Quality assurance method and system for intensity modulated radiotherapy plan based on virtual prediction

Abstract

The present invention discloses a quality assurance method and system for intensity modulated radiotherapy plans based on virtual prediction, including: extracting complexity parameters from the patient's radiotherapy plan RTplan, and extracting three-dimensional dose distribution characteristics from RTdose, combining the pre-processed data of the TPS calculation flux map, and designing and training an end-to-end convolutional neural network model. The model comprehensively analyzes the plan complexity, dosimetric characteristics, and flux image characteristics to predict virtual measured flux and gamma pass rate. Through multi-dimensional data fusion, the present invention significantly improves prediction accuracy and QA efficiency, and can automatically evaluate the quality of radiotherapy plans in advance, helping to accelerate the radiotherapy process and improve treatment accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing and radiotherapy, and in particular to a method and system for quality assurance of intensity modulated radiotherapy plans for tumor radiotherapy patients based on virtual prediction. Background Art

[0002] Radiotherapy has become one of the three major methods for treating malignant tumors. About 70% of cancer patients need radiotherapy during treatment. In the past two decades, radiotherapy technology has made great progress, from the initial conventional radiotherapy and three-dimensional conformal radiotherapy to the current intensity-modulated radiotherapy, volumetric rotation intensity-modulated radiotherapy, and more advanced proton and heavy ion radiotherapy. Intensity-modulated radiotherapy can deliver a higher dose to the radiotherapy target area while protecting adjacent normal tissues by adjusting the rotation of the gantry, the movement of the multi-leaf collimator and the dose rate. It is a tumor radiotherapy technology currently commonly used by radiotherapy institutions.

[0003] Quality Assurance (QA) of IMRT for patients is a key method to ensure accurate dose delivery. The common method is to transmit the radiotherapy plan to a measurement device (such as a diode or ionization chamber array, electronic portal imaging system or film, etc.), and then compare and analyze it with the calculated dose of the treatment planning system.

[0004] Gamma analysis combines dose difference and distance to agreement (DTA) and is widely used in current patient-specific quality assurance. The American Association of Physics in Medicine Task Force 218 recommends that for intensity-modulated radiotherapy plans, a gamma passing rate (GPR) of 3% / 2mm to reach 95% is clinically acceptable.

[0005] There are many sources of error that affect the accuracy of treatment planning and delivery, including beam modeling uncertainty in treatment planning, spatial uncertainty of the delivery system, and intrinsic uncertainty in measurement-based QA [3]. Highly modulated treatment plans usually involve more restrictive machine parameters, resulting in greater uncertainty in dose calculation and treatment delivery. These machine parameters include gantry rotation speed, multileaf collimator (MLC) leaf position, multileaf collimator leaf speed, jaw position, dose rate, and machine jump number MU per control point. Some planning parameters, such as beam energy, number of control points, minimum segment area and width, also have an impact on accuracy. In addition, due to the complexity of measurement problems, measurement systems such as electronic portal imaging systems and two-dimensional or three-dimensional dose verification systems themselves also have systematic deviations. These deviations will also affect the results of radiotherapy plan verification. Summary of the invention

[0006] In view of the defects of the prior art, the present invention provides a quality assurance method and system for intensity modulated radiotherapy plans based on virtual prediction. Starting from the complexity of the radiotherapy plan, the omics characteristics of the dose spatial distribution, and the flux image of the plan verification, a deep learning method is used to predict the measured flux image and the final gamma pass rate. This virtual quality assurance method for intensity modulated radiotherapy plans can not only warn of quality control failures of patient radiotherapy plans in advance, but also help analyze the causes of failure, thereby reducing cumbersome clinical work and optimizing the radiotherapy process.

[0007] In order to achieve the above invention object, the technical solution adopted by the present invention is as follows:

[0008] A quality assurance method for intensity modulated radiotherapy plan based on virtual prediction includes the following steps:

[0009] S1: Extract information including gantry, jaw, MLC and MU from the patient's radiotherapy plan RTplan, calculate and screen complexity parameters related to the Gamma pass rate.

