CN117717719A

CN117717719A - Method for calculating target region DVH in particle radiation dose model

Info

Publication number: CN117717719A
Application number: CN202311150486.2A
Authority: CN
Inventors: 姜冠群; 赵毅
Original assignee: Shandong Zhuoye Medical Technology Co ltd
Current assignee: Shandong Zhuoye Medical Technology Co ltd
Priority date: 2023-09-07
Filing date: 2023-09-07
Publication date: 2024-03-19

Abstract

The invention relates to the technical field of medical clinic, in particular to a method for calculating a target region DVH in a particle radiation dose model. The method comprises the following steps: performing coordinate mapping and coordinate evaluation screening on the historical dose contour data and the historical target region contour data to generate contour feature coordinate data; performing validity screening and analysis processing on the CT image area position data to generate screening area position data; coordinate integration is carried out on the vertical axis coordinates of the screening area position data serving as outline characteristic coordinate data, and pre-calibration space coordinate data are generated; error screening is carried out on the pre-calibrated space coordinate data, and space coordinate data are generated; establishing and optimizing a three-dimensional point cloud model of the dose-target area according to the space coordinate data, and generating an optimized dose-target area model; and (5) performing dose histogram numerical calculation of the target region on the optimized dose-target region model, and generating target region DVH data. The method and the device for calculating the target region DVH are accurate and uniform.

Description

Method for calculating target region DVH in particle radiation dose model

Technical Field

The invention relates to the technical field of medical clinic, in particular to a method for calculating a target region DVH in a particle radiation dose model.

Background

In radiation therapy, it is important to calculate and evaluate the distribution of the dose over the target area. In calculating the three-dimensional radiation dose of the radioactive particles according to AAPM report No. 43 (american society of medical and physical), computer-aided dose calculation software is required. These software calculate the absorbed dose using a more complex mathematical model from AAPM report number 43, calculate the three-dimensional dose distribution using dose calculation software, which will calculate dose values throughout the region of interest (e.g., tumor surrounding and normal tissue), and generate a dose-volume histogram (DVH) as well as a dose profile. However, the conventional calculation of the target volume and the ratio between the dose volumes is required when calculating the DVH of the target volume, wherein the accuracy of the calculation depends on manually set parameters such as voxel size, iteration number, etc., and different manually set parameters may affect the result of the dose calculation, so that the target DHV calculation method is not uniform.

Disclosure of Invention

Based on this, the present invention provides a method for calculating a target DVH in a particle radiation dose model, solving at least one of the above-mentioned technical problems.

To achieve the above object, a method of calculating a target volume DVH in a particle radiation dose model, comprising the steps of: step S1: acquiring historical dose profile data and historical target profile data in a particle radiation dose model; performing spatial mapping processing on the historical dose profile data and the historical target profile data by using a deep convolutional neural network algorithm to generate profile feature spatial mapping data; carrying out coordinate mapping and coordinate evaluation screening on the contour feature space mapping data to generate contour feature coordinate data; step S2: acquiring CT image region position data of a cloud; performing validity screening and analysis processing on the CT image region position data to generate analyzed screening region position data; step S3: coordinate integration processing is carried out on the screening area position data serving as the vertical axis coordinates of the outline characteristic coordinate data so as to generate pre-calibrated space coordinate data; error screening is carried out on the pre-calibrated space coordinate data, and space coordinate data are generated; step S4: establishing and optimizing a three-dimensional point cloud model of the dose-target area according to the space coordinate data, and generating an optimized dose-target area model; step S5: performing model triangulation on the optimized dose-target region model to generate a triangulation model; carrying out triangulation volume parameter extraction and optimization on the triangulation model to generate optimized dose-target area subdivision volume parameters; and (3) performing dose histogram numerical calculation of the target region on the optimized dose-target region subdivision volume parameter, and generating target region DVH data.

According to the invention, the historical dose profile data and the historical target profile data are subjected to spatial mapping processing through the deep convolutional neural network algorithm, so that the original data can be converted into profile characteristic spatial mapping data, profile characteristic information can be extracted, and the spatial relationship between the dose and the target is captured. The contour feature space mapping data is subjected to coordinate mapping, the contour feature space mapping data is converted into contour feature coordinate data, the space position information of the contour is converted into specific coordinate values, and the quality evaluation can be carried out on the generated coordinate data through coordinate evaluation screening, so that low-quality or invalid data are eliminated, and the accuracy of subsequent calculation is improved. By acquiring the position data of the CT image area of the cloud, the image data related to dose calculation and target area analysis can be acquired, and the acquired position data of the CT image area is subjected to effectiveness screening and analysis processing to generate analyzed position data of the screening area, incomplete or inaccurate data are eliminated, and the reliability of subsequent processing is improved to obtain more accurate space coordinate data. And carrying out coordinate integration processing on the screening area position data serving as the vertical axis coordinates of the contour feature coordinate data, and integrating the screening area position data with the contour feature coordinate data to obtain pre-calibrated space coordinate data, so that the consistency of the space coordinate data and the actual CT image data can be ensured, and the calculation accuracy is improved. The pre-calibrated space coordinate data is subjected to error screening, the quality and the accuracy of the space coordinate data can be improved by screening and removing data points with larger errors, errors introduced in the data acquisition or processing process are removed, the accurate space coordinate data can better describe the position and morphological characteristics of a target area, and the accuracy and the reliability of a calculation result are improved. And establishing a three-dimensional point cloud model of the dose-target area by using the space coordinate data, wherein the point cloud model is based on the space coordinate, and correlating the dose value with the corresponding target area position to form a three-dimensional point cloud representation. By optimizing the point cloud model, the accuracy and precision of the dose-target model are improved, the generated dose-target model can provide more accurate dose distribution information, and the model can be used for further dose evaluation, dose optimization and treatment planning. By triangulating the optimized dose-target model, a triangulated model can be generated, which is a process of expressing the dose-target model as a set of connected triangles, so that the surface of the model is smoother and more continuous, and such a triangulated model can better describe the morphology and geometry of the dose-target model. On the basis of the triangulation model, triangulation volume parameters can be extracted, wherein the parameters comprise the area, the volume and the geometric characteristics of each triangle, are used for describing the volume distribution and the morphological characteristics of the dose-target model, and the representation and the calculation accuracy of the dose-target model can be improved by optimizing the triangulation volume parameters. With optimized dose-target subdivision volume parameters, a dose histogram (DVH) value calculation of the target may be performed, where DVH is used to describe the volume distribution of the target at different dose levels, and by calculating the volume percentages of the target at different dose levels, detailed information about the dose distribution may be provided, and indicators such as dose coverage and dose uniformity may be assessed. Therefore, the method for calculating the target volume DVH of the invention maps the dose in the particle radiation dose model with the target volume, considers the target area of the CT image of the patient in the coordinates, the calculation accuracy does not depend on manually set parameters such as voxel size, iteration times and the like, the result of dose calculation cannot be influenced by different manually set parameters, and the calculated target volume DVH is accurate and uniform.

Preferably, step S1 comprises the steps of: step S11: acquiring historical dose profile data and historical target profile data in a particle radiation dose model; step S12: performing time series association processing on the historical dose profile data and the historical target profile data to generate historical profile association data; step S13: performing contour feature extraction and spatial mapping on the historical contour associated data by using a deep convolutional neural network algorithm to generate contour feature spatial mapping data; step S14: carrying out coordinate mapping on the contour feature space mapping data according to preset coordinate planning information to generate initial contour feature coordinate data; step S15: and carrying out coordinate clustering evaluation screening on the initial contour feature coordinate data by using a K-means algorithm so as to obtain the contour feature coordinate data.

