CN114566251A - Radiation therapy plan evaluation method and device based on unsupervised learning - Google Patents

Radiation therapy plan evaluation method and device based on unsupervised learning Download PDF

Info

Publication number
CN114566251A
CN114566251A CN202210085979.1A CN202210085979A CN114566251A CN 114566251 A CN114566251 A CN 114566251A CN 202210085979 A CN202210085979 A CN 202210085979A CN 114566251 A CN114566251 A CN 114566251A
Authority
CN
China
Prior art keywords
plan
anatomical structure
category
data
algorithm
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210085979.1A
Other languages
Chinese (zh)
Inventor
鞠垚
汪倩倩
姚毅
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Suzhou Linatech Medical Science And Technology
Original Assignee
Suzhou Linatech Medical Science And Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Suzhou Linatech Medical Science And Technology filed Critical Suzhou Linatech Medical Science And Technology
Priority to CN202210085979.1A priority Critical patent/CN114566251A/en
Publication of CN114566251A publication Critical patent/CN114566251A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H20/00ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
    • G16H20/40ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to mechanical, radiation or invasive therapies, e.g. surgery, laser therapy, dialysis or acupuncture
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/23Clustering techniques
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/23Clustering techniques
    • G06F18/232Non-hierarchical techniques
    • G06F18/2321Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0012Biomedical image inspection
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/60Analysis of geometric attributes
    • G06T7/62Analysis of geometric attributes of area, perimeter, diameter or volume
    • GPHYSICS
    • G16INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16HHEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H70/00ICT specially adapted for the handling or processing of medical references
    • G16H70/20ICT specially adapted for the handling or processing of medical references relating to practices or guidelines
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10072Tomographic images
    • G06T2207/10081Computed x-ray tomography [CT]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20092Interactive image processing based on input by user
    • G06T2207/20104Interactive definition of region of interest [ROI]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30004Biomedical image processing
    • G06T2207/30096Tumor; Lesion

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Epidemiology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Public Health (AREA)
  • Primary Health Care (AREA)
  • Probability & Statistics with Applications (AREA)
  • Geometry (AREA)
  • Bioethics (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Surgery (AREA)
  • Urology & Nephrology (AREA)
  • Radiation-Therapy Devices (AREA)

Abstract

The invention discloses a radiotherapy plan evaluation method and a device based on unsupervised learning, wherein the radiotherapy plan evaluation method comprises the following steps: s1: aiming at any disease type, obtaining a corresponding anatomical structure vector according to data of n patients; s2: reducing the dimension of the high-dimensional anatomical structure vector by using a dimension reduction algorithm; s3: clustering the data according to the anatomical structure by using a clustering algorithm to obtain k categories; s4: analyzing and processing the clinically implemented high-quality radiotherapy plan result in the data corresponding to each category according to the k categories to obtain a plan scoring template corresponding to each category; s5: calculating new data of any patient and judging the category of the new data; s6: and adopting a plan scoring template of the corresponding category to guide the design of a radiation treatment plan corresponding to new data and scoring the dose distribution result of the radiation treatment plan. The invention realizes the individual evaluation of the cases with different geometrical structure complexity and improves the quality and efficiency of the radiotherapy plan.

