Tumor position monitoring method and system

Abstract

The invention discloses a tumor position monitoring method and system, and is applied to the technical field of medical treatment, and the method comprises the steps: obtaining a CT image of a patient, carrying out the data enhancement of the CT image, and obtaining a three-dimensional image and a body surface depth image; inputting a floating image in the CT image and the body surface depth image into an optical body surface inside and outside correlation model for training to obtain a trained optical body surface inside and outside correlation model; wherein the optical body surface internal and external correlation model is composed of a UNet network and a Kolmogorov-Arnod network; acquiring a body surface depth image of a target patient, inputting the body surface depth image of the target patient into the trained optical body surface internal and external correlation model for processing, and outputting a 3D tumor position corresponding to the body surface depth image of the target patient; and real-time prediction and tracking of the tumor position are realized in a non-invasive manner.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a tumor position monitoring method and system. Background Art

[0002] During radiotherapy, the target area position will change due to physiological factors such as respiratory movement and cardiac movement. In order to ensure that the treatment dose is accurately delivered to the target area and avoid damage to normal tissue, the target area position needs to be monitored in real time. Among the existing image-guided radiotherapy methods: MRI-based image guidance lacks electron density information and is relatively expensive. Ultrasound-based image guidance has limited indications and is difficult to be directly used for image-guided radiotherapy of lung cancer. X-ray image-guided radiotherapy can provide rich internal anatomical structure information, but it is radioactive and cannot frequently irradiate patients. In contrast, optical surface-based image-guided radiotherapy is non-invasive, radiation-free, and can achieve real-time monitoring. However, it lacks internal anatomical information and cannot directly reflect internal tumor changes.

[0003] In order to overcome these defects, the present application proposes a tumor position monitoring method and system to achieve truly non-invasive and radiation-free tumor tracking and improve the safety and effectiveness of radiotherapy. Summary of the invention

[0004] The purpose of this application is to provide a tumor position monitoring method and system, aiming to solve the problem that optical surface image-guided radiotherapy lacks internal anatomical information and is difficult to directly reflect changes in tumor position.

[0005] To achieve the above objectives, this application provides the following technical solutions:

[0006] In a first aspect, the present application provides a method for monitoring tumor location, comprising:

[0007] Acquire a CT image of a patient, perform data enhancement on the CT image, and obtain a three-dimensional image and a body surface depth image;

[0008] Based on the floating image in the CT image and the body surface depth image, an optical body surface internal and external correlation model is input for training to obtain a trained optical body surface internal and external correlation model; wherein the optical body surface internal and external correlation model is composed of a UNet network and a Kolmogorov-Arnold network;

[0009] A body surface depth image of a target patient is acquired, the body surface depth image of the target patient is input into a trained optical body surface internal and external correlation model for processing, and a 3D tumor position corresponding to the body surface depth image of the target patient is output.

[0010] In a second aspect, the present application provides a system for a tumor location monitoring method, comprising:

[0011] Preprocessing module: acquiring a CT image of a patient, performing data enhancement on the CT image, and obtaining a three-dimensional image and a body surface depth image;

[0012] Training module: inputting the optical body surface internal and external correlation model for training based on the floating image in the CT image and the body surface depth image, so as to obtain a trained optical body surface internal and external correlation model; wherein the optical body surface internal and external correlation model is composed of a UNet network and a Kolmogorov-Arnold network;

[0013] Test module: obtain the target patient's body surface depth image, input the target patient's body surface depth image into the trained optical body surface internal and external correlation model for processing, and output the 3D tumor position corresponding to the target patient's body surface depth image.

[0014] In a third aspect, the present application provides a device, comprising a processor and a memory coupled to the processor, wherein the memory stores program instructions for implementing a tumor location monitoring method; and the processor is used to execute the program instructions stored in the memory to implement a tumor location monitoring method.

[0015] In a fourth aspect, the present application provides a storage medium storing program instructions executable by a processor, wherein the program instructions are used to execute a tumor location monitoring method.

[0016] The present application provides a tumor location monitoring method and system, which has the following beneficial effects:

[0017] (1) The optical body surface internal and external correlation model proposed in this application obtains the optical depth image of the patient's body surface, avoiding the disadvantage of the traditional method of frequently using radioactive means such as X-rays for monitoring, significantly reducing the patient's radiation exposure, and has the advantages of being non-invasive and radiation-free;

[0018] (2) Combining UNet with Kolmogorov-Arnold network (KAN) fully captures the nonlinear correlation between the surface and internal tumor movement, achieving more accurate tumor location prediction and tracking. Compared with traditional methods with limited markers or linear dimensionality reduction models, the optical surface internal and external correlation model significantly improves tracking accuracy, ensuring the accuracy and safety of radiotherapy.