The present invention collects historical dose profile data and historical target profile data in a particle radiation dose model, the historical dose profile data being dose distribution data of radiation treatment plans previously performed for the target, the historical target profile data being data describing target shape and position. And performing time series association processing on the historical dose profile data and the historical target profile data, and matching the dose profile data at the same time point with the target profile data to ensure that the dose profile data and the target profile data come from the same treatment plan or time point, wherein the time series association processing can establish a corresponding relation between the dose and the target, so that an accurate data basis is provided for subsequent analysis and calculation. The historical contour associated data is processed by using a deep convolutional neural network algorithm, the algorithm can extract contour characteristic information from the data, and the characteristic information can be subjected to space mapping, namely, two-dimensional or three-dimensional contour data are mapped into a specific space, so that the generated contour characteristic space mapping data can better represent the relation between the dose and the target area. According to preset coordinate planning information, mapping the outline feature space mapping data into another coordinate space from an original space, so that the data can be represented according to planned coordinates, and more consistent coordinate references can be provided for subsequent data processing and analysis through coordinate mapping. The K-means algorithm is used for carrying out clustering evaluation screening on the initial contour feature coordinate data, the K-means algorithm can divide data points into different clustering clusters, each cluster represents a group of similar data points, and contour feature coordinate data with representative and similar features can be selected through the clustering evaluation screening, so that the accuracy and the efficiency of subsequent analysis and calculation are improved.

Preferably, step S13 comprises the steps of: step S131: carrying out data division on the historical contour association data on a time sequence to respectively generate a contour association training set and a contour association test set; step S132: establishing feature mapping relations between dose contours and target contours of different layers based on a depth convolution neural network algorithm to generate an initial contour feature model; step S133: performing model training treatment on the initial contour feature model by using a contour association training set to generate a contour feature model; step S134: carrying out model contour feature learning optimization on the contour feature model by using a distributed learning strategy to generate an optimized contour feature model; step S135: and transmitting the profile-associated test set to an optimized profile feature model for profile feature extraction, and performing spatial mapping in the optimized profile feature model to generate profile feature spatial mapping data.

According to the invention, the historical contour associated data is divided according to the time sequence, so that the training set and the test set can reasonably train and evaluate the model without overlapping data, and the generalization capability and accuracy of the model can be verified by dividing the training set and the test set. The depth convolution neural network algorithm is used for establishing a characteristic mapping relation between the dose profile and the target area profile, and the characteristics of the dose profile and the target area profile are extracted and mapped through the hierarchical structure of the network, so that the correlation characteristics between the dose and the target area can be captured, and a foundation is provided for subsequent model training and optimization. And (3) carrying out model training treatment on the initial contour feature model by using a contour association training set, and inputting contour association data in the training set into the model, wherein the model can adapt to the feature and association rule of the data through learning and optimization, so that the learning capability of the model on the dose contour and the target contour feature is improved. The contour feature model is subjected to model contour feature learning optimization by using a distributed learning strategy, the distributed learning can distribute calculation tasks to a plurality of calculation units for parallel processing, so that the learning speed and the optimization effect of the model are accelerated, and the learning and optimization capability of the contour feature model on the dose contour and the target contour can be further improved by applying the distributed learning strategy, so that a more accurate and reliable optimized contour feature model is generated. The contour features extracted through the model can capture the association information between the dose and the target region, and can also be used for carrying out space mapping in an optimized contour feature model, so that feature data are mapped into a specific space, the expression and the calculation effect of the feature data are further improved, and the generated contour feature space mapping data can provide a more accurate and reliable data basis for the subsequent steps.

Preferably, step S2 comprises the steps of: step S21: acquiring CT image region position data of a cloud; step S22: performing time sequence extraction processing on the profile feature space mapping data to generate profile time sequence data; step S23: and carrying out effectiveness screening treatment on the CT image region position data according to the contour time sequence data, analyzing the CT image region position data, and generating analyzed screening region position data.

The method for acquiring the position data of the CT image area in the cloud can provide image information required by radiotherapy planning and evaluation, and provides a basis for subsequent data processing and analysis. The profile time series data are generated, the characteristic data of different time points can be corresponding to form a time series data set, so that the change and the association between the dose and the target area at different time points can be captured, and an accurate data basis is provided for subsequent calculation and analysis. And (3) carrying out validity screening and analysis processing on the CT image region position data according to the contour time sequence data, wherein invalid or inaccurate data points can be eliminated, and the reliability of the CT image region position data is ensured. Meanwhile, the original data can be converted into a format with more readability and operability by analyzing, and the generated analyzed screening area position data can be used as input data in a subsequent step, so that an accurate and reliable basis is provided for calculation and analysis.

Preferably, step S3 comprises the steps of: step S31: coordinate integration processing is carried out on the screening area position data serving as the vertical axis coordinates of the outline characteristic coordinate data so as to generate pre-calibrated space coordinate data; step S32: performing error calculation on the pre-calibration space coordinate data by using a radiation dose target space coordinate calibration formula to generate error data of the pre-calibration space coordinate data; step S33: performing threshold comparison processing on error data by using a preset target space coordinate error threshold, and eliminating pre-calibrated space coordinate data corresponding to the error data when the error data is larger than the target space coordinate error threshold; and when the error data is not greater than the target space coordinate error threshold value, marking the pre-calibrated space coordinate data corresponding to the error data as space coordinate data.

According to the method, the screening area position data and the contour feature coordinate data are integrated, and the pre-calibrated space coordinate data can be generated, so that the consistency of the space coordinate data and the position and morphological features of an actual target area can be ensured, and the calculation accuracy is improved. By comparing the difference between the pre-calibrated spatial coordinate data and the spatial coordinates of the radiation dose target region, the error of the pre-calibrated spatial coordinate data can be obtained, the deviation between the pre-calibrated spatial coordinate data and the real target region position is evaluated, and accurate error information is provided for subsequent data processing and analysis. Comparing the error data with a threshold value, if the error data is larger than the target space coordinate error threshold value, indicating that the data deviate greatly and possibly have larger errors, and eliminating the errors; if the error data is not greater than the target space coordinate error threshold, then this data is indicated as having a small error and may be marked as valid space coordinate data.

Preferably, the formula for calibrating the spatial coordinates of the target volume of the radiation dose in step S32 is as follows:；

in the method, in the process of the invention,error data represented as pre-calibrated spatial coordinate data,the horizontal axis coordinates expressed as pre-calibrated spatial coordinate data,the radiation dose output power expressed in abscissa axis coordinates,expressed as the rate at which the radiation dose is absorbed by the target volume,represented as vertical axis coordinates of pre-calibrated spatial coordinate data,vertical axis coordinates expressed as pre-calibrated spatial coordinate data,expressed as the initial output power of the radiation dose,expressed as the initial rate at which the radiation dose is absorbed by the target,an outlier represented as error data.

