Method and device for estimating organ movement change in radiotherapy process

Abstract

The invention discloses a method and a device for estimating organ motion change in a radiotherapy process, relates to the technical field of medical radiotherapy, and solves the problem that future motion conditions of organs in the radiotherapy process cannot be accurately estimated in the prior art. The method comprises the following steps: extracting deformation field data of a historical CT (Computed Tomography) image; the method comprises the following steps: acquiring deformation field data, taking the deformation field data as input, predicting deformation field data at the current moment in a pre-trained image deformation field prediction model, generating a CT image at the next moment, and evaluating the displacement condition by comparing the similarity between the CT images at the current moment and the CT image at the next moment. According to the method, the change between the images can be captured more accurately, so that the prediction precision is improved, and the possible position and shape of the organ at the future moment can be predicted by analyzing the deformation field data and the pre-trained image deformation field prediction model.

Description

Technical Field

[0001] The present application relates to the technical field of medical radiotherapy, and in particular to a method and device for estimating changes in organ motion during radiotherapy. Background Art

[0002] During radiotherapy, human body movement or physiological changes may cause organ displacement, which in turn affects the effect of radiotherapy. For example, the filling of the bladder causes the position of organs to change, which increases the risk of normal tissue damage and has always been a clinical concern. In order to reduce the impact of organ movement on radiotherapy and ensure that the radiation dose is accurately delivered to the tumor tissue while minimizing damage to surrounding healthy tissues, the commonly used technologies are:

[0003] Respiratory gating technology: By monitoring the patient's breathing pattern, radiation therapy is only delivered during specific phases of the respiratory cycle (such as exhalation or inspiration) to reduce organ movement caused by respiratory motion.

[0004] Positioning technology: Use special fixation devices (such as vacuum pads, thermoplastic films, etc.) to keep the patient's position unchanged during treatment. This helps reduce organ movement caused by patient movement.

[0005] Real-time tracking systems: Use external sensors or internal markers to track the position of an organ in real time and adjust the radiation therapy device in real time to compensate for the movement of the organ.

[0006] Image-guided radiation therapy (IGRT): Uses imaging techniques (such as CT, MRI, or X-rays) during treatment to obtain real-time information about the position of organs, allowing precise adjustment of radiation therapy equipment to ensure accurate delivery of radiation doses.

[0007] Among the above-mentioned existing technologies, the first two are open-loop control methods before radiotherapy, and their accuracy cannot be guaranteed; the latter two are real-time monitoring methods during radiotherapy, which cannot predict future changes in organs and have high requirements for real-time adjustment of radiotherapy equipment. Summary of the invention

[0008] The purpose of this application is to overcome the problem in the prior art that the future movement of an organ during radiotherapy cannot be accurately predicted, and to provide a method and device for predicting changes in organ movement during radiotherapy.

[0009] In a first aspect, a method for estimating organ motion changes during radiotherapy is provided, comprising:

[0010] Get CT images of historical moments;

[0011] Extract deformation field data of historical CT images;

[0012] Inputting the deformation field data of the historical CT image into a pre-trained image deformation field prediction model to predict the deformation field data at the current moment;

[0013] Calculate the CT image at the next moment according to the deformation field data at the current moment and the CT image at the current moment;

[0014] The displacement is evaluated by comparing the similarity of the CT images at the current moment and the next moment.

[0015] In some possible implementations, the historical CT images are CT images taken from time 0 to time t after the patient is fixed in position using a position fixation device, wherein time 0 is the initial time and time t is the current time.

[0016] In some possible implementations, extracting deformation field data of historical CT images includes:

[0017] Constructing an image registration network, wherein the image registration network includes an encoder and a decoder, the encoder is used to extract image features of an input CT image, and the decoder is used to decode and reconstruct the image features output by the encoder into deformation field data;

[0018] The CT image I 0 and CT images I 1 Input into the image registration network to obtain the deformation field data f 0 , the CT image I 1 and CT images I 2 Input into the image registration network to obtain the deformation field data f 1 , and so on, the CT image I t-1 and CT images I t Input into the image registration network to obtain the deformation field data f t-1 , until the deformation field data f is obtained 0 to f t-1 , where CT image I 0 is the CT image taken at the initial time 0, CT image I t is the CT image taken at the current time t, CT image I t-1 It is the CT image taken at the previous time t-1 before the current time t.

