A helical tomotherapy plan evaluation method, system and electronic device

Abstract

The application relates to the technical field of artificial intelligence, and provides a helical tomotherapy plan evaluation method, which comprises the following steps: acquiring an RTPlan file, a CT image file and an RTStruct file corresponding to a helical tomotherapy plan; extracting model input characteristic parameters from the RTPlan file, the CT image file and the RTStruct file; inputting the model input characteristic parameters into a gamma pass rate prediction model for calculation to obtain a gamma pass rate prediction result; and evaluating the helical tomotherapy plan based on the gamma pass rate prediction result, so that the evaluation efficiency of the radiotherapy plan is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to a helical tomotherapy plan evaluation method and system, an electronic device and a computer readable storage medium. BACKGROUND

[0002] Helical tomotherapy (HT) is an advanced radiotherapy technique that provides highly conformal dose distribution by using intensity-modulated radiation therapy. Helical tomotherapy plan is a cancer radiotherapy planning scheme relying on a helical tomotherapy (TOMO) device.

[0003] Gamma passing rate (GPR) is an index used to evaluate the accuracy of dose distribution in radiotherapy plans. It is used to compare the differences between the dose received by the patient during actual treatment and the planned dose distribution. In radiotherapy, ensuring accurate dose delivery is crucial for treatment effectiveness, so accurate prediction of gamma passing rate helps improve the accuracy and efficiency of treatment planning.

[0004] In traditional technology, the method for obtaining gamma passing rate usually involves the following steps:

[0005] 1. Plan system dose calculation: using the dose calculation algorithm of the clinical patient treatment application to perform plan system dose calculation.

[0006] 2. Dose verification: dose verification includes point dose verification and dose distribution verification. The point dose verification includes selecting a point dose measurement probe suitable for IMRT dose verification and a corresponding phantom. The principle of measuring point position selection is to make the difference between the maximum dose and the minimum dose in the probe volume as small as possible. Dose distribution verification includes using film, probe array, EPID, machine log file based calculation, etc. to verify the dose distribution.

[0007] 3. Gamma analysis: by comparing the differences between the measured dose distribution and the calculated dose distribution to determine the gamma passing rate.

[0008] It is found in the research that the accuracy of the phantom positioning, the correctness of the device parameters, the accuracy of the probe array and EPID measurement, etc. have a great influence on the results of dose distribution verification. In addition, the use of film, probe array, EPID, etc. has a very high occupancy rate on the helical tomotherapy linear accelerator treatment device. Therefore, the traditional verification analysis process not only needs to ensure the accuracy of each verification operation, but also has a high device occupancy rate, which leads to a large number of clinical working hours occupied by physicists and low efficiency. SUMMARY

[0009] To improve the evaluation efficiency of a radiotherapy plan, an embodiment of the present application provides a helical tomotherapy plan evaluation method, system, electronic device and computer storage medium.

[0010] The helical tomotherapy plan evaluation method provided by the present application comprises obtaining an RTPlan file, a CT image file and an RTStruct file corresponding to a helical tomotherapy plan; extracting model input feature parameters from the RTPlan file, the CT image file and the RTStruct file; inputting the model input feature parameters into a gamma pass rate prediction model for calculation to obtain a gamma pass rate prediction result; and evaluating the helical tomotherapy plan based on the gamma pass rate prediction result.

[0011] Based on the above technical solution, the gamma pass rate corresponding to the radiotherapy plan can be quickly obtained based on the gamma pass rate prediction model, thereby improving the evaluation efficiency of the radiotherapy plan. Meanwhile, in the prediction, the patient CT image file is introduced, and a more accurate prediction result can be obtained, thereby improving the accuracy of the radiotherapy plan evaluation.

[0012] In an implementation, the gamma pass rate prediction model is constructed based on an XGBoost algorithm.

[0013] In an implementation, the method for obtaining the training sample of the gamma pass rate prediction model comprises: constructing a case data set, wherein the case data set comprises a target number of completed radiotherapy case data, and the case data comprises a historical RTPlan file, a historical CT image file and a historical RTStruct file; extracting initial feature combination data from the historical RTPlan file, the historical CT image file and the historical RTStruct file of the case data according to a feature extraction strategy; pre-processing the initial feature combination data corresponding to each case data; performing feature screening analysis based on the pre-processed initial feature combination data to determine an optimal feature combination; and constructing the training sample based on the optimal feature combination.