[0019] (3) Through the 2D / 3D registration mechanism, the body surface movement is mapped to the position change of the internal tumor, realizing a non-invasive tumor tracking method; it not only reduces the dependence on markers in the body, reduces the patient's physical burden and tracking instability, but also improves the real-time and personalization level of radiotherapy. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1A schematic diagram of a flow chart of a tumor location monitoring method according to Example 1 of the present application;

[0021] Figure 2 A schematic diagram of a tumor location monitoring method according to Example 1 of the present application;

[0022] Figure 3 This is a schematic diagram of the structure of the optical body surface internal and external correlation model of Example 1 of the present application;

[0023] Figure 4 This is a schematic diagram of the tumor registration result of the patient data from the end of exhalation to the end of inspiration in Example 1 of the present application;

[0024] Figure 5 A schematic diagram of a tumor motion curve in the X direction of the tumor mass center of a patient in Example 1 of the present application;

[0025] Figure 6 A schematic diagram of a tumor motion curve in the Y direction of the tumor mass center of a patient in Example 1 of the present application;

[0026] Figure 7 A schematic diagram of a tumor motion curve in the Z direction of the tumor mass center of the patient in Example 1 of the present application;

[0027] Figure 8 This is a schematic diagram of the tumor registration result of the digital phantom from the end of exhalation to the end of inspiration in Example 1 of the present application;

[0028] Fig. 9 A schematic diagram of a tumor motion curve in the X direction of the tumor mass center of the digital phantom in Example 1 of the present application;

[0029] Fig.10 A schematic diagram of a tumor motion curve in the Y direction of the tumor mass center of the digital phantom in Example 1 of the present application;

[0030] Fig.11 A schematic diagram of a tumor motion curve in the Z direction of the tumor mass center of the digital phantom in Example 1 of the present application;

[0031] Fig.12 This is a schematic diagram of the structure of a tumor location monitoring system according to Example 2 of the present application;

[0032] Fig.13 This is a schematic diagram of the device structure of Example 3 of the present application;

[0033] Fig.14 This is a schematic diagram of the storage medium structure of Example 4 of the present application. DETAILED DESCRIPTION

[0034] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0035] The following is an analysis of the solutions in the prior art in combination with related technologies.

[0036] As a new non-invasive, radiation-free real-time monitoring method, surface-guided radiotherapy has attracted much attention. Due to the expansion and contraction of the chest and the involvement of a series of tissues in the body when the human body breathes, respiratory movement will cause the joint movement of the chest and abdominal surface and the tumor in the body. Therefore, many researchers have established internal and external correlation models based on optical surface. For example, Torshabi et al. studied the association modeling of 3 external LED surface markers and tumors in the body based on fuzzy logic. However, the algorithm is sensitive to the initial input data; Iizuka et al. determined the internal and external correlation model by monitoring 4 infrared markers and gold markers implanted in the tumor; Dou Meng et al. of Soochow University used three external markers and adopted the idea of ​​probability optimization to establish a respiratory movement association model. However, it is difficult for limited markers to accurately reflect the movement law of tumors caused by respiratory movement. In addition, for the movement of tumors in the body, the above studies were all obtained by implanting metal markers, which did not achieve true non-invasiveness. Moreover, as the respiratory movement progresses, the markers may migrate or even disappear, reducing the safety and accuracy of radiotherapy.

[0037] In addition, many researchers have used deep learning-based models to establish tumor tracking models based on specific patients, and used principal component analysis (PCA) to reduce dimensions, reduce computational complexity, and improve model efficiency. Unfortunately, PCA is a linear dimensionality reduction model, and the human anatomical structure is complex, tissues vary greatly, and the respiratory motion of the surface and target areas inside the body has a complex nonlinear relationship. Therefore, based on PCA as a linear dimensionality reduction method, it is difficult to accurately describe the nonlinear motion association between the surface and the tumor inside the body, and the accuracy is limited.

[0038] In summary, the shortcomings of existing technologies include: Limited number of markers: Existing methods mostly use a small number of body surface markers, such as 3-4 LEDs or infrared dots, to establish the association between the body surface and the tumor in the body, which makes it difficult to fully capture the complex nonlinear motion caused by breathing, resulting in insufficient model accuracy. Reliance on invasive markers: Many existing solutions rely on implanted metal markers to locate the location of tumors in the body. Although the accuracy of the association is increased, the markers may migrate or disappear with respiratory movement, which not only increases the discomfort and risk of patients, but also reduces the safety and stability of radiotherapy. Limitations of linear dimensionality reduction models: Some deep learning-based body surface-tumor association models use linear dimensionality reduction methods such as principal component analysis (PCA) to simplify calculations, but due to the complex structure of the human body and the nonlinear characteristics of the motion caused by breathing, linear dimensionality reduction is difficult to accurately describe the relationship between the body surface and the tumor position, limiting the tracking accuracy of the model. Radiation exposure: Although X-ray-based image guidance can provide rich internal anatomical information, frequent use will increase patient radiation exposure and is not suitable for high-frequency real-time monitoring.