The invention utilizes a radiation dose target space coordinate calibration formula which fully considers the transverse axis coordinate of the pre-calibration space coordinate dataRadiation dose output power on horizontal axis coordinatesRate of radiation dose absorbed by the target areaVertical axis coordinates of pre-calibrated spatial coordinate dataVertical axis coordinates of pre-calibrated spatial coordinate dataInitial output power of radiation doseInitial rate of radiation dose absorbed by target areaAnd interactions between functions to form a functional relationship: that is to say,the error related to the radiation dose output in the pre-calibrated space coordinate data is corrected through the radiation dose output power and the radiation dose initial output power of the horizontal axis coordinate, and the accuracy of the radiation treatment plan can be improved through the calibration of the radiation dose output, so that the accurate dose distribution is ensured; errors related to the target absorption radiation dose in the pre-calibrated spatial coordinate data can be corrected by the rate of the target absorption radiation dose and the initial rate of the target absorption radiation dose, and the target absorption dose can be predicted more accurately by calibrating the target absorption rate, so that the accuracy of treatment planning is improved; the horizontal axis, the vertical axis and the vertical axis of the pre-calibrated spatial coordinate data are calculated, and the spatial coordinate data can be more accurately calibrated by considering the influence of the spatial position, so that the consistency of the dose distribution and the target area position is improved. By calculating the error data, accuracy information of the pre-calibrated spatial coordinate data can be obtained, helping to assess the quality and accuracy of the radiation treatment plan. Abnormal adjustment value using error data The functional relation is adjusted and corrected, and the error influence caused by abnormal data or error items is reduced, so that error data of the pre-calibrated space coordinate data can be accurately generatedThe accuracy and the reliability of error calculation on the pre-calibrated space coordinate data are improved. Meanwhile, the adjustment value in the formula can be adjusted according to actual conditions and is applied to different pre-calibrated space coordinate data, so that the flexibility and applicability of the algorithm are improved.

Preferably, step S4 comprises the steps of: step S41: carrying out hierarchical spatial coding processing on the spatial coordinate data to generate a spatial coding tree; step S42: establishing a three-dimensional point cloud model of the dose-target area according to the space coding tree to generate a dose-target area model; step S43: and (3) performing model optimization on the dose-target region model by using a GANs technology to generate an optimized dose-target region model.

The hierarchical spatial coding processing can effectively organize and represent the spatial coordinate data, and convert the spatial coordinate data into the spatial coding tree with a hierarchical structure, and the coding mode can provide higher-level abstraction and representation of spatial information, so that the storage and processing of the data are more efficient, the spatial characteristics and the association of a dose-target area can be better captured by generating the spatial coding tree, and a foundation is provided for the establishment and optimization of a subsequent model. And establishing a three-dimensional point cloud model of the dose-target area according to the space coding tree, and mapping the dose and target area data into the three-dimensional point cloud model to more intuitively show the distribution and change condition of the dose in the target area. The model provides a basis for subsequent model optimization and analysis. Model optimization of the dose-target model by using a Generated Antagonism Network (GANs) technology can further improve the accuracy and reliability of the model, and the generated dose-target model can be more in line with the distribution and characteristics of real data by training a generator and a discriminator model, so that the learning ability of the model on the dose distribution and the target characteristics is improved, and the model is better suitable for different radiotherapy scenes and case demands.

Preferably, step S42 comprises the steps of: step S421: extracting space feature vectors from the space coding tree by using a principal component analysis method to generate space feature vectors; step S422: performing dimension reduction mapping processing on the space feature vector to generate a point cloud popularity of the space feature vector; step S423: performing topology optimization on the point cloud popularity by using a topology data analysis method to generate topology point cloud parameters; step S424: and carrying out three-dimensional point cloud model reconstruction processing according to the topological point cloud parameters to generate a dose-target region model.

The invention utilizes the principal component analysis method to extract the space feature vector of the space coding tree, the Principal Component Analysis (PCA) is a common data dimension reduction technology, can be used for extracting the space feature vector of the space coding tree, can identify the most representative space feature through the PCA, and maps the most representative space feature into a feature vector space with lower dimension, thus reducing the dimension of data, removing unimportant features and improving the expression efficiency and the calculation efficiency of the data. Through dimension reduction mapping, similarity relations among the space feature vectors can be reserved and visualized as point cloud popularity, so that the structure and distribution of the data can be understood more intuitively. The topological data analysis method can be used for analyzing and optimizing the topological structure of the point cloud popularity, and by identifying key topological features in the point cloud popularity, such as persistence topological features, point cloud parameters with stable topological structures can be extracted, the topological point cloud parameters can better capture the structure and shape features in the dose-target model, and the accuracy and reliability of the model are improved. And reconstructing the three-dimensional point cloud model according to the topological point cloud parameters to generate a dose-target area model, and reconstructing the three-dimensional point cloud model to more accurately express the dose distribution and the spatial characteristics of the target area, thereby providing a more visual dose-target area model and helping the radiotherapy professional to carry out treatment planning and evaluation.

Preferably, step S5 comprises the steps of: step S51: performing model triangulation on the optimized dose-target region model by using a Delaunay triangulation algorithm to generate a triangulation model; step S52: performing three-dimensional model mesh optimization on the triangulation model to generate a mesh triangulation model; step S53: performing dose and target area triangulation volume parameter extraction processing on the mesh triangulation model to generate dose-target area triangulation volume parameters; step S54: optimizing the dose-target subdivision volume parameter by using a parallel computing technology to generate an optimized dose-target subdivision volume parameter; step S55: and performing target volume dose histogram numerical calculation on the optimized dose-target volume subdivision volume parameter by using a triangulation target volume DVH numerical calculation formula, and generating target volume DVH data.

The invention can divide the optimized dose-target area model into a group of non-overlapping triangles through the Delaunay triangulation algorithm to form the triangulation model, can provide more detailed and accurate expression of the dose-target area model, so that the model is more suitable for subsequent processing and analysis, and the triangulation model can provide more detailed geometric information, thereby facilitating further data analysis and visualization. By adjusting and optimizing the mesh of the triangulation model, the model can be smoother and more continuous, irregularities and sharp edges in the model are reduced, and the visualization effect and analysis accuracy of the model are improved. The mesh triangulation model is processed, and triangulation volume parameters of the dose and the target area can be extracted, wherein the parameters can comprise information such as the dose and the target area volume of each triangle, and the extraction of the triangulation volume parameters can help to quantify the spatial relationship between the dose and the target area and provide a basis for further dose analysis and evaluation. The dose-target area subdivision volume parameters are optimized by applying a parallel computing technology, the parallel computing can improve the processing speed and efficiency, the optimizing process is more efficient, the accuracy and the reliability of data can be further improved by optimizing the dose-target area subdivision volume parameters, and therefore a more reliable basis is provided for radiotherapy planning and evaluation. According to the optimized dose-target subdivision volume parameters, a dose volume histogram of the target can be calculated by using a triangulation target DVH numerical calculation formula, the target DVH data provides a detailed description of the target dose distribution, the coverage, uniformity and various dose parameters of the dose can be measured, and the dose distribution calculation method is very important for evaluating the effect of the radiotherapy plan and the dose distribution of the target, and is helpful for quantifying and comparing the quality and effect of different treatment plans.

Preferably, the triangulation target region DVH numerical calculation formula in step S55 is as follows:

the method comprises the steps of carrying out a first treatment on the surface of the In the method, in the process of the invention,represented as target volume DVH data,expressed as target doseThe time coefficient of the histogram is calculated,the number of dose-related triangles represented as mesh triangulation models,represented as a given dose threshold value,denoted as the firstThe area of the triangle with respect to each dose,represented as the intersection of the dose triangle with the target triangle,the number of target-related triangles represented as mesh triangulation models,expressed as the maximum dose in the target volume,denoted as the firstThe triangular area associated with each target region,an abnormal adjustment value expressed as target DVH data.