[0014] Based on the above technical solution, the features in the case data set are comprehensively extracted and screened, and then the optimal feature combination is screened, so as to improve the accuracy of the prediction model.

[0015] In an implementation, the feature extraction strategy comprises: extracting treatment planning system related features, absolute leaf opening time statistical quantity related features, relative leaf opening time statistical quantity related features, geometric sine map features and modulation sine map features from the historical RTPlan file; and extracting imageomics features from the historical CT image file and the historical RTStruct file.

[0016] In an implementation, the treatment planning system related features include modulation factors, number of projections, number of projections per rotation, number of rotations, projection time, gantry period, treatment time, field width, tilt, couch movement distance, couch speed, target length, treatment time over fraction dose; absolute leaf opening time statistic related features include LOT mean, LOT standard deviation, LOT median, LOT mode, LOT kurtosis, LOT skewness, LOT maximum, LOT minimum, proportion of LOT less than 100, proportion of LOT less than 50, proportion of LOT less than 30, proportion of LOT less than 20, and proportion of LOT with projection time interval less than 20 ms; the relative leaf opening time statistic related features include Fractional LOT mean, Fractional LOT standard deviation, Fractional LOT median, Fractional LOT mode, Fractional LOT maximum, Fractional LOT minimum, proportion of Fractional LOT less than 5, proportion of Fractional LOT less than 10, proportion of Fractional LOT less than 25, proportion of Fractional LOT less than 50, proportion of Fractional LOT less than 75, proportion of Fractional LOT less than 90, proportion of Fractional LOT less than 95; the geometric sinogram features include 0 nearest neighbor open leaf, 1 nearest neighbor open leaf, 2 nearest neighbor open leaf, number of connected components, connected component length, treatment region, proportion of discontinuous projections, fraction of closed leaves, fraction of closed leaves in treatment region, centroid; the modulation sinogram features include number of openings and closings during treatment, planned sinogram time variation coefficient, expanded planned sinogram time variation coefficient, leaf opening time variation coefficient, expanded leaf opening time variation coefficient, modulation index, mean of sinogram, mode of sinogram, standard deviation of sinogram, average sinogram symmetry.

[0017] Based on the above technical solution, by selecting the feature parameters to construct the initial feature combination data, comprehensive extraction of related features can be ensured.

[0018] In an implementation, the pre-processing of the initial feature combination data corresponding to each of the case data includes down-sampling the initial feature combination data under the sample grouping corresponding to the regression task and the classification task, respectively.

[0019] Based on the technical scheme, by constructing the training samples corresponding to the classification task and the regression task, the model can output the gamma passing rate and the classification result under the preset threshold after training, so that the user can intuitively and quickly obtain the evaluation result. At the same time, by the downsampling operation, the balanced processing of the sample data can be realized, the data distribution is ensured to be relatively balanced, and overfitting of the prediction is prevented.

[0020] In an implementation, the feature screening analysis based on the preprocessed initial feature combination data to determine the optimal feature combination includes: performing a first screening on the features in the initial feature combination based on correlation analysis to obtain a first feature combination; performing a second screening on the first feature combination based on a recursive feature elimination cross-validation algorithm to obtain a second feature combination; and performing SHAP value evaluation selection on the second feature combination to determine the optimal feature combination.

[0021] Based on the technical scheme, the optimal feature combination is obtained through multiple screenings, which can ensure that the optimal feature combination selected can improve the prediction accuracy and performance of the model.

[0022] Based on the same inventive concept, the embodiment of the present application also provides a helical tomotherapy plan evaluation system, which is used to implement the helical tomotherapy plan evaluation method.

[0023] In addition, the embodiment of the present application also provides an electronic device, which includes a processor, a memory, and a program or instruction stored on the memory and executable on the processor, and the program or instruction is executed by the processor to implement the above method.