Description

Radiation therapy plan evaluation method and device based on unsupervised learning
Technical Field
The invention belongs to the technical field of radiotherapy, and particularly relates to a radiotherapy plan evaluation method and device based on unsupervised learning.
Background
Tumor radiotherapy has unique advantages as one of the main means of tumor therapy, and the main goal is to protect the surrounding normal tissues as much as possible while ensuring that the target area reaches a specific dose. In clinical applications, radiation therapy delivery requires specialized physicians and physicists to plan the radiation therapy. The Planning of radiotherapy plan can be divided into three steps of Target Volume (PTV) and Organs At Risk (OARs) delineation, plan design and plan evaluation.
How to evaluate the plan effectively and objectively is a question to be discussed.
Dose Volume Histograms (DVH) are important tools for designing, formulating, and evaluating radiotherapy plans because they can reflect the Dose-Volume relationship in three-dimensional radiotherapy plans. In clinical work, the radiotherapist evaluates DVH manually, which takes a long time and is inefficient. In order to simplify the operation, a researcher directly reads or customizes a formula to read a relevant dosimetry statistical index of the region of interest based on script function design software provided by a radiotherapy Planning System (TPS) in the market, compares the relevant dosimetry statistical index with a prescription requirement of a doctor, and judges whether the plan meets clinical requirements or not. Clinical objectives given by a radiotherapy doctor are referred to the existing academic consensus and guideline, and a more qualified authority customizes a modified set of evaluation criteria on the basis of the existing academic consensus and guideline.
Clinical studies have shown that the achievable dose for a patient is closely related to its geometrical anatomical characteristics. The clinical cases are statistically sorted, the association model of the geometrical anatomical structures between organs and the corresponding radiotherapy plan dosimetry information is constructed, and personalized evaluation criteria can be made for the radiotherapy plan of each patient before plan design.
The existing radiotherapy plan evaluation has the following problems:
first, the existing radiotherapy plan evaluation method refers to the existing academic consensus and guideline, and further qualifies the authoritative institution or defines a modified set of evaluation criteria on the basis of the academic consensus and guideline. The method has certain subjectivity and can cause instability of plan quality.
Second, if the clinical plans are subject to uniform normative criteria, it can result in clinically set objectives that are often difficult to achieve for cases with complex geometries, whereas for cases with simple geometries, clinical objectives are very easy to achieve, and if further optimization is stopped at this time, a radiotherapy plan with suboptimal quality may result.
Thirdly, the existing professional software for radiotherapy planning evaluation simulates and calculates dose attenuation according to the geometric relationship between the organs at risk and the target area, so that the method for quantitatively calculating the irradiated dose only considers the physical properties of radiation in human tissues when predicting the target area and the organs at risk, does not consider the performance of a linear accelerator and a grating, the mutual restriction properties of other normal organs at risk and the like, and therefore the predicted target area and the organs at risk dose are the most ideal results on the whole and cannot well meet the actual conditions.
Disclosure of Invention
In order to solve the technical problem, the invention provides a radiation therapy plan evaluation method and device based on unsupervised learning.
In order to achieve the purpose, the technical scheme of the invention is as follows:
in one aspect, the invention discloses a radiation therapy plan evaluation method based on unsupervised learning, which comprises the following steps:
s1: for any disease species, obtaining an OVH curve, position information and volume information of the interested region among the interested regions according to the data of n patients, and respectively forming corresponding anatomical structure vectors;
s2: reducing the dimension of the high-dimensional anatomical structure vector obtained in the step S1 by using a dimension reduction algorithm of unsupervised learning;
s3: clustering the low-dimensional anatomical structure vector obtained by the data set according to S1 and the low-dimensional attribute vector obtained after the data set is processed by S2 by using a clustering algorithm of unsupervised learning to obtain k categories;
s4: analyzing and processing the clinically implemented high-quality radiotherapy plan result in the data corresponding to each category according to the k categories obtained in the step S3 to obtain a plan scoring template corresponding to each category;
s5: calculating new data of any patient and judging the category of the patient;
s6: and adopting a plan scoring template of the corresponding category to guide the design of a radiation treatment plan corresponding to new data and scoring the dose distribution result of the radiation treatment plan.