The invention utilizes a triangulation target region DVH numerical calculation formula which fully considers the time coefficient of the target region dose histogramExpressed as the number of dose-dependent triangles of the mesh triangulation modelIt is indicated that the number of the elements is,given dose thresholdExpressed as the firstTriangular area of individual dose dependenceExpressed as the intersection of a dose triangle with a target triangleExpressed as the number of target-related triangles of the mesh triangulation modelExpressed as the maximum dose in the target area Expressed as the firstTriangular area associated with each target regionExpressed as and the interaction relationship between the functions to form a functional relationship: that is to say,analysis of target DVH values based on patient specific conditions may be performed by a given dose threshold, maximum dose in target, thDose-dependent triangular area and thThe triangular area related to each target is used for quantifying the influence of different characteristics of the dose and the target on the dose distribution, controlling and limiting the sensitivity and the contribution degree of the dose so as to meet the accuracy and safety requirements of treatment, and determining the treatment plan design can furthestImproving the effect of radiation treatment, thereby calculating the preliminary relation between the target area and the radiation dose; considering the intersection of the dose triangle with the target triangle enables accurate calculation of the dose contribution within the target, which helps to distinguish the effect of the dose within the target from outside the target, providing more accurate dose analysis and assessment results; the time coefficient of the target region dose histogram is used for weighting treatment, the dose distribution can change along with the time in radiation treatment, and the relation between the dose and the time can be considered by introducing the time coefficient, so that the description and analysis capability of the dose distribution are further enhanced. By means of the functional relation, the dose distribution of the target area can be described from a quantitative angle, the relation between the dose and the time and the overlapping area of the dose and the target area are considered, meanwhile, the influence of different characteristics of the dose and the target area on the dose distribution is quantified, a more comprehensive, accurate and personalized dose analysis result is provided, and more powerful support is provided for the decision and optimization of radiotherapy. Abnormal adjustment value using target region DVH data The functional relation is adjusted and corrected, and the error influence caused by abnormal data or error items is reduced, so that the target region DVH data is generated more accuratelyThe accuracy and the reliability of the calculation of the dose histogram value of the target region for the optimized dose-target region subdivision volume parameter are improved. Meanwhile, the adjustment value in the formula can be adjusted according to actual conditions and is applied to different optimized dose-target area subdivision volume parameters, so that the flexibility and applicability of the algorithm are improved.

The method has the beneficial effects that the method carries out space mapping processing on the historical dose profile data and the historical target profile data through the deep convolutional neural network algorithm, and converts the historical dose profile data and the historical target profile data into profile feature space mapping data, so that the processing can capture the space relation between the dose and the target more accurately, thereby improving the accuracy and the reliability of a radiation dose model, and effectively integrating and characterizing the dose information and the target information in the historical data. And acquiring position data of the CT image area of the cloud, screening and analyzing the position data to generate analyzed position data of the screening area, wherein the position data provides a basis for subsequent space coordinate processing, so that accurate position information of the target area is ensured to be accurate, and the accurate position information of the target area is vital to the accuracy of dose analysis and treatment planning. And carrying out coordinate integration processing on the position data of the screening area serving as the vertical axis coordinates of the outline characteristic coordinate data to generate pre-calibrated space coordinate data, carrying out error calculation on the pre-calibrated space coordinate data by utilizing a radiation dose target space coordinate calibration formula to generate error data of the pre-calibrated space coordinate data, screening the error data by setting a target space coordinate error threshold value, and eliminating the pre-calibrated space coordinate data with the error exceeding the threshold value, wherein the accuracy and the reliability of the space coordinate data can be ensured in the pre-calibration and screening process. Through carrying out error calculation and screening on the pre-calibrated space coordinate data, the data with larger errors can be eliminated, accurate space coordinate data are generated, personalized evaluation can be carried out on the dose distribution according to a preset target space coordinate error threshold value, and the specific situation of a patient can be better known through the personalized evaluation, so that a treatment plan suitable for the individual needs of the patient is formulated. By means of hierarchical spatial coding and three-dimensional model reconstruction, an accurate dose-target model can be generated, which can reflect the relationship between the dose and the target, and provide a more accurate dose analysis and assessment tool, and the accurate dose-target model is helpful for knowing the effect and possible side effects of the treatment plan, and further guiding the formulation and optimization of the treatment plan. The dose-target volume model is triangulated and optimized, and dose-target volume parameters can be extracted and further used to calculate dose histogram values for the target volume, generating accurate target volume DVH data that provides an accurate description of the extent to which the target volume is affected at different dose levels, which is important for assessing the effectiveness, dose coverage and uniformity of the treatment plan, helping the physician to make accurate clinical decisions and optimization of the treatment plan.

FIG. 1 is a flow chart illustrating the steps of a method for calculating a target volume DVH in a particle radiation dose model according to the present invention; FIG. 2 is a detailed flowchart illustrating the implementation of step S1 in FIG. 1; FIG. 3 is a flowchart illustrating the detailed implementation of step S4 in FIG. 1; FIG. 4 is a flowchart illustrating the detailed implementation of step S42 in FIG. 3; FIG. 5 is a flowchart illustrating the detailed implementation of step S5 in FIG. 1; the achievement of the objects, functional features and advantages of the present invention will be further described with reference to the accompanying drawings, in conjunction with the embodiments.

Detailed Description

The following is a clear and complete description of the technical method of the present patent in conjunction with the accompanying drawings, and it is evident that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, are intended to fall within the scope of the present invention.

Furthermore, the drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus a repetitive description thereof will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. The functional entities may be implemented in software or in one or more hardware modules or integrated circuits or in different networks and/or processor methods and/or microcontroller methods.

It will be understood that, although the terms "first," "second," etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, a first element could be termed a second element, and, similarly, a second element could be termed a first element, without departing from the scope of example embodiments. The term "and/or" as used herein includes any and all combinations of one or more of the associated listed items.

To achieve the above object, referring to fig. 1 to 5, the present invention provides a method for calculating a target DVH in a particle radiation dose model, comprising the steps of: step S1: acquiring historical dose profile data and historical target profile data in a particle radiation dose model; performing spatial mapping processing on the historical dose profile data and the historical target profile data by using a deep convolutional neural network algorithm to generate profile feature spatial mapping data; carrying out coordinate mapping and coordinate evaluation screening on the contour feature space mapping data to generate contour feature coordinate data; step S2: acquiring CT image region position data of a cloud; performing validity screening and analysis processing on the CT image region position data to generate analyzed screening region position data; step S3: coordinate integration processing is carried out on the screening area position data serving as the vertical axis coordinates of the outline characteristic coordinate data so as to generate pre-calibrated space coordinate data; error screening is carried out on the pre-calibrated space coordinate data, and space coordinate data are generated; step S4: establishing and optimizing a three-dimensional point cloud model of the dose-target area according to the space coordinate data, and generating an optimized dose-target area model; step S5: performing model triangulation on the optimized dose-target region model to generate a triangulation model; carrying out triangulation volume parameter extraction and optimization on the triangulation model to generate optimized dose-target area subdivision volume parameters; and (3) performing dose histogram numerical calculation of the target region on the optimized dose-target region subdivision volume parameter, and generating target region DVH data.

