Internal and external irradiation dose superposition method, system, terminal and storage medium

Abstract

The present application relates to a method, system, terminal and storage medium for superimposing internal and external irradiation doses, and relates to the technical field of cervical cancer radiotherapy. The method includes obtaining external irradiation positioning images and internal irradiation positioning images; slicing all internal irradiation positioning images, inputting all slice images into a segmentation network model, identifying target features and segmenting, and obtaining segmentation results; filling the target part in the slice image according to a specified value to generate an auxiliary slice image; inputting the corresponding slice image, the segmentation result and the auxiliary slice image into a synthesis network model to generate a synthetic slice image; integrating the corresponding synthetic slice images to obtain a synthetic positioning image; selecting one of them as a reference positioning image, and obtaining a first deformation field and a second deformation field through a deformation registration algorithm; collecting external irradiation doses according to the first deformation field, collecting corresponding internal irradiation doses according to the second deformation field, and calculating equivalent biological doses. The present application can accurately calculate the superimposition results of internal and external irradiation doses.

Description

Technical Field

[0001] The present application relates to the field of cancer radiotherapy technology, and in particular to a method, system, terminal and storage medium for superimposing internal and external irradiation doses. Background Art

[0002] Radiotherapy is a treatment method that uses radiation to irradiate tumor cells, interrupt the DNA chain of tumor cells, and control the growth of tumor cells. It is one of the main treatment methods for cervical cancer. Radiotherapy is not only the first choice for treatment of advanced cervical cancer, but also an important means of postoperative adjuvant treatment of early cervical cancer and treatment of recurrent tumors. Radiotherapy includes external irradiation and intracavitary close-range irradiation. When treating patients, external irradiation combined with close-range radiotherapy is usually used to control the toxic and side effects of tumors and surrounding normal tissues.

[0003] It is understood that since external irradiation uses X-rays generated by a linear accelerator to accurately strike the pelvic involvement and metastatic areas of cervical cancer from outside the body, only one simulation positioning CT image needs to be collected when formulating an external radiation therapy plan. Internal irradiation is to place the source applicator or interpolation needle into the human body to irradiate the primary area of ​​cervical cancer at close range. Each time internal radiation therapy is performed, the implantation position and angle of the source applicator or interpolation needle will be different, and the filling of the patient's rectum and bladder will also be different, which will affect the patient's tumor location and shape. Therefore, before each internal radiation therapy, the doctor needs to collect a simulation positioning CT image to re-formulate the internal radiation therapy plan based on the current positioning image and the dose of the target area and organs at risk during the previous radiation therapy. When calculating the internal and external irradiation doses, deformation registration technology is usually used for dose superposition analysis.

[0004] However, in the dose superposition analysis based on registration, since the applicator or interpolation needle is implanted in the patient's body and the position and direction of the applicator or interpolation needle are different each time, the registration is more difficult and the registration effect is poor, which in turn affects the accuracy of calculating the internal and external irradiation doses. Summary of the invention

[0005] In order to accurately calculate the internal and external radiation doses, the present application proposes an internal and external radiation dose superposition method, system, terminal and storage medium.

[0006] In a first aspect of the present application, a method for superimposing internal and external irradiation doses is provided, the method comprising:

[0007] Acquire an external irradiation positioning image and n internal irradiation positioning images within a first preset time range, where n is an integer greater than 1, and both the external irradiation positioning image and the internal irradiation positioning image are three-dimensional images;

[0008] Processing each of the internal irradiation positioning images to obtain a plurality of slice images;

[0009] Inputting all the slice images into a segmentation network model, identifying target features in the slice images, and performing segmentation to obtain a plurality of segmentation results corresponding to the slice images, wherein the target features are applicators or interpolation needles;

[0010] Filling the image information of the target part in all the slice images according to the specified value to generate a plurality of auxiliary slice images corresponding to each slice image;

[0011] Inputting the corresponding slice image, the auxiliary slice image and the segmentation result into a synthetic network model to generate a plurality of synthetic slice images corresponding to each slice image, wherein the synthetic slice image is a slice image that makes the slice image of the original applicator or interpolation needle part have a pixel value close to that of the surrounding tissue;

[0012] Integrating all the synthetic slice images to obtain n synthetic positioning images corresponding to the n internal irradiation positioning images, wherein the synthetic positioning images are three-dimensional images;

[0013] Determine one of the n synthetic positioning images as a reference positioning image, and obtain a first deformation field of the external illumination positioning image relative to the reference positioning image through a deformation registration algorithm, and obtain a second deformation field of other synthetic positioning images in the n synthetic positioning images relative to the reference positioning image;

[0014] An external radiation dose map is collected according to the first deformation field to obtain an external radiation dose in the reference positioning image coordinate system; a corresponding internal radiation dose is collected according to the second deformation field to obtain an internal radiation dose in the reference positioning image coordinate system; and an equivalent biological dose in the reference positioning image coordinate system is calculated according to the external radiation dose and the internal radiation dose.

[0015] By adopting the above technical solution, the segmentation network model can identify the applicator or interpolation needle in the slice image, and extract the images of the applicator and the interpolation needle therefrom. Through the segmentation network model, all slice images can be automatically segmented. Furthermore, the slice image, the corresponding auxiliary slice image and the segmentation result are input into the synthesis network model, and the synthetic slice image can be automatically generated through the synthesis network model, and by combining the synthetic slice images, a synthetic positioning image can be obtained. The synthetic positioning image is an image from which the influence of the applicator or interpolation needle is removed. By replacing the internal irradiation positioning image with the synthetic positioning image, and then aligning the synthetic positioning image with the external irradiation positioning image and calculating the superimposed dose, the accuracy of calculating the internal and external irradiation doses can be improved.

[0016] In a possible implementation, the segmentation result is a binary image, in which the value belonging to the source applicator or the interpolation needle position is 1, and the value not belonging to the source applicator or the interpolation needle position is 0.

[0017] In a possible implementation, the filling of the image information belonging to the target part in all the slice images according to the specified value includes: filling the image belonging to the target part in the slice image with the minimum pixel value in the area within a preset range from the edge of the target part.

[0018] By adopting the above technical solution, the information of the applicator or interpolation needle can be directly removed from the auxiliary slice image, making it close to the slice image of the internal irradiation positioning image of the surrounding part without the applicator or interpolation needle, thereby making the synthetic slice image synthesized by the synthetic network model more accurate.

[0019] In a possible implementation manner: training a synthetic network model; the training step of the synthetic network model includes: acquiring a sample image, wherein the sample image includes the corresponding slice image, the segmentation result and the auxiliary slice image;

[0020] Inputting the sample image into a generator to obtain the synthetic slice image;

[0021] The synthetic slice image is input as a negative sample image to the discriminator, the corresponding auxiliary slice image is input as a negative sample image to the discriminator, and the image of the relevant part in the external irradiation positioning image without an applicator or an interpolation needle is input as a positive sample image to the discriminator to obtain a discrimination result, and a discrimination loss function and a total loss function are calculated based on the discrimination result; based on the discrimination loss function and the total loss function, the parameters of the generator and the parameters of the discriminator are adjusted to obtain the synthetic network model.

[0022] By adopting the above technical solution and using the generator and the discriminator to train the synthetic network model, the synthetic slice image can be made closer to the slice image of the internal irradiation image without the applicator and the interpolation needle.

[0023] In a possible implementation manner: adjusting the parameters of the generator and the parameters of the discriminator based on the discriminant loss function and the total loss function includes:

[0024] Setting parameters of the generator to fixed values;

[0025] Adjusting the parameters of the discriminator based on the discriminant loss function so that the discriminant loss function reaches a preset first convergence condition, where the first convergence condition is that the discriminant loss function no longer decreases;

[0026] Setting the parameters of the discriminator to fixed values ​​that enable the discriminant loss function to reach the first convergence condition;

[0027] Adjusting the parameters of the generator based on the discriminant loss function and the total loss function so that the total loss function reaches a preset second convergence condition, where the second convergence condition is that the total loss function no longer decreases;

[0028] Repeat the above optimization process until the set optimization rounds are completed.

[0029] In a possible implementation manner: the calculating the discrimination loss function and the total loss function based on the discrimination result includes:

[0030] The discriminant loss function is calculated according to the following formula:

[0031]

[0032] Among them, X i represents the input image to the discriminator, i represents the number of the training sample, Disc(X i ) represents the discrimination result output by the discriminator, and the discriminator is used to distinguish whether the input image is a synthetic image or a real image; the synthetic image is a category label with a discrimination value of 0, and the real image is a category label with a discrimination value of 1, C i Represents the input image X i The category label of .

[0033] In a possible implementation manner: the calculating the discrimination loss function and the total loss function based on the discrimination result includes:

[0034] The total loss function is calculated according to the following formula:

[0035]

[0036] Ω i ={(m,n)Y i (m,n)=0}

[0037] Among them, Ω i Indicates the non-applicator or interpolation needle area, I i represents the slice image, M i represents the auxiliary slice image, represents the segmented image, represents the synthetic slice image, and L1 represents the absolute value of the error.

[0038] By adopting the above technical scheme, when calculating the total loss function, the proximity of the pixel values ​​of the non-applicator or interpolation needle area in the synthetic slice image to the slice image of the internal irradiation positioning image and the ability of the synthetic slice image to confuse the discriminator are taken into account, so that the synthetic network model can be optimized.

[0039] In a second aspect of the present application, a method system for superimposing internal and external irradiation doses is provided, the system comprising:

[0040] An acquisition module, used to acquire an external irradiation positioning image and n internal irradiation positioning images within a first preset time range, where n is an integer greater than 1, and both the external irradiation positioning image and the internal irradiation positioning image are three-dimensional images;

[0041] A processing module, used for processing each of the internal irradiation positioning images to obtain a plurality of slice images;

[0042] A segmentation module is used to input all the slice images into a segmentation network model, identify target features in the slice images, and perform segmentation to obtain multiple segmentation results corresponding to the slice images, wherein the target features are applicators or interpolation needles; an auxiliary module is used to fill the image information belonging to the target part in all the slice images according to the specified value, and generate multiple auxiliary slice images corresponding to each slice image one by one;

[0043] A synthesis module, used for inputting the corresponding slice image, the auxiliary slice image and the segmentation result into a synthesis network model to generate a plurality of synthetic slice images corresponding to each slice image; the synthetic slice image is a slice image that makes the slice image of the original applicator or interpolation needle part have a pixel value close to that of the surrounding tissue;

[0044] A generating module, used for integrating all the synthetic slice images to obtain n synthetic positioning images corresponding to the n internal irradiation positioning images one by one, wherein the synthetic positioning images are three-dimensional images;

[0045] A registration module, used to determine one of the n synthetic positioning images as a reference positioning image, and obtain a first deformation field of the external illumination positioning image relative to the reference positioning image through a deformation registration algorithm, and obtain a second deformation field of other synthetic positioning images in the n synthetic positioning images relative to the reference positioning image;

[0046] an acquisition module, used for acquiring an external irradiation dose map according to the first deformation field to obtain an external irradiation dose in the reference positioning image coordinate system, and acquiring a corresponding internal irradiation dose according to the second deformation field to obtain an internal irradiation dose in the reference positioning image coordinate system; and a calculation module, used for calculating an equivalent biological dose in the reference positioning image coordinate system according to the external irradiation dose and the internal irradiation dose.

[0047] In a third aspect of the present application, a smart terminal is provided. The smart terminal includes a memory and a processor. The memory stores a computer program that can be loaded by the processor and executes the above-mentioned internal and external irradiation dose superposition method.

[0048] In a fourth aspect of the present application, a computer-readable storage medium is provided, storing a computer program that can be loaded by a processor and execute any of the above-mentioned internal and external radiation dose superposition methods.

[0049] In summary, the present application includes at least one of the following beneficial technical effects:

[0050] In the present application, the segmentation network model can identify the applicator or interpolation needle in the slice image, and extract the images of the applicator and the interpolation needle therefrom. Through the segmentation network model, all slice images can be automatically segmented. Furthermore, the slice image, the corresponding auxiliary slice image and the segmentation result are input into the synthesis network model, and the synthetic slice image can be automatically generated through the synthesis network model, and by combining the synthetic slice images, a synthetic positioning image can be obtained. The synthetic positioning image is an image from which the influence of the applicator or interpolation needle is removed. By replacing the internal irradiation positioning image with the synthetic positioning image, and then aligning the synthetic positioning image with the external irradiation positioning image and calculating the superimposed dose, the accuracy of calculating the internal and external irradiation doses can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of a method for superimposing internal and external irradiation doses according to one embodiment of the present application.

[0052] Figure 2 It is a flow chart of a segmentation network model according to one embodiment of the present application.

[0053] Figure 3 It is a structural diagram of a segmentation network model according to one embodiment of the present application.

[0054] Figure 4 It is a flow chart of a synthetic network model according to one embodiment of the present application.

[0055] Figure 5 It is a system schematic diagram of an internal and external irradiation dose superposition system according to one embodiment of the present application.

[0056] Figure 6 It is a schematic diagram of the structure of a smart terminal according to one embodiment of the present application.

[0057] In the figure, 21, acquisition module; 22, processing module; 23, segmentation module; 24, auxiliary module; 25, synthesis module; 26, generation module; 27, registration module; 28, acquisition module; 29, calculation module; 301, CPU; 302, ROM; 303, RAM; 304, bus; 305, I / O interface; 306, input part; 307, output part; 308, storage part; 309, communication part; 310, drive; 311, removable medium. DETAILED DESCRIPTION

[0058] The present application is further described in detail below in conjunction with the accompanying drawings.

[0059] This specific embodiment is merely an explanation of the present application and is not a limitation of the present application. After reading this specification, those skilled in the art may make modifications to the present embodiment without any creative contribution as needed, but such modifications are protected by the patent law as long as they are within the scope of the claims of the present application.

[0060] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0061] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0062] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification.

[0063] The embodiment of the present application provides a method for superimposing internal and external radiation doses, which is applied in the process of performing radiotherapy on patients. Before formulating a radiotherapy plan for the patient, the external radiation positioning image and the internal radiation positioning image of each fraction can be accurately aligned to accurately calculate the sum of the internal radiation dose and the external radiation dose accumulated at the patient's lesion location, organs at risk and normal tissues.

[0064] It is understood that radiation therapy includes external beam radiation and intracavitary brachytherapy.

[0065] The principle of external irradiation is to use the X-rays generated by a linear accelerator to accurately strike the pelvic involvement and metastatic areas of cervical cancer from outside the body, so that the dose in the target area is high enough, while the radiation dose to normal tissues is lower. External irradiation mainly targets the paracervical area and lymphatic drainage area.

[0066] The principle of internal irradiation is to take advantage of the natural favorable conditions of cervical cancer, place the radioactive source in the natural lumen of the human body, and irradiate the primary area of ​​cervical cancer at close range. Specifically, the radioactive source can be placed in the human body by pre-implanting an applicator or an interpolation needle.

[0067] It can be seen that external irradiation combined with brachytherapy can combine the advantages of both, and can control the toxicity and side effects of surrounding normal tissues while controlling the tumor. Usually, when patients with cervical cancer undergo radiotherapy, they will undergo 25-30 IMRT / VMAT external irradiation treatments, and then 3-5 6Gy / time brachytherapy. In this process, it is particularly important to calculate the superimposed dose of external irradiation and internal irradiation.

[0068] It is also known that for external beam radiation therapy, when formulating a radiation therapy plan, the patient needs to collect simulated positioning CT images, namely the external beam radiation positioning images mentioned above, under the condition of vacuum cushion or thermoplastic film to assist in fixing the body position. Then, the doctor needs to outline the target area and organs at risk of cervical cancer patients on the external beam radiation positioning images, and then determine the target area dose and organ at risk dose according to the prescribed dose, so as to calculate the average according to the number of external beam radiation treatments to obtain each external beam radiation treatment plan.

[0069] For internal radiation therapy, when formulating a radiation therapy plan, the patient also needs to use a vacuum cushion or thermoplastic film to assist in fixing the body position. After the doctor inserts the applicator or interpolation needle into the uterine cavity, the internal radiation positioning image is collected. Then, the doctor needs to outline the target area and organs at risk of cervical cancer patients on the external radiation positioning image, and then determine the internal radiation therapy plan based on the prescribed dose. It should be noted that each time internal radiation therapy is performed, the implantation position and angle of the applicator or interpolation needle, as well as the patient's rectum and bladder filling conditions will be different, which may cause the patient's tumor position and shape to change, resulting in the need to collect internal radiation positioning images before each internal radiation therapy and to re-formulate the internal radiation therapy plan.

[0070] Furthermore, since external radiation therapy uses X-rays generated by a linear accelerator and does not require an applicator or interpolation needle, the external radiation positioning image and the internal radiation positioning image differ under the influence of the applicator and the interpolation needle, which makes it difficult to calculate the total dose after the multiple external radiation doses and the multiple internal radiation doses are superimposed.

[0071] At present, there are two methods to superimpose external radiation dose and internal radiation dose. The first method is based on the CEC-ESTRO recommendation, which assumes that the irradiation of organs at risk in external radiation is uniform and equivalent to the prescribed dose. Based on this, the internal radiation dose or the DVH parameter of the internal radiation dose is directly superimposed with the external radiation prescription dose. The DVH parameter can be, for example, the DVH of the target area. 90 , organ-threatening D 2cc And other DVH related indicators. 90 This means the minimum dose that uniformly covers 90% of the target volume, D 2cc That is, it means the lowest dose within the 2cc volume with the highest dose. In this process, in order to make the internal irradiation dose distribution compatible and comparable with the external irradiation dose distribution, it is necessary to select the corresponding parameters of the internal irradiation dose according to different tissue properties to calculate the equivalent biological dose EQD2 of 2Gy fractions. The calculation formula of equivalent biological dose is:

[0072]

[0073] The parameter n represents the number of treatment fractions, d represents the single dose, in Gy. α is the starting slope of the cell inventory curve, and β is the superlinear part of the radiation effect. Set to 10, if the organ is at risk Set to 3. However, IMRT or VMAT technology used in external beam radiation emphasizes conformality rather than uniformity, because it will lead to uneven dose distribution in the organs at risk during treatment, thus making this dose superposition analysis method inaccurate.

[0074] The second method is to superimpose the actual DVH parameters of internal and external irradiation, specifically using the parameters of the actual distribution of DVH of each fractional dose for superposition. However, this method requires the assumption that the dose distribution area of ​​each dose evaluation index is consistent during each fractional treatment, that is, the area with high external irradiation dose has high internal irradiation dose accordingly. Since organ deformation and the inhomogeneity of dose distribution are not taken into account, it cannot directly reflect the actual overall irradiation dose.

[0075] In addition, the stacking dose can also be calculated based on the registration technology. Registration technology is a technology that establishes a spatial transformation relationship between multiple data sources and is often used to process medical images. When calculating the stacking dose, it can be divided into the following two processing methods according to whether to remove the influence of the internal irradiation source or the interpolation needle.

[0076] The first method is to remove the applicator and interpolation needle parts in the internal irradiation positioning image and assign a reasonable value to obtain the internal irradiation positioning image without the applicator and interpolation needle parts, and then refer to the processed internal irradiation positioning image of a certain fraction to align the external irradiation positioning image and the internal irradiation positioning images of other fractions, unify the dose to the reference internal irradiation positioning image coordinate system, and then calculate the superposition results of the internal irradiation dose and the external irradiation dose in all the areas of interest, and then statistically analyze the DVH parameter indicators of the superposition results.

[0077] The second method is to ignore the influence of the presence or absence of internal and external irradiation applicators, and directly refer to the internal irradiation positioning image with an applicator or interpolation needle in a certain fraction, align the external irradiation positioning image and the internal irradiation positioning images with applicators or interpolation needles in other fractions, and map their respective doses to the reference internal irradiation positioning image coordinate system and superimpose them, and analyze the DVH parameter indicators of the superposition results.

[0078] The image registration method in this paper is to solve the three-dimensional reference image I ref With 3D Moving Image I mov The spatial mapping relationship G between the reference image I ref The upper coordinate (x, y, z) and the moving image I mov (G x (x,y,z),G y (x,y,z),G z The information at the (x,y,z) position corresponds to the anatomical position of the same patient:

[0079]

[0080] Among them, the symbol Indicates one-to-one correspondence between spatial positions.

[0081] According to the analysis requirements, a positioning image of a certain fraction of internal irradiation or external irradiation can be selected as a reference, and the deformation registration technology can be used to register all other fractions of internal irradiation positioning images that remove the influence of the applicator or interpolation needle, as well as all other fractions of external irradiation positioning images. Then, the dose distribution results of all fractions are mapped to the same positioning image coordinate system, and then the doses are superimposed point by point to analyze the dose distribution parameters after superposition. The following takes the selection of a certain fraction of internal irradiation positioning image as a reference image as an example. Dose can also be unified to the reference internal irradiation image coordinate system using the spatial mapping relationship:

[0082]

[0083] j=1,2...,nBT;j≠ref

[0084] Where nBT represents the internal radiation dose per fraction; Dose BTrefIndicates the Dose result of the selected reference internal irradiation positioning image corresponding to the fractionated treatment; Dose EBRT Indicates the Dose result corresponding to the external irradiation positioning image; Dose BTj Indicates the Dose results of fractionated treatment corresponding to other non-reference internal irradiation positioning images; G EBRT G represents the spatial mapping relationship obtained by registering the external irradiation positioning image with the selected internal irradiation positioning image; j It represents the spatial mapping relationship obtained by registering other j-th fractionated internal irradiation positioning images with reference to the selected internal irradiation positioning image.

[0085] Currently, the internal irradiation positioning image I without the source applicator or interpolation needle BTj The process of '(x,y,z) can be expressed mathematically as:

[0086] j=1,2,...,nBT

[0087] Among them, Ω applicator Indicates the area where the applicator or interpolation needle is implanted in the patient's body; I padding is an artificial value used to fill the source or interpolation needle area so that I BTj '(x,y,z) looks like a real captured image. In the related art, Ω applicator It can be obtained by threshold segmentation, but the pixel values ​​of different applicator materials vary greatly in CT or MR positioning images, and a certain threshold may only be applicable to one material. Moreover, due to the influence of other treatment tubes implanted by the patient, the segmentation effect of this method is not very stable. In addition, due to I padding The filling value is set artificially. The filling value in multiple internal irradiation positioning images is prone to be too uniform, which may affect the registration effect.

[0088] For the internal irradiation positioning image after removing the influence of the applicator or interpolation needle in the internal irradiation image, the Dose space mapping relationship can be expressed as:

[0089]

[0090] Among them, G' EBRT The reference image I is removed. BTref 'Spatial mapping relationship obtained by registering external irradiation positioning images; G' j The reference image I is removed. BTref 'Register the other j-th steps to remove the image that affects the positioning I BTj 'Obtained spatial mapping relationship.

[0091] The process of dose superposition after registration can be expressed as:

[0092]

[0093] Furthermore, according to the dose superposition results of all positions, DVH is statistically generated, and parameter analysis of the DVH of the dose superposition results is completed.

[0094] In summary, the dose superposition analysis of internal and external irradiation based on direct DVH superposition introduces too many limitations on the uniformity of dose distribution in the target area or organs at risk, and completely ignores the problem that the position and morphology of the structural anatomy of the internal and external irradiation patient cannot be directly corresponded, and it is impossible to obtain accurate dose analysis results. In the dose superposition analysis based on registration, it is ignored that the internal irradiation positioning image has more applicators or interpolation needles implanted in the patient's body than the external irradiation positioning image. These differences may increase the difficulty of registration and cause poor registration effects in local areas, especially between different fractions of internal irradiation. Although there are applicators or interpolation needles, their positions and directions are significantly different. Direct registration without any processing will lead to obvious errors. At present, in the method of adding and removing the processing step of the applicator or interpolation needle, the segmentation method of the applicator or interpolation needle is relatively simple, which may be affected by other implanted pipes in the patient's body, and may only be suitable for the processing of applicators or interpolation needles of one material, and cannot adapt to the removal operation of applicators or interpolation needles of different materials. At the same time, it is currently only applicable to internal irradiation positioning images using CT images, and cannot support the situation of internal irradiation positioning images using nuclear magnetic images.

[0095] Furthermore, the removal of the applicator or interpolation needle requires manual specification of the filling value, which cannot ensure that the image after filling is similar or close to the surrounding tissue, and may introduce new registration problems.

[0096] Based on the above-mentioned problems, an embodiment of the present application provides a method for superimposing internal and external irradiation doses, and the main process is described as follows.

[0097] like Figure 1 As shown:

[0098] Step S100: acquiring an external irradiation positioning image and n internal irradiation positioning images within a first preset time range, where n is an integer greater than 1.

[0099] Among them, the external irradiation positioning image is a CT image, which is mainly obtained by a CT simulation positioning machine or a DR simulation positioning machine. The internal irradiation positioning image can be a CT image or an MR image, which is mainly obtained by nuclear medicine imaging equipment such as positron emission tomography or single photon emission computed tomography. Both the external irradiation positioning image and the internal irradiation positioning image are three-dimensional images.

[0100] It is understood that during the treatment process, the patient needs to undergo multiple external radiation therapy and multiple internal radiation therapy, and the cumulative total radiation dose needs to be calculated in order to adjust the subsequent treatment plan. Therefore, when calculating the cumulative total radiation dose, a first preset time range needs to be set, and the first preset time range changes as the number of radiation therapy increases. The number of internal radiation therapy within the first preset time range is the number of internal radiation positioning images.

[0101] Step S200: Process each of the internal irradiation positioning images to obtain a plurality of slice images.

[0102] Since the internal irradiation positioning image is a three-dimensional image, a plurality of two-dimensional slice images can be obtained when the internal irradiation positioning image is sliced.

[0103] Step S300: inputting all the slice images into a segmentation network model, identifying target features in the slice images, and performing segmentation to obtain a plurality of segmentation results corresponding to the slice images.

[0104] The segmentation network model is used to identify target features in the slice image and remove the target features from the slice image to obtain a segmentation result. In the embodiment of the present application, the target feature is an applicator or an interpolation needle.

[0105] It is understandable that when different segmentation network models are used, the segmentation results obtained are different. In some specific embodiments, the segmentation result is a binary image, where the value belonging to the applicator or interpolation needle position is 1, and the value not belonging to the applicator or interpolation needle position is 0.

[0106] Step S400: Filling the image information of the target part in all the slice images according to the specified value to generate a plurality of auxiliary slice images corresponding to each slice image.

[0107] In order to make the auxiliary slice image closer to the slice image of the internal irradiation positioning image without the applicator and the interpolation needle, it is necessary to assign a reasonable value, i.e., a filling value, to the area with the same position as the applicator or the interpolation needle in the slice image according to the segmentation result, so as to generate the auxiliary slice image. The filling value can be selected from the minimum pixel value in the area within a preset range from the edge of the target feature in the slice image.

[0108] Specifically, the auxiliary slice image can be obtained by filling the image belonging to the target part in the internal illumination positioning image with the minimum pixel value in the area where the edge of the target part is within a preset range.

[0109] In a specific example, assume that region B in image A is the region where the target feature is located, and region C is the region within a preset range from the edge of region B. There are three pixel values ​​a, b, and c in the portion of region C that does not belong to region B. Among them, the pixel value of a is the lowest. Therefore, region B is preferably filled with the pixel value corresponding to a.

[0110] The preset range can select the most appropriate size according to actual needs.

[0111] In order to segment the source applicator or interpolation needle in the slice image, other segmentation network structures such as FCN network structure, U-Net network structure, SegNet network structure or DeepLab network structure can be selected. However, in the embodiment of the present application, the slice image is processed mainly by modifying the SwinTransformer network.

[0112] Reference Figure 2 and Figure 3 Specifically, the segmentation network model consists of multiple stages. Each stage is responsible for processing image features of different scales and levels. The division of stages is based on patch merging and feature transformation. At the beginning of each stage, the image or feature map is divided into a series of patches, and then converted into feature dimensions acceptable to the Transformer through linear embedding or convolution operations. Then, these features are processed by multiple Swin Transformer Blocks, each of which contains components such as a self-attention mechanism based on a shifted window and a feedforward network.

[0113] As the stages progress, the network gradually merges adjacent patches, reducing the number of tokens and increasing the feature dimension. This operation helps to produce a hierarchical representation, allowing the model to capture information at different scales. For example, in stage 2, the patch merging layer merges pixel blocks within a 2×2 domain, and a linear layer is used to operate on the merged features, thereby reducing the number of tokens and adjusting the feature dimension. This operation is repeated in subsequent stages to produce feature representations at different levels.

[0114] Furthermore, the principle of the segmentation network model is as follows: input the H×W positioning image, divide it into continuous non-overlapping 4×4 sub-blocks, flatten the pixel values ​​in each sub-block into a 16-dimensional vector and generate a 64-dimensional vector after a uniform linear transformation. At this time, the entire image is represented as A vector sequence of length 64. After the vector sequence is processed by several consecutive SwinTransformer modules, the output A vector sequence of 64 in length. Through sub-block fusion, the 2*2 vector sequence at continuous positions in space is synthesized into a new vector twice as long as the original one after sub-block fusion. Sub-block fusion is to link the 4 adjacent vectors into an intermediate vector of 4 times the length. After a fully connected linear transformation, a new vector of twice the length of the original one is generated, thereby reducing the number of vectors in the vector sequence and doubling the length of each vector. Repeating the above process, four different sequences can be generated using the encoder module. The sequence at the bottom contains the least number of vectors, but the length of a single vector is the longest.

[0115] The SwinTransformer module consists of two in-window self-attention processing modules with different window division methods. In-window self-attention means that each input vector is not calculated with all other vectors, but with some vectors in the window where it is located. This can greatly reduce the amount of calculation and improve the processing speed. According to the different ways of window division, it is divided into regular window division and offset window division. Regular window division means that after the positions of the corresponding blocks of the vectors are directly arranged, they are divided into continuous non-overlapping N*N windows. Offset window means that after the vectors are arranged, they are moved backward in two different directions in a loop. Then it is divided into continuous non-overlapping N*N windows. Figure 3 The symbol z in is a set of vector sequences, and the symbol l-1 represents the layer number of the SwinTransformer input vector sequence. A SwinTransformer module actually has two layers of in-window self-attention with different window division methods, so the layer number of its output result is l+1.

[0116] The output of the bottom layer of the encoder of the segmentation network model is A vector sequence z of length 512 28 , after rearranging it, it is converted to a channel size of 512 The feature map is processed into a size of 512 using the bilinear upsampling module in the convolutional neural network. This feature map is obtained by reverse rearrangement. A vector sequence z of length 512 29 At this time z 29 With z 23 The number of vectors is the same, but the lengths of the vectors are different. After the linking operation in the vector dimension direction, the two vector sequences are combined into a vector sequence with a length of The vector sequence z with dimension 768 30 .z 30 After linear transformation, the vector dimension is reduced to 256, forming a vector sequence z 31 .z 31Then processed by two-level SwinTransformer modules, the number of The dimension of the new vector sequence z is 256 35 .z 35 After z 28 The number of vectors in the new sequence is doubled, and the length of the vector is gradually reduced, and finally restored to the number of The dimension of the vector sequence z is 64 49 . Then, the vector sequence z 49 Directly rearrange to 64 channels The feature map of the image is directly linked to the feature map of the input image after one-level convolution operation after two-level bilinear upsampling, and then passes through the output layer of 1*1 convolution and Sigmoid to obtain the final segmentation result.

[0117] In order to enable the segmentation network model to process the slice image more accurately, the segmentation network model adopts the following training method.

[0118] The training method of the segmentation network model is:

[0119] Acquire slice images with applicator or interpolation needle I i , and the annotations Y on these slice images i A sample consists of an image and an annotation (I i ,Y i ), where i represents the number of the training sample.

[0120] The actual output of the segmentation network model is

[0121] The training process of the segmentation network model is the same as the conventional method of neural network, so I will not go into details here.

[0122] However, in order to make the output of the segmentation network model as close to the annotation as possible, the loss function is designed as:

[0123]

[0124] The loss function is divided into two parts: sample average BCE loss and sample average DSC loss. Parameters α and β are non-negative numbers. Parameter α is used to control the contribution of BCE loss to the loss function, and parameter β is used to control the contribution of DSC loss to the loss function. By minimizing the loss function, Adam or other optimizers can be used to complete the segmentation network model parameter learning.

[0125] In an embodiment of the present application, when a slice image is input into a trained segmentation network model, a segmentation result and an auxiliary slice image can be obtained at the same time.

[0126] In order to remove the influence of the applicator or interpolation needle in each internal irradiation positioning image, each slice image of each internal irradiation positioning image needs to be processed one by one by the segmentation network model, and each slice image needs to be bound to the corresponding segmentation result and auxiliary slice image.

[0127] Step S500: inputting the corresponding slice image, the auxiliary slice image and the segmentation result into a synthetic network model to generate a plurality of synthetic slice images corresponding to each slice image.

[0128] Reference Figure 4 , wherein the synthetic network model is used to generate synthetic slice images. The synthetic slice image is a slice image without the influence of the applicator and the interpolation needle, which is a slice image in which the original applicator or interpolation needle part of the slice image has a pixel value close to that of the surrounding tissue. After the synthetic slice image is generated, the internal irradiation positioning image composed of the synthetic slice image can replace the original internal irradiation positioning image and perform registration operation with the external irradiation positioning image or other internal irradiation positioning images.

[0129] It can be understood that although the applicator and interpolation needle have been removed from the auxiliary slice image, there is still a difference between the internal irradiation positioning image synthesized from the auxiliary slice image and the real internal irradiation positioning image without the applicator and interpolation needle. If the internal irradiation positioning image composed of the auxiliary slice image is directly used as the processed internal irradiation positioning image for registration, the registration effect may be poor, thereby affecting the accuracy of the superposition of the internal irradiation dose and the external irradiation dose. Therefore, it is necessary to generate a more realistic slice image without the applicator and interpolation needle through a synthetic network model.

[0130] In order to generate more realistic slice images without applicators and interpolation needles, other network structures such as convolutional neural networks, variational autoencoders, and deep fusion networks can be used.

[0131] This application uses a generative adversarial network based on the basic architecture of the SwinTransformer network.

[0132] It can be understood that the Generative Adversarial Networks (GAN) consists of two main parts: the generator and the discriminator. The two parts, the generator and the discriminator, evolve together through mutual competition and cooperation, thereby generating fake data that is closer and closer to the real data. Among them, the task of the generator is to learn the distribution of real data and try to generate new, seemingly real data. It receives random noise as input and then outputs the generated data. During the training process, the goal of the generator is to deceive the discriminator as much as possible so that it misjudges the generated data as real data. The task of the discriminator is to distinguish whether the input data is real or generated by the generator. It receives data as input and outputs a probability value indicating the possibility that the data is real data. During the training process, the goal of the discriminator is to judge the authenticity of the data as accurately as possible, that is, to correctly distinguish between real data and generated data.

[0133] It is understandable that when training the synthetic network model in the application scenario of the present application, there is a lack of real internal irradiation positioning images and corresponding slice images of the same patient without an applicator and an interpolation needle to guide the training. Therefore, it is necessary to set up a discriminator to distinguish the generated synthetic slice images from the slice images of the real internal irradiation positioning images of the same modality without an applicator and an interpolation needle. The generator needs to confuse the discriminator to make the synthetic slice images as close as possible to the slice images of the real internal irradiation positioning images without an applicator and an interpolation needle.

[0134] Specifically, the training steps of the synthetic network model are as follows:

[0135] First, obtain a sample image.

[0136] The sample image includes the corresponding slice image, the segmentation result and the auxiliary slice image. Of course, in other embodiments, if other neural network structures are used to generate synthetic slice images, one or two images can be selected from the slice image, the segmentation image and the auxiliary slice image as the sample image according to actual needs.

[0137] Secondly, the sample image is input into the generator to obtain the synthetic slice image.

[0138] The generator can be implemented using the Transformer network structure. The specific implementation method is the existing technical means in the relevant field, which will not be described in detail here. Of course, when other network structures are selected to implement the function of generating adversarial networks, the corresponding network structure can be selected as the generator according to the selected network structure.

[0139] Then, the synthetic slice image is input into the discriminator as a negative sample image, the corresponding auxiliary slice image is input into the discriminator as a negative sample image, and the image of the relevant part in the external irradiation positioning image without an applicator or an interpolation needle is input into the discriminator as a positive sample image to obtain a discrimination result, and the discrimination loss function and the total loss function are calculated based on the discrimination result.

[0140] It is understandable that because the function of the synthetic network model is implemented based on the discriminator, when training the synthetic network model, it is necessary to pre-train the discriminator so that the discriminator has the ability to distinguish between real slice images without applicators and interpolation needles and synthetic slice images without applicators and interpolation needles, and then train the synthetic network model.

[0141] When training the discriminator, it is necessary to input the synthetic slice image as a negative sample image into the discriminator, input the corresponding auxiliary slice image as a negative sample image into the discriminator, and train the images of other patients' similar parts as positive sample images. The images of other patients' similar parts are the images of related parts. Each sample can be expressed as (X i ,C i ), where X i is the input image, C i is the category label.

[0142] When training a neural network, it is necessary to calculate the corresponding loss function based on the samples, and then adjust the network parameters according to the loss function until the loss function reaches the convergence condition, so that the training can be completed. Therefore, when training the discriminator, it is necessary to first calculate the discriminant loss function, and then adjust the discriminator parameters based on the discriminant loss function, so that the discriminant loss function reaches the preset first convergence condition.

[0143] The first convergence condition is that the discriminant loss function no longer decreases. Of course, the first convergence condition can also be set to the number of training times reaching a certain threshold.

[0144] Specifically, the discriminant loss function can be calculated according to the following formula:

[0145]

[0146] Among them, X i represents the input image to the discriminator, i represents the number of the training sample, Disc(X i ) represents the discriminant result output by the discriminator. The discriminator is used to distinguish whether the input image is a synthetic image or a real image; the synthetic image is a category label with a discriminant value of 0, and the real image is a category label with a discriminant value of 1. i Represents the input image X i The category label of .

[0147] Furthermore, after the training of the discriminator is completed, the network parameters of the discriminator are set to fixed values ​​that make the loss function reach the first convergence condition to continue training the synthetic network model.

[0148] When training the synthetic network model, it is necessary to adjust the parameters of the generator based on the discriminant loss function and the total loss function so that the total loss function reaches a preset second convergence condition.

[0149] The second convergence condition is that the total loss function does not decrease any more. Of course, the first convergence condition can also be set to that the number of training times reaches a certain threshold.

[0150] Specifically, the total loss function can be calculated according to the following formula:

[0151]

[0152] Ω i ={(m,n)Y i (m,n)=0}

[0153] Among them, Ω i Indicates the non-applicator or interpolation needle area, I i represents the slice image, M i represents the auxiliary slice image, represents the segmented image, represents the synthetic slice image, and L1 represents the absolute value of the error.

[0154] It is worth mentioning that the total loss function consists of two parts. The first part is to make the pixel value of the image generated in the non-applicator or interpolation needle area as close to the input internal irradiation image as possible. Therefore, when calculating the total loss function, the L1 loss function is introduced. The L1 loss function is the mean absolute error (MAE), which is used to measure the absolute value of the difference between the predicted value and the true value, and then calculate the average error amplitude. Its calculation range is from 0 to positive infinity. The second part is the opposite of the discriminator loss, that is, the synthetic slice image needs to confuse the discriminator. In order to control the increase in the proportion of the discriminant loss function in the total loss function, the parameter γ is also set. The parameter γ is a positive number. In some specific embodiments, the parameter γ can be set to 20.

[0155] Finally, after obtaining the discriminant loss function and the total loss function, the parameters of the generator and the discriminator can be adjusted based on the discriminant loss function and the total loss function, and the above optimization process is repeated until the set optimization rounds are completed, thereby obtaining a synthetic network model.

[0156] Each slice image is processed using the trained synthetic network model to obtain a synthetic slice image.

[0157] Step S600: Integrate all the synthesized slice images to obtain n synthesized positioning images corresponding one-to-one to the n internal irradiation positioning images.

[0158] Since only the applicator and interpolation needle are removed from the synthetic slice image, the spatial position information represented by the synthetic slice image does not change. Therefore, by combining the synthetic slice images, a three-dimensional synthetic positioning image can also be obtained. The synthetic positioning image can accurately reflect the lesion site without the influence of the applicator and interpolation needle, and can replace the internal irradiation positioning image, which is conducive to accurate registration and calculation of the superposition dose.

[0159] Step S700: Determine one of the n synthetic positioning images as a reference positioning image, and obtain a first deformation field of the external illumination positioning image relative to the reference positioning image through a deformation registration algorithm, and obtain a second deformation field of other synthetic positioning images in the n synthetic positioning images relative to the reference positioning image.

[0160] Step S800: collecting an external radiation dose map according to the first deformation field to obtain an external radiation dose in the reference positioning image coordinate system, and collecting a corresponding internal radiation dose according to the second deformation field to obtain an internal radiation dose in the reference positioning image coordinate system.

[0161] Step S900: Calculating the equivalent biological dose in the reference positioning image coordinate system according to the external irradiation dose and the internal irradiation dose.

[0162] Steps S700 to S900 are a registration method and a method for calculating a superposition dose. This part is a prior art in the related field, so it will not be described in detail here.

[0163] In some specific embodiments, in addition to selecting a synthetic positioning image as a reference positioning image and registering it with other synthetic positioning images and external irradiation positioning images, an external irradiation positioning image can also be selected as a reference positioning image to be registered with each synthetic positioning image. It should also be noted that when selecting a synthetic positioning image, in order to facilitate the calculation of the superposition dose each time, the synthetic positioning image of the first internal irradiation treatment can be selected as the reference positioning image.

[0164] Figure 5 An internal and external irradiation dose superposition system is provided for one embodiment of the present application.

[0165] The internal and external radiation dose superposition method provided in the embodiment of the present application can be well applied to the above-mentioned scenario and can achieve the effect of accurately calculating the internal and external radiation doses.

[0166] like Figure 5The internal and external irradiation dose superposition system shown includes an acquisition module 21, a processing module 22, a segmentation module 23, an auxiliary module 24, a synthesis module 25, a generation module 26, a registration module 27, a collection module 28 and a calculation module 29, wherein: the acquisition module 21 is used to acquire an external irradiation positioning image and n internal irradiation positioning images within a first preset time range, n is an integer greater than 1, and the external irradiation positioning image and the internal irradiation positioning image are both three-dimensional images.

[0167] The processing module 22 is used to process each of the internal irradiation positioning images to obtain multiple slice images.

[0168] The segmentation module 23 is used to input all the slice images into the segmentation network model, identify the target features in the slice images, and perform segmentation to obtain multiple segmentation results corresponding to the slice images, where the target features are source applicators or interpolation needles.

[0169] The auxiliary module 24 is used to fill the image information belonging to the target part in all the slice images according to the specified value, and generate a plurality of auxiliary slice images corresponding to each slice image.

[0170] The synthesis module 25 is used to input the corresponding slice image, the auxiliary slice image and the segmentation result into the synthesis network model to generate a plurality of synthetic slice images corresponding to each slice image; the synthetic slice image is a slice image that makes the slice image of the original applicator or interpolation needle part have a pixel value close to that of the surrounding tissue.

[0171] The generating module 26 is used to integrate all the synthetic slice images to obtain n synthetic positioning images corresponding to the n internal irradiation positioning images one by one, and the synthetic positioning images are three-dimensional images.

[0172] The registration module 27 is used to determine one of the n synthetic positioning images as a reference positioning image, and obtain a first deformation field of the external illumination positioning image relative to the reference positioning image through a deformation registration algorithm, and obtain a second deformation field of other synthetic positioning images in the n synthetic positioning images relative to the reference positioning image.

[0173] The acquisition module 28 is used to acquire an external radiation dose map according to the first deformation field to obtain the external radiation dose in the reference positioning image coordinate system, and to acquire the corresponding internal radiation dose according to the second deformation field to obtain the internal radiation dose in the reference positioning image coordinate system.

[0174] The calculation module 29 is used to calculate the equivalent biological dose in the reference positioning image coordinate system according to the external irradiation dose and the internal irradiation dose.

[0175] This embodiment is only one possible implementation method and does not limit other possible implementation methods.

[0176] Figure 6 A schematic diagram of the structure of an intelligent terminal suitable for implementing the embodiments of the present application is shown.

[0177] like Figure 6 As shown, the intelligent terminal includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part to the random access memory (RAM) 303. In the RAM 303, various programs and data required for system operation are also stored. The CPU 301, ROM 302 and RAM 303 are connected to each other through a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0178] The following components are connected to the I / O interface 305: an input section 306 including a keyboard, a mouse, etc.; an output section 307 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed, so that a computer program read therefrom is installed into the storage section 308 as needed.

[0179] In particular, according to an embodiment of the present application, the above reference flow chart Figure 1 The described process can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a machine-readable medium, and the computer program contains program code for executing the method shown in the flow chart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 309, and / or installed from the removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, the above-mentioned functions defined in the system of the present application are executed.

[0180] It should be noted that the computer-readable medium shown in the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0181] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the aforementioned module, a program segment or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0182] The units or modules involved in the embodiments described in the present application may be implemented by software or by hardware. The units or modules described may also be set in a processor, for example, may be described as: a processor includes: an acquisition module 21, a processing module 22, a segmentation module 23, an auxiliary module 24, a synthesis module 25, a generation module 26, a registration module 27, an acquisition module 28 and a calculation module 29. The names of these units or modules do not, in some cases, constitute limitations on the units or modules themselves. For example, the acquisition module 21 may also be described as "a module for acquiring external irradiation positioning images and n internal irradiation positioning images within a first preset time range".

[0183] As another aspect, the present application further provides a computer-readable storage medium, which may be included in the smart terminal described in the above embodiment; or may exist independently and not be assembled into the smart terminal. The above computer-readable storage medium stores one or more programs, and when the above programs are used by one or more processors to execute the internal and external irradiation dose superposition method described in the present application.

[0184] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of application involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the aforementioned application concept. For example, the above features are replaced with (but not limited to) technical features with similar functions applied in the present application.

Claims

1. A method for superimposing internal and external irradiation doses, characterized in that: The method comprises: Acquire an external irradiation positioning image and n internal irradiation positioning images within a first preset time range, where n is an integer greater than 1, and both the external irradiation positioning image and the internal irradiation positioning image are three-dimensional images; Processing each of the internal irradiation positioning images to obtain a plurality of slice images; Inputting all the slice images into a segmentation network model, identifying target features in the slice images, and performing segmentation to obtain a plurality of segmentation results corresponding to the slice images, wherein the target features are applicators or interpolation needles; Filling the image information of the target part in all the slice images according to the specified value to generate a plurality of auxiliary slice images corresponding to each slice image; Inputting the corresponding slice image, the auxiliary slice image and the segmentation result into a synthetic network model to generate a plurality of synthetic slice images corresponding to each slice image, wherein the synthetic slice image is a slice image that makes the slice image of the original applicator or interpolation needle part have a pixel value close to that of the surrounding tissue; Integrating all the synthetic slice images to obtain n synthetic positioning images corresponding to the n internal irradiation positioning images, wherein the synthetic positioning images are three-dimensional images; Determine one of the n synthetic positioning images as a reference positioning image, and obtain a first deformation field of the external illumination positioning image relative to the reference positioning image through a deformation registration algorithm, and obtain a second deformation field of other synthetic positioning images in the n synthetic positioning images relative to the reference positioning image; An external radiation dose map is collected according to the first deformation field to obtain an external radiation dose in the reference positioning image coordinate system; a corresponding internal radiation dose is collected according to the second deformation field to obtain an internal radiation dose in the reference positioning image coordinate system; and an equivalent biological dose in the reference positioning image coordinate system is calculated according to the external radiation dose and the internal radiation dose.

2. The internal and external irradiation dose superposition method according to claim 1, characterized in that: The segmentation result is a binary image, where the value belonging to the source applicator or the interpolation needle position is 1, and the value not belonging to the source applicator or the interpolation needle position is 0.

3. The internal and external irradiation dose superposition method according to claim 2, characterized in that: The filling process of the image information belonging to the target part in all the slice images according to the specified value includes: filling the image belonging to the target part in the slice images with the minimum pixel value in the area within a preset range from the edge of the target part.

4. The method for superimposing internal and external irradiation doses according to claim 1, characterized in that: Train synthetic network models; The training steps of the synthetic network model include: Acquire a sample image, wherein the sample image includes the corresponding slice image, the segmentation result and the auxiliary slice image; Inputting the sample image into a generator to obtain the synthetic slice image; The synthetic slice image is input as a negative sample image to the discriminator, the corresponding auxiliary slice image is input as a negative sample image to the discriminator, and the image of the relevant part in the external irradiation positioning image without an applicator or an interpolation needle is input as a positive sample image to the discriminator to obtain a discrimination result, and a discrimination loss function and a total loss function are calculated based on the discrimination result; based on the discrimination loss function and the total loss function, the parameters of the generator and the parameters of the discriminator are adjusted to obtain the synthetic network model.

5. The method for superimposing internal and external irradiation doses according to claim 4, characterized in that: The adjusting the parameters of the generator and the parameters of the discriminator based on the discriminant loss function and the total loss function comprises: Setting parameters of the generator to fixed values; Adjusting the parameters of the discriminator based on the discriminant loss function so that the discriminant loss function reaches a preset first convergence condition, where the first convergence condition is that the discriminant loss function no longer decreases; Setting the parameters of the discriminator to fixed values ​​that enable the discriminant loss function to reach the first convergence condition; Adjusting the parameters of the generator based on the discriminant loss function and the total loss function so that the total loss function reaches a preset second convergence condition, where the second convergence condition is that the total loss function no longer decreases; Repeat the above optimization process until the set optimization rounds are completed.

6. The method for superimposing internal and external irradiation doses according to claim 4, characterized in that: The calculation of the discrimination loss function and the total loss function based on the discrimination result comprises: The discriminant loss function is calculated according to the following formula: Among them, X i represents the input image to the discriminator, i represents the number of the training sample, Disc(X i ) represents the discrimination result output by the discriminator, and the discriminator is used to distinguish whether the input image is a synthetic image or a real image; the synthetic image is a category label with a discrimination value of 0, and the real image is a category label with a discrimination value of 1, C i Represents the input image X i The category label of .

7. The method for superimposing internal and external irradiation doses according to claim 6, characterized in that: The calculation of the discrimination loss function and the total loss function based on the discrimination result comprises: The total loss function is calculated according to the following formula: Ω i ={(m,n)|Y i (m,n)=0} Among them, Ω i Indicates the non-applicator or interpolation needle area, I i represents the slice image, M i represents the auxiliary slice image, represents the segmentation result, represents the synthetic slice image, and L1 represents the absolute value of the error.

8. An internal and external irradiation dose superposition system, characterized in that: include, An acquisition module (21) is used to acquire an external irradiation positioning image and n internal irradiation positioning images within a first preset time range, where n is an integer greater than 1, and both the external irradiation positioning image and the internal irradiation positioning image are three-dimensional images; A processing module (22) is used to process each of the internal irradiation positioning images to obtain a plurality of slice images; A segmentation module (23) is used to input all the slice images into a segmentation network model, identify target features in the slice images, and perform segmentation to obtain a plurality of segmentation results corresponding to the slice images, wherein the target features are applicators or interpolation needles; An auxiliary module (24) is used to fill in the image information belonging to the target part in all the slice images according to the specified value, and generate a plurality of auxiliary slice images corresponding to each slice image one by one; A synthesis module (25) is used to input the corresponding slice image, the auxiliary slice image and the segmentation result into a synthesis network model to generate a plurality of synthetic slice images corresponding to each slice image; the synthetic slice image is a slice image that makes the slice image of the original applicator or interpolation needle part have a pixel value close to that of the surrounding tissue; A generating module (26) is used to integrate all the synthetic slice images to obtain n synthetic positioning images corresponding one to one to the n internal irradiation positioning images, wherein the synthetic positioning images are three-dimensional images; A registration module (27) is used to determine one of the n synthetic positioning images as a reference positioning image, and obtain a first deformation field of the external irradiation positioning image relative to the reference positioning image through a deformation registration algorithm, and obtain a second deformation field of other synthetic positioning images in the n synthetic positioning images relative to the reference positioning image; an acquisition module (28), configured to acquire an external radiation dose map according to the first deformation field to obtain an external radiation dose in the reference positioning image coordinate system, and to acquire a corresponding internal radiation dose according to the second deformation field to obtain an internal radiation dose in the reference positioning image coordinate system; as well as, A calculation module (29) is used to calculate the equivalent biological dose in the reference positioning image coordinate system according to the external irradiation dose and the internal irradiation dose.

9. An intelligent terminal, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and execute the internal and external radiation dose superposition method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: A computer program is stored which can be loaded by a processor and executes the internal and external radiation dose superposition method as claimed in any one of claims 1 to 7.

Images