On the basis of the technical scheme, the following improvements can be made:
preferably, S1 specifically includes the following steps:
s1.1: obtaining region-of-interest delineation information of n patients, and performing interpolation in the layer thickness direction;
s1.2: calculating the mutual OVH curve among the interested regions;
s1.3: calculating position information and volume information of the region of interest according to the CT image, the delineation contour information of each organ and the target area of the human body;
s1.4: and respectively forming corresponding anatomical structure vectors according to the OVH curves among the interested regions obtained in the step S1.2 and the position information and the volume information of the interested regions obtained in the step S1.3.
Preferably, in S1.2, a morphological dilation-erosion method or a direct distance calculation method is used to calculate the mutual OVH curve between the regions of interest.
Preferably, S2 specifically includes the following steps, wherein the high-dimensional anatomical structure vector is an OVH vector:
s2.1: for OVH attribute vectors of n patients, after m-dimensional features of all attributes are extracted, the OVH attribute vectors are arranged into m rows and n columns of matrixes, namely m-dimensional OVH vectors;
s2.2: and reducing the m-dimensional OVH vector to t dimension by using a dimension reduction algorithm of unsupervised learning.
As a preferable scheme, the unsupervised learning dimensionality reduction algorithm is one or more of a principal component analysis algorithm, a singular value decomposition algorithm and a t-distribution field embedding algorithm.
Preferably, S3 specifically includes the following:
and clustering the data set to k categories according to the extracted and processed F characteristic attribute vectors including all the OVH vectors, organ absolute volume vectors and organ relative position vectors after dimensionality reduction by using a clustering algorithm of unsupervised learning.
As a preferable scheme, the clustering algorithm of unsupervised learning is one or more of a K-means clustering algorithm, a hierarchical clustering algorithm, a mean shift algorithm and a density-based clustering algorithm.
Preferably, S4 specifically includes the following: on the basis of clinical evaluation indexes, the dose results of the clinically implemented high-quality radiotherapy plans in the data corresponding to each category are statistically analyzed, hardware information and software information used for obtaining corresponding results are obtained, the dose results corresponding to the score ranges of different index items in a radiotherapy plan score table are calculated according to different levels of plan effects, and a plan score template suitable for the category data is generated.
Preferably, S5 specifically includes the following steps:
s5.1: acquiring an anatomical structure vector of new data;
s5.2: reducing the dimension of the high-dimensional anatomical structure vector;
s5.3: and (5) calculating the similarity between the reduced-dimension vector obtained in the step (5.2) and the low-dimension anatomical structure vector obtained in the step (5.1) and each cluster central point, and judging the category of the new data.
In another aspect, the present invention discloses a radiotherapy plan evaluation device based on unsupervised learning, comprising:
the anatomical structure vector obtaining module is used for obtaining an OVH curve, position information and volume information of each region of interest according to the data of n patients aiming at any disease type and respectively forming corresponding anatomical structure vectors;
the dimension reduction processing module is used for reducing the dimension of the high-dimensional anatomical structure vector obtained by the anatomical structure vector obtaining module by using a dimension reduction algorithm of unsupervised learning;
the clustering processing module is used for clustering the low-dimensional anatomical structure vector obtained by the data set according to the anatomical structure vector obtaining module and the low-dimensional attribute vector obtained after the processing of the dimension reduction processing module by using a clustering algorithm of unsupervised learning to obtain k categories;
the plan scoring template generating module is used for analyzing and processing the clinically implemented high-quality radiotherapy plan results in the data corresponding to each category according to the k categories obtained by the clustering processing module to obtain a plan scoring template corresponding to each category;
the category judgment module is used for calculating the new data of any patient and judging the category of the patient;
and the radiation treatment plan evaluation module is used for adopting the plan grading template of the corresponding category to guide the design of the radiation treatment plan corresponding to the new data and grading the dose distribution result.
