阅读说明：本技术 一种肿瘤患者精准剂量验证方法、装置及设备 (Accurate dose verification method, device and equipment for tumor patient ) 是由 金献测 谢聪颖 于 2019-10-21 设计创作，主要内容包括：本发明提供了一种肿瘤患者精准剂量验证方法、装置及设备,包括建立肿瘤患者信息数据库；对肿瘤患者信息数据库中患者样本数据进行分类,建立精准放疗剂量验证评估模型,得到基于患者DVH剂量验证的归一化QA指标；通过参数分析算法分析肿瘤患者信息数据库数据,得到肿瘤患者的关键计划参数；以所述肿瘤患者信息数据库样本的关键计划参数为输入参数,输入到所述的患者精准放疗剂量验证评估模型,并通过递归算法校正所述患者精准放疗剂量验证评估模型,构建精准放疗计划个体化QA自动预测模型。本发明通过提取关联计划参数和算法实现对精准放疗剂量验证评估模型的优化,通过递归算法优化得到精准放疗计划预测QA指标,实现精准的QA预测且评估效率高。(The invention provides a method, a device and equipment for verifying accurate dose of a tumor patient, which comprises the steps of establishing a tumor patient information database; classifying patient sample data in a tumor patient information database, establishing an accurate radiotherapy dose verification and evaluation model, and obtaining a normalized QA index based on patient DVH dose verification; analyzing the data of the tumor patient information database through a parameter analysis algorithm to obtain key plan parameters of the tumor patient; and inputting the key plan parameters of the tumor patient information database samples as input parameters into the patient accurate radiotherapy dose verification and evaluation model, correcting the patient accurate radiotherapy dose verification and evaluation model through a recursive algorithm, and constructing an accurate radiotherapy plan individualized QA automatic prediction model. According to the method, the optimization of the accurate radiotherapy dose verification and evaluation model is realized by extracting the associated plan parameters and the algorithm, the accurate radiotherapy plan prediction QA index is obtained by recursive algorithm optimization, the accurate QA prediction is realized, and the evaluation efficiency is high.)

1. A precise dose verification method for tumor patients is characterized by comprising

Establishing an information database of a tumor patient;

classifying patient sample data in a tumor patient information database, establishing an accurate radiotherapy dose verification and evaluation model, and obtaining a normalized QA index;

analyzing data of the tumor patient information database through a parameter analysis algorithm to obtain key plan parameters of the tumor patient;

and inputting the key plan parameters of the tumor patient information database samples as input parameters into the patient accurate radiotherapy dose verification and evaluation model, and correcting the patient accurate radiotherapy dose verification and evaluation model through a recursive algorithm to obtain an automatic QA prediction model for accurate radiotherapy plan prediction.

2. The precise dose verification method for tumor patients according to claim 1, wherein analyzing the data in the tumor patient information database by a parameter analysis algorithm to obtain the key planning parameters of the tumor patients comprises:

extracting characteristic parameters from data in a tumor patient information database, and generating association rules through the extracted characteristic parameters;

constructing an association relation network of the extracted feature parameters through the extracted feature parameters and association rules;

and selecting the characteristic parameters, and judging the influence weight on the dose verification result by calculating the information gain of the selected characteristic parameters to obtain the key plan parameters.

3. The precise dose verification method for tumor patients according to claim 2, wherein the information gain calculation formula is:

Gain(A,B)＝Entropy(A)-Entropy(A/B)

a is a characteristic parameter, Gain (A, B) is the degree of uncertainty reduction of B (overdue probability) when the characteristic parameter A is known to be constant, and Encopy (A/B) is the conditional Entropy when the characteristic X is fixed.

4. The method for accurate dose verification of tumor patients according to claim 1, wherein classifying the patient sample data in the tumor patient information database to establish an accurate radiotherapy dose verification and evaluation model comprises:

acquiring patient sample data in a tumor patient information database, extracting characteristic information in the tumor patient information database from a three-dimensional dose verification error distribution map in the patient sample data, and classifying and storing the extracted characteristic information;

carrying out data preprocessing on the extracted characteristic information and carrying out data normalization processing;

dividing the extracted data in the tumor patient information database into a training set and a verification set, training a radiotherapy dose verification and evaluation model by using the training set to obtain the accurate radiotherapy dose verification and evaluation model, and verifying through the verification set.