[0024] The embodiment of the present application further provides a computer-readable storage medium storing a computer program, and the computer program is executed by a processor to implement the above method. BRIEF DESCRIPTION OF DRAWINGS

[0025] The accompanying drawings, which form a part of the present application, are intended to provide further understanding of the present application, and the illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute improper limitations on the present application.

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0027] Figure 1 A flow chart of the helical tomotherapy plan evaluation method provided by the embodiment of the present application is shown.

[0028] Figure 2 A flow chart of a method for constructing a gamma pass rate prediction model in an embodiment of the present application is shown.

[0029] Figure 3 A flow chart of a method for obtaining training samples in an embodiment of the present application is shown.

[0030] Figure 4 A flow chart of a preprocessing method in an embodiment of the present application is shown.

[0031] Figure 5 A flow chart of a feature screening analysis method in an embodiment of the present application is shown.

[0032] Figure 6 A structural schematic diagram of a helical tomotherapy plan evaluation system provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0033] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0034] In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more, and "first", "second", and various numerical numbers are only used for distinguishing purposes, and do not limit the scope of the embodiments of the present application.

[0035] The features, structures or characteristics in the present application can be combined in any suitable manner in one or more embodiments. In various embodiments of the present application, the magnitude of the serial number of each process does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0036] Some optional features in the embodiments of the present application can be implemented independently in some scenarios without relying on other features, to solve corresponding technical problems and achieve corresponding effects, or can be combined with other features according to needs in some scenarios.

[0037] In the present application, unless otherwise specified, the same or similar parts between various embodiments can be mutually referred. In various embodiments of the present application, if there is no special specification and no logical conflict, the terms and / or descriptions between different embodiments are consistent and can be mutually referred, and the technical features in different embodiments can be combined to form new embodiments according to their inherent logical relationship. The implementation manners of the present application do not constitute a limitation on the protection scope of the present application.

[0038] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0039] The embodiments of the present application provide a helical tomotherapy plan evaluation method, which realizes automatic prediction of gamma pass rate by combining artificial intelligence technology, does not need to occupy any experimental verification equipment, improves the efficiency of obtaining gamma pass rate, and saves the clinical working time of a physicist.

[0040] Please refer to Figure 1 The method provided by the embodiments of the present application specifically includes the following steps.

[0041] S101, acquire an RTPlan file, a CT image file and an RTStruct file corresponding to a helical tomotherapy plan.

[0042] In implementation, the evaluation system can acquire the RTPlan file, the CT image file and the RTStruct file corresponding to the helical tomotherapy plan to be evaluated by interfacing with a radiotherapy planning system or receiving user input, wherein the CT image file is a CT image file in patient treatment data.

[0043] S102, extract model input feature parameters from the RTPlan file and the CT image file and the RTStruct file.

[0044] The model input feature parameters include a combination of multiple feature parameters, which are determined based on the input layer of the gamma pass rate prediction model and will be described below.

[0045] In an implementation, the corresponding feature parameters can be extracted from the RTPlan file based on a matlab programming environment and by using an open-source matlab library TCoMx. The corresponding feature parameters can be extracted from the CT image file and the RTStruct file based on a python programming environment and by using an open-source python library pyRadomics.

[0046] S103, input the model input feature parameters into a gamma pass rate prediction model for calculation to obtain a gamma pass rate prediction result.

[0047] In the embodiments of the present application, the prediction result output by the model can include dose distribution global 3% / 2mm, local 3% / 2mm, global 2% / 2mm and local 2% / 2mm, gamma pass rates corresponding to the four evaluation standards, and classification results under a preset threshold: pass or not pass.

[0048] S104, evaluate the helical tomotherapy plan based on the gamma pass rate prediction result.

[0049] Therefore, based on the method provided in the embodiments of the present application, the user only needs to provide the RTPlan file, the CT image file and the RTStruct file corresponding to the helical tomotherapy plan, and the gamma pass rate prediction result can be directly obtained to evaluate the helical tomotherapy plan, which is not only efficient but also does not occupy device resources; further, the multi-dimensional feature parameters are introduced into the input layer of the gamma pass rate prediction model provided in the embodiments of the present application, and the imageomic features and other information in the patient treatment data are effectively utilized, so as to provide more comprehensive and multi-dimensional data basis for the prediction of the gamma pass rate, which helps to ensure the accuracy of the prediction result of the model.