[0010] S2: Extract the shape, texture, statistics and high-order features of the three-dimensional dose distribution from the patient's radiotherapy dose RTdose, and screen the important dosimetric features related to the pass rate.

[0011] S3: Obtain the patient's TPS calculation flux map and preprocess the flux map.

[0012] S4: Design an end-to-end convolutional neural network model, including a network encoder and decoder. The input is the complexity parameter feature, flux image feature and dosimetry feature, and the output is the virtual measured flux. The network weights are adjusted by the loss function.

[0013] S5: Randomly select a certain proportion of data sets to form a training set to train the convolutional neural network model, and then select a certain data set called a validation set to verify the trained virtual flux prediction model through evaluation indicators.

[0014] S6: Evaluate the accuracy of the model using indicators including mean absolute error (MAE), mean square error (MSE), consistency distance (DTA), maximum dose deviation, and gamma pass rate. Save the optimized model to facilitate prediction of virtual measurement flux and gamma pass rate for new patients.

[0015] Preferably, in step S1, planning complexity parameters with a high correlation with the Gamma pass rate or a large contribution to the importance of predicting the Gamma pass rate are selected, including blade stroke LT, average aperture area to front door area ratio AAJA, aperture circumference variation APV, modulation degree M, LT modulation complexity score LTMCS and blade stroke and modulation complexity parameter LAAM.

[0016] Preferably, in step S2, the important dose-omics features include: grayscale co-occurrence matrix features, grayscale size region matrix features and grayscale dependence matrix features.

[0017] Preferably, in step S3, the calculated and measured flux maps are reconstructed with a resolution of 1 mm×1 mm, and the beam intensity distribution is normalized to the central axis of the EPID to eliminate the impact of daily machine output changes.

[0018] Preferably, in step S4, the encoder and the decoder use a U-Net architecture, and the complexity parameter features and the dosimetric features are added to the flux image features to form a joint feature for predicting the virtual measured flux.

[0019] Preferably, in step S5, 60% of the data set data is used for training, 20% is used to verify the quality of the model, and the remaining 20% ​​of the independent test set data is used to evaluate the accuracy of the model. During the training process, online data enhancement methods are used, including translation, rotation and cropping.

[0020] Preferably, in step S6, the Gamma pass rate includes: using 3% maximum dose deviation and 3mm consistency distance, 3% maximum dose deviation and 2mm consistency distance and 2% maximum dose deviation and 2mm consistency distance.

[0021] The present invention also discloses an intensity modulated radiotherapy plan quality assurance system, which can be used to implement the above-mentioned intensity modulated radiotherapy plan quality assurance method, specifically including:

[0022] Data input module: used to input the patient's radiotherapy plan RTplan, radiotherapy dose RTdose and TPS calculation flux map data. This module also includes data preprocessing functions.

[0023] Complexity parameter extraction and screening module: extracts the complexity parameters of the radiotherapy plan from RTplan, calculates related parameters, and screens out important complexity parameter features related to the Gamma pass rate.

[0024] Dosimetric feature extraction and screening module: extracts the shape, texture, statistics and high-order features of the three-dimensional dose distribution from RTdose, and screens out key dosimetric features related to the gamma pass rate.

[0025] Flux image preprocessing module: obtains and preprocesses TPS calculated flux images, including resolution reconstruction and normalization of beam intensity distribution to eliminate the impact of machine output changes.

[0026] Convolutional neural network model module: Design and implement an end-to-end convolutional neural network model. This module includes a network encoder and decoder, inputs complexity parameter features, flux image features and dosimetry features, outputs virtual measurement flux, and optimizes model weights through a loss function.

[0027] Training and verification module: This module randomly selects data sets to form training sets and verification sets, trains and verifies the convolutional neural network model, and uses evaluation indicators to verify the accuracy of the model.

[0028] Model evaluation and saving module: Use evaluation indicators to perform final evaluation on the model, optimize and save the model for prediction of virtual measurement flux and gamma pass rate of new patients.

[0029] Patient data storage module: used to store the patient's radiotherapy plan data, radiotherapy dose data, TPS flux map, and derived complexity parameters and dose-omics characteristics, and radiotherapy plan pass rate information.