According to the invention, the historical dose profile data and the historical target profile data are subjected to spatial mapping processing through the deep convolutional neural network algorithm, so that the original data can be converted into profile characteristic spatial mapping data, profile characteristic information can be extracted, and the spatial relationship between the dose and the target is captured. The contour feature space mapping data is subjected to coordinate mapping, the contour feature space mapping data is converted into contour feature coordinate data, the space position information of the contour is converted into specific coordinate values, and the quality evaluation can be carried out on the generated coordinate data through coordinate evaluation screening, so that low-quality or invalid data are eliminated, and the accuracy of subsequent calculation is improved. By acquiring the position data of the CT image area of the cloud, the image data related to dose calculation and target area analysis can be acquired, and the acquired position data of the CT image area is subjected to effectiveness screening and analysis processing to generate analyzed position data of the screening area, incomplete or inaccurate data are eliminated, and the reliability of subsequent processing is improved to obtain more accurate space coordinate data. And carrying out coordinate integration processing on the screening area position data serving as the vertical axis coordinates of the contour feature coordinate data, and integrating the screening area position data with the contour feature coordinate data to obtain pre-calibrated space coordinate data, so that the consistency of the space coordinate data and the actual CT image data can be ensured, and the calculation accuracy is improved. The pre-calibrated space coordinate data is subjected to error screening, the quality and the accuracy of the space coordinate data can be improved by screening and removing data points with larger errors, errors introduced in the data acquisition or processing process are removed, the accurate space coordinate data can better describe the position and morphological characteristics of a target area, and the accuracy and the reliability of a calculation result are improved. And establishing a three-dimensional point cloud model of the dose-target area by using the space coordinate data, wherein the point cloud model is based on the space coordinate, and correlating the dose value with the corresponding target area position to form a three-dimensional point cloud representation. By optimizing the point cloud model, the accuracy and precision of the dose-target model are improved, the generated dose-target model can provide more accurate dose distribution information, and the model can be used for further dose evaluation, dose optimization and treatment planning. By triangulating the optimized dose-target model, a triangulated model can be generated, which is a process of expressing the dose-target model as a set of connected triangles, so that the surface of the model is smoother and more continuous, and such a triangulated model can better describe the morphology and geometry of the dose-target model. On the basis of the triangulation model, triangulation volume parameters can be extracted, wherein the parameters comprise the area, the volume and the geometric characteristics of each triangle, are used for describing the volume distribution and the morphological characteristics of the dose-target model, and the representation and the calculation accuracy of the dose-target model can be improved by optimizing the triangulation volume parameters. With optimized dose-target subdivision volume parameters, a dose histogram (DVH) value calculation of the target may be performed, where DVH is used to describe the volume distribution of the target at different dose levels, and by calculating the volume percentages of the target at different dose levels, detailed information about the dose distribution may be provided, and indicators such as dose coverage and dose uniformity may be assessed. Therefore, the method for calculating the target volume DVH of the invention maps the dose in the particle radiation dose model with the target volume, and considers the target area of the CT image of the patient in the coordinates, thereby eliminating the physical characteristic difference of the traditional radiation therapy, enabling the DVH calculation method to be applied to the particle radiation dose model, and ensuring that the calculated target volume DVH is accurate and uniform.

In an embodiment of the present invention, as described with reference to fig. 1, which is a schematic flow chart of steps of a method for calculating a target DVH in a particle radiation dose model according to the present invention, in this embodiment, the method for calculating the target DVH in the particle radiation dose model includes the following steps: step S1: acquiring historical dose profile data and historical target profile data in a particle radiation dose model; performing spatial mapping processing on the historical dose profile data and the historical target profile data by using a deep convolutional neural network algorithm to generate profile feature spatial mapping data; carrying out coordinate mapping and coordinate evaluation screening on the contour feature space mapping data to generate contour feature coordinate data;

in an embodiment of the present invention, historical dose profile data and historical target profile data in a particle radiation dose model are obtained, where the historical dose profile data refers to dose distribution data of a radiation treatment plan previously performed on the target, and the historical target profile data refers to data describing the shape and position of the target. The deep convolutional neural network algorithm is implemented using a deep learning framework TensorFlow. The method comprises the steps of loading historical dose contour data and historical target contour data as input, defining a network structure, and carrying out spatial mapping on the data based on the characteristic of a convolution layer of a convolution network to obtain contour feature spatial mapping data after spatial mapping processing. Mapping the contour feature space mapping data to a specified coordinate system using a coordinate mapping technique, such as affine transformation or non-ragid deformation, and performing coordinate evaluation screening on the mapped data to remove possible abnormal or invalid data points, thereby obtaining processed contour feature coordinate data, wherein the contour feature coordinate data comprises an abscissa represented by dose contour data, an ordinate represented by target contour data and other features. Step S2: acquiring CT image region position data of a cloud; performing validity screening and analysis processing on the CT image region position data to generate analyzed screening region position data;

In the embodiment of the invention, the CT image region position data stored in the cloud is acquired by connecting with the cloud, and the data may be stored in a file form, for example, in DICOM format, a DICOM file is read by using a medical image processing library, or CT image data in different formats is processed by using other suitable tools and libraries, so as to obtain usable CT image region position data. And (3) carrying out validity screening on the CT image region position data to remove invalid or incomplete data points possibly existing, wherein the method comprises the steps of checking the integrity of the data, whether the data format meets the requirements and whether necessary key information is contained. The effective CT image region position data is analyzed, and key information required by people, such as position coordinates, voxel sizes and the like of the region, is extracted, and the method involves the steps of analyzing DICOM labels, analyzing image head information, extracting the region of interest by using an image processing algorithm and the like. Step S3: coordinate integration processing is carried out on the screening area position data serving as the vertical axis coordinates of the outline characteristic coordinate data so as to generate pre-calibrated space coordinate data; error screening is carried out on the pre-calibrated space coordinate data, and space coordinate data are generated;

In the embodiment of the invention, the screening area position data is used as the vertical axis coordinates of the contour feature coordinate data, the contour feature coordinate data comprises the horizontal coordinates represented by the dose contour data, the vertical coordinates represented by the target area contour data and other features, and the screening area position data and the feature coordinates are integrated to form new coordinate data. And performing error screening on the generated pre-calibrated spatial coordinate data, wherein the error screening comprises checking the accuracy and consistency of the data and removing abnormal values or wrong data points. Error screening may be performed using statistical analysis, threshold detection, or other suitable screening algorithms to obtain error-screened spatial coordinate data. These spatial coordinate data contain the screening area position data as pre-calibration data for vertical axis coordinates that can be used for subsequent processing and analysis. Step S4: establishing and optimizing a three-dimensional point cloud model of the dose-target area according to the space coordinate data, and generating an optimized dose-target area model;

in the embodiment of the invention, the space coordinate data is used as input to build a three-dimensional point cloud model of the dose-target area, the space coordinate data is converted into a three-dimensional point cloud form, each point represents a space position, and a dose value associated with the position is represented by color or other attributes. Using a three-dimensional visualization library, such as Matplotlib in VTK, paraView, or Python, to create a three-dimensional point cloud model, the library-provided functions and tools may be used to convert the spatial coordinate data into a point cloud and set a corresponding dose value attribute for each point, and then we perform optimization of the dose-target model, including model rendering of the point cloud model, to obtain a more accurate and reliable dose-target model. Step S5: performing model triangulation on the optimized dose-target region model to generate a triangulation model; carrying out triangulation volume parameter extraction and optimization on the triangulation model to generate optimized dose-target area subdivision volume parameters; and (3) performing dose histogram numerical calculation of the target region on the optimized dose-target region subdivision volume parameter, and generating target region DVH data.

In the embodiment of the invention, the optimized dose-target area model is subjected to model triangulation. This divides the model surface into a number of triangles forming a triangulation model, which can be performed using three-dimensional modeling software or a computational geometry library. And carrying out triangulation volume parameter extraction on the triangulation model, wherein the triangulation volume parameter extraction comprises the steps of calculating parameters such as the area, normal vector and the like of each triangle, and extracting the volume and position information of each voxel, wherein the parameters are used for subsequent optimization and calculation. The dose-target volume subdivision volume parameter is optimized, which may involve operations such as adjusting the volume accuracy, subdivision volume triangle meshing, etc., to obtain more accurate and reliable volume parameters. And calculating the dose histogram value of the target region according to the optimized dose-target region subdivision volume parameter, calculating the dose value of each voxel by using a mathematical formula, and distributing the dose value to a corresponding dose interval to generate target region DVH data.