[0019] In some possible implementations, the encoder is composed of a stack of blocks of a Swin-transformer structure to extract and fuse image information at two moments; the decoder is composed of a deconvolution block in a conventional convolutional neural network to reconstruct the fused feature information into deformation field data of the images at two moments.

[0020] In some possible implementations, the pre-trained image deformation field prediction model uses a CA_RNN structure network, which can directly input two-dimensional matrix data and output two-dimensional matrix data, wherein the input gate in each unit processes the data as follows: the hidden state two-dimensional data h transmitted at the previous moment is t-1 And the two-dimensional deformation field data x input at this moment t All of them are extracted through convolution blocks, then spliced ​​between channels, and then fused through the channel attention mechanism to obtain the hidden state two-dimensional data h t ,h t Used to pass to the next unit or the input gate of this unit.

[0021] In some possible implementations, the calculation formula for the CT image at a moment is:

[0022] I t+1 =I t *f t

[0023] Among them, I t+1 is the CT image at the next time t+1, I t is the CT image at the current time t, f t is the predicted deformation field data at the current moment, and * is the matrix dot product.

[0024] In some possible implementations, during the process of evaluating the displacement, a structural similarity index is used to measure the similarity between images, and if the evaluated displacement is greater than a preset threshold, an early warning prompt is issued.

[0025] In a second aspect, a device for estimating organ motion changes during radiotherapy is provided, comprising:

[0026] An image acquisition module, used to acquire CT images of historical moments;

[0027] Image deformation field extraction module, used to extract deformation field data of historical CT images;

[0028] An image deformation field prediction module is used to input the deformation field data of the historical CT image into a pre-trained image deformation field prediction model to predict the deformation field data at the current moment;

[0029] An image generation module is used to calculate a CT image at the next moment based on the deformation field data at the current moment and the CT image at the current moment;

[0030] The motion displacement evaluation module is used to evaluate the displacement by comparing the similarity between the CT images at the current moment and the next moment.

[0031] In a third aspect, a computer-readable storage medium is provided, wherein the computer-readable medium stores a program code for execution by a device, wherein the program code includes steps for executing the method in any one of the implementations of the first aspect described above.

[0032] In a fourth aspect, an electronic device is provided, comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements a method as in any one of the implementations in the first aspect above.

[0033] This application has the following beneficial effects:

[0034] 1. This application uses deformation field as an intermediate medium, which can more accurately capture the changes between images, thereby improving the accuracy of prediction. By analyzing the deformation field data, the pre-trained image deformation field prediction model can predict the possible position and shape of the organ at future moments, thereby providing a more accurate reference for treatment.

[0035] 2. This application adopts a new CA_RNN (convolutional recurrent neural network) structure, which enables the original RNN unit to effectively process two-dimensional input data, thereby better adapting to the needs of medical image processing. Through this improvement, the pre-trained image deformation field prediction model can more accurately capture and analyze the spatiotemporal characteristics in the data, and can more accurately predict the future movement of organs during radiotherapy. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application.

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0038] Figure 1 is a flow chart of a method for estimating organ motion changes during radiotherapy according to Example 1 of the present application;

[0039] Figure 2 is a schematic diagram of extracting deformation field data in a method for estimating organ motion changes during radiotherapy according to Example 1 of the present application;

[0040] Figure 3 This is a schematic diagram of the original RNN unit;

[0041] Figure 4 This is a schematic diagram of the calculation of the input gate of the CA_RNN unit in the method for estimating organ motion changes during radiotherapy in Example 1 of the present application;

[0042] Figure 5 is an overall flow chart of the method for estimating organ motion changes during radiotherapy according to Example 1 of the present application;

[0043] Figure 6 is an expanded flow chart of the method for estimating organ motion changes during radiotherapy according to Example 1 of the present application;

[0044] Figure 7 is a schematic diagram of calculating a CT image at the next moment in the method for estimating organ motion changes during radiotherapy according to Example 1 of the present application;

[0045] Figure 8 is a structural block diagram of a device for estimating organ motion changes during radiotherapy according to Embodiment 2 of the present application;

[0046] Fig. 9 It is a schematic diagram of the internal structure of the electronic device of Example 4 of the present application.

[0047] Reference numerals:

[0048] 100, image acquisition module; 200, image deformation field extraction module; 300, image deformation field prediction module; 400, image generation module; 500, motion displacement evaluation module. DETAILED DESCRIPTION

[0049] 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.

[0050] Example 1

[0051] like Figure 1 As shown, a method for estimating organ motion changes during radiotherapy according to Example 1 of the present application includes:

[0052] S100, obtaining a CT image of a historical moment;

[0053] It should be noted that CT (Computed Tomography) is an electronic computer tomography scan, which uses an X-ray beam to scan a layer of a certain thickness of the human body examination part. The detector receives the X-rays that pass through the human tissue in different directions on the layer, and inputs them into the computer through analog / digital conversion. After computer processing, a digital matrix of tissue attenuation coefficients of the scanned layer is obtained, and then the values ​​in the matrix are converted from digital to analog and displayed on the fluorescent screen with different gray levels of black and white, which constitutes a CT image. The CT image can be obtained from a CT machine. In order to minimize the amplitude of organ movement during radiotherapy, the historical CT image is a CT image taken from time 0 to time t after the patient is fixed in position using a body fixation device, which can help reduce organ movement caused by patient movement, thereby reducing the impact of organ movement on radiotherapy, ensuring that the radiation dose is accurately delivered to the tumor tissue, and minimizing damage to surrounding healthy tissues. Time 0 is the initial time and time t is the current time.

[0054] S200, extracting deformation field data of historical CT images;

[0055] In this embodiment, an image registration network based on an encoder-decoder architecture is used to extract deformation field data, such as Figure 2 As shown in the figure, the image configuration network includes an encoder and a decoder. The encoder is used to extract the image features of the input CT image, and the decoder is used to decode and reconstruct the image features output by the encoder into deformation field data. When extracting deformation field data, the image at the current moment t and the image at the next moment t+1 are first channel-joined and sent to the encoder to extract the image features, and then the above features are decoded and reconstructed into the image deformation field by the decoder. Among them, the encoder is composed of a block stack of the Swin-transformer structure to realize the extraction and fusion of image information at two moments; the decoder is composed of a deconvolution block in a conventional convolutional neural network, and the fused feature information is reconstructed into deformation field data related to the images at two moments.

[0056] Specifically, firstly, the CT image I 0 and CT images I 1 Input into the image registration network to obtain the deformation field data f 0 , the CT image I 1 and CT images I 2 Input into the image registration network to obtain the deformation field data f 1 , and so on, the CT image I t-1 and CT images I t Input into the image registration network to obtain the deformation field data f t-1 , until the deformation field data f is obtained 0 to ft-1 , where CT image I 0 is the CT image taken at the initial time 0, CT image I t is the CT image taken at the current time t, CT image I t-1 It is the CT image taken at the previous time t-1 before the current time t.

[0057] It should be noted that Swin Transformer is a deep learning model architecture for computer vision tasks, proposed by the research team of Microsoft Research Asia in 2021. It is based on the Transformer architecture, but unlike the traditional Transformer, Swin Transformer has a unique hierarchical feature map and window self-attention mechanism, which makes this structure particularly suitable for processing image data to extract and fuse contextual feature information of different scales, and performs well in the field of computer vision (such as image classification, object detection and semantic segmentation). The deconvolution network is a way to reconstruct low-dimensional feature maps in the image model of the encoder-decoder architecture.

[0058] S300, inputting the deformation field data of the historical CT image into the pre-trained image deformation field prediction model to predict the deformation field data at the current moment;

[0059] In time series data analysis, recurrent neural networks (RNN) and long short-term memory networks (LSTM) are commonly used structures. RNN has a simple structure and is suitable for processing sequences of different lengths, but long sequences may cause gradient problems. LSTM solves this problem through a gating mechanism and is suitable for learning long-term dependencies. However, in specific scenarios, such as the fields involved in this application, only recent data needs to be paid attention to, so the gradient problem of RNN is not significant. This makes RNN more advantageous in simplifying the structure and improving computational efficiency, and training is simpler. Therefore, this application adopts the CA_RNN structure based on RNN improvement for spatiotemporal sequence prediction.

[0060] like Figure 3 As shown in the figure, each unit input gate and output gate in the original RNN network adopts a fully connected linear layer structure, as shown in the following formula:

[0061]

[0062] Among them, h is the hidden state, x is the input, and the bias term b h , b y is a one-dimensional vector, W hh , W xh , W hy is the linear layer weight matrix, f i 、f 0 is the nonlinear activation function of the input gate and output gate.

[0063] The original RNN unit is only suitable for processing single-dimensional time series data, not for processing spatiotemporal sequence image type data. In this application, the input gate and output gate in the RNN network are improved to obtain a CA_RNN structure network, so that it can directly input two-dimensional matrix data and output two-dimensional data. Among them, the processing flow of the input gate in each unit for data is as follows Figure 4 As shown in the figure, the CA_RNN structure network can directly input two-dimensional matrix data and output two-dimensional matrix data. The processing process of the input gate in each unit for data is as follows: the hidden state two-dimensional data h transmitted at the previous moment is t-1 And the two-dimensional deformation field data x input at this moment t All of them are extracted through convolution blocks, then spliced ​​between channels, and then fused through the channel attention mechanism to obtain the hidden state two-dimensional data h t ,h t It is used to pass to the next unit or the input gate of this unit. Its calculation process can be expressed by the following formula:

[0064]

[0065] Among them, Convb() represents the convolution operation, Attention() represents the operation through the channel attention mechanism, and Conv1*1() represents the convolution operation with a convolution kernel size of 1*1.

[0066] The same CA_RNN unit output gate converts the original linear layer operation in the RNN into Figure 4 The convolution operation in is adapted to the requirements of this embodiment.

[0067] like Figure 5 and Figure 6 As shown in the figure, the "zipper" structure is adopted, which can be expanded according to the length of the input image sequence, without limiting the length of the input sequence, and the actual application process is flexible. Among them, the deformation field extraction module and the deformation field prediction module share the model weights in each time step. First, the deformation field data is extracted according to the current moment image and the next moment image to obtain the deformation field matrix data at the current moment, such as Figure 5 and Figure 6 The deformation field matrix f in 1 In the pre-trained image deformation field prediction model, the deformation field matrix data f at the current moment 1 As input, the hidden state data matrix information h transmitted from the previous moment is integrated 0 , calculate the pre-trained image deformation field prediction model passed to the next moment. In the last moment of the image sequence, the hidden state data matrix information h 3After the convolution operation, the deformation field matrix information f at the next moment is predicted 4 Finally, f 4 and I 4 The image to be predicted is obtained after matrix dot multiplication. 5 .

[0068] In the above end-to-end neural network, the similarity between the image predicted by the model and the label image is used as the training loss to learn the weight parameters in the deformation field extraction module and the prediction module. The specific loss function is shown in the following formula: f Neutralization I m is the image of the next moment and the current moment in the time series corresponding to the fixed image and the moving image, and f is the deformation field corresponding to the current moment image. The loss function consists of the similarity loss L sim and smoothness L smooth It consists of two parts, L sim The loss is expressed as the root mean square error between the fixed image and the moving image*deformation field, L smooth The loss function is to limit the smoothness of the deformation field. The deformation in the time series should be relatively small and smooth, which is represented by the first norm of the deformation field and the second norm of its gradient.

[0069]

[0070]

[0071]

[0072] In addition, in the model training process, a self-supervised training method is adopted, which has low requirements on training data. 1 ,I 2 ,I 3 ,I 4 ,I 5 ), first (I 1 ,I 2 ) Prediction I 3 As a training sample, then (I 1 ,I 2 ,I 3 ) prediction as training samples, and finally by (I 1 ,I 2 ,I 3 ,I 4 ) prediction, and so on.

[0073] S400, calculating a CT image at the next moment according to the deformation field data at the current moment and the CT image at the current moment;

[0074] In this step, the image at the next moment is calculated based on the deformation field data predicted by the model and the image data at the current moment. Figure 7 As shown, the calculation process involves multiplying the corresponding elements in the two matrices element by element to generate the image data for the next moment. The calculation process is shown in the following formula:

[0075] I t+1 =I t *f t

[0076] Among them, I t+1 is the CT image at the next time t+1, I t is the CT image at the current time t, f t is the predicted deformation field data at the current moment, and * is the matrix dot product.

[0077] S500 , evaluating the displacement by comparing the similarity between the CT images at the current moment and the next moment.

[0078] In this step, the displacement is evaluated by comparing the similarity between the initial organ image and the predicted image. If a large displacement is detected, the system will trigger an early warning prompt. In this embodiment, the structural similarity (SSIM) indicator is used to measure the similarity between images. The SSIM indicator is a comprehensive image quality assessment method that comprehensively considers the three key visual dimensions of image brightness, contrast and structure to accurately evaluate and compare the similarity between images. Through this multi-dimensional comparison, SSIM can provide more accurate and reliable image similarity assessment results, thereby effectively monitoring and warning of abnormal displacement.

[0079] Exemplarily, in the radiotherapy treatment of bladder cancer, after determining the target area and the organ-at-risk area before treatment, the dose distribution is calculated, and then the radiotherapy equipment parameters are determined. However, during the radiotherapy treatment, the bladder may be displaced due to physiological reasons, resulting in the original set equipment parameters being inapplicable, the target area dose being insufficient, and the organ-at-risk dose being too high, resulting in increased toxic and side effects. Using the method of the present application, the image displacement situation of the next moment can be predicted and warned by using the CT image information of the historical moment during the treatment process. Specifically, the deformation field data of the first and second moments can be obtained by extracting the deformation field data by inputting the CT images of the first, second, and third moments. The deformation field data of the third moment can be predicted by the pre-trained image deformation field prediction model, and then the CT image of the fourth moment can be obtained by spatial transformation of the CT image of the third moment + the deformation field. The fourth CT image and the first CT image are calculated for similarity to determine whether there is a displacement risk. If there is, a warning is issued. Otherwise, the CT images of the fifth, sixth, seventh... moments are predicted by analogy to continue to determine whether there is a displacement risk.

[0080] During radiotherapy, the motion monitoring of the patient's organs is crucial to ensure the accuracy and safety of treatment. Traditional real-time monitoring methods can capture the movement of organs instantly during treatment, but cannot predict the future motion state. In order to improve the accuracy of prediction and achieve more effective early warning, this application proposes an innovative end-to-end overall network structure. Use the deformation field as an intermediate medium instead of directly predicting specific images at future moments. The advantage of this method is that the deformation field can more accurately capture the changes between images, thereby improving the accuracy of the prediction. By analyzing the deformation field data, the network can predict the possible position and shape of the organ at future moments, thereby providing a more accurate reference for treatment.

[0081] Secondly, a new type of CA_RNN (convolutional recurrent neural network) structure is innovatively proposed in this application. This structure enables the original RNN unit to effectively process two-dimensional input data, thereby better adapting to the needs of medical image processing. Through this improvement, the spatiotemporal features in the data can be more accurately captured and analyzed, providing a more advanced and effective technical means for patient monitoring during radiotherapy.

[0082] Example 2

[0083] like Figure 8 As shown, a device for estimating organ motion changes during radiotherapy according to Embodiment 2 of the present application includes:

[0084] An image acquisition module 100, used to acquire a CT image at a historical moment;

[0085] An image deformation field extraction module 200 is used to extract deformation field data of historical CT images;

[0086] The image deformation field prediction module 300 is used to input the deformation field data of the historical CT image into the pre-trained image deformation field prediction model to predict the deformation field data at the current moment;

[0087] An image generation module 400 is used to calculate a CT image at the next moment based on the deformation field data at the current moment and the CT image at the current moment;

[0088] The motion displacement evaluation module 500 is used to evaluate the displacement by comparing the similarity between the CT images at the current moment and the next moment.

[0089] It should be noted that other specific implementations of the device for estimating organ motion changes during radiotherapy in this embodiment can refer to the specific implementations of the method for estimating organ motion changes during radiotherapy mentioned above, and will not be repeated here to avoid redundancy.

[0090] Example 3

[0091] A computer-readable storage medium involved in Embodiment 3 of the present application, the computer-readable storage medium stores a program code for execution by a device, the program code including steps for executing a method in any one of the implementations in Embodiment 1 of the present application;

[0092] Among them, the computer-readable storage medium can be a read-only memory (ROM), a static storage device, a dynamic storage device or a random access memory (RAM); the computer-readable storage medium can store program code, and when the program stored in the computer-readable storage medium is executed by the processor, the processor is used to execute the steps of the method in any one of the implementation methods in Example 1 of the present application.

[0093] Example 4

[0094] like Fig. 9 As shown, an electronic device involved in Embodiment 4 of the present application includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor, and when the program or instruction is executed by the processor, a method in any one of the implementations in Embodiment 1 of the present application is implemented;

[0095] Among them, the processor can adopt a general-purpose central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), a graphics processing unit (GPU) or one or more integrated circuits to execute relevant programs to implement the method in any one of the implementation methods in Example 1 of the present application.

[0096] The processor may also be an integrated circuit electronic device with signal processing capability. In the implementation process, each step of the method in any implementation of Embodiment 1 of the present application may be completed by an integrated logic circuit of hardware in the processor or by instructions in software form.

[0097] The above-mentioned processor may also be a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application may be implemented or executed. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application may be directly embodied as being executed by a hardware decoding processor, or may be executed by a combination of hardware and software modules in a decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in a memory, and the processor reads the information in the memory, and completes the functions required to be performed by the unit included in the data processing device of the embodiment of the present application in combination with its hardware, or executes the method in any one of the implementation modes in the embodiment 1 of the present application.

[0098] The above are only preferred specific implementations of the present application; however, the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical solution and its improved ideas of the present application within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for estimating organ motion changes during radiotherapy, characterized in that: include: Get CT images of historical moments; Extract deformation field data of historical CT images; Inputting the deformation field data of the historical CT image into a pre-trained image deformation field prediction model to predict the deformation field data at the current moment; Calculate the CT image at the next moment according to the deformation field data at the current moment and the CT image at the current moment; The displacement is evaluated by comparing the similarity of the CT images at the current moment and the next moment.

2. The method for estimating organ motion changes during radiotherapy according to claim 1, characterized in that: The historical CT images are CT images taken from time 0 to time t after the patient's position is fixed using a position fixing device, wherein time 0 is the initial time and time t is the current time.

3. The method for estimating organ motion changes during radiotherapy according to claim 2, characterized in that: Extract deformation field data from historical CT images, including: Constructing an image registration network, wherein the image registration network includes an encoder and a decoder, the encoder is used to extract image features of an input CT image, and the decoder is used to decode and reconstruct the image features output by the encoder into deformation field data; Input CT images I0 and I1 into the image registration network to obtain deformation field data f0, input CT images I1 and I2 into the image registration network to obtain deformation field data f1, and so on. t-1 and CT images I t Input into the image registration network to obtain the deformation field data f t-1 , until the deformation field data f0 to f are obtained t-1 , where CT image I0 is the CT image taken at the initial time 0, and CT image I t is the CT image taken at the current time t, CT image I t-1 It is the CT image taken at the previous time t-1 before the current time t.

4. The method for estimating organ motion changes during radiotherapy according to claim 3, characterized in that: The encoder is composed of a stack of blocks of a Swin-transformer structure to extract and fuse image information at two moments; the decoder is composed of deconvolution blocks in a conventional convolutional neural network to reconstruct the fused feature information into deformation field data of images at two moments.

5. The method for estimating organ motion changes during radiotherapy according to claim 1, characterized in that: The pre-trained image deformation field prediction model adopts a CA_RNN structure network, which can directly input two-dimensional matrix data and output two-dimensional matrix data. The input gate in each unit processes the data as follows: the hidden state two-dimensional data h transmitted at the previous moment is t-1 And the two-dimensional deformation field data x input at this moment t All of them are extracted through convolution blocks, then spliced ​​between channels, and then fused through the channel attention mechanism to obtain the hidden state two-dimensional data h t ,h t Used to pass to the next unit or the input gate of this unit.

6. The method for estimating organ motion changes during radiotherapy according to claim 1, characterized in that: The calculation formula for the CT image at the next moment is: I t+1 =I t *f t Among them, I t+1 is the CT image at the next time t+1, I t is the CT image at the current time t, f t is the predicted deformation field data at the current moment.

7. The method for estimating organ motion changes during radiotherapy according to claim 1, characterized in that: In the process of evaluating the displacement, the structural similarity index is used to measure the similarity between images. If the evaluated displacement is greater than the preset threshold, an early warning prompt is issued.

8. A device for estimating organ motion changes during radiotherapy, characterized in that: include: An image acquisition module, used to acquire CT images of historical moments; Image deformation field extraction module, used to extract deformation field data of historical CT images; An image deformation field prediction module is used to input the deformation field data of the historical CT image into a pre-trained image deformation field prediction model to predict the deformation field data at the current moment; An image generation module is used to calculate a CT image at the next moment based on the deformation field data at the current moment and the CT image at the current moment; The motion displacement evaluation module is used to evaluate the displacement by comparing the similarity between the CT images at the current moment and the next moment.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program code for execution by a device, wherein the program code includes steps for executing the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: The electronic device comprises a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction implements the method as claimed in any one of claims 1 to 7 when executed by the processor.

Images