5. The precise dose verification method for tumor patients according to claim 4, wherein the training set is 80% of the data in the extracted tumor patient information database, and the verification set is 20% of the data in the extracted tumor patient information database.

6. The precise dose verification method for tumor patients according to claim 1, wherein the recursive algorithm is decision tree or multi-layer perceptive neural network.

7. The precise dose verification method for tumor patients according to claim 1, wherein the database of information about tumor patients comprises treatment plan complexity parameters and plan quality parameters.

8. The precise dose verification method for tumor patients according to claim 7, wherein the parameter analysis algorithm is a rule-off analysis algorithm comprising:

performing feature selection and dimension reduction processing on the treatment plan complexity parameter and plan quality parameter in the tumor patient information database;

discretizing the screened treatment plan complexity parameters and plan quality parameters, and converting the parameters into transaction data sets;

searching a frequent item set in the transactional data set, calculating support degree, generating an association rule, and setting a minimum support degree and a confidence threshold;

and constructing an incidence relation network of the treatment plan complexity parameter and the plan quality parameter by taking the screened treatment plan complexity parameter and the plan quality parameter as points and the generated incidence rule as a directed edge.

9. An accurate dose testing device for tumor patients, which is characterized in that,

the information database module is used for establishing an information database of the tumor patient;

the accurate radiotherapy dose verification and evaluation model establishing module is used for classifying patient sample data in the tumor patient information database, establishing an accurate radiotherapy dose verification and evaluation model, obtaining a normalized QA index and establishing an individualized QA automatic prediction model;

the key plan parameter analysis module analyzes the data of the tumor patient information database through a parameter analysis algorithm to obtain key plan parameters of the tumor patient;

and the accurate radiotherapy dose verification and evaluation model optimization module takes the key plan parameters of the tumor patient information database samples as input parameters, inputs the input parameters into the accurate radiotherapy dose verification and evaluation model of the patient, and corrects the accurate radiotherapy dose verification and evaluation model of the patient through a recursive optimization algorithm to obtain an accurate radiotherapy plan prediction QA index.

10. An oncology patient precision dose verification device, comprising: a processor and a machine-readable storage medium storing machine-executable instructions executable by the processor; the processor is configured to execute machine-executable instructions to implement the method of any of claims 1-8.

Technical Field

The invention relates to the field of intelligent control algorithms, in particular to a method, a system and equipment for verifying accurate dose of a tumor patient.

Background

Precision radiation therapy is a constantly sought goal in the field of radiotherapy, namely to ensure that the tumor region is exposed to an accurate and sufficient dose of radiation during the radiation treatment, while avoiding excessive exposure of the surrounding organs at risk. The emerging techniques of precision radiotherapy, such as Intensity-modulated radiation therapy (IMRT), are well suited to this point in the dose distribution. However, due to the high degree of IMRT conformality and complexity of the planning design process, especially with the emergence of the emerging IMRT technique, Volume Modulated Arc Therapy (VMAT), the plan has more dynamically varying parameters during design and execution, increasing the uncertainty of the planning implementation. Therefore, pre-treatment dose verification becomes a critical step in the quality assurance process (QA) of accurate radiotherapy technology-specific patient planning, such as IMRT.

Due to individual patient differences, different patient-to-patient planning complexity and quality differences, the planning will also generate dose deviations of different characteristics during clinical delivery. Thus, an ideal patient-specific IMRT plan QA objective would be to ensure that the clinically delivered dose to each patient treatment plan differs from the planned dose by an amount that meets the relevant clinically allowable error criteria. However, due to objective limitations, existing dose verification methods often fail to detect dose errors sensitively and/or effectively, and the results of planning QA are not good or good, so that the development of precise radiotherapy is limited by the bottleneck.