According to the radio therapy plan evaluation method and device based on unsupervised learning, according to clinical guidelines and technical reports, combined with high-quality radiotherapy plan results which are clinically implemented, each clustered category data obtained based on the unsupervised learning method obtains a corresponding plan scoring template, so that personalized evaluation of cases with different geometrical structure complexity is realized, and the problem that the difference between individual patients is difficult to meet under the condition that the plan evaluation meets clinical specifications is solved; meanwhile, the restriction of factors such as accelerator and grating performance on the radiotherapy planning effect is considered by combining a scoring table of a high-quality planning dose result which is clinically implemented; meanwhile, the ideal dose parameters corresponding to the radiotherapy plan scoring table are applied to the design of the inverse intensity modulated plan, so that the plan quality difference caused by the abundant degree of clinical experience of a physicist and the consumable time difference is improved, and the quality and the efficiency of the radiotherapy plan are improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings needed to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained according to the drawings without inventive efforts.
FIG. 1 is a flowchart of a method for unsupervised learning-based radiation treatment plan evaluation according to an embodiment of the present invention;
FIG. 2 is a flowchart of an algorithm for calculating an OVH (overlap volume histogram);
FIG. 3 is a flowchart of an algorithm for PCA dimension reduction of an anatomical structure vector according to an embodiment of the present invention;
FIG. 4 illustrates anatomical features selected during clustering of a cervical cancer database according to an embodiment of the present invention;
fig. 5 is a flowchart of an algorithm of K-means clustering in unsupervised learning according to an embodiment of the present invention.
Detailed Description
Preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. 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.
The expression "comprising" an element is an "open" expression which merely means that a corresponding component or step is present and should not be interpreted as excluding additional components or steps.
In order to achieve the object of the present invention, in some embodiments of a radiation therapy planning evaluating method and apparatus based on unsupervised learning, which take cervical cancer as an example, as shown in fig. 1, the radiation therapy planning evaluating method includes the following steps:
s1: obtaining an OVH curve, position information and volume information of the interested region among the interested regions according to data of n patients, and respectively forming corresponding anatomical structure vectors, wherein the data comprises: the CT image, each organ of the human body and the target area draw outline information;
s2: reducing the dimension of the high-dimensional anatomical structure vector obtained in the step S1 by using a dimension reduction algorithm of unsupervised learning;
s3: clustering the low-dimensional anatomical structure vector obtained by the data set according to S1 and the low-dimensional attribute vector obtained after the data set is processed by S2 by using a clustering algorithm of unsupervised learning to obtain k categories;
s4: analyzing and processing the clinically implemented high-quality radiotherapy plan result in the data corresponding to each category according to the k categories obtained in the step S3 to obtain a plan scoring template corresponding to each category;
s5: calculating new data of any patient and judging the category of the patient;
s6: and adopting a plan scoring template of the corresponding category to guide the design of a radiation treatment plan corresponding to new data and scoring the dose distribution result of the radiation treatment plan.
The invention relates to a radiation therapy plan evaluation method based on unsupervised learning, which comprises the steps of firstly obtaining anatomical structure characteristic attributes of a data set to form vectors, reducing the dimension of high-dimensional anatomical structure vectors, then clustering data according to the reduced vectors and original low-dimensional anatomical structure vectors by using a clustering algorithm, and generating different cluster plan scoring templates according to clustering results.
The invention adopts a method for searching most similar data in a database by adopting an anatomical structure vector, wherein the anatomical structure vector comprises but is not limited to Histogram information (OVH) of overlapped Volume of a region of interest, relative position information and Volume information.
Further, on the basis of the above embodiment, S1 specifically includes the following steps:
s1.1: obtaining region-of-interest delineation information of n patients from a cervical cancer database, including a target region, a rectum, a bladder, a small intestine, a spinal cord, a left femoral head and a right femoral head, and performing interpolation in a layer thickness direction, wherein gray linear interpolation is used in the embodiment;
s1.2: calculating the mutual OVH curve among the interested regions;
s1.3: calculating position information and volume information of the region of interest according to the CT image, the delineation contour information of each organ and the target area of the human body;