[0050] In an example, the gamma pass rate prediction model provided in the embodiments of the present application can be constructed based on the XGBoost (eXtreme Gradient Boosting) algorithm, wherein the XGBoost is a high-efficiency, flexible and scalable machine learning algorithm based on the gradient boosting framework, and the construction of the gamma pass rate prediction model based on the XGBoost has the advantages of high efficiency, good prediction performance, good interpretability, high flexibility and strong stability.

[0051] For details, please refer to Figure 2 The method for constructing the gamma pass rate prediction model in the embodiments of the present application specifically includes the following steps:

[0052] S201, obtaining a training sample.

[0053] For details, please refer to Figure 3 The method for obtaining the training sample in the embodiments of the present application specifically includes the following steps.

[0054] S301, constructing a case data set.

[0055] The case data set includes a target number of completed radiotherapy case data, and each case data includes a historical RTPlan file, a historical CT image file, a historical RTStruct file and a gamma pass rate of a measured dose distribution.

[0056] In an example, 300 cases of radiotherapy completed in a Tomo system can be selected, mainly including head and neck, abdominal and pelvic tumors, and all case data can be collected. The gamma pass rate of the measured dose distribution is the true gamma pass rate obtained after treatment, wherein the gamma pass rate of the measured dose distribution includes the gamma pass rate under 4 evaluation standards of global 3% / 2mm, local 3% / 2mm, global 2% / 2mm and local 2% / 2mm, which is used as the labeled data in the training sample.

[0057] S302, extracting initial feature combination data from the historical RTPlan file, the historical CT image file and the historical RTStruct file of the case data according to a feature extraction strategy.

[0058] Specifically, the features possibly related to the gamma pass rate can be preliminarily determined according to the research and analysis on the parameter data contained in each file, and then the corresponding feature extraction strategy is generated to extract the initial feature combination data.

[0059] In an implementation, the feature extraction strategy includes extracting 66-dimension complexity parameter features from the historical RTPlan file, wherein the complexity parameter features include treatment planning system related features, absolute leaf opening time statistics related features, relative leaf opening time statistics related features, geometric sinogram features and modulation sinogram features, and extracting imageomic features from the historical CT image file and the historical RTStruct file.

[0060] In an example, for each patient in the case data set, according to the RTPlan file in the patient case data, based on the matlab programming environment, using the open source matlab library TCoMx, 63-dimensional complexity parameter features are extracted, including 13 treatment planning system related feature parameters such as Modulation Factor (MF), projection times, projection times per rotation, rotation times, projection time, Gantry period, Treatment Time (TT), Field Width (FW), Pitch, treatment couch movement distance, treatment couch speed, target length, and treatment time exceeding the fraction dose; 13 absolute leaf opening time statistics related parameters such as LOT (Leaf Opening Time) average, LOT standard deviation, LOT median, LOT mode, LOT kurtosis, LOT skewness, LOT maximum, LOT minimum, proportion of LOT less than 100, proportion of LOT less than 50, proportion of LOT less than 30, proportion of LOT less than 20, and proportion of LOT with projection time interval less than 20 milliseconds; 13 relative leaf opening time statistics related parameters such as Fractional LOT average, Fractional LOT standard deviation, Fractional LOT median, Fractional LOT mode, Fractional LOT maximum, Fractional LOT minimum, proportion of Fractional LOT less than 5, proportion of Fractional LOT less than 10, proportion of Fractional LOT less than 25, proportion of Fractional LOT less than 50, proportion of Fractional LOT less than 75, proportion of Fractional LOT less than 90, and proportion of Fractional LOT less than 95; 13 geometric sinogram parameters such as 0 nearest neighbor open leaf, 1 nearest neighbor open leaf, 2 nearest neighbor open leaf, connected component number, connected component length, treatment region, discontinuous projection proportion, closed leaf fraction, closed leaf fraction in the treatment region (CLS in , CLS in,disc , CLS in,area , CLS in,area,disc ), and centroid; and 11 modulation sinogram parameters such as treatment period open and close times, planned sinogram time variation coefficient, expanded planned sinogram time variation coefficient, leaf opening time variation coefficient, expanded leaf opening time variation coefficient, modulation index, sinogram average, sinogram mode, sinogram standard deviation, and average sinogram symmetry.