[0030] QA Result Analysis Module: Analyzes and displays the gamma pass rate of the virtual measurement flux and the reasons for QA failure, including visualization images showing the dose deviation of the measurement points in the X and Y directions and the low pass rate locations.

[0031] User interaction interface module: provides users with an interactive interface to display the current patient's TPS calculation flux map, radiotherapy dose distribution, virtual measurement flux prediction results, and provide text and image analysis of QA results.

[0032] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned quality assurance method for intensity modulated radiotherapy plan when executing the program.

[0033] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the above-mentioned intensity modulated radiotherapy plan quality assurance method is implemented.

[0034] Compared with the prior art, the advantages of the present invention are:

[0035] 1. Comprehensive analysis of multi-dimensional information: Compared with the existing technology that only relies on plan complexity to predict the Gamma pass rate, this invention not only combines the complexity characteristics of the radiotherapy plan, but also integrates the dosimetry characteristics and the flux map characteristics of TPS calculation. This integration of multi-dimensional information takes into account the spatial dose distribution characteristics of the radiotherapy plan and the inherent disturbance problems of the plan verification equipment, greatly improving the accuracy and redundancy of the prediction and ensuring more reliable QA results.

[0036] 2. Virtual measurement flux prediction and QA analysis: By predicting the virtual measurement flux, the present invention can not only calculate the virtual Gamma pass rate, but also analyze the reasons for QA failure in detail. The system can provide visualization of the dose deviation of the measurement points in the X and Y directions and the specific locations with low pass rates, so that physicists can quickly identify and deal with problems in the radiotherapy plan in a targeted manner, thereby improving the efficiency and accuracy of QA analysis.

[0037] 3. Automation and efficiency: The present invention can predict the QA results of the patient's radiotherapy plan before actual measurement, realizing the automation of radiotherapy plan verification. This automation improves the work efficiency of physicists, simplifies the radiotherapy process, and shortens the patient's waiting time. At the same time, the system can be deployed in portable devices, expanding the application scenarios and increasing the flexibility and convenience of use. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flow chart of a virtual IMRT plan quality assurance method in an embodiment of the present invention.

[0039] Figure 2 It is a schematic diagram of the process of extracting the complexity parameter features of the radiotherapy plan in an embodiment of the present invention.

[0040] Figure 3 It is a schematic diagram of the process of extracting dosimetric features of radiotherapy dose distribution in an embodiment of the present invention.

[0041] Figure 4 It is a training graph of the virtual measurement flux prediction model in the embodiment of the present invention.

[0042] Figure 5 It is a diagram of a user interaction interface of a portable system in an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and examples.

[0044] like Figures 1 to 4 As shown, the present invention provides a quality assurance method for intensity modulated radiotherapy plan based on virtual prediction, comprising:

[0045] S1 Extracts radiotherapy plan (RTplan) complexity parameter information

[0046] S11 The present invention uses the Pydicom software package to read the RTplan of patients undergoing tumor radiotherapy (including patients with lung cancer, liver cancer, brain metastasis, bone metastasis, etc.), and uses self-compiled python code to obtain information such as the field area, aperture circumference, leaf movement, machine jump number, field modulation complexity, etc. of different sub-fields in the intensity-modulated radiotherapy plan.

[0047] S12 uses the 2% / 2mm standard Gamma pass rate (>=95%) as the threshold to classify the patient's QA results into two categories: pass and fail. A random forest (RF) model is established, and the aforementioned complexity parameters are used as the input of the random forest model to set the standard Gamma pass rate as the output of the RF model.

[0048] S13 establishes training set, validation set and test set. The above model is trained by random parameter search to obtain the best RF regression model (five-fold cross validation). The test set data is input into the above model to predict the Gamma pass rate and select the more important complexity parameters, including blade stroke LT, average aperture area to front door area ratio AAJA, aperture perimeter variation APV, modulation degree M, LT modulation complexity score LTMCS and LAAM.

[0049]

[0050] In the formula, k represents the control point, N represents the number of MLCs, ApertureArea represents the aperture area under the k control point, JawArea represents the area surrounded by the jaw under the k control point, Perimeter represents the aperture perimeter under the k control point, and P max Represents the maximum distance between two sets of leaves.

[0051] S2 Extraction of dosimetric features

[0052] S21 obtains the patient's radiotherapy dose RTdose file, uses the Pydicom software package to read the data in the three-dimensional dose grid, and normalizes the data to 1×1×1mm 3 The Pyradiomics software package was used to extract seven types of dosimetric features, including shape features (2D / 3D), first-order features, gray-level co-occurrence matrix features (GLCM), gray-level size zone matrix features (GLSZM), gray-level run length matrix features (GLRLM), neighborhood gray-level difference matrix features (NGTDM), and gray-level dependence matrix features (GLDM).

[0053] S22 uses the 2% / 2mm standard Gamma pass rate (>=95%) as the threshold to classify the patient's QA results into two categories: pass and fail. Establish training sets, validation sets, and test sets. Input the training data set into the random forest model. Then, calculate the importance of each feature in the data set and sort it in descending order to use as a feature selection method.

[0054] S23 selects important dosimetric features related to the pass rate, including grayscale co-occurrence matrix features, grayscale size area matrix features, grayscale dependence matrix and other features.

[0055] S3 obtains the patient's TPS calculation flux and performs preprocessing such as resolution normalization on the flux map.

[0056] For each patient, S31 uses the Portal dose Image prediction algorithm in the radiotherapy planning system TPS to calculate the PDIP flux for each field. The PDIP image is a dose distribution image generated by TPS similar to the 2D EPID. When the QA plan is performed on a medical linear accelerator, a true measured flux map based on the EPID is obtained. Prior to measurement, the electronic portal imaging device (EPID) is absolutely calibrated in calibration units (CU) using 100MU and an SSD (source skin distance) of 100 cm to irradiate an open field of 10×10 cm.

[0057] S32 The calculated flux based on TPS and the measured flux based on EPID usually have different resolutions. The present invention resamples both to a resolution of 1mm×1mm. The beam intensity distribution is normalized to the central axis of EPID to eliminate the impact of daily machine output changes.

[0058] S4 builds a convolutional neural network

[0059] The encoder and decoder of the convolutional neural network use an architecture similar to U-Net. The data input to the network includes complexity parameter features, dosimetric features, and flux images. However, these feature spaces have different dimensions, and the above feature dimensions still need to be filled in according to the subsequent deep prediction model dimensions. These features are then combined into the input of the deep prediction model. The output of the network is a virtual measured flux map.

[0060] S5 model training

[0061] S51 60% of the data set is used for training, 20% is used to verify the quality of the model, and the remaining 20% ​​of the independent test set data is used to evaluate the accuracy of the model. The data set of the external center is used as an external independent test set to evaluate the generalization ability of the model. During the training process, we used online data enhancement methods, including translation, rotation, and cropping.

[0062] During S52 training, MAE loss is used to calculate the difference between the reference flux map and the predicted flux map, and the hyperparameters of the network model are optimized based on the results.

[0063] S6 generates virtual measurement flux and predicts gamma pass rate

[0064] S61 We use the distance to agreement (DTA) and the maximum dose deviation and the Gamma pass rate to evaluate the deviation between the predicted virtual flux and the actual measured flux. The commonly used Gamma pass rate uses 3% maximum dose deviation with 3mm agreement distance, 3% maximum dose deviation with 2mm agreement distance and 2% maximum dose deviation with 2mm agreement distance. The present invention uses the standard of 2% maximum dose deviation with 2mm agreement distance.

[0065] S62 The present invention provides medical physicists with model interpretability through the heat map of the attention mechanism.

[0066] In one embodiment of the present invention, a quality assurance system for intensity modulated radiotherapy plans is provided. The system can be used to implement the above-mentioned quality assurance method for intensity modulated radiotherapy plans, and specifically includes:

[0067] Patient radiotherapy data information library module: used to store and manage patients' radiotherapy-related data, including RTplan, RTdose, radiotherapy flux map, and complexity parameters, dosimetry characteristics and radiotherapy plan pass rate information derived from these data. This module ensures that all data can be efficiently accessed and analyzed within the system.