s1.4: and respectively forming corresponding anatomical structure vectors according to the OVH curves among the interested regions obtained in the step S1.2 and the position information and the volume information of the interested regions obtained in the step S1.3.
Further, on the basis of the above embodiment, as shown in fig. 2, in S1.2, the mutual OVH curve between the regions of interest is calculated by using a morphological dilation-erosion method or a direct distance calculation method.
Further, based on the above embodiment, S2 specifically includes the following steps, where the high-dimensional anatomical structure vector is an OVH vector:
s2.1: for OVH attribute vectors of n patients, after m-dimensional features of all attributes are extracted, the OVH attribute vectors are arranged into m rows and n columns of matrixes, namely m-dimensional OVH vectors;
s2.2: and reducing the m-dimensional OVH vector to t dimension by using a dimension reduction algorithm of unsupervised learning.
Further, on the basis of the above embodiment, the unsupervised learning dimension reduction algorithm is one or more of a principal component analysis algorithm, a singular value decomposition algorithm, and a t-distribution domain embedding algorithm.
As shown in fig. 3, the following describes the dimensionality reduction by taking the principal component analysis algorithm, PCA method as an example.
A1: for OVH attribute vectors of n patients, after m-dimensional features of all attributes are extracted, the OVH attribute vectors are arranged into m rows and n columns of matrixes, namely m-dimensional OVH vectors;
a2: sorting the eigenvectors of the covariance matrix of each matrix from big to small according to eigenvalues, unitizing the first t eigenvectors, and transposing to form a projection space;
a3: calculating the data after dimensionality reduction according to the projection space and the original matrix;
a4: and (4) carrying out PCA inverse transformation on the data subjected to the dimensionality reduction, and comparing the data with the original data to judge the selection of the t value.
It is noted that the scope of the present invention is not limited to only PCA dimension reduction algorithms.
Further, on the basis of the above embodiment, S3 specifically includes the following contents:
and clustering the data set to k categories according to the extracted and processed F characteristic attribute vectors including all the OVH vectors, organ absolute volume vectors and organ relative position vectors after dimensionality reduction by using a clustering algorithm of unsupervised learning.
Further, on the basis of the above embodiment, the clustering algorithm of unsupervised learning is one or more of a K-means clustering algorithm, a hierarchical clustering algorithm, a mean shift algorithm, and a density-based clustering algorithm.
Fig. 4 shows 8 feature attributes selected in this embodiment, and the data in the data set is clustered based on these feature attributes.
As shown in fig. 5, the following takes a K-means clustering algorithm as an example to perform a specific description of clustering.
B1: calculating the similarity of each characteristic attribute among the data, wherein the cosine value of the included angle among the anatomical structure vectors is used for representing to obtain a similarity matrix;
b2: finding out isolated data points according to the similarity density parameter, and taking the isolated data points as a class independently;
b3: sorting the data points according to the similarity density of the data points to form a candidate center matrix;
b4: for a data set containing n cases of data, the number of clusters k is from 1 to
Figure BDA0003487987190000091
Circulating, wherein the clustering center of each k value and the square sum of the intra-class errors at the time are calculated in the circulating process;
b5: and analyzing the error sum of squares in the class under different k values to determine the clustering number.
In B4, the initial clustering center of each k value is obtained, the similarity between each data and the initial clustering center is calculated and clustered, the center point of each cluster is recalculated, the similarity between each data and the new clustering center is calculated, clustering is performed again, and the steps are repeated until the clustering center is unchanged.
It is noted that the scope of the present invention is not limited to the K-means clustering algorithm.
Further, on the basis of the above embodiment, S4 specifically includes the following contents: on the basis of clinical evaluation indexes, the dose results of the clinically implemented high-quality radiotherapy plan in the data corresponding to each category are statistically analyzed, hardware information and software information used for obtaining corresponding results are obtained, the dose results corresponding to the score ranges of different index items in a radiotherapy plan score table are calculated according to different levels of plan effects, and a cervical cancer plan score template suitable for the category data is generated.
When different cluster plan scoring templates are generated according to the clustering results, besides the clinical dose requirement is considered, executed high-quality radiotherapy plan information is analyzed and extracted, possible dose results are divided according to different grades (different accelerators, TPS and the like), and corresponding grade scores are calculated.