As an example of the present invention, referring to fig. 2, a detailed implementation step flow diagram of step S1 in fig. 1 is shown, where step S1 includes: step S11: acquiring historical dose profile data and historical target profile data in a particle radiation dose model;

in an embodiment of the present invention, historical dose profile data and historical target profile data in a particle radiation dose model are obtained, where the historical dose profile data refers to dose distribution data of a radiation treatment plan previously performed on the target, and the historical target profile data refers to data describing the shape and position of the target. Step S12: performing time series association processing on the historical dose profile data and the historical target profile data to generate historical profile association data;

in the embodiment of the invention, the historical dose profile data and the historical target profile data are matched according to the corresponding time stamp so as to ensure that the dose of each time point or each case corresponds to the target profile data, and the historical profile associated data is generated. Step S13: performing contour feature extraction and spatial mapping on the historical contour associated data by using a deep convolutional neural network algorithm to generate contour feature spatial mapping data;

In the embodiment of the invention, a deep learning framework such as TensorFlow or PyTorch is used for designing and training a convolutional neural network model, history contour associated data is used as input, features are extracted through operations such as convolution and pooling, and spatially mapped contour feature data is output. Step S14: carrying out coordinate mapping on the contour feature space mapping data according to preset coordinate planning information to generate initial contour feature coordinate data;

in the embodiment of the invention, the contour feature space mapping data is mapped to the corresponding coordinate position in the coordinates according to preset coordinate planning information, such as the size of the coordinate scaling after the space mapping, the corresponding relation of the mapped coordinates and the like, so as to form initial contour feature coordinate data. Step S15: and carrying out coordinate clustering evaluation screening on the initial contour feature coordinate data by using a K-means algorithm so as to obtain the contour feature coordinate data.

In the embodiment of the invention, the initial contour feature coordinate data is divided into a plurality of clusters by using a K-means algorithm, each cluster represents a group of coordinate data with similar features, and contour feature coordinate data with better features is screened out by evaluating the aggregation degree of the coordinate data in the clusters and the separation degree among the clusters.

In the embodiment of the invention, the historical contour association data is subjected to data division on a time sequence to generate a contour association training set and a contour association testing set, for example, a group of dose contour data containing different time points and corresponding target contour data are divided into the training set and the testing set, the training set is used for training a model, and the testing set is used for evaluating the performance of the model. The method comprises the steps of establishing characteristic mapping relations between dose contours and target contours of different layers based on a deep convolutional neural network algorithm, generating an initial contour characteristic model, taking the dose contours and the target contours as input through the convolutional neural network, extracting key characteristic representations through a plurality of convolutional layers and pooling layers of the network, capturing the relations between dose and the target, and forming the initial contour characteristic model. Model training is carried out on the initial contour feature model by utilizing a contour association training set to generate a contour feature model, for example, dose contour and target contour data in the training set are used as input, and parameters of the model are adjusted through a back propagation algorithm and an optimization algorithm, so that the feature mapping relation between the dose and the target can be accurately predicted, and the trained contour feature model is obtained. And performing model contour feature learning optimization on the contour feature model by using a distributed learning strategy to generate an optimized contour feature model, for example, distributing the contour feature model to a plurality of computing nodes by using a distributed computing technology, wherein each node is responsible for processing a part of data. The nodes are communicated and cooperated to jointly optimize the contour feature learning process of the model. The distributed learning strategy can accelerate the training speed of the model and improve the accuracy of the model, and an optimized contour feature model is generated.

In the embodiment of the present invention, the CT image region position data stored in the cloud is obtained, and these data may be stored in a file format, for example, in DICOM format, and the DICOM file is read by using a medical image processing library, or CT image data in different formats is processed by using other suitable tools and libraries, so as to obtain the available CT image region position data. Extracting feature representations associated with each point in time from the profile feature space map data may be accomplished by chronologically accessing feature data at different points in time in the dataset, one feature vector or feature matrix for each point in time. And screening the position data of the CT image area according to the contour time sequence data of each time point, only retaining the position of the effective area corresponding to the characteristic data, analyzing the screened position data, and converting the position data into an understandable and processable format, such as coordinate points or area boundary information.

In the embodiment of the invention, the screening area position data is used as the vertical axis coordinates of the contour feature coordinate data, the contour feature coordinate data comprises the horizontal coordinates represented by the dose contour data, the vertical coordinates represented by the target area contour data and other features, and the screening area position data and the feature coordinates are integrated to form new coordinate data. According to a given radiation dose target space coordinate calibration formula, pre-calibration space coordinate data are substituted into the formula, and the error of each coordinate point is calculated, wherein the error data represent the difference between each pre-calibration space coordinate data and the actual target space coordinate. And carrying out threshold comparison processing on the error data by using a preset target space coordinate error threshold. When the error data is larger than the target space coordinate error threshold value, the corresponding pre-calibrated space coordinate data is removed; when the error data is not greater than the target space coordinate error threshold, the corresponding pre-calibrated space coordinate data is marked as space coordinate data, for example, the preset target space coordinate error threshold is 0.1, if the error of a certain pre-calibrated space coordinate exceeds 0.1, the data corresponding to the coordinate is eliminated, otherwise, the coordinate is marked as valid space coordinate data.

Preferably, the formula for calibrating the spatial coordinates of the target volume of the radiation dose in step S32 is as follows:the method comprises the steps of carrying out a first treatment on the surface of the In the method, in the process of the invention,error data represented as pre-calibrated spatial coordinate data,the horizontal axis coordinates expressed as pre-calibrated spatial coordinate data,the radiation dose output power expressed in abscissa axis coordinates,expressed as the rate at which the radiation dose is absorbed by the target volume,represented as vertical axis coordinates of pre-calibrated spatial coordinate data,vertical axis coordinates expressed as pre-calibrated spatial coordinate data,expressed as the initial output power of the radiation dose,expressed as the initial rate at which the radiation dose is absorbed by the target,an outlier represented as error data.

The invention utilizes a radiation dose target space coordinate calibration formula which fully considers the transverse axis coordinate of the pre-calibration space coordinate dataRadiation dose output power on horizontal axis coordinatesRate of radiation dose absorbed by the target areaVertical axis coordinates of pre-calibrated spatial coordinate dataVertical axis coordinates of pre-calibrated spatial coordinate dataInitial output power of radiation doseInitial rate of radiation dose absorbed by target areaAnd interactions between functions to form a functional relationship: that is to say,the error related to the radiation dose output in the pre-calibrated space coordinate data is corrected through the radiation dose output power and the radiation dose initial output power of the horizontal axis coordinate, and the accuracy of the radiation treatment plan can be improved through the calibration of the radiation dose output, so that the accurate dose distribution is ensured; errors related to the target absorption radiation dose in the pre-calibrated spatial coordinate data can be corrected by the rate of the target absorption radiation dose and the initial rate of the target absorption radiation dose, and the target absorption dose can be predicted more accurately by calibrating the target absorption rate, so that the accuracy of treatment planning is improved; the horizontal axis, the vertical axis and the vertical axis of the pre-calibrated spatial coordinate data are calculated, and the spatial coordinate data can be more accurately calibrated by considering the influence of the spatial position, so that the consistency of the dose distribution and the target area position is improved. By calculating the error data, accuracy information of the pre-calibrated spatial coordinate data can be obtained, helping to assess the quality and accuracy of the radiation treatment plan. Abnormal adjustment value using error data The functional relation is adjusted and corrected, and the error influence caused by abnormal data or error items is reduced, therebyMore accurate generation of error data for pre-calibrated spatial coordinate dataThe accuracy and the reliability of error calculation on the pre-calibrated space coordinate data are improved. Meanwhile, the adjustment value in the formula can be adjusted according to actual conditions and is applied to different pre-calibrated space coordinate data, so that the flexibility and applicability of the algorithm are improved.