The current gamma index is used as a common method for evaluating radiation treatment plan and QA at present, and mainly compares an actually measured dose with a dose calculated by a planning system, wherein the three parameters comprise a Percent Dose Difference (PDD), a distance to consistency (DTA) and a gamma passing rate. The percent dose difference method is extremely sensitive in the region with large dose gradient change, and a small position offset may cause a large error, so that the percent dose difference method is suitable for the region with gentle dose distribution; the DTA method is particularly sensitive in a region with a gentle dose gradient, and a small dose deviation can cause a large error, so that the DTA method is suitable for a region with a steep dose distribution; the gamma analysis method combines the advantages of the first two methods, and becomes a daily routine dose verification method applicable to both steep and gentle dose regions. Common gamma throughput measurement benchmarks include a 3% maximum dose deviation and a 3mm distance deviation (3%/3 mm benchmark), a 2% maximum dose deviation and a 2mm distance deviation (2%/2 mm benchmark), and a 1% maximum dose deviation and a 1mm distance deviation (commonly referred to as a 1%/1 mm benchmark), wherein the 3%/3 mm benchmark is more suitable for daily patient dose verification, and the 2%/2 mm benchmark and the 1%/1 mm benchmark are more suitable for scientific research due to strictness.

Dose calculations are performed in a three-dimensional network matrix, and the display of two-dimensional and three-dimensional dose distributions is actually a two-dimensional and three-dimensional representation of the dose distributions of the three-dimensional network matrix elements, etc., so that the dose in a certain region of interest, e.g., the target region, how much volume of an important organ is irradiated at what dose level can be calculated and represented. Plan evaluation based on dose-volume histogram (DVH) is an important tool for daily patient QA, and can quantitatively evaluate dose deviation generated after plan execution, and is more clinically relevant in dose verification.

However, two important assessment indicators for dose validation are currently available: both the gamma index and the dose volume histogram, DVH, are significantly deficient. The calculation of the gamma index is based on pixel points, and regional image segmentation of tumor target areas, surrounding organs at risk and the like is not considered, so that the gamma index cannot be associated with the tumor control rate and normal tissue complication probability of a tumor patient after radiotherapy to guide clinical work. Three-dimensional dose verification based on patient anatomical structures is an important development trend of radiotherapy plan quality assurance at present, independent three-dimensional dose algorithms carried by quality control software are utilized, or original dose distribution is disturbed to reconstruct three-dimensional dose distribution, dose-volume information corresponding to various anatomical structures is provided through DVH, physical teachers can be helped to find and know error-prone problems possibly existing in the radiotherapy process more clearly, but at present, no unified judgment standard exists in clinic, evaluation is mainly carried out on DVH matrix errors of calculated and actually executed doses by doctors and physical teachers, related data size is large and tedious, and evaluation efficiency is low.

The gamma index high pass rate commonly used for dose verification of the current accurate radiotherapy plan cannot guarantee that the calculated dose of a patient is accurately executed, and the gamma index cannot reflect the dose-volume relation between a clinical target area and surrounding organs at risk and possible clinical effects. Patient-based DVH dose verification can improve the correlation between the verification results and the accuracy of actual clinical performance, but currently, there is no unified standard, and the method mainly depends on the subjective judgment of doctors. The quality assessment of different doctors and patients is greatly different due to many factors such as the clinical knowledge, experience, energy, state and the like of doctors.

Disclosure of Invention

The invention mainly aims to provide an accurate dose verification method for tumor patients, which is used for providing the tumor patients with QA (quality assurance) indexes for individually making radiotherapy plans aiming at different disease conditions of the tumor patients. The optimization of the accurate radiotherapy dose verification and evaluation model is realized by extracting the associated plan parameters and the algorithm so as to improve the efficiency of metering verification and the consistency of results, and the individualized QA automatic prediction model is constructed by recursive algorithm optimization so as to realize automatic and accurate QA prediction and have high evaluation efficiency.

In order to achieve the above object, according to one aspect of the present invention, there is provided a method for verifying a precise dose of a tumor patient, comprising

Establishing an information database of a tumor patient;

classifying patient sample data in a tumor patient information database, establishing an accurate radiotherapy dose verification and evaluation model, and obtaining a normalized QA index;

analyzing data of the tumor patient information database through a parameter analysis algorithm to obtain key plan parameters of the tumor patient;

and inputting the key plan parameters of the tumor patient information database samples as input parameters into the patient accurate radiotherapy dose verification and evaluation model, and correcting the patient accurate radiotherapy dose verification and evaluation model through a recursive algorithm to obtain an automatic QA prediction model for accurate radiotherapy plan prediction.