Further, on the basis of the above embodiment, S5 specifically includes the following steps:
s5.1: acquiring an anatomical structure vector of the new data;
s5.2: reducing the dimension of the high-dimensional anatomical structure vector;
s5.3: and (5) calculating the similarity between the reduced-dimension vector obtained in the step (5.2) and the low-dimension anatomical structure vector obtained in the step (5.1) and each cluster central point, and judging the category of the new data.
The embodiment of the invention also discloses a radiotherapy plan evaluation device based on unsupervised learning, which comprises:
the anatomical structure vector obtaining module is used for obtaining an OVH curve, position information and volume information of each region of interest according to the data of n patients aiming at cervical cancer and respectively forming corresponding anatomical structure vectors;
the dimension reduction processing module is used for reducing the dimension of the high-dimensional anatomical structure vector obtained by the anatomical structure vector obtaining module by using a dimension reduction algorithm of unsupervised learning;
the clustering processing module is used for clustering the low-dimensional anatomical structure vector obtained by the data set according to the anatomical structure vector obtaining module and the low-dimensional attribute vector obtained after the processing of the dimension reduction processing module by using a clustering algorithm of unsupervised learning to obtain k categories;
the plan scoring template generating module is used for analyzing and processing the clinically implemented high-quality radiotherapy plan results in the data corresponding to each category according to the k categories obtained by the clustering processing module to obtain a plan scoring template corresponding to each category;
the category judgment module is used for calculating the new data of any patient and judging the category of the patient;
and the radiation treatment plan evaluation module is used for adopting the plan grading template of the corresponding category to guide the design of the radiation treatment plan corresponding to the new data and grading the dose distribution result.
The invention relates to a radiotherapy plan evaluation method and a radiotherapy plan evaluation device based on unsupervised learning, which have the following beneficial effects:
firstly, the invention extracts each data anatomical structure vector to carry out cluster analysis, and compares the anatomical structure vector of new data with the cluster result to find the most similar data point, thereby obtaining the corresponding plan scoring template, effectively realizing the individual evaluation of data with different geometric structure complexity, and effectively solving the problems that the clinical target is difficult to reach for the data with more complex geometric structure, and the clinical target is easy to reach for the data point with simple geometric structure
Second, the present invention addresses the problem of planning assessments to be difficult to meet patient-to-patient variation in the context of meeting clinical norms. Aiming at the clustering result, the invention combines the clinical index and the executed high-quality clinical plan to obtain a plan scoring template suitable for various data, and does not use a single clinical index any more. Meanwhile, the dose parameters contained in the radiotherapy plan scoring table take actual physical constraints into consideration, the scoring table designed by combining the clinically implemented high-quality plan dose result takes constraints of factors such as accelerator and grating performance on the radiotherapy plan effect into consideration, and the ideal dose parameters corresponding to the radiotherapy plan scoring table are applied to the reverse intensity modulated plan design, so that the plan quality difference caused by the abundant degree of clinical experience of a physicist and the consumable time difference can be improved, and the efficiency and the quality of the radiotherapy plan design are improved.
It should be understood that the various techniques described herein may be implemented in connection with hardware or software or, alternatively, with a combination of both. Thus, the methods and apparatus of the present invention, or certain aspects or portions thereof, may take the form of program code (i.e., instructions) embodied in tangible media, such as floppy diskettes, CD-ROMs, hard drives, or any other machine-readable storage medium, wherein, when the program is loaded into and executed by a machine, such as a computer, the machine becomes an apparatus for practicing the invention.
The above embodiments are merely illustrative of the technical concept and features of the present invention, and the purpose thereof is to enable those skilled in the art to understand the content of the present invention and implement the present invention, and not to limit the scope of the present invention, and all equivalent changes or modifications made according to the spirit of the present invention should be covered in the scope of the present invention.