[0061] Based on the python programming environment, the multi-dimensional image feature is extracted by using the open source python library pyRadomics, including shape features, first-order features, gray level co-occurrence matrix features, gray level size zone matrix features, gray run length matrix features, adjacent gray difference matrix features, and gray dependence matrix features.

[0062] It is worth noting that the feature parameters in the above examples are selected based on the record information in the current files and the correlation analysis. It can be understood that when the application target or file standard changes, the feature extraction strategy can be set according to the actual situation, so that the initial feature data proposed can cover the initial parameters related to the gamma prediction rate as much as possible.

[0063] S303, pre-processing the initial feature combination data corresponding to each case data.

[0064] Specifically, referring to Figure 4 , the pre-processing method specifically includes the following steps.

[0065] S401, removing the initial feature combination data containing missing values to ensure the integrity of the data.

[0066] S402, grouping the remaining initial feature combination data.

[0067] Specifically, according to the training target, the training samples of the accurate regression task and the training samples of the classification task, in an example, all the remaining initial feature combination data can be divided into two groups, i.e. the initial feature combination data corresponding to the regression task and the classification task are the same. In another example, the remaining initial feature data can also be randomly divided into regression task groups or classification task groups.

[0068] S403, down-sampling the grouped initial feature combination data.

[0069] In an example, the down-sampling method for the regression task group includes: selecting 94% as the threshold when regressing 3% / 2mm target, selecting 85% as the threshold when regressing 2% / 2mm target, respectively counting the number of initial feature combination data less than the threshold and greater than or equal to the threshold, and randomly selecting a part of data from the data with more quantity, the number of which is 3 times that of the data with less quantity. The down-sampling method for the classification task group includes: selecting 10 different thresholds from 90% to 99%, performing binary classification on the data under different thresholds, setting the target value greater than or equal to the threshold as 1 and the target value less than the threshold as 0, counting the number of 0 and 1, and determining the part with less quantity, and randomly selecting a part of data from the part with more quantity, so that the number is the same as the part with less quantity.

[0070] Balancing the data through these grouping and downsampling operations ensures that the data distribution is relatively balanced, preventing overfitting in prediction.

[0071] S404, the initial feature combination parameters after downsampling are optimized.

[0072] Specifically, if there is only one unique value in a certain parameter in the initial feature combination after downsampling, the feature can be deleted. At the same time, repeated initial feature combination data is deleted.

[0073] After the above method is completed, the following steps are included.

[0074] S304, feature selection analysis is performed based on the initial feature combination data after preprocessing to determine the optimal feature combination.

[0075] Please refer to Figure 5 In this step, the feature selection analysis method specifically includes the following steps:

[0076] S501, based on correlation analysis, the first feature combination is obtained by performing the first screening on the features in the initial feature combination.

[0077] In implementation, the correlation between different features and the correlation between each feature and the target variable can be analyzed first, and features with high correlation with the target variable and low correlation with other features are screened out to construct the first feature combination.

[0078] When performing regression and classification prediction, if there is high correlation between features, it will lead to instability and inaccuracy of the prediction model, so as to reduce redundant features and improve the training efficiency and prediction performance of the model.

[0079] In a specific example, the correlation between each feature and the target variable can be calculated first to obtain a correlation coefficient matrix, and each feature can be sorted based on the correlation coefficient matrix, and then the correlation matrix of the sorted feature matrix is calculated to obtain the correlation between different features. According to the set threshold 0.9, find the feature pair with high correlation, screen the index with correlation greater than 0.9 and less than or equal to 1, and delete the features with high correlation.

[0080] S502, based on the recursive feature elimination cross-validation algorithm, the second feature combination is obtained by performing secondary screening on the first feature combination.

[0081] In implementation, based on 5-fold cross-validation, recursive feature elimination cross-validation can be performed with mean absolute error as evaluation index to further refine the feature set and obtain the second feature combination.