Analyzing data in the tumor patient information database through a parameter analysis algorithm to obtain key plan parameters of the tumor patient, wherein the key plan parameters comprise:

extracting characteristic parameters from the data in the tumor patient information database, and generating association rules by extracting the characteristic parameters;

constructing an association relation network for extracting the characteristic parameters by extracting the characteristic parameters and association rules;

and selecting the characteristic parameters, and judging the influence weight on the dose verification result by calculating the information gain of the selected characteristic parameters to obtain the key plan parameters.

The information gain calculation formula is as follows:

Gain(A,B)＝Entropy(A)-Entropy(A/B)

a is a characteristic parameter, Gain (A, B) is the degree of uncertainty reduction of B (overdue probability) when the characteristic parameter A is known to be constant, and Encopy (A/B) is the conditional Entropy when the characteristic X is fixed.

Classifying patient sample data in a tumor patient information database, and establishing an accurate radiotherapy dose verification and evaluation model, wherein the accurate radiotherapy dose verification and evaluation model comprises the following steps:

acquiring patient sample data in a tumor patient information database, extracting characteristic information in the tumor patient information database from a three-dimensional dose verification error distribution map in the patient sample data, and classifying and storing the extracted characteristic information;

carrying out data preprocessing on the extracted characteristic information and carrying out data normalization processing;

dividing the extracted data in the tumor patient information database into a training set and a verification set, training the radiotherapy dose verification evaluation model by using the training set to obtain the accurate radiotherapy dose verification evaluation model, and verifying through the verification set.

The training set accounts for 80% of the data in the extracted tumor patient information database, and the validation set accounts for 20% of the data in the extracted tumor patient information database.

The recursive algorithm is a decision tree or a multi-layered perceptive neural network.

The information database of a tumor patient includes treatment plan complexity parameters and plan quality parameters.

The parameter analysis algorithm is a rule-related analysis algorithm, and comprises the following steps:

performing feature selection and dimension reduction processing on the treatment plan complexity parameter and the plan quality parameter in the tumor patient information database;

discretizing the screened treatment plan complexity parameters and plan quality parameters, and converting the parameters into transaction data sets;

searching a frequent item set in the transactional data set, calculating support degree, generating an association rule, and setting a minimum support degree and a confidence degree threshold value;

and constructing an incidence relation network of the treatment plan complexity parameter and the plan quality parameter by taking the screened treatment plan complexity parameter and the plan quality parameter as points and the generated incidence rule as a directed edge.

An accurate dosage verification device for tumor patients,

the information database module is used for establishing an information database of the tumor patient;

the accurate radiotherapy dose verification and evaluation model establishing module is used for classifying patient sample data in the tumor patient information database, establishing an accurate radiotherapy dose verification and evaluation model and obtaining a normalized QA index;

the key plan parameter analysis module analyzes the data of the tumor patient information database through a parameter analysis algorithm to obtain key plan parameters of the tumor patient;

and the accurate radiotherapy dose verification and evaluation model optimization module takes the key plan parameters of the tumor patient information database samples as input parameters, inputs the input parameters into the accurate radiotherapy dose verification and evaluation model of the patient, corrects the accurate radiotherapy dose verification and evaluation model of the patient through a recursive optimization algorithm, and obtains an accurate radiotherapy plan QA automatic prediction model.

A oncology patient precision dose verification device, comprising: a processor and a machine-readable storage medium storing machine-executable instructions executable by the processor; the processor is configured to execute machine executable instructions to perform the method of any of the above methods.

According to the tumor patient accurate dose verification method, the optimization of the accurate radiotherapy dose verification evaluation model is realized by extracting the associated plan parameters and the algorithm, so that the efficiency of metering verification and the result consistency are improved, an accurate personalized radiotherapy plan QA index is provided for the tumor patient, the automatic and accurate QA prediction is realized, and the evaluation efficiency is high.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this application, illustrate embodiments of the invention and, together with the description, serve to explain the invention and not to limit the invention. In the drawings:

FIG. 1 is a flow chart of a method of the present invention;

FIG. 2 is a schematic diagram of an enhanced decision tree classification algorithm according to the present invention.

Detailed Description

It should be noted that the embodiments and features of the embodiments in the present application may be combined with each other without conflict. The present invention will be described in detail below with reference to the embodiments with reference to the attached drawings.

It is noted that, unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

Referring to fig. 1, the present invention discloses a method for verifying accurate dose of tumor patients, comprising:

101. and establishing an information database of the tumor patient.

Based on the existing clinical data, such as IMRT/VMAT plan QA of the patient with common clinical tumor such as nasopharyngeal carcinoma and prostatic carcinoma, the information data of the tumor patient is analyzed subjectively and objectively, and the information data is raw data, measurement data and analysis data.

① raw data including patient basic data (age, sex, tumor type, prescription amount), treatment plan, three-dimensional dose distribution, three-dimensional dose difference distribution image, phantom plan, phantom dose, organs at risk structure, contour information of regions of interest (ROIs), DVH matrix, plan parameter matrix, accelerator PDD and Profiles, post-treatment clinical follow-up data, etc., wherein the ROIs include tumor target volume and important organs at risk, such as ROIs of nasopharyngeal carcinoma including GTV, PTV, brainstem, spinal cord, right and left parotid, etc., ROIs of prostate cancer include tumor target volume (GTV), clinical tumor target volume (PTV), normal prostate, bladder, rectum, and femoral head, etc.

② measurement data acquisition of pre-treatment and on-line two-dimensional and three-dimensional data using measurement software tools.

③ analyzing data, analyzing and reconstructing software by using rotating intensity modulated verification motif ArcCHECK, 3DVH, electronic radio-field image system, Edose analysis and reconstruction software, combining with DICOM-CT in radiotherapy planning system (Pininacle, Monaco, etc.), radiotherapy plan (RTplan), radiotherapy dose (RTdose), radiotherapy structure (RTstructure), etc., obtaining dose verification and reconstruction dose distribution diagram, reconstructing DVH matrix, error distribution diagram, etc., verifying data and preprocessing, analyzing target area, 2D/3D gamma passage rate (2%/2 mm, 3%/3 mm, 10% threshold), using MATLAB to obtain planning complexity parameter matrix of planning system (including intensity Modulation Index (MI), conformal intensity modulation complexity score (modular plexisc, MCS), machine jump number (MU) and area per control point, small area (<3), MU area sub-field (2), percentage of conformal intensity modulation complexity score (modular plexisc, MU), and quality index (including maximum learning quality index, minimum learning index, MU coverage index, minimum learning index, and the like), and analyzing the subsequent learning index and the input of the minimum learning index, the MU.

And establishing an information database of the patient based on the three types of data of the tumor patient.

103. And classifying patient sample data in the tumor patient information database, establishing an accurate radiotherapy dose verification and evaluation model, and obtaining a normalized QA index.

The method specifically comprises the following steps:

1. importing raw data and analysis data of a patient from an information database of a tumor patient, wherein the raw data and the analysis data can specifically comprise raw CT data and contour data of an existing case, a calculated three-dimensional dose distribution of a VMAT plan of the case, and a reconstructed three-dimensional dose distribution image acquired by a measuring device, and the raw data and the analysis data are divided into two groups according to labels: pass group (dose error is within clinically acceptable range) and fail group (dose error has exceeded clinically acceptable range);

2. data preprocessing, namely checking data consistency, processing invalid values, missing values and the like, normalizing the included information data, improving the resolution of structures such as organs at risk and the like, facilitating the processing such as segmentation and the like in the next step, and converting image data such as CT (computed tomography) and the like into mask images with the same size by adopting original gray images; for three-dimensional planning dose distribution and reconstruction three-dimensional dose distribution, the value can be interpolated to the size of CT pixels; in addition, for the same type of case, PTV doses were normalized to a uniform prescribed dose;

3. training and verifying an accurate radiotherapy dose verification and evaluation model: the included tumor patient information data is divided into a training set and a verification set, wherein 80% of the training set and 20% of the verification set can be adopted, the training set data and the verification set data are set according to the proportion according to the twenty-eight golden section method, the accurate radiotherapy dose verification and evaluation model can be trained through enough data, and the verification and evaluation model is more accurate due to the enough verification data. For example, 200 prostate cases, 160 cases were selected as the training set, and 40 cases were selected as the verification set, the data in the training set was used for training, and the data in the verification set was used for verifying the weights, preventing overfitting. During training, various loss functions can be adopted, such as a norm of predicted dose and reconstructed three-dimensional dose, or mean square error.

4. And (3) testing a model: and predicting a new patient by adopting the model obtained by the previous training, comparing the model with a dose verification result label evaluated by an expert doctor, and testing the accuracy of the model.

105. And analyzing the data of the tumor patient information database through a parameter analysis algorithm to obtain the key plan parameters of the tumor patient.

The method comprises the steps of comprehensively evaluating the difference between a radiotherapy plan before treatment and dose verification reconstructed three-dimensional dose distribution of each patient by clinicians and physicists, further combing, analyzing and dose verification through related radiotherapy plan complexity parameters and plan quality parameters (DVH matrix), analyzing information database data of tumor patients by using a parameter analysis algorithm of a deep learning technology, wherein the parameter analysis algorithm can be principal component analysis or association rule analysis and the like, analyzing the influence rule of the radiotherapy complexity parameters and the plan quality parameters on QA indexes, manually screening the currently common parameters by experts, screening key influence parameters through a normalized index model trained by a deep neural network, mining the leading key plan parameters behind the parameters, analyzing the association among the key plan parameters, breaking through the limitation of few samples and statistical methods in the past, the method discloses a complex chain among a plurality of key plan parameters and the synergistic influence of the complex chain on the QA index, and quantifies the influence weight and the key influence mechanism of the plan parameters on the QA index change by using an information gain algorithm, so as to seek a new breakthrough of effectively improving the clinical performability of the accurate radiotherapy treatment plan clinically.

According to the previous person and the research results, correlation research is carried out on IMRT/VMAT treatment plan complexity parameters and plan quality parameters collected in a tumor patient information database, such as MI, MCS, MU and area per control point, percentage of small MU (<3), small area subfield (<50mm2) and the like, target area coverage rate V95 of PTV, maximum and minimum dose points, and each DVH matrix parameter corresponding to each different OAR and the like.

Extracting characteristic parameters from data in a tumor patient information database, and generating association rules through the extracted characteristic parameters;

constructing an association relation network of the extracted feature parameters through the extracted feature parameters and association rules;

selecting the characteristic parameters, and judging the influence weight on the dose verification result by calculating the information gain of the selected characteristic parameters to obtain the key plan parameters;

and dividing the influence weight of the judged dose verification result by using a dose verification result obtained by an accurate radiotherapy dose verification evaluation model.

For example, an association rule analysis technology is adopted, the association relationship among the plurality of key plan parameters is mined, and an association relationship network is constructed based on the association relationship, and the method specifically comprises the following steps:

①, performing feature selection (algorithm such as LASSO linear model) and dimension reduction (PCA \ LDA algorithm) on the planning parameters in the data, and removing a plurality of redundant or irrelevant features (parameters);

② discretizing the screened plan parameters and converting the parameters into transaction data sets;

③ searching the frequent item set in the data set and calculating the support degree thereof, further generating association rules, setting the minimum support degree and confidence threshold value, and screening out the association rules with clinical guidance significance;

④ the key plan parameters are points, the association rules among the key plan parameters are directed edges, and the association relationship network of the key plan parameters is constructed.

Information gain is used as an indicator for feature selection. Information gain is one of the important indicators for feature selection, and represents how much information a feature can bring to a classification system: the more information is brought (i.e. the greater the information gain), the more important the feature is, and the calculation formula is:

Gain(A,B)＝Entropy(A)-Entropy(A/B) (2)

wherein A is a characteristic parameter, Gain (A, B) is the reduction degree of uncertainty of B (overdue probability) under the condition that the information Gain is certain, and Encopy (A/B) is the conditional Entropy when the characteristic X is fixed.

The first formula, Encopy (A), is to calculate the information entropy of the data set A, pi represents the probability of the ith class, and n represents the number of classes in the sample set. The second formula calculates the change of entropy (namely information gain) of the data set after division by using the characteristics, the dose verification result obtained by the accurate radiotherapy dose verification and evaluation model is used as the division characteristics, if the change of the plan parameter value changes to cause the drastic change of the dose verification result, the information gain corresponding to the characteristics (parameters) is larger, otherwise, the information gain is smaller.

107. And (3) inputting the key plan parameters of the tumor patient information database samples as input parameters into the patient accurate radiotherapy dose verification and evaluation model to improve the efficiency of metering verification and the consistency of results, and correcting the patient accurate radiotherapy dose verification and evaluation model through a recursive algorithm to obtain an accurate radiotherapy plan prediction QA index.

By machine learning recursive algorithm (decision tree, multilayer perception neural network, etc.), referring to fig. 2, the recursive algorithm of the decision tree is adopted to calculate the information gain rate attribute by attribute, a large amount of treatment schemes with quality control of dose are taken as training samples, the basic data of a patient, the determined QA index (model output) and the determined key plan parameters (model input) are included, and a QA prediction model of a precise radiotherapy plan of a tumor patient such as nasopharyngeal carcinoma and prostate cancer is constructed to analyze the passing rate of the radiotherapy plan of a new case. On the basis, new clinical cases are continuously added into the model as test samples, so that a classifier model is strengthened, the optimization model is continuously updated in an iterative manner, QA results obtained by the prediction model are input into intensity-modulated radiotherapy dose verification software, DVH indexes of all organs and target areas are analyzed, each DVH index and prediction result are compared with expert decisions in a statistical analysis mode, the prediction accuracy is improved through the optimization model, and finally the accurate radiotherapy plan personalized QA prediction model is constructed.

The method has the advantages that the problem of correlation between the tumor control rate and the probability of normal tissue complications after radiotherapy of a tumor patient by using the current gamma index and the dose volume histogram is solved, the calculated dose of the patient cannot be accurately executed by using the gamma index high pass rate commonly used for dose verification of the current accurate radiotherapy plan, and the gamma index cannot reflect the dose-volume relation between a clinical target area and surrounding organs at risk and possible clinical effects. Meanwhile, the problems that the conventional QA depends on doctors and physicists to manually evaluate the DVH matrix errors of calculated and actually executed doses, the data volume is large and tedious, the evaluation efficiency is low and the like are solved, and the quality evaluation of different doctors and different patients is greatly different according to the clinical knowledge, experience, energy, state and other factors of the doctors. A brand-new objective accurate dose verification and evaluation model is developed on the basis of patient DVH dose verification in the past by utilizing artificial intelligence technologies such as deep learning, a support vector machine, association rule analysis, information gain, decision trees and the like and combining large clinical data of nasopharyngeal carcinoma and prostatic carcinoma IMRT/VMAT plan QA.

The invention also comprises a device applying the method, which comprises the following steps:

the information database module is used for establishing an information database of the tumor patient;

the accurate radiotherapy dose verification and evaluation model establishing module is used for classifying patient sample data in the tumor patient information database, establishing an accurate radiotherapy dose verification and evaluation model and obtaining a normalized QA index;

the key plan parameter analysis module analyzes the data of the tumor patient information database through a parameter analysis algorithm to obtain key plan parameters of the tumor patient;

and the accurate radiotherapy dose verification and evaluation model optimization module takes the key plan parameters of the tumor patient information database samples as input parameters, inputs the input parameters into the accurate radiotherapy dose verification and evaluation model of the patient, corrects the accurate radiotherapy dose verification and evaluation model of the patient through a recursive optimization algorithm, and obtains an accurate radiotherapy plan QA automatic prediction model.

Still include accurate dose verification equipment of tumour patient, include: a processor and a machine-readable storage medium storing machine-executable instructions executable by the processor; the processor is configured to execute machine executable instructions to perform the method of any of the above methods. The technical content not disclosed by the invention adopts the known technology in the field.

It is to be understood that the above-described embodiments are only a few, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular is intended to include the plural unless the context clearly dictates otherwise, and it should be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of features, steps, operations, devices, components, and/or combinations thereof.

It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.