Claims (10)

1. A radiation therapy plan evaluation method based on unsupervised learning is characterized by comprising the following steps:
s1: for any disease species, obtaining an OVH curve, position information and volume information of the interested region among the interested regions according to the data of n patients, and respectively forming corresponding anatomical structure vectors;
s2: reducing the dimension of the high-dimensional anatomical structure vector obtained in the step S1 by using a dimension reduction algorithm of unsupervised learning;
s3: clustering the low-dimensional anatomical structure vector obtained by the data set according to S1 and the low-dimensional attribute vector obtained after the data set is processed by S2 by using a clustering algorithm of unsupervised learning to obtain k categories;
s4: analyzing and processing the clinically implemented high-quality radiotherapy plan result in the data corresponding to each category according to the k categories obtained in the step S3 to obtain a plan scoring template corresponding to each category;
s5: calculating new data of any patient and judging the category of the patient;
s6: and adopting a plan scoring template of the corresponding category to guide the design of a radiation treatment plan corresponding to new data and scoring the dose distribution result of the radiation treatment plan.
2. The radiation therapy plan evaluation method according to claim 1, wherein S1 specifically includes the steps of:
s1.1: obtaining region-of-interest delineation information of n patients, and performing interpolation in the layer thickness direction;
s1.2: calculating mutual OVH curves among the regions of interest;
s1.3: calculating position information and volume information of the region of interest according to the CT image, the delineation contour information of each organ and the target area of the human body;
s1.4: and respectively forming corresponding anatomical structure vectors according to the OVH curves among the interested regions obtained in the step S1.2 and the position information and the volume information of the interested regions obtained in the step S1.3.
3. The radiation treatment plan evaluation method of claim 2, wherein in S1.2, the mutual OVH curves between the regions of interest are calculated by using a morphological dilation-erosion method or a direct distance calculation method.
4. The radiation treatment plan evaluation method of claim 1, wherein S2 particularly includes the step of, wherein the high-dimensional anatomical structure vector is an OVH vector:
s2.1: for OVH attribute vectors of n patients, after m-dimensional features of all attributes are extracted, the OVH attribute vectors are arranged into m rows and n columns of matrixes, namely m-dimensional OVH vectors;
s2.2: and reducing the m-dimensional OVH vector to t dimension by using a dimension reduction algorithm of unsupervised learning.
5. The radiation treatment plan evaluation method of claim 4, wherein the unsupervised learning dimension reduction algorithm is one or more of a principal component analysis algorithm, a singular value decomposition algorithm, and a t-distribution domain embedding algorithm.
6. The radiation treatment plan evaluation method according to claim 1, wherein S3 includes the following contents: and clustering the data set to k categories according to the extracted and processed F characteristic attribute vectors including all the OVH vectors, organ absolute volume vectors and organ relative position vectors after dimensionality reduction by using a clustering algorithm of unsupervised learning.
7. The radiation treatment plan evaluation method of claim 6, wherein the unsupervised learning clustering algorithm is one or more of a K-means clustering algorithm, a hierarchical clustering algorithm, a mean shift algorithm, a density-based clustering algorithm.
8. The radiation treatment plan evaluation method according to claim 1, wherein S4 includes the following contents: on the basis of clinical evaluation indexes, the dose results of the clinically implemented high-quality radiotherapy plan in the data corresponding to each category are statistically analyzed, hardware information and software information used for obtaining corresponding results are obtained, the dose results corresponding to the score ranges of different index items in a radiotherapy plan score table are calculated according to different levels of plan effects, and a plan score template suitable for the category data is generated.
9. The radiation treatment plan evaluation method as set forth in claim 1, wherein S5 particularly includes the steps of:
s5.1: acquiring an anatomical structure vector of the new data;
s5.2: reducing the dimension of the high-dimensional anatomical structure vector;
s5.3: and (5) calculating the similarity between the reduced-dimension vector obtained in the step (5.2) and the low-dimension anatomical structure vector obtained in the step (5.1) and each cluster central point, and judging the category of the new data.
10. An unsupervised learning-based radiation treatment plan evaluation device, comprising:
the anatomical structure vector obtaining module is used for obtaining an OVH curve, position information and volume information of each region of interest according to the data of n patients aiming at any disease type and respectively forming corresponding anatomical structure vectors;
the dimension reduction processing module is used for reducing the dimension of the high-dimensional anatomical structure vector obtained by the anatomical structure vector obtaining module by using a dimension reduction algorithm of unsupervised learning;
the clustering processing module is used for clustering the low-dimensional anatomical structure vector obtained by the data set according to the anatomical structure vector obtaining module and the low-dimensional attribute vector obtained after the processing of the dimension reduction processing module by using a clustering algorithm of unsupervised learning to obtain k categories;
the plan scoring template generating module is used for analyzing and processing the clinically implemented high-quality radiotherapy plan results in the data corresponding to each category according to the k categories obtained by the clustering processing module to obtain a plan scoring template corresponding to each category;
the category judgment module is used for calculating the new data of any patient and judging the category of the patient;
and the radiation treatment plan evaluation module is used for adopting the plan grading template of the corresponding category to guide the design of the radiation treatment plan corresponding to the new data and grading the dose distribution result.
CN202210085979.1A 2022-01-25 2022-01-25 Radiation therapy plan evaluation method and device based on unsupervised learning Pending CN114566251A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210085979.1A CN114566251A (en) 2022-01-25 2022-01-25 Radiation therapy plan evaluation method and device based on unsupervised learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210085979.1A CN114566251A (en) 2022-01-25 2022-01-25 Radiation therapy plan evaluation method and device based on unsupervised learning