[0082] S503, SHAP value evaluation selection is performed on the second feature combination to determine the optimal feature combination.

[0083] In implementation, the SHAP importance of each feature can be extracted and calculated first, the features are ranked based on the SHAP values, the number of features used is gradually increased according to the feature importance ranking, the XGBoost model is trained, and the data is segmented using different seeds, the prediction error is calculated, and the optimal feature combination is found.

[0084] S305, training samples are constructed based on the optimal feature combination.

[0085] After the optimal feature combination is determined, the parameter values of each feature can be obtained based on the corresponding initial combination feature data, and the corresponding measured gamma pass rate and classification under different thresholds are used as labels to generate training samples.

[0086] In an example, the optimal feature combination of the embodiments of the present application includes:

[0087] For the regression task of 3% / 2mm, the optimal feature combination is: LOT sharpness, the proportion of LOT less than 100, the average of the sinogram, the mode of the sinogram, the inclination, the treatment bed speed, the projection number, the rotation number, and the 40-dimensional image feature.

[0088] For the regression task of 2% / 2mm, the optimal feature combination is: LOT sharpness, the average of the sinogram, the mode of the sinogram, the inclination, the treatment bed speed, the projection number, the rotation number, and the 34-dimensional image feature.

[0089] For the classification task of 3% / 2mm, the optimal feature combination is: LOT mode, LOT minimum value, the proportion of LOT less than 100, Fractional LOT mode, Fractional LOT maximum value, the average of the sinogram, the mode of the sinogram, treatment time, target length, treatment bed moving distance, projection number, rotation number, and 48-dimensional image feature.

[0090] For the classification task of 2% / 2mm, the optimal feature combination is: Fractional LOT standard deviation, Fractional LOT maximum value, the number of opening and closing during treatment, modulation factor, the average of the sinogram, the mode of the sinogram, planned sinogram time variation coefficient, expanded leaf opening time variation coefficient, the proportion of Fractional LOT less than 75, and 33-dimensional image feature.

[0091] It is worth noting that in the process of extracting the model input feature parameters in the above step S102, the corresponding feature parameters are extracted from each file for reference to the optimal feature combination.

[0092] S202, determine the model optimal parameters.

[0093] In an example, the parameter search can be performed through the Hyperopt library, and the optimal parameters of the gamma pass rate prediction model are found through Bayesian optimization.

[0094] S203, model initialization.

[0095] Based on the optimal feature combination and the optimal parameters, the gamma pass rate prediction model is initialized.

[0096] S204, model training and evaluation.

[0097] In implementation, in order to evaluate the stability of the model, the dataset can be split, trained and evaluated multiple times through multiple random seeds.

[0098] In an example, a test set can be randomly divided from all training sample sets, wherein the number of test sets accounts for 20% of the total amount; the remaining training sample sets are randomly divided into a training set and a validation set, wherein the validation set accounts for 10% of the remaining amount, and the training set accounts for 90% of the remaining amount.

[0099] After determining the training set, the validation set and the test set, the initialized prediction model is first trained using the training set, and the trained prediction model is verified based on the validation set. Then, the feature parameters in the test set are input into the trained prediction model to obtain the output prediction result, and then the prediction result is compared with the true value corresponding to the feature parameters to determine the accuracy of the prediction result. In this way, after multiple splitting and evaluation, the average value of each evaluation index is calculated, the mean absolute error (MAE), the root mean square error (RMSE), and the classification accuracy are calculated to verify the performance of the gamma pass rate prediction model.

[0100] S205, verify the model based on the multi-center data set.

[0101] In implementation, the multi-center data can be used for model testing to evaluate the performance of the prediction model on different data sets to ensure the generalization ability, stability and reliability of the model. The multi-center data set can be constructed by obtaining case data from different regions and institutions.

[0102] Specifically, first, the feature variables and the target variables are extracted from the multi-center data set, then the extracted feature variables and the target variables are preprocessed to be consistent with the training sample format, then the trained prediction model is used to predict the feature variables, and the mean absolute error (MAE), the root mean square error (RMSE), and the classification accuracy are calculated to verify and evaluate the performance of the prediction model.

