Remote radiotherapy target delineation system based on cloud computing

Abstract

The present invention discloses a remote radiotherapy target area delineation system based on cloud computing, which relates to the technical field of radiotherapy target area delineation, and includes an image data acquisition module, a data analysis module, a deviation distribution abnormal signal determination module and a software abnormality determination module; by comparing a revised delineation image corresponding to a newly acquired initial delineation image with an existing target area deviation distribution vector, a deviation distribution abnormal signal is generated in time to remind an operator that the deviation distribution is abnormal, and further verify whether there is an error in the software generating the corresponding disease radiotherapy target area delineation image or an error in the expert modification, which can ensure the accuracy and safety of radiotherapy, avoid radiotherapy errors caused by undetected errors, thereby reducing potential risks to patient health, and helping to discover and correct possible errors.

Description

Technical Field

[0001] The present invention relates to the technical field of radiotherapy target area delineation, and in particular to a remote radiotherapy target area delineation system based on cloud computing. Background Art

[0002] Remote radiotherapy target delineation connects the data of radiotherapy doctors, physicists and patients through the network to achieve remote collaboration and accurate target delineation. It uses powerful cloud computing capabilities to quickly process large amounts of medical imaging data and shorten the time for target delineation. Radiotherapy contouring software is used to quickly and automatically delineate radiotherapy target delineation images for different lesions. Experts then review and revise the radiotherapy target delineation images, and use the reviewed and revised radiotherapy target delineation images as reference images for the final radiotherapy. The radiotherapy contouring software automatically shortens the entire delineation process to less than 2 minutes, reducing the heavy workload of radiotherapy doctors in target delineation and effectively improving their work efficiency.

[0003] However, when experts in different lesion fields correct the radiotherapy target area delineation images automatically drawn by the radiotherapy contour delineation software, different experts will have different correction results. But after the experts correct the automatically drawn radiotherapy target area delineation images, they are unable to determine whether there is any abnormality in the corrected image differences, which leads to risks in the accuracy and safety of radiotherapy, and even causes radiotherapy errors. Based on this, a remote radiotherapy target area delineation system based on cloud computing is proposed. Summary of the invention

[0004] The purpose of the present invention is to provide a remote radiotherapy target area delineation system based on cloud computing, which solves the technical problem that after the experts correct the automatically delineated radiotherapy target area delineation image, they are unable to determine whether there is any abnormality in the corrected image difference, which leads to risks in the accuracy and safety of radiotherapy.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] The cloud computing-based remote radiotherapy target delineation system includes:

[0007] The image data acquisition module defines the radiotherapy target area delineation image generated by the radiotherapy contour delineation software as the initial delineation image, and the radiotherapy target area delineation image modified by the expert as the revised delineation image, and acquires the initial delineation image and revised delineation image of different types of lesions;

[0008] A data analysis module is used to compare and analyze the initial outline images and the revised outline images of different types of lesions to obtain target area deviation distribution vectors corresponding to different types of lesions;

[0009] The deviation distribution abnormal signal determination module is used to obtain the corrected outline image corresponding to the initial outline image in real time during real-time radiotherapy for analysis, obtain the real-time deviation distribution vector and the reference deviation distribution vector, calculate the distribution similarity between the real-time deviation distribution vector and the reference deviation distribution vector, and generate the deviation distribution abnormal signal according to the distribution similarity determination;

[0010] The software abnormality judgment module, when a deviation distribution abnormality signal is generated, uses the same input data to run the radiotherapy contouring software again, and again judges the deviation distribution abnormality signal on the corresponding initial contouring image and the revised contouring image. If the deviation distribution abnormality signal is still generated, a software abnormality signal is generated.

[0011] As a further solution of the present invention: the specific method of obtaining the target area deviation distribution vectors corresponding to different types of lesions is:

[0012] Step S1: randomly selecting a type of lesion from different types of lesions as the analysis type lesion;

[0013] Step S2: obtaining the regional deviation values ​​corresponding to the initial outline image and the modified outline image in each group of control images in the tumor area, clinical target area, internal target area and planned target area in each partition in n groups of control images of the analyzed lesions, and vectorizing them to express them as the regional deviation vector Rn(Gn, Cn, I n, Pn) of each group of control images, wherein a group of control images includes an initial outline image and its corresponding modified outline image, and n is a positive integer, 60≥n>1;

[0014] Step S3: The mean of Gn, Cn, In and Pn in the regional deviation vector of each group of control images is used as the calibration vector DB (BG, BC, BI, BP). The vector similarity Fn between the regional deviation vector of each group of control images and the calibration vector is calculated by the vector similarity calculation formula, and the vector similarity Fn is analyzed to obtain the target deviation distribution vector K1 (EG1, EC1, EI 1, EP1) of the analysis class lesion.