[0039] Therefore, while ensuring non-invasiveness and radiation-free, the present application achieves high-precision, real-time tumor position monitoring by establishing a nonlinear in vivo and in vitro correlation model, thereby improving radiotherapy effects and patient safety.

[0040] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only 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 making creative work are within the scope of protection of this application.

[0041] Example 1

[0042] See also Figure 1 , is a flow chart of a tumor location monitoring method according to Example 1 of the present application; the steps include:

[0043] S1: Acquire a CT image of a patient, perform data enhancement on the CT image, and obtain a three-dimensional image and a body surface depth image.

[0044] In this embodiment, based on the patient's 4DCT image data, a certain phase is selected as a floating image, and other phases are selected as reference images. Through traditional image registration methods, deformation information between different phases is extracted, and thin plate spline interpolation is used to simulate deformation within the respiratory cycle. Random weights are superimposed on the deformation between phases and within phases to generate a large number of mixed deformation fields, thereby generating diversified three-dimensional images covering each expiratory phase of the lungs. This step provides rich training data for the model and improves its adaptability to respiratory motion.

[0045] In the process of data enhancement, the body surface contour or depth information is extracted based on the CT image through image processing technology to generate a body surface depth image. In order to establish an accurate association between the body surface and the internal tumor movement, the body surface depth image needs to be registered with the three-dimensional image. The body surface movement is mapped to the position change of the internal tumor through registration.

[0046] S2: Based on the floating image in the CT image and the body surface depth image, an optical body surface internal-external correlation model is input for training to obtain a trained optical body surface internal-external correlation model; wherein the optical body surface internal-external correlation model is composed of a UNet network and a Kolmogorov-Arnold network.

[0047] In this embodiment, during the model training stage, the floating image and body surface depth image in the CT image will be used as the input of the optical body surface inside-outside correlation model, where the three-dimensional image is used as the gold standard to obtain the loss function for training the optical body surface inside-outside correlation model.

[0048] The residual convolution block is used to extract the features of the input body surface depth image, and the channel layer is used as the third dimension feature to obtain a three-dimensional feature map. The extracted three-dimensional feature map is input into the UNet-KAN network. The UNet network is used to extract multi-scale spatial features from the body surface depth map to help the model better capture the overall and local contextual relationship. The Kolmogorov-Arnold network (KAN) is embedded in the bottleneck part of the encoder-decoder path to enhance the network's nonlinear feature extraction capabilities. Inside the UNet-KAN network, through a 2D / 3D registration mechanism based on deep learning, the body surface motion is mapped to the position change of the internal tumor, achieving accurate association between the body surface and the internal tumor.

[0049] Through the patient's UNet-KAN structure, the model generates a real-time 3D deformation field to describe the dynamic deformation of the patient's body surface at a specific moment. This deformation field reflects the nonlinear relationship between body surface movement and changes in internal tumor position.

[0050] By using the spatial transformation module to apply the generated 3D deformation field to the planned CT image, the 3D image at the current moment and the corresponding target area position can be obtained, thus achieving non-invasive and radiation-free tumor tracking.

[0051] During the entire training process, the model will continuously adjust parameters to minimize the difference between the predicted results and the actual results. Through iterative training and optimization, a trained optical surface internal and external correlation model is finally obtained.

[0052] Specifically, the UNet network is an encoder and decoder structure; the encoder downsamples the input body surface depth image to extract multi-scale spatial features. The decoder jumps and connects to fuse the multi-scale spatial features extracted during the encoding process, and upsamples and restores them to the original resolution to obtain a feature representation containing spatial information of different scales. The local and overall geometric structure in the body surface depth image is captured to help the optical body surface internal and external correlation model better capture the overall and local contextual relationship, especially the detection of subtle structural changes.

[0053] S3: Acquire a body surface depth image of the target patient, input the body surface depth image of the target patient into a trained optical body surface internal and external correlation model for processing, and output a 3D tumor position corresponding to the body surface depth image of the target patient.

[0054] In this embodiment, a 3D deformation field is generated based on a trained optical body surface internal and external correlation model, and the 3D deformation field is the dynamic deformation of the target patient at the current moment; the 3D deformation field is applied to the body surface depth image of the target patient through a spatial transformation module to obtain the 3D image at the current moment and the corresponding 3D tumor position.

[0055] Furthermore, this application was validated using the TCIA4D lung dataset;

[0056] See also Figure 4 , is a schematic diagram of the tumor registration results of the patient data from the end of exhalation to the end of inspiration in Example 1 of the present application; it shows the 3D results of the patient at different phases, and it can be seen that relatively good results were obtained for both the lungs and the tumors.