Publications (1)

Publication Number Publication Date
CN114566251A true CN114566251A (en) 2022-05-31

Family

ID=81713621

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210085979.1A Pending CN114566251A (en) 2022-01-25 2022-01-25 Radiation therapy plan evaluation method and device based on unsupervised learning

Country Status (1)

Country Link
CN (1) CN114566251A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115223683A (en) * 2022-08-19 2022-10-21 中山大学肿瘤防治中心(中山大学附属肿瘤医院、中山大学肿瘤研究所) Radiotherapy plan evaluation method, radiotherapy plan evaluation device, computer equipment and medium
CN117116421A (en) * 2023-10-24 2023-11-24 福建自贸试验区厦门片区Manteia数据科技有限公司 Method and device for determining radiotherapy plan

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115223683A (en) * 2022-08-19 2022-10-21 中山大学肿瘤防治中心(中山大学附属肿瘤医院、中山大学肿瘤研究所) Radiotherapy plan evaluation method, radiotherapy plan evaluation device, computer equipment and medium
CN115223683B (en) * 2022-08-19 2023-08-25 中山大学肿瘤防治中心(中山大学附属肿瘤医院、中山大学肿瘤研究所) Radiation treatment plan evaluation method, radiation treatment plan evaluation device, computer equipment and medium
CN117116421A (en) * 2023-10-24 2023-11-24 福建自贸试验区厦门片区Manteia数据科技有限公司 Method and device for determining radiotherapy plan
CN117116421B (en) * 2023-10-24 2024-01-16 福建自贸试验区厦门片区Manteia数据科技有限公司 Method and device for determining radiotherapy plan

Similar Documents

Publication Publication Date Title
US11735309B2 (en) Method and system for automated quality assurance in radiation therapy
Men et al. More accurate and efficient segmentation of organs‐at‐risk in radiotherapy with convolutional neural networks cascades
Soltaninejad et al. Automated brain tumour detection and segmentation using superpixel-based extremely randomized trees in FLAIR MRI
US8976929B2 (en) Automatic generation of patient-specific radiation therapy planning parameters
Pukala et al. Benchmarking of five commercial deformable image registration algorithms for head and neck patients
Field et al. Machine learning applications in radiation oncology
CN108770373B (en) pseudo-CT generation from MR data using a feature regression model
CN108778416B (en) Systems, methods, and media for pseudo-CT generation from MR data using tissue parameter estimation
Zhong et al. A preliminary experience of implementing deep-learning based auto-segmentation in head and neck cancer: a study on real-world clinical cases
WO2021036366A1 (en) Standardized artificial intelligence automatic radiotherapy planning method and system
US10346593B2 (en) Methods and systems for radiotherapy treatment planning
CN114566251A (en) Radiation therapy plan evaluation method and device based on unsupervised learning
Jia et al. Beam orientation optimization for intensity modulated radiation therapy using adaptive l2, 1-minimization
CN111028914B (en) Artificial intelligence guided dose prediction method and system
Zhou et al. A method of using deep learning to predict three‐dimensional dose distributions for intensity‐modulated radiotherapy of rectal cancer
US10342994B2 (en) Methods and systems for generating dose estimation models for radiotherapy treatment planning
US11724126B2 (en) Methods and systems for quality-aware continuous learning for radiotherapy treatment planning
CN115083567A (en) Rule-based automatic radiotherapy plan obtaining method and device
Duan et al. Contouring quality assurance methodology based on multiple geometric features against deep learning auto‐segmentation
van Rooij et al. Strategies to improve deep learning-based salivary gland segmentation
CN116433967B (en) Personalized target spot selection method oriented to noninvasive nerve regulation technology
Bai et al. Comparing of two dimensional and three dimensional fully convolutional networks for radiotherapy dose prediction in left-sided breast cancer
Stoll et al. Comparison of safety margin generation concepts in image guided radiotherapy to account for daily head and neck pose variations
Mahesh et al. MRI techniques using image processing Brain Tumor Detection
Mehta et al. Autosegmentation of lung computed tomography datasets using deep learning U-Net architecture

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination