the dynamic characteristics of beam dose rate deviation that may change with the mechanical motion state of the machine. This makes the prediction model no longer static, but adaptable to complex scenarios such as dynamic intensity-modulated radiotherapy with continuous gantry rotation during treatment, significantly improving the applicability and prediction accuracy of the method in real-world clinical environments.

[0049] Using mean squared error (MSE) as the cost function is beneficial for the rapid and stable convergence of the training process due to its mathematical properties. By minimizing the mean squared error between the predicted dose rate and the actual dose rate, the optimization of model parameters can be driven towards minimizing the overall prediction error. This ensures that the trained S-NET model has good generalization ability and can provide stable and reliable prediction results for new and unseen treatment plans, avoiding excessive prediction bias caused by individual outliers.

[0050] The high-precision prediction of beam dose rate by the S-NET model is a crucial prerequisite for the subsequent accurate calculation of the predicted actual dose distribution (Dose-B). This approach, through the specific training and construction methods described above, ensures the high reliability of the predicted dose rate (Dq) output by S-NET. This guarantees the accuracy and reliability of the source data in the entire "intelligent prediction-dose calculation-deviation evaluation" chain, ultimately improving the overall quality and clinical reference value of the generated dose evaluation report.

[0051] Optionally, in the intelligent prediction-based radiotherapy planning dose evaluation method of this embodiment, The neural network model M-NET was trained as follows: The planned and actual positions of each blade of the grating, recorded in the historical operation log of the medical accelerator, were used as the basis for training. The neural network is trained using training data.

[0052] Specifically, the planned and actual positions of each blade of the grating, recorded in the medical accelerator's operation log over the past month or a user-specified time period, are used as training data for model training.

[0053] like Figure 3 As shown, the network structure of the M-NET neural network model in this embodiment includes an input layer, a hidden layer, and an output layer.

[0054] The input parameters of the neural network model M-NET include: the planned positions of each blade of the grating. Frame angle Ang and blade speed ; The output of the neural network model M-NET is the predicted actual position of each blade of the grating. When training the neural network model M-NET, the cost function used is the mean squared error (MSE), which is calculated as follows:

[0055] Where n is the number of radiation fields and m is the number of grating blades. For the first Ge Ziye Di The position of each leaf Actual value For the first Ge Ziye Di Predicted position values ​​for each leaf.

[0056] This embodiment presents a radiotherapy planned dose evaluation method based on intelligent prediction. By using the planned and actual positions of each leaf of the grating recorded in the historical operation log of a medical accelerator as training data, the M-NET model can accurately learn the mechanical characteristics, response delay, and systematic errors of the multi-leaf grating system under different operating conditions of a specific device. This method achieves accurate prediction of the actual arrival position of the leaves during complex dynamic treatment processes, providing crucial geometric accuracy for subsequent dose calculations.

[0057] The model innovatively incorporates blade speed With planned location The frame angle Ang is used as an input parameter. This design allows the model to capture the dynamic effects of blade inertia and response delay at different movement speeds, greatly enhancing the accuracy of blade position prediction under complex modes such as dynamic intensity-modulated therapy. This design makes the prediction model closer to the actual physical operating mechanism of the equipment, significantly improving its applicability and reliability in real clinical scenarios.

[0058] The mean squared error (MSE) cost function employed covers the prediction errors for all firing fields (n) and all blades (m). This joint optimization of the positional accuracy of all blades forces the model to learn not only the motion patterns of individual blades but also the cooperative motion relationships between blades. As a result, the trained M-NET model can accurately predict the complete grating shape, avoiding the problem of models only accurately predicting some blades while ignoring the overall shape fidelity, thus ensuring the comprehensiveness and robustness of the prediction results.

[0059] The high-precision prediction of the grating blade position by the M-NET model directly determines the accuracy of the radiation field shape in subsequent dose calculations. Even small deviations in the radiation field shape can cause significant changes in dose distribution, especially at the edge of the target area and near organs at risk. Therefore, the blade position prediction accuracy ensured by the above method is the core foundation for the entire system to accurately assess the impact of machine execution errors on clinical dosing, fundamentally guaranteeing the reliability and clinical value of the final dose evaluation report.

[0060] Furthermore, in this embodiment, a method for evaluating radiotherapy planning dose based on intelligent prediction, The dose calculation based on the predicted actual beam dose rate of each subfield and the actual position of each grating blade in each subfield, to obtain the predicted actual dose distribution data Dose-B, includes: The predicted actual beam dose rate for each subfield { } and the actual positions of each blade of the grating in each subfield { The data is input as a parameter to the dose calculation engine, which performs the calculation and outputs the predicted actual dose distribution data Dose-B. The dose calculation engine employs one of the following dose calculation algorithms: a convolution-based pen-beam model algorithm, a Monte Carlo algorithm, or a linear Boltzmann equation solving algorithm. The geometric model on which the dose calculation is based is a three-dimensional anatomical structure reconstructed from the patient's CT image data. The predicted actual dose distribution data Dose-B is a three-dimensional dose distribution matrix, which is used to characterize the spatial distribution of radiation energy deposited in the patient's body under the predicted actual machine operating parameters. When calculating the Dose-B, the dose calculation engine combines one or more parameters from the field direction, field shape, and wedge angle.

[0061] The specific implementation of "obtaining predicted actual dose distribution data Dose-B based on prediction parameters" in this technical solution brings the following significant and specific technical effects: 1. A precise mapping relationship from machine error to clinical dosage was established.

[0062] By using neural networks to predict actual machine operating parameters (beam dose rate) } and the position of the grating blades { The data was input into a professional dose calculation engine, successfully constructing a complete quantitative mapping chain from "device execution error to dose distribution within the patient's body." This allows the evaluation of treatment plans to move beyond ideal machine parameters and proactively predict the actual dose distribution the patient will receive under actual device operation, greatly enhancing the clinical relevance and practical value of dose evaluation.

[0063] 2. It ensures the scientific validity and reliability of dose prediction.

[0064] Employing clinically validated high-precision dose calculation algorithms, such as convolution-based pen-beam model algorithms, Monte Carlo algorithms, or linear Boltzmann equation solving algorithms, ensures that the physical accuracy of the dose calculation itself is maintained even when the input parameters are predicted values. This provides a reliable technical foundation for subsequent deviation analysis, ensuring that the entire prediction system incorporates both the uncertainties of equipment execution and the scientific rigor of dose calculation.

[0065] 3. It achieves truly individualized dose deviation prediction.

[0066] Dose calculation is based on the reconstructed three-dimensional anatomical structure from the patient's CT image data, and the output is also a three-dimensional dose distribution matrix, Dose-B. This allows the predicted dose distribution to accurately reflect the deposition characteristics of radiation energy in the specific patient's anatomical structure. It fully considers the impact of tissue density heterogeneity on dose distribution, thus achieving personalized dose deviation prediction for each patient, rather than approximate assessment based on a universal phantom.

[0067] 4. A complete clinical treatment scenario simulation was constructed.

[0068] By incorporating key treatment parameters such as field direction, field shape, and wedge angle during dose calculation, the predicted dose distribution can fully replicate the real clinical treatment scenario. This comprehensive approach ensures that the predicted actual dose distribution data from Dose-B accurately reflects the dose distribution characteristics under specific treatment techniques (such as IMRT and VMAT), providing clinicians with highly realistic dose prediction results.

[0069] 5. It provides intuitive and reliable data support for clinical decision-making.

[0070] The final output, the three-dimensional dose distribution matrix Dose-B, visually represents the spatial distribution of radiation energy deposited within the patient's body under the predicted machine's actual operating parameters. This provides physicians with direct, quantitative evidence to assess the robustness of treatment plans during actual execution. This allows physicians to identify potential dose deviation risk areas before treatment, enabling more informed clinical decisions.

[0071] Optionally, in a method for evaluating radiotherapy planned dose based on intelligent prediction in this embodiment, calculating the first deviation data DELTA-A between the planned dose distribution data Dose-A and the predicted actual dose distribution data Dose-B includes: A quantitative comparison is performed on Dose-A and Dose-B to generate the first deviation data DELTA-A, which is used to quantify the difference between the two.

[0072] The first deviation data DELTA-A includes a global dose deviation index, which is one or more of the following: Mean, standard deviation, maximum, or median dose difference of voxels throughout the body or in a specific region of interest; The percentage difference in volume at key dose levels in the dose-volume histogram.

[0073] And / or, the first deviation data DELTA-A includes a spatial consistency assessment index, which is obtained through three-dimensional gamma analysis; the three-dimensional gamma analysis simultaneously considers dose difference criteria and distance difference criteria, and outputs gamma pass rate.

[0074] And / or, the first deviation data DELTA-A includes deviation indicators based on clinical objectives, including: For the tumor target area, calculate the difference between Dose-A and Dose-B in the volume percentage of the tumor that received the prescribed dose or more. For organs at risk, calculate the difference between Dose-A and Dose-B in terms of the percentage of volume that received a specific tolerable dose.

[0075] Optionally, in a method for evaluating radiotherapy planned dose based on intelligent prediction in this embodiment, calculating the second deviation data DELTA-P between the predicted dose distribution data Dose-B and the prescribed dose includes: The Dose-B is compared with a predefined clinical prescription dosage protocol to generate the second deviation data DELTA-P.

[0076] The second bias data DELTA-P includes bias indices for the tumor target area, which are one or more of the following: The difference between the volume percentage of the target area receiving the prescribed dose and the clinical target value; The difference between the maximum, minimum, or average dose within the target area and the prescribed dose; The difference between the dose uniformity index of the target area and the target value.

[0077] And / or, the second deviation data DELTA-P includes a deviation index for at least one organ at risk, said index being any one or more of the following: The volume percentage of the organ at risk receiving a dose exceeding its tolerance level; The difference between the maximum dose to which an organ is at risk and its tolerable dose; The difference between the average dose to the organ at risk and its tolerable dose.

[0078] In this embodiment, the second deviation data DELTA-P is calculated based on the dose-volume histogram generated from the predicted dose distribution data Dose-B.

[0079] Furthermore, this embodiment provides a method for evaluating radiotherapy planning dose based on intelligent prediction. include: The indicators in the second deviation data DELTA-P are compared with the preset clinically acceptable thresholds; When any indicator exceeds the clinically acceptable threshold, a prompt message is generated in the dose evaluation report, which includes a suggestion to optimize the treatment plan.

[0080] Optionally, in the intelligent prediction-based radiotherapy planning dose evaluation method of this embodiment, The step of generating a dose evaluation report based on the first deviation data DELTA-A and the second deviation data DELTA-P includes: integrating the DELTA-A and DELTA-P to generate a structured dose evaluation report.

[0081] The dose evaluation report includes one or more of the following elements: The planned dose distribution Dose-A and the predicted actual dose distribution Dose-B are displayed side-by-side on the same set of patient anatomical images; A dose difference distribution map used to illustrate the difference in dose between Dose-A and Dose-B; A dose-volume histogram comparison curve used to illustrate the relationship between Dose-A, Dose-B and prescription dose; A table containing the specific values ​​of the first deviation data DELTA-A and the second deviation data DELTA-P.

[0082] Optionally, in the intelligent prediction-based radiotherapy planning dose evaluation method of this embodiment, The acquisition of the treatment plan formulated by the radiotherapy planning system and the calculation of the corresponding planned dose distribution data Dose-A include: The treatment plan data can be directly retrieved from the database of the radiotherapy planning system, or imported by parsing the digital treatment plan file. The calculation yields the corresponding planned dose distribution data Dose-A, including: The radiotherapy planning system performs dose calculations based on the treatment plan and outputs the planned dose distribution data Dose-A; or, the dose calculation engine within the radiotherapy planning system recalculates the planned dose distribution data Dose-A based on machine parameters obtained from the treatment plan and the patient's anatomical model. The planned dose distribution data Dose-A is a three-dimensional dose distribution matrix, which is consistent with the predicted actual dose distribution data Dose-B in terms of spatial dimension and grid resolution.

[0083] like Figure 4 As shown, this embodiment also provides a radiotherapy planning dose evaluation system based on intelligent prediction, including: The planned dose distribution calculation module acquires the treatment plan formulated by the radiotherapy planning system and calculates the corresponding planned dose distribution data Dose-A. The beam dose rate prediction module uses a trained neural network model S-NET to predict the actual beam dose rate of each subfield. The grating blade position prediction module uses a trained neural network model M-NET to predict the actual position of each grating blade in each subfield. The dose calculation module performs dose calculation based on the predicted actual beam dose rate of each subfield and the actual position of each grating blade in each subfield to obtain the predicted actual dose distribution data Dose-B. The dose distribution data evaluation module calculates the first deviation data DELTA-A between the planned dose distribution data Dose-A and the predicted actual dose distribution data Dose-B, and calculates the second deviation data DELTA-P between the predicted dose distribution data Dose-B and the prescription dose. The dose evaluation report module generates a dose evaluation report based on the first deviation data DELTA-A and the second deviation data DELTA-P.

[0084] This embodiment also provides a computer-readable storage medium storing a computer-executable program, which, when executed, implements the control method for a radiotherapy system as described above.

[0085] The computer-readable storage medium described in this embodiment may include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable storage medium may also be any readable medium other than a readable storage medium, capable of transmitting, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0086] This embodiment also provides an electronic device, including a processor and a memory, wherein the memory is used to store a computer-executable program, and when the computer program is executed by the processor, the processor executes the aforementioned method for evaluating radiotherapy planning dose based on intelligent prediction.

[0087] The electronic device is manifested in the form of a general-purpose computing device. It may contain one or more processors that work collaboratively. This invention also does not preclude distributed processing, meaning that processors may be distributed across different physical devices. The electronic device of this invention is not limited to a single entity, but may also be the sum of multiple physical devices.

[0088] The memory stores a computer-executable program, typically machine-readable code. The computer-readable program can be executed by the processor to enable the electronic device to perform the method of the present invention, or at least some steps of the method.

[0089] The memory includes volatile memory, such as random access memory (RAM) and / or cache memory, and may also be non-volatile memory, such as read-only memory (ROM).

[0090] It should be understood that the electronic device of the present invention may also include elements or components not shown in the examples above. For example, some electronic devices also include display units such as a display screen, and some electronic devices also include human-computer interaction elements such as buttons and keyboards. Any electronic device capable of executing a computer-readable program in its memory to implement the method of the present invention or at least some steps of the method can be considered as an electronic device covered by the present invention.

[0091] From the above description of the embodiments, those skilled in the art will readily understand that the present invention can be implemented by hardware capable of executing specific computer programs, such as the system of the present invention, and the electronic processing unit, server, client, mobile phone, control unit, processor, etc. included in the system. The present invention can also be implemented by computer software executing the methods of the present invention, for example, by control software executed by a microprocessor, electronic control unit, client, server, etc. However, it should be noted that the computer software executing the methods of the present invention is not limited to execution in one or a specific set of hardware entities; it can also be implemented in a distributed manner by unspecified hardware. For computer software, the software product can be stored in a computer-readable storage medium (such as a CD-ROM, USB flash drive, portable hard drive, etc.) or distributed across a network, as long as it enables electronic devices to execute the methods according to the present invention.

[0092] The above embodiments are only used to illustrate the present invention and are not intended to limit the technical solutions described herein. Although the present invention has been described in detail with reference to the above embodiments, the present invention is not limited to the specific embodiments described above. Therefore, any modifications or equivalent substitutions to the present invention, as well as all technical solutions and improvements that do not depart from the spirit and scope of the invention, are covered within the scope of the claims of the present invention.

Claims

1. A method for evaluating radiotherapy planned dose based on intelligent prediction, characterized in that, include: Obtain the treatment plan formulated by the radiotherapy planning system and calculate the corresponding planned dose distribution data Dose-A; The actual beam dose rate of each subfield is predicted using the trained neural network model S-NET. Using the trained neural network model M-NET, the actual positions of each blade of the grating in each subfield are predicted; Based on the predicted actual beam dose rate of each subfield and the actual position of each grating blade in each subfield, dose calculation is performed to obtain the predicted actual dose distribution data Dose-B. Calculate the first deviation data DELTA-A between the planned dose distribution data Dose-A and the predicted actual dose distribution data Dose-B; Calculate the second deviation data DELTA-P between the predicted dose distribution data Dose-B and the prescribed dose; A dose evaluation report is generated based on the first deviation data DELTA-A and the second deviation data DELTA-P.

2. The method for evaluating radiotherapy planning dose based on intelligent prediction according to claim 1, Its features are, The neural network model S-NET was trained as follows: The planned dose rate and actual dose rate recorded in the historical operation log of the medical accelerator were used as training data. According to this, the neural network is trained; The input parameters of the neural network model S-NET include the planned dose rate Dp and the gantry angle Ang; The output of the neural network model S-NET is the predicted actual dose rate Dq. When training the neural network model S-NET, the cost function used is the mean squared error (MSE), which is calculated as follows: Where n is the number of shooting fields, For the first The actual dose rate of each subfield For the first Predicted dose rate for individual fields.

3. The method for evaluating radiotherapy planning dose based on intelligent prediction according to claim 1, Its features are, The neural network model M-NET was trained as follows: The planned and actual positions of each blade of the grating, recorded in the historical operation log of the medical accelerator, were used as the basis for training. The neural network is trained using training data; The input parameters of the neural network model M-NET include: the planned positions of each blade of the grating. Frame angle Ang and blade speed ; The output of the neural network model M-NET is the predicted actual position of each blade of the grating. When training the neural network model M-NET, the cost function used is the mean squared error (MSE), which is calculated as follows: Where n is the number of radiation fields and m is the number of grating blades. For the first Ge Ziye Di The position of each leaf Actual value For the first Ge Ziye Di Predicted position values ​​for each leaf.

4. The method for evaluating radiotherapy planning dose based on intelligent prediction according to claim 1, Its features are, The dose calculation based on the predicted actual beam dose rate of each subfield and the actual position of each grating blade in each subfield, to obtain the predicted actual dose distribution data Dose-B, includes: The predicted actual beam dose rate for each subfield { } and the actual positions of each blade of the grating in each subfield { The data is input as a parameter to the dose calculation engine, which performs the calculation and outputs the predicted actual dose distribution data Dose-B. The dose calculation engine employs one of the following dose calculation algorithms: a convolution-based pen-beam model algorithm, a Monte Carlo algorithm, or a linear Boltzmann equation solving algorithm. The geometric model on which the dose calculation is based is a three-dimensional anatomical structure reconstructed from the patient's CT image data. The predicted actual dose distribution data Dose-B is a three-dimensional dose distribution matrix, which is used to characterize the spatial distribution of radiation energy deposited in the patient's body under the predicted actual machine operating parameters. When calculating the Dose-B, the dose calculation engine combines one or more parameters from the field direction, field shape, and wedge angle.

5. The method for evaluating radiotherapy planning dose based on intelligent prediction according to claim 1, Its features are, The calculation of the first deviation data DELTA-A between the planned dose distribution data Dose-A and the predicted actual dose distribution data Dose-B includes: A quantitative comparison is performed on Dose-A and Dose-B to generate the first deviation data DELTA-A, which is used to quantify the difference between the two. The first deviation data DELTA-A includes a global dose deviation index, which is one or more of the following: Mean, standard deviation, maximum, or median dose difference of voxels throughout the body or in a specific region of interest; The percentage difference in volume at key dose levels in the dose-volume histogram; And / or, the first deviation data DELTA-A includes a spatial consistency assessment index, which is obtained through three-dimensional gamma analysis; the three-dimensional gamma analysis simultaneously considers dose difference criteria and distance difference criteria, and outputs gamma pass rate; And / or, the first deviation data DELTA-A includes deviation indicators based on clinical objectives, including: For the tumor target area, calculate the difference between Dose-A and Dose-B in the volume percentage of the tumor that received the prescribed dose or more. For organs at risk, calculate the difference between Dose-A and Dose-B in terms of the percentage of volume that received a specific tolerable dose.

6. The method for evaluating radiotherapy planning dose based on intelligent prediction according to claim 1, Its features are, The calculation of the second deviation data DELTA-P between the predicted dose distribution data Dose-B and the prescribed dose includes: The Dose-B is compared with a predefined clinical prescription dosage protocol to generate the second deviation data DELTA-P; The second bias data DELTA-P includes bias indices for the tumor target area, which are one or more of the following: The difference between the volume percentage of the target area receiving the prescribed dose and the clinical target value; The difference between the maximum, minimum, or average dose within the target area and the prescribed dose; The difference between the dose uniformity index of the target area and the target value; And / or, the second deviation data DELTA-P includes a deviation index for at least one organ at risk, said index being any one or more of the following: The volume percentage of the organ at risk receiving a dose exceeding its tolerance level; The difference between the maximum dose to which an organ is at risk and its tolerable dose; The difference between the average dose to the organ at risk and its tolerable dose.

7. A method for evaluating radiotherapy planning dose based on intelligent prediction as described in claim 6. Its features are, include: The indicators in the second deviation data DELTA-P are compared with the preset clinically acceptable thresholds; When any indicator exceeds the clinically acceptable threshold, a prompt message is generated in the dose evaluation report, which includes a suggestion to optimize the treatment plan.

8. The method for evaluating radiotherapy planning dose based on intelligent prediction according to claim 1, Its features are, The generation of a dose evaluation report based on the first deviation data DELTA-A and the second deviation data DELTA-P includes: Integrate the DELTA-A and DELTA-P to generate a structured dose evaluation report; The dose evaluation report includes one or more of the following elements: The planned dose distribution Dose-A and the predicted actual dose distribution Dose-B are displayed side-by-side on the same set of patient anatomical images; A dose difference distribution map used to illustrate the difference in dose between Dose-A and Dose-B; A dose-volume histogram comparison curve used to illustrate the relationship between Dose-A, Dose-B and prescription dose; A table containing the specific values ​​of the first deviation data DELTA-A and the second deviation data DELTA-P.

9. The method for evaluating radiotherapy planning dose based on intelligent prediction according to claim 1, Its features are, The acquisition of the treatment plan formulated by the radiotherapy planning system and the calculation of the corresponding planned dose distribution data Dose-A include: The treatment plan data can be directly retrieved from the database of the radiotherapy planning system, or imported by parsing the digital treatment plan file. The calculation yields the corresponding planned dose distribution data Dose-A, including: The radiotherapy planning system performs dose calculations based on the treatment plan and outputs the planned dose distribution data Dose-A; or, the dose calculation engine within the radiotherapy planning system recalculates the planned dose distribution data Dose-A based on machine parameters obtained from the treatment plan and the patient's anatomical model. The planned dose distribution data Dose-A is a three-dimensional dose distribution matrix, which is consistent with the predicted actual dose distribution data Dose-B in terms of spatial dimension and grid resolution.

10. A radiotherapy planning dose evaluation system based on intelligent prediction, characterized in that, include: The planned dose distribution calculation module acquires the treatment plan formulated by the radiotherapy planning system and calculates the corresponding planned dose distribution data Dose-A. The beam dose rate prediction module uses a trained neural network model S-NET to predict the actual beam dose rate of each subfield. The grating blade position prediction module uses a trained neural network model M-NET to predict the actual position of each grating blade in each subfield. The dose calculation module performs dose calculation based on the predicted actual beam dose rate of each subfield and the actual position of each grating blade in each subfield to obtain the predicted actual dose distribution data Dose-B. The dose distribution data evaluation module calculates the first deviation data DELTA-A between the planned dose distribution data Dose-A and the predicted actual dose distribution data Dose-B, and calculates the second deviation data DELTA-P between the predicted dose distribution data Dose-B and the prescription dose. The dose evaluation report module generates a dose evaluation report based on the first deviation data DELTA-A and the second deviation data DELTA-P.

Images