[0015] Step S4: Repeat steps S1-S3 to obtain target area deviation distribution vectors corresponding to different types of lesions, Ki(EG i, EC i, EI i, EP i), where i refers to different types of lesions and is a positive integer.

[0016] As a further solution of the present invention, in the n groups of control images of the analyzed lesions, the specific method of obtaining the regional deviation values ​​corresponding to the initial outline image and the modified outline image in each group of control images in the tumor area, clinical target area, internal target area and planned target area in each partition is:

[0017] Image processing software was used to obtain the areas of each partition of the tumor area, clinical target area, internal target area and planned target area in the initial outline image and the revised outline image in each group of control images, and the absolute value of the difference between the areas of each partition of the initial outline image and the revised outline image in each group of control images was marked as the regional deviation value corresponding to each partition of each group of control images.

[0018] As a further solution of the present invention: the specific method of analyzing the target area deviation distribution vector of the lesion-like focus is:

[0019] The number of vector similarities Fn that is greater than the preset value Q1 is obtained, and the number a is compared and analyzed with the preset value Q2. When the number a is greater than or equal to the preset value Q2, the calibration vector DB is used as the target deviation distribution vector K1 of the analysis lesion. When the number a is less than the preset value Q2, the mean of the sum of the maximum and minimum values ​​corresponding to Gn, Cn, In and Pn in the regional deviation vector Rn(Gn, Cn, In, Pn) is obtained, and then the target deviation distribution vector K1(EG1, EC1, EI 1, EP1) of the analysis lesion is generated. The specific value of the preset value Q1 is 0.88, where a is the number of Fn that is greater than the preset value Q1, and n≥a≥1.

[0020] As a further solution of the present invention: the specific method of generating the deviation distribution abnormal signal is:

[0021] In real-time radiotherapy, when the revised outline image corresponding to the initial outline image is obtained in real time, the same method as step S2 is adopted to obtain the real-time regional deviation values ​​corresponding to the revised outline image in each partition, and obtain the real-time deviation distribution vector. Then, according to the type of lesion corresponding to the initial outline image, the target area deviation distribution vector of the corresponding type of lesion is obtained from the data analysis module, and it is used as the benchmark deviation distribution vector. The same vector similarity calculation method as step S3 is adopted to calculate the distribution similarity between the real-time deviation distribution vector and the benchmark deviation distribution vector. When the distribution similarity is less than the preset value Q3, a deviation distribution abnormality signal is generated. Otherwise, no processing is performed.

[0022] As a further solution of the present invention: the specific method of generating the software abnormality signal is:

[0023] When a deviation distribution abnormality signal is generated, the same input data is used to run the radiotherapy contouring software again to obtain a revised contouring image corresponding to the initial contouring image, and the deviation distribution abnormality signal judgment module is used to perform judgment and analysis on the revised contouring image corresponding to the initial contouring image. If a deviation distribution abnormality signal is still generated, a software abnormality signal is generated; otherwise, no processing is performed.

[0024] Beneficial effects of the present invention:

[0025] The present invention compares the revised outline image corresponding to the newly acquired initial outline image with the existing target area deviation distribution vector in real-time radiotherapy, and promptly generates a deviation distribution abnormality signal to remind the operator that the deviation distribution is abnormal and to further verify whether there is an error in the software generating the corresponding disease radiotherapy target area outline image or an error in the expert modification. This can ensure the accuracy and safety of radiotherapy, avoid radiotherapy errors caused by undetected errors, thereby reducing potential risks to patient health, helping to improve the level of radiotherapy quality control, promptly discover and correct possible errors, help optimize the radiotherapy process, enhance the reliability and stability of the entire radiotherapy system, and promote continuous improvement and perfection of radiotherapy technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The present invention will be further described below in conjunction with the accompanying drawings.

[0027] Figure 1 It is a schematic diagram of the system framework structure of the present invention;

[0028] Figure 2 It is a schematic diagram of the framework structure of the method of the present invention. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] Embodiment 1

[0031] See also Figure 1-Figure 2 As shown, the present invention is a remote radiotherapy target delineation system based on cloud computing, comprising:

[0032] An image data acquisition module is used to acquire radiotherapy target area delineation images generated by radiotherapy delineation software and radiotherapy target area delineation images modified by experts for different types of lesions, and define the radiotherapy target area delineation images generated by radiotherapy delineation software as initial delineation images, and define the radiotherapy target area delineation images modified by experts as revised delineation images;

[0033] It should be noted that the different lesions include lung cancer, colorectal cancer, liver cancer, prostate cancer, and nasopharyngeal cancer;

[0034] The data analysis module is used to compare and analyze the radiotherapy target area delineation images generated by the radiotherapy contour delineation software for different types of lesions with the radiotherapy target area delineation images modified by experts, and obtain the target area deviation distribution vectors corresponding to different types of lesions according to the analysis results. The specific method is as follows:

[0035] Step S1: randomly selecting a type of lesion from different types of lesions as the analysis type lesion;

[0036] Step S2: obtaining n groups of control images of the analyzed lesions, wherein a group of control images includes an initial delineation image and a corresponding modified delineation image, where n is a positive integer, 60≥n>1;

[0037] Obtain the regional deviation values ​​corresponding to the initial outline image and the revised outline image in each group of control images in each group of control images in GTV (tumor area), CTV (clinical target area), I TV (internal target area) and PTV (planned target area) in each partition, and vectorize them to express them as the regional deviation vector Rn (Gn, Cn, I n, Pn) of each group of control images;

[0038] It should be noted that when delineating the radiotherapy target volume, GTV (tumor volume), CTV (clinical target volume), I TV (internal target volume), and PTV (planned target volume) are generally delineated. GTV represents the tumor tissue and metastatic lymph node area seen on the image. CTV is based on GTV, taking into account subclinical lesions and the possible invasion range of the tumor. PTV is based on CTV, taking into account factors such as positional repetitive deviation, respiratory movement, and changes in urine retention during treatment. ITV is an area that takes into account organ movement and positional changes between fractionated treatments. It is usually obtained by external expansion based on GTV or CTV. GTV (tumor volume), CTV (clinical target volume), I TV (internal target volume), and PTV (planned target volume) are obtained through a series of external expansions, just like the process of expanding from two rings to three rings, and then to four and five rings.

[0039] The specific method for obtaining the regional deviation values ​​corresponding to the initial outline image and the modified outline image in each group of control images in the GTV (tumor area), CTV (clinical target area), I TV (internal target area) and PTV (planned target area) is as follows:

[0040] The area of ​​each partition of GTV (tumor area), CTV (clinical target area), I TV (internal target area) and PTV (planned target area) in the initial outline image and the revised outline image in each group of control images is obtained by using image processing software. This can be achieved by identifying and measuring the number of pixels in each partition. The existing and mature calculation factors are not described again. Then, the absolute value of the difference between the area of ​​each partition of the initial outline image and the revised outline image in each group of control images is marked as the regional deviation value corresponding to each partition of each group of control images;

[0041] Step S3: taking the mean of Gn, Cn, In and Pn in the regional deviation vector Rn(Gn, Cn, In, Pn) of each group of control images as the calibration vector DB(BG, BC, BI, BP), and calculating the vector similarity Fn between the regional deviation vector Rn(Gn, Cn, In, Pn) of each group of control images and the calibration vector DB(BG, BC, BI, BP) by using the vector similarity calculation formula;

[0042] The specific formula for calculating vector similarity is:

[0043] in,

[0044] Obtain the number and value of vector similarity Fn greater than a preset value Q1, mark it as and mark it as Fa, where a is the number of vector similarities Fn greater than the preset value Q1, n≥a≥1, and the specific value of the preset value Q1 is 0.88;

[0045] The number a is compared and analyzed with the preset value Q2. When the number a is greater than or equal to the preset value Q2, it means that there are many vectors in the regional deviation vector Rn whose vector similarity with the calibration vector DB is greater than the preset value Q2, and the vector difference between most of the regional deviation vectors and the calibration vector DB is not large. Then the calibration vector DB is used as the target deviation distribution vector K1 (EG1, EC1, EI 1, EP1) of the analysis lesion. When the number a is less than the preset value Q2, it means that there are few vectors in the regional deviation vector Rn whose vector similarity with the calibration vector DB is greater than the preset value Q2. Then, the mean of the sum of the maximum and minimum values ​​corresponding to Gn, Cn, In and Pn in the regional deviation vector Rn (Gn, Cn, In, Pn) is obtained respectively, and then the target deviation distribution vector K1 (EG1, EC1, EI 1, EP1) of the analysis lesion is generated, that is, EG1 = (Gmax + Gmin) / 2, EC1 = (Cmax + Cmin) / 2, EI 1=(Imax+Imin) / 2 and EP1=(Pmax+Pmin) / 2, Gmax and Gmin, Cmax and Cmin, Imax and Imin, and Pmax and Pmin are the maximum and minimum values ​​of Gn, Cn, In and Pn respectively, and the specific value of the preset value Q2 is formulated by relevant personnel according to actual needs;