As an example of the present invention, referring to fig. 3, a detailed implementation step flow diagram of step S4 in fig. 1 is shown, where step S42 includes: step S41: carrying out hierarchical spatial coding processing on the spatial coordinate data to generate a spatial coding tree;

in the embodiment of the invention, the space coordinate data is subjected to layering processing, a space coding tree is constructed according to different hierarchical structures, the space coordinate data can be divided into different areas by using a space division algorithm such as octree or R tree, and the space coding tree of the hierarchical structure is established. Step S42: establishing a three-dimensional point cloud model of the dose-target area according to the space coding tree to generate a dose-target area model;

in the embodiment of the invention, the spatial coordinate data corresponding to each node is obtained by traversing the nodes in the spatial coding tree, and the spatial coordinate points are converted into a three-dimensional point cloud model according to the dose data and the target area data to represent the relation between the dose and the target area. Step S43: and (3) performing model optimization on the dose-target region model by using a GANs technology to generate an optimized dose-target region model.

In the embodiment of the invention, a generator network and a discriminator network are trained by using a generated countermeasure network (GANs), the generator network takes a dose-target model as input to generate an optimized dose-target model, the discriminator network evaluates the difference between the generated model and a real model, the quality of the generated model is gradually improved by repeatedly training and optimizing the dose-target model by using the trained generated countermeasure network, so as to obtain a more accurate and reliable optimized dose-target model.

As an example of the present invention, referring to fig. 4, a detailed implementation step flow diagram of step S42 in fig. 3 is shown, where step S42 includes: step S421: extracting space feature vectors from the space coding tree by using a principal component analysis method to generate space feature vectors;

in the embodiment of the invention, for each node of the spatial coding tree, the center position, the size and the direction of the node bounding box can be extracted as a part of the spatial feature vector, and the spatial feature vector is further enriched by calculating the spatial relationship between the nodes, such as the distance, the angle and the like. Step S422: performing dimension reduction mapping processing on the space feature vector to generate a point cloud popularity of the space feature vector;

in the embodiment of the invention, the space feature vector is extracted through the space coding tree after the principal component analysis method, the data in the space feature vector is mapped and ordered according to a time sequence and other modes to generate a point cloud popularity, wherein each point represents one space feature vector, and the coordinates of the point represent the feature value after dimension reduction. Step S423: performing topology optimization on the point cloud popularity by using a topology data analysis method to generate topology point cloud parameters;

In the embodiment of the invention, the topological structure of the point cloud fashion is analyzed by using a topological data analysis technology such as persistent coherence, space filling curves and the like, and topological point cloud parameters such as topological connectivity, hole number and the like can be extracted by identifying the persistent features and the topological features in the fashion so as to describe the morphological features of the dose-target model. Step S424: and carrying out three-dimensional point cloud model reconstruction processing according to the topological point cloud parameters to generate a dose-target region model.

In the embodiment of the invention, the three-dimensional point cloud model can be reconstructed by utilizing the information of the topological point cloud parameters, and the characteristic point cloud is mapped to a proper position and form to generate a more accurate dose-target area model with topological characteristics for subsequent dose calculation and target area DVH analysis.

As an example of the present invention, referring to fig. 5, a detailed implementation step flow diagram of step S5 in fig. 1 is shown, where step S5 includes: step S51: performing model triangulation on the optimized dose-target region model by using a Delaunay triangulation algorithm to generate a triangulation model;

in the embodiment of the invention, by inputting the point cloud data in the dose-target region model into the Delaunay triangulation algorithm, a group of continuous and non-overlapping triangles can be automatically generated and used for representing the geometric shape of the model, so as to generate the triangulation model. Step S52: performing three-dimensional model mesh optimization on the triangulation model to generate a mesh triangulation model;

in the embodiment of the invention, the triangulation model can be smoother and finer by applying the grid optimization technology such as grid smoothing, surface fineness and the like, so that the geometric quality and the accuracy of the model are improved, and the grid triangulation is generated. Step S53: performing dose and target area triangulation volume parameter extraction processing on the mesh triangulation model to generate dose-target area triangulation volume parameters;

in the embodiment of the invention, the volume, the dose value and the target area attribute of each triangle are calculated, so that the dose-target area subdivision volume parameters, such as the dose distribution, the dose maximum value and the like, can be extracted to describe the dose characteristics of the model and generate the dose-target area subdivision volume parameters. Step S54: optimizing the dose-target subdivision volume parameter by using a parallel computing technology to generate an optimized dose-target subdivision volume parameter;

According to the embodiment of the invention, the calculation and optimization process of the dose-target area subdivision volume parameter can be accelerated through a parallel calculation technology, the calculation efficiency and precision are improved, and the optimized dose-target area subdivision volume parameter is generated. Step S55: and performing target volume dose histogram numerical calculation on the optimized dose-target volume subdivision volume parameter by using a triangulation target volume DVH numerical calculation formula, and generating target volume DVH data.

In the embodiment of the invention, according to the dose-target subdivision volume parameter and the target DVH calculation formula, the volume proportion in each dose interval can be calculated and expressed as the dose histogram of the target so as to reflect the dose distribution condition of the target and generate target DVH data. Preferably, the triangulation target region DVH numerical calculation formula in step S55 is as follows:

；

in the method, in the process of the invention,represented as target volume DVH data,represented as a time coefficient of the target volume dose histogram,the number of dose-related triangles represented as mesh triangulation models,represented as a given dose threshold value,denoted as the firstThe area of the triangle with respect to each dose,represented as the intersection of the dose triangle with the target triangle,the number of target-related triangles represented as mesh triangulation models, Expressed as the maximum dose in the target volume,denoted as the firstThe triangular area associated with each target region,an abnormal adjustment value expressed as target DVH data.

The invention utilizes a triangulation target region DVH numerical calculation formula which fully considers the time coefficient of the target region dose histogramExpressed as the number of dose-dependent triangles of the mesh triangulation modelExpressed as, a given dose thresholdExpressed as the firstTriangular area of individual dose dependenceExpressed as the intersection of a dose triangle with a target triangleExpressed as the number of target-related triangles of the mesh triangulation modelExpressed as the maximum dose in the target areaExpressed as the firstTriangular area associated with each target regionExpressed as and the interaction relationship between the functions to form a functional relationship: that is to say,analysis of target DVH values based on patient specific conditions may be performed by a given dose threshold, maximum dose in target, thDose-dependent triangular area and thThe triangular areas associated with each target are used to quantify the effect of dose and different features of the target on dose distribution, and to control and limit the sensitivity and extent of contribution of the dose to meet the accuracy and safety requirements of the treatment, determine the treatment plan The effect of radiotherapy can be improved to the greatest extent by design, so that the preliminary relation between the target area and the radiation dose is calculated; considering the intersection of the dose triangle with the target triangle enables accurate calculation of the dose contribution within the target, which helps to distinguish the effect of the dose within the target from outside the target, providing more accurate dose analysis and assessment results; the time coefficient of the target region dose histogram is used for weighting treatment, the dose distribution can change along with the time in radiation treatment, and the relation between the dose and the time can be considered by introducing the time coefficient, so that the description and analysis capability of the dose distribution are further enhanced. By means of the functional relation, the dose distribution of the target area can be described from a quantitative angle, the relation between the dose and the time and the overlapping area of the dose and the target area are considered, meanwhile, the influence of different characteristics of the dose and the target area on the dose distribution is quantified, a more comprehensive, accurate and personalized dose analysis result is provided, and more powerful support is provided for the decision and optimization of radiotherapy. Abnormal adjustment value using target region DVH dataThe functional relation is adjusted and corrected, and the error influence caused by abnormal data or error items is reduced, so that the target region DVH data is generated more accurately The accuracy and the reliability of the calculation of the dose histogram value of the target region for the optimized dose-target region subdivision volume parameter are improved. Meanwhile, the adjustment value in the formula can be adjusted according to actual conditions and is applied to different optimized dose-target area subdivision volume parameters, so that the flexibility and applicability of the algorithm are improved.