[0103] The performance evaluation results of the gamma pass rate prediction model constructed based on the above method are as follows:

[0104] For the regression task of 3% / 2mm, the performance evaluation results are as follows: the MAE on the single-center test set is 1.3342, and the RMSE is 2.2096; the MAE on the multi-center test set is 1.4608, and the RMSE is 2.2096.

[0105] For the regression task of 2% / 2mm, the performance evaluation results are as follows: the MAE on the single-center test set is 2.7612, and the RMSE is 4.0668; the MAE on the multi-center test set is 3.1458, and the RMSE is 3.8531.

[0106] For the classification task of 3% / 2mm, the performance evaluation results are as follows: the classification accuracy on the single-center test set is 80%; the classification accuracy on the multi-center test set is 90%.

[0107] For the classification task of 2% / 2mm, the performance evaluation results are as follows: the classification accuracy on the single-center test set is 90%; the classification accuracy on the multi-center test set is 85%.

[0108] As can be seen, based on the above technical solution, according to the treatment data such as the CT image file, the RTPlan file, and the RTStruct file of the patient, comprehensive complexity parameter features and imageomic features are extracted, after feature preprocessing, effective features are systematically screened out through correlation analysis, recursive feature elimination cross-validation, and SHAP value evaluation, and the model parameter optimization is performed by using the Bayesian optimization algorithm, which significantly improves the prediction performance and stability of the gamma pass rate prediction model, and through multi-center data verification, the generalization ability on different data sets is verified, which can effectively improve the accuracy and efficiency of the radiotherapy plan evaluation.

[0109] Based on the same inventive concept, the present application also provides a helical tomotherapy plan evaluation system for executing the above method, please refer to Figure 6 The helical tomotherapy plan evaluation system includes an input module 61, a data processing module 62, an analysis and calculation module 63, and an output module 64.

[0110] The input module 61 is configured to acquire an RTPlan file, a CT image file and an RTStruct file corresponding to the spiral tomotherapy plan to be evaluated, and perform integrity verification on the files. After the verification is passed, the files are sent to the data processing module 62. The acquisition of the files includes receiving files uploaded by a user or requesting the files from a radiotherapy planning system based on a user operation.

[0111] The data processing module 62 is configured to extract model input feature parameters from the RTPlan file, the CT image file and the RTStruct file, and send the input feature parameters to the analysis and calculation module 63.

[0112] The analysis and calculation module 63 is configured to input the model input feature parameters into a gamma pass rate prediction model to obtain a gamma pass rate prediction result, and display the result to a user through the output module 64.

[0113] In addition, the embodiments of the present application further provide an electronic device, which includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor. When the program or instruction is executed by the processor, the method in any one of the embodiments of the present application is implemented. The processor can be a general central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), a graphics processing unit (GPU), or one or more integrated circuits, which is used to execute the related program to implement the method in any one of the embodiments of the present application.

[0114] The processor can also be an integrated circuit electronic device with signal processing capability. In the implementation process, each step of the method in any one of the embodiments of the present application can be completed by the integrated logic circuit of hardware or the instruction in the form of software in the processor.

[0115] The above processor can also be a general processor, a digital signal processor, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as hardware code processor execution or executed by a combination of hardware and software modules in the code processor.

[0116] The software module can be located in a random access memory, a flash memory, a read only memory, a programmable read only memory or an electrically erasable programmable memory, a register, and other mature storage media in the art. The storage medium is located in the storage memory, and the processor reads the information in the storage memory, and combines the functions required by the units included in the device for data processing of the embodiments of the present application to be executed by the hardware, or executes the method in any one of the implementation manners of the embodiments of the present application.

[0117] Another embodiment of the present application relates to a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to realize the above-mentioned method embodiments.

[0118] Those skilled in the art can understand that all or part of the steps of the above-mentioned method embodiments can be completed by programs instructing related hardware. The programs are stored in a storage medium, and include a plurality of instructions for making a device (which can be a single chip microcomputer, a chip, etc.) or a processor execute all or part of the steps of the method of each embodiment of the present application. The above-mentioned storage medium includes: a U disk, a mobile hard disk, a read only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media which can store program codes.