[0046] Step S4: Repeat steps S1-S3 to obtain target deviation distribution vectors corresponding to different types of lesions, Ki(EG i, EC i, EI i, EP i), where EG i, EC i, EI i and EP i represent the deviations of different types of lesions in GTV (tumor area), CTV (clinical target area), I TV (internal target area) and PTV (planned target area), respectively, and i refers to different types of lesions, and i is a positive integer;

[0047] Through the target deviation distribution vectors corresponding to different types of lesions, we can have a deeper understanding of the difference characteristics between the images of different types of lesions generated by radiotherapy contour delineation software and the images modified by experts, which is conducive to evaluating the accuracy of radiotherapy target delineation images of different types of lesions generated by radiotherapy software;

[0048] The deviation distribution abnormal signal determination module, in real-time radiotherapy, when obtaining the corrected outline image corresponding to the initial outline image in real time, adopts the same method as step S2 to obtain the real-time regional deviation values ​​corresponding to the corrected outline image in each partition, and obtains the real-time deviation distribution vector, and then obtains the target area deviation distribution vector of the corresponding type of lesion from the data analysis module according to the type of lesion corresponding to the initial outline image, and uses it as the reference deviation distribution vector, adopts the same vector similarity calculation method as step S3, calculates the distribution similarity between the real-time deviation distribution vector and the reference deviation distribution vector, and generates a deviation distribution abnormal signal when the distribution similarity is less than the preset value Q3, otherwise, no processing is performed;

[0049] In real-time radiotherapy, the revised outline image corresponding to the newly acquired initial outline image is compared with the existing target area deviation distribution vector, and a deviation distribution abnormality signal is generated in time to remind the operator that the deviation distribution is abnormal. Please further verify whether there is an error in the software generating the corresponding disease radiotherapy target area outline image or an error in the expert modification. This can ensure the accuracy and safety of radiotherapy, avoid radiotherapy errors caused by undetected errors, thereby reducing potential risks to patient health, helping to improve the level of radiotherapy quality control, and promptly discover and correct possible errors, which helps to optimize the radiotherapy process, improve the reliability and stability of the entire radiotherapy system, and promote the continuous improvement and perfection of radiotherapy technology.

[0050] Embodiment 2

[0051] As the second embodiment of the present invention, when the present application is implemented, compared with the first embodiment, the technical solution of the present embodiment is different from the first embodiment only in that the present embodiment further includes a software abnormality determination module;

[0052] The software abnormality determination module, when generating a deviation distribution abnormality signal, uses the same input data to run the radiotherapy contour delineation software again to obtain a revised contour image corresponding to the initial contour image, and again uses the deviation distribution abnormality signal determination module to determine and analyze the revised contour image corresponding to the initial contour image. If the deviation distribution abnormality signal is still generated, it is determined that there is a problem with the radiotherapy contour delineation software and a software abnormality signal is generated. Otherwise, no processing is performed;

[0053] It should be noted that the experts who make modifications each time are random and not fixed.

[0054] Embodiment 3

[0055] As the third embodiment of the present invention, when the present application is specifically implemented, compared with the first and second embodiments, the technical solution of this embodiment is to combine the solutions of the first and second embodiments mentioned above for implementation.

[0056] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0057] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A remote radiotherapy target delineation system based on cloud computing, characterized in that: include: The image data acquisition module defines the radiotherapy target area delineation image generated by the radiotherapy contour delineation software as the initial delineation image, and the radiotherapy target area delineation image modified by the expert as the revised delineation image, and acquires the initial delineation image and revised delineation image of different types of lesions; A data analysis module is used to compare and analyze the initial outline images and the revised outline images of different types of lesions to obtain target area deviation distribution vectors corresponding to different types of lesions; The deviation distribution abnormal signal determination module is used to obtain the corrected outline image corresponding to the initial outline image in real time during real-time radiotherapy for analysis, obtain the real-time deviation distribution vector and the reference deviation distribution vector, calculate the distribution similarity between the real-time deviation distribution vector and the reference deviation distribution vector, and generate the deviation distribution abnormal signal according to the distribution similarity determination; The software abnormality determination module, when a deviation distribution abnormality signal is generated, uses the same input data to run the radiotherapy contouring software again, and again performs deviation distribution abnormality signal determination on the corresponding initial contouring image and the revised contouring image. If the deviation distribution abnormality signal is still generated, a software abnormality signal is generated; The specific method of obtaining the target deviation distribution vectors corresponding to different types of lesions is as follows: Step S1: randomly selecting a type of lesion from different types of lesions as the analysis type lesion; Step S2: obtaining the regional deviation values ​​corresponding to the initial outline image and the modified outline image in each group of control images in the tumor area, clinical target area, internal target area and planned target area in each partition in n groups of control images of the analyzed lesions, and vectorizing them to express them as the regional deviation vector Rn (Gn, Cn, In, Pn) of each group of control images. A group of control images includes an initial outline image and its corresponding modified outline image, n is a positive integer, 60≥n>1; Step S3: The mean of Gn, Cn, In and Pn in the regional deviation vector of each group of control images is used as the calibration vector DB (BG, BC, BI, BP). The vector similarity Fn between the regional deviation vector of each group of control images and the calibration vector is calculated by the vector similarity calculation formula, and the vector similarity Fn is analyzed to obtain the target deviation distribution vector K1 (EG1, EC1, EI1, EP1) of the analysis class lesion. Step S4: Repeat steps S1-S3 to obtain target area deviation distribution vectors corresponding to different types of lesions, Ki (EGi, ECi, EIi, EPi), where EGi, ECi, EIi and EPi represent the deviations of different types of lesions in the tumor area, clinical target area, internal target area and planned target area, respectively, i refers to different types of lesions, and i is a positive integer; The specific method of obtaining the regional deviation values ​​corresponding to the initial outline image and the modified outline image in each group of control images in the tumor area, clinical target area, internal target area and planned target area in each partition is as follows: The area of ​​each partition of the tumor area, clinical target area, internal target area and planned target area in the initial outline image and the revised outline image in each group of control images is obtained using image processing software, and the absolute value of the difference between the area of ​​each partition of the initial outline image and the revised outline image in each group of control images is marked as the regional deviation value corresponding to each partition of each group of control images; The specific method of analyzing the target deviation distribution vector of the lesion is: Obtain the number of vector similarities Fn that is greater than the preset value Q1, compare and analyze the number a with the preset value Q2, and when the number a is greater than or equal to the preset value Q2, use the calibration vector DB as the target deviation distribution vector K1 of the analysis lesion; when the number a is less than the preset value Q2, obtain the mean of the sum of the maximum and minimum values ​​corresponding to Gn, Cn, In and Pn in the regional deviation vector Rn (Gn, Cn, In, Pn), and then generate the target deviation distribution vector K1 (EG1, EC1, EI1, EP 1), that is, EG1=(Gmax+Gmin) / 2, EC1=(Cmax+Cmin) / 2, EI1=(Imax+Imin) / 2 and EP1=(Pmax+Pmin) / 2, Gmax and Gmin, Cmax and Cmin, Imax and Imin, and Pmax and Pmin are the maximum and minimum values ​​of Gn, Cn, In and Pn respectively, and the specific value of the preset value Q1 is 0.88, where a is the number of Fn greater than the preset value Q1, and n≥a≥1; The specific method of generating the deviation distribution anomaly signal is: In real-time radiotherapy, when the revised outline image corresponding to the initial outline image is obtained in real time, the same method as step S2 is adopted to obtain the real-time regional deviation values ​​corresponding to the revised outline image in each partition, and obtain the real-time deviation distribution vector. Then, according to the type of lesion corresponding to the initial outline image, the target area deviation distribution vector of the corresponding type of lesion is obtained from the data analysis module, and it is used as the benchmark deviation distribution vector. The same vector similarity calculation method as step S3 is adopted to calculate the distribution similarity between the real-time deviation distribution vector and the benchmark deviation distribution vector. When the distribution similarity is less than the preset value Q3, a deviation distribution abnormality signal is generated. Otherwise, no processing is performed.

2. The cloud computing-based remote radiotherapy target delineation system according to claim 1, characterized in that: The specific method of generating software exception signal is: When a deviation distribution abnormality signal is generated, the same input data is used to run the radiotherapy contouring software again to obtain a revised contouring image corresponding to the initial contouring image, and the deviation distribution abnormality signal judgment module is used to perform judgment and analysis on the revised contouring image corresponding to the initial contouring image. If a deviation distribution abnormality signal is still generated, a software abnormality signal is generated; otherwise, no processing is performed.

3. The cloud computing-based remote radiotherapy target delineation system according to claim 1, characterized in that: The specific formula for calculating vector similarity is: ; in, ; ； 。

Images