[0068] Data input module: used to input and display the TPS calculated flux, radiotherapy plan RTplan and radiotherapy dose RTdose of the current patient. This module not only displays the image distribution of these data, but also provides input functions to facilitate users to view and analyze the radiotherapy data of the current patient.

[0069] Data extraction and processing module: extracts key information from the input data, such as field area, aperture circumference, leaf motion, machine hop count (MU), field modulation complexity, three-dimensional dose distribution, shape, texture, statistics and high-order dosimetric features, and resolution normalization of flux maps.

[0070] Complexity parameter analysis module: Calculate and screen the complexity parameters of radiotherapy plans related to the Gamma pass rate, including leaf stroke LT, average aperture area to front door area ratio AAJA, aperture perimeter variation APV, modulation degree M, LT modulation complexity score LTMCS and LAAM.

[0071] Feature selection and modeling module: Based on the extracted complexity parameters and dosimetric features, the random forest (RF) model was used for feature selection, and the optimal RF regression model was established through five-fold cross validation.

[0072] Virtual flux prediction module: Using a designed convolutional neural network (CNN) with a U-Net architecture, the complexity parameters, dosimetric features, and flux image features are combined to predict the virtual measured flux and gamma pass rate.

[0073] Model training and validation module: responsible for creating training sets, validation sets, and test sets, using data augmentation technology to train and validate the model, and evaluating model performance through indicators such as MAE, MSE, consistency distance DTA, and Gamma pass rate.

[0074] QA Analysis and Gamma Pass Rate Calculation Module: Calculates the Gamma pass rate of the patient's radiotherapy plan, analyzes and displays QA results, including pass or fail classification. If QA fails, the system will provide detailed text and image analysis to help users identify and correct the problem.

[0075] Result output and visualization module: displays TPS calculated flux diagram, radiotherapy dose distribution and virtual measured flux predicted by network model. This module is also responsible for visual display of results, providing clear analysis reports to help users make decisions.

[0076] User interaction interface module: provides a friendly user interface to support user interaction with the system, such as data input, result display, QA analysis, etc. Figure 5 The design in the package intuitively displays all key functions and data analysis results.

[0077] In one embodiment of the present invention, a terminal device is provided, the terminal device includes a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding functions; the processor described in the embodiment of the present invention can be used for the operation of the quality assurance method of intensity-modulated radiotherapy plan.

[0078] In one embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It is understandable that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one disk storage.

[0079] One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the corresponding steps of the quality assurance method for intensity-modulated radiotherapy plans in the above-mentioned embodiment; one or more instructions in the computer-readable storage medium are loaded and executed by a processor.

[0080] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0081] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0082] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0083] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0084] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the implementation methods of the present invention, and should be understood that the protection scope of the present invention is not limited to such special statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A quality assurance method for intensity modulated radiotherapy plan based on virtual prediction, comprising the following steps: S1: Extract information including gantry, jaw, MLC and MU from the patient's radiotherapy plan RTplan, calculate and screen complexity parameters related to Gamma pass rate; S2: Extract the shape, texture, statistics and high-order features of the three-dimensional dose distribution from the patient's radiotherapy dose RTdose, and screen the important dosimetric features related to the pass rate; S3: obtaining the patient's TPS calculation flux map and preprocessing the flux map; S4: Design an end-to-end convolutional neural network model, including a network encoder and a decoder; the input is complexity parameter features, flux image features and dosimetry features, and the output is a virtual measured flux; adjust the network weights through a loss function; S5: Randomly select a certain proportion of data sets to form a training set to train the convolutional neural network model, and then select a certain data set called a validation set to verify the trained virtual flux prediction model through evaluation indicators; S6: Evaluate the accuracy of the model using indicators including mean absolute error (MAE), mean square error (MSE), consistency distance (DTA), maximum dose deviation, and gamma pass rate. Save the optimized model to facilitate prediction of virtual measurement flux and gamma pass rate for new patients.