[0119] The above are the preferred embodiments of the present application, not the limitation of the protection scope of the present application, so: all equivalent changes made on the basis of the structure, shape, principle of the present application should be covered in the protection scope of the present application.

Claims

1. A helical tomotherapy plan evaluation method, characterized by, The method comprises the steps of: obtaining an RTPlan file, a CT image file and an RTStruct file corresponding to a helical tomotherapy plan; extracting model input feature parameters from the RTPlan file, the CT image file and the RTStruct file, the model input feature parameters being corresponding feature parameters extracted from each file with reference to an optimal feature combination; inputting the model input feature parameters into a gamma pass rate prediction model for calculation to obtain a gamma pass rate prediction result, the gamma pass rate prediction model being constructed based on an XGBoost algorithm and initialized based on an optimal feature combination and optimal parameters, the optimal parameters being searched through a Hyperopt library and found through Bayesian optimization; evaluating the helical tomotherapy plan based on the gamma pass rate prediction result; the method for obtaining training samples of the gamma pass rate prediction model comprises: constructing a case data set, the case data set including a target number of completed radiotherapy case data, the case data including historical RTPlan files, historical CT image files and historical RTStruct files; extracting initial feature combination data from the historical RTPlan files, the historical CT image files and the historical RTStruct files of the case data according to a feature extraction strategy; preprocessing the initial feature combination data corresponding to each case data, the preprocessing steps comprising: down-sampling the initial feature combination data under the sample grouping corresponding to the regression task and the classification task, respectively; performing a first screening on the features in the initial feature combination based on correlation analysis to obtain a first feature combination; performing a second screening on the first feature combination based on a recursive feature elimination cross-validation algorithm to obtain a second feature combination; performing SHAP value evaluation selection on the second feature combination to determine the optimal feature combination; constructing the training samples based on the optimal feature combination; the feature extraction strategy comprises: extracting treatment planning system related features, absolute leaf opening time statistical quantity related features, relative leaf opening time statistical quantity related features, geometric sinusoidal map features and modulation sinusoidal map features from the historical RTPlan files; extracting imageomics features from the historical CT image files and the historical RTStruct files; The treatment planning system related features include modulation factors, number of projections, number of projections per rotation, number of rotations, projection time, gantry period, treatment time, field width, tilt, couch movement distance, couch speed, target length, treatment time over fraction dose; absolute leaf opening time statistic related features include LOT mean, LOT standard deviation, LOT median, LOT mode, LOT kurtosis, LOT skewness, LOT maximum, LOT minimum, proportion of LOT less than 100, proportion of LOT less than 50, proportion of LOT less than 30, proportion of LOT less than 20, and proportion of LOT with projection time interval less than 20 ms; the relative leaf opening time statistic related features include Fractional LOT mean, Fractional LOT standard deviation, Fractional LOT median, Fractional LOT mode, Fractional LOT maximum, Fractional LOT minimum, proportion of Fractional LOT less than 5, proportion of Fractional LOT less than 10, proportion of Fractional LOT less than 25, proportion of Fractional LOT less than 50, proportion of Fractional LOT less than 75, proportion of Fractional LOT less than 90, proportion of Fractional LOT less than 95; the geometric sinogram features include 0 nearest neighbor open leaf, 1 nearest neighbor open leaf, 2 nearest neighbor open leaf, number of connected components, connected component length, treatment region, proportion of discontinuous projections, fraction of closed leaves, fraction of closed leaves in treatment region, centroid; the modulation sinogram features include number of open and closed during treatment, planned sinogram time variation coefficient, expanded planned sinogram time variation coefficient, leaf opening time variation coefficient, expanded leaf opening time variation coefficient, modulation index, mean of sinogram, mode of sinogram, standard deviation of sinogram, average sinogram symmetry.

2. A helical tomotherapy plan evaluation system, characterized by, The system is used to implement the helical tomotherapy plan evaluation method of claim 1.

3. An electronic device, comprising: The system comprises a processor, a memory, and a program or instructions stored on the memory and executable on the processor, and the program or instructions are executed by the processor to implement the method of claim 1.

4. A computer readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to implement the method of claim 1. The computer program is executed by the processor to implement the method of claim 1.

Images