The present embodiments are, therefore, to be considered in all respects as illustrative and not restrictive, the scope of the invention being indicated by the appended claims rather than by the foregoing description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

The foregoing is only a specific embodiment of the invention to enable those skilled in the art to understand or practice the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method of calculating a target volume DVH in a particle radiation dose model, comprising the steps of:

step S1: acquiring historical dose profile data and historical target profile data in a particle radiation dose model; performing spatial mapping processing on the historical dose profile data and the historical target profile data by using a deep convolutional neural network algorithm to generate profile feature spatial mapping data; carrying out coordinate mapping and coordinate evaluation screening on the contour feature space mapping data to generate contour feature coordinate data;

Step S2: acquiring CT image region position data of a cloud; performing validity screening and analysis processing on the CT image region position data to generate analyzed screening region position data;

step S3: coordinate integration processing is carried out on the screening area position data serving as the vertical axis coordinates of the outline characteristic coordinate data so as to generate pre-calibrated space coordinate data; error screening is carried out on the pre-calibrated space coordinate data, and space coordinate data are generated;

step S4: establishing and optimizing a three-dimensional point cloud model of the dose-target area according to the space coordinate data, and generating an optimized dose-target area model;

step S5: performing model triangulation on the optimized dose-target region model to generate a triangulation model; carrying out triangulation volume parameter extraction and optimization on the triangulation model to generate optimized dose-target area subdivision volume parameters; and (3) performing dose histogram numerical calculation of the target region on the optimized dose-target region subdivision volume parameter, and generating target region DVH data.

2. A method of calculating a target DVH in a particle radiation dose model according to claim 1, wherein step S1 comprises the steps of:

step S11: acquiring historical dose profile data and historical target profile data in a particle radiation dose model;

Step S12: performing time series association processing on the historical dose profile data and the historical target profile data to generate historical profile association data;

step S13: performing contour feature extraction and spatial mapping on the historical contour associated data by using a deep convolutional neural network algorithm to generate contour feature spatial mapping data;

step S14: carrying out coordinate mapping on the contour feature space mapping data according to preset coordinate planning information to generate initial contour feature coordinate data;

step S15: and carrying out coordinate clustering evaluation screening on the initial contour feature coordinate data by using a K-means algorithm so as to obtain the contour feature coordinate data.

3. A method of calculating a target DVH in a particle radiation dose model according to claim 2, wherein step S13 comprises the steps of:

step S131: carrying out data division on the historical contour association data on a time sequence to respectively generate a contour association training set and a contour association test set;

step S132: establishing feature mapping relations between dose contours and target contours of different layers based on a depth convolution neural network algorithm to generate an initial contour feature model;

Step S133: performing model training treatment on the initial contour feature model by using a contour association training set to generate a contour feature model;

step S134: carrying out model contour feature learning optimization on the contour feature model by using a distributed learning strategy to generate an optimized contour feature model;

step S135: and transmitting the profile-associated test set to an optimized profile feature model for profile feature extraction, and performing spatial mapping in the optimized profile feature model to generate profile feature spatial mapping data.

4. A method of calculating a target DVH in a particle radiation dose model according to claim 3, wherein step S2 comprises the steps of:

step S21: acquiring CT image region position data of a cloud;

step S22: performing time sequence extraction processing on the profile feature space mapping data to generate profile time sequence data;

step S23: and carrying out effectiveness screening treatment on the CT image region position data according to the contour time sequence data, analyzing the CT image region position data, and generating analyzed screening region position data.

5. A method of calculating a target DVH in a particle radiation dose model according to claim 4, wherein step S3 comprises the steps of:

Step S31: coordinate integration processing is carried out on the screening area position data serving as the vertical axis coordinates of the outline characteristic coordinate data so as to generate pre-calibrated space coordinate data;

step S32: performing error calculation on the pre-calibration space coordinate data by using a radiation dose target space coordinate calibration formula to generate error data of the pre-calibration space coordinate data;

step S33: performing threshold comparison processing on error data by using a preset target space coordinate error threshold, and eliminating pre-calibrated space coordinate data corresponding to the error data when the error data is larger than the target space coordinate error threshold; and when the error data is not greater than the target space coordinate error threshold value, marking the pre-calibrated space coordinate data corresponding to the error data as space coordinate data.

6. A method of calculating a target DVH in a particle radiation dose model according to claim 5, wherein the radiation dose target spatial coordinates calibration formula in step S32 is as follows:

；

in the method, in the process of the invention,error data, denoted pre-calibrated spatial coordinate data, ">Horizontal axis coordinates expressed as pre-calibrated spatial coordinate data,/->Radiation dose output expressed in abscissa, < > >Expressed as the rate at which the target absorbs radiation dose, < >>Vertical axis coordinates expressed as pre-calibrated spatial coordinate data,/->Vertical axis coordinates expressed as pre-calibrated spatial coordinate data,/->Expressed as initial output power of radiation dose, +.>Expressed as the initial rate at which the target absorbs radiation dose, < >>An outlier represented as error data.

7. A method of calculating a target DVH in a particle radiation dose model according to claim 6, wherein step S4 comprises the steps of:

step S41: carrying out hierarchical spatial coding processing on the spatial coordinate data to generate a spatial coding tree;

step S42: establishing a three-dimensional point cloud model of the dose-target area according to the space coding tree to generate a dose-target area model;

step S43: and (3) performing model optimization on the dose-target region model by using a GANs technology to generate an optimized dose-target region model.

8. A method of calculating a target DVH in a particle radiation dose model according to claim 7, wherein step S42 comprises the steps of:

step S421: extracting space feature vectors from the space coding tree by using a principal component analysis method to generate space feature vectors;

step S422: performing dimension reduction mapping processing on the space feature vector to generate a point cloud popularity of the space feature vector;

Step S423: performing topology optimization on the point cloud popularity by using a topology data analysis method to generate topology point cloud parameters;

step S424: and carrying out three-dimensional point cloud model reconstruction processing according to the topological point cloud parameters to generate a dose-target region model.

9. A method of calculating a target DVH in a particle radiation dose model according to claim 8, wherein step S5 comprises the steps of:

step S51: performing model triangulation on the optimized dose-target region model by using a Delaunay triangulation algorithm to generate a triangulation model;

step S52: performing three-dimensional model mesh optimization on the triangulation model to generate a mesh triangulation model;

step S53: performing dose and target area triangulation volume parameter extraction processing on the mesh triangulation model to generate dose-target area triangulation volume parameters;

step S54: optimizing the dose-target subdivision volume parameter by using a parallel computing technology to generate an optimized dose-target subdivision volume parameter;

step S55: and performing target volume dose histogram numerical calculation on the optimized dose-target volume subdivision volume parameter by using a triangulation target volume DVH numerical calculation formula, and generating target volume DVH data.

10. A method of calculating a target DVH in a particle radiation dose model according to claim 9, wherein the triangulation target DVH value in step S55 is calculated as:

；

in the method, in the process of the invention,expressed as target DVH data, +.>Time coefficient expressed as target dose histogram, < >>Dose-dependent number of triangles expressed as mesh triangulation model,/->Expressed as a given dose threshold,/->Denoted as +.>Triangle area of individual dose dependence ∈>Expressed as the intersection of the dose triangle with the target triangle,/->Target-related number of triangles expressed as mesh triangulation model, +.>Expressed as maximum dose in the target area, < >>Denoted as +.>Triangular area associated with each target region, +.>An abnormal adjustment value expressed as target DVH data.