2. The method for quality assurance of intensity modulated radiotherapy plan according to claim 1, characterized in that: In step S1, plan complexity parameters with a high correlation with the Gamma pass rate or a large contribution to the importance of predicting the Gamma pass rate are selected, including blade stroke LT, average aperture area to front door area ratio AAJA, aperture perimeter variation APV, modulation degree M, LT modulation complexity score LTMCS and LAAM.

3. The method for quality assurance of intensity modulated radiotherapy plan according to claim 1, characterized in that: In step S2, important dosimetric features include: grayscale co-occurrence matrix features, grayscale size region matrix features and grayscale dependence matrix features.

4. The method for quality assurance of intensity modulated radiotherapy plan according to claim 1, characterized in that: In step S3, the calculated and measured flux maps are reconstructed with a resolution of 1 mm × 1 mm, and the beam intensity distribution is normalized to the central axis of the EPID to eliminate the impact of daily machine output changes.

5. The method for quality assurance of intensity modulated radiotherapy plan according to claim 1, characterized in that: In step S4, the encoder and decoder use the U-Net architecture, and the complexity parameter features and dosimetric features are added to the flux image features to form a joint feature for predicting the virtual measured flux.

6. The method for quality assurance of intensity modulated radiotherapy plan according to claim 1, characterized in that: In step S5, 60% of the data set is used for training, 20% is used to verify the quality of the model, and the remaining 20% ​​of the independent test set data is used to evaluate the accuracy of the model; During training, online data augmentation methods are used, including translation, rotation, and cropping.

7. The method for quality assurance of intensity modulated radiotherapy plan according to claim 1, characterized in that: In step S6, the gamma pass rate includes: using 3% maximum dose deviation and 3mm consistency distance, 3% maximum dose deviation and 2mm consistency distance and 2% maximum dose deviation and 2mm consistency distance.

8. An intensity modulated radiotherapy plan quality assurance system, characterized by: The system can be used to implement the quality assurance method for intensity modulated radiotherapy plan according to any one of claims 1 to 7, specifically comprising: Data input module: used to input the patient's radiotherapy plan RTplan, radiotherapy dose RTdose and TPS calculation flux map data; this module also includes data preprocessing function; Complexity parameter extraction and screening module: extracts the complexity parameters of the radiotherapy plan from RTplan, calculates related parameters, and screens out important complexity parameter features related to the Gamma pass rate; Dosimetric feature extraction and screening module: extracts the shape, texture, statistics and high-order features of the three-dimensional dose distribution from RTdose, and screens out key dosimetric features related to the gamma pass rate; Flux image preprocessing module: obtains and preprocesses TPS calculated flux images, including resolution reconstruction and normalization of beam intensity distribution to eliminate the impact of machine output changes; Convolutional neural network model module: Design and implement an end-to-end convolutional neural network model; this module includes a network encoder and decoder, inputs complexity parameter features, flux image features and dosimetry features, outputs virtual measurement flux, and optimizes model weights through a loss function; Training and verification module: This module randomly selects data sets to form training sets and verification sets, trains and verifies the convolutional neural network model, and uses evaluation indicators to verify the accuracy of the model; Model evaluation and saving module: Use evaluation indicators to perform final evaluation on the model, optimize and save the model for prediction of virtual measurement flux and gamma pass rate of new patients; Patient data storage module: used to store the patient's radiotherapy plan data, radiotherapy dose data, TPS flux map, and derived complexity parameters and dose group characteristics, and radiotherapy plan pass rate information; QA result analysis module: Analyzes and displays the gamma pass rate of the virtual measurement flux and the reasons for QA failure, including visualization images showing the dose deviation of the measurement points in the X and Y directions and the low pass rate positions; User interaction interface module: provides users with an interactive interface to display the current patient's TPS calculation flux map, radiotherapy dose distribution, virtual measurement flux prediction results, and provide text and image analysis of QA results.

9. A computer device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for quality assurance of intensity modulated radiotherapy plan as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the method for quality assurance of intensity modulated radiotherapy plan described in one of claims 1 to 7 is implemented.

Images