[0057] See also Figure 5 , 6 , 7 are schematic diagrams of tumor motion curves in X, Y, and Z directions of the tumor centroid of the patient in Example 1 of the present application, respectively; the horizontal line is the gold standard, and the dots are the prediction results based on the optical body surface internal and external correlation model, and the tumor centroid error is within 1 mm.

[0058] This application has also been tested on the XCAT digital body film, see Figure 8 , 9 , 10, 11. According to the 3D result display and tumor centroid motion curve at different phases, it can be seen that the method of this application is highly consistent with the gold standard. Therefore, both the patient and the digital model have obtained good registration results. Therefore, the internal and external correlation model of the optical body surface proposed in this application is feasible, which reduces the radiation exposure of patients and can become a safe and efficient solution for radiotherapy target monitoring.

[0059] In summary, Example 1 of the present application realizes high-precision tumor position prediction and tracking under 2D / 3D registration by combining UNet with Kolmogorov-Arnold network (KAN). First, obtain the patient's three-dimensional image and body surface depth image, extract multi-scale spatial features through the UNet network, and capture the contextual relationship between the body surface and the tumor; use the KAN network to enhance the nonlinear feature extraction capability and establish a nonlinear association model between the body surface and the tumor. The real-time 3D deformation field is output by the optical body surface internal and external association model to describe the dynamic deformation of the body surface, and act on the planned CT image to obtain the 3D image and target area position at the current moment. The method proposed in this application is non-invasive, radiation-free, reduces the burden on patients, improves the effect and safety of radiotherapy, and has broad application prospects.

[0060] Example 2

[0061] See also Fig.12 , is a structural diagram of a tumor location monitoring system according to Example 2 of the present application; the specific contents include:

[0062] Preprocessing module: acquiring a CT image of a patient, performing data enhancement on the CT image, and obtaining a three-dimensional image and a body surface depth image;

[0063] Training module: inputting the optical body surface internal and external correlation model for training based on the floating image in the CT image and the body surface depth image, so as to obtain a trained optical body surface internal and external correlation model; wherein the optical body surface internal and external correlation model is composed of a UNet network and a Kolmogorov-Arnold network;

[0064] Test module: obtain the target patient's body surface depth image, input the target patient's body surface depth image into the trained optical body surface internal and external correlation model for processing, and output the 3D tumor position corresponding to the target patient's body surface depth image.

[0065] Example 3

[0066] See also Fig.13 , is a schematic diagram of the device structure of Embodiment 3 of the present application. The device 50 includes a processor 51 and a memory 52 coupled to the processor 51 .

[0067] The memory 52 stores program instructions for implementing the above-mentioned tumor position monitoring method.

[0068] The processor 51 is used to execute program instructions stored in the memory 52 to implement tumor position monitoring.

[0069] The processor 51 may also be referred to as a CPU (Central Processing Unit).

[0070] The processor 51 may be an integrated circuit chip with signal processing capabilities. The processor 51 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0071] Example 4

[0072] See also Fig.14, which is a schematic diagram of the structure of the storage medium of Example 4 of the present application. The storage medium of the embodiment of the present application stores a program file 61 that can implement all the above methods, wherein the program file 61 can be stored in the above storage medium in the form of a software product, including a number of instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or a computer, a server, a mobile phone, a tablet, and other devices.

[0073] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0074] The above description is only a preferred embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

[0075] Although the embodiments of the present application have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present application, and that the scope of the present application is defined by the appended claims and their equivalents.

[0076] Of course, the present invention may have many other implementations. Based on this implementation, other implementations obtained by ordinary technicians in this field without any creative work are all within the scope of protection of the present invention.

Claims

1. A method for monitoring tumor location, characterized in that: include: Acquire a CT image of a patient, perform data enhancement on the CT image, and obtain a three-dimensional image and a body surface depth image; Based on the floating image in the CT image and the body surface depth image, an optical body surface internal and external correlation model is input for training to obtain a trained optical body surface internal and external correlation model; wherein the optical body surface internal and external correlation model is composed of a UNet network and a Kolmogorov-Arnold network; A body surface depth image of a target patient is acquired, the body surface depth image of the target patient is input into a trained optical body surface internal and external correlation model for processing, and a 3D tumor position corresponding to the body surface depth image of the target patient is output.

2. A tumor location monitoring method according to claim 1, characterized in that: The steps of acquiring a CT image of a patient, performing data enhancement on the CT image, and obtaining a three-dimensional image and a body surface depth image specifically include the following steps: Extracting a three-dimensional image from the CT image by image processing technology; A body surface depth image is extracted from the CT image using a depth image acquisition method.

3. A tumor location monitoring method according to claim 1, characterized in that: The optical body surface internal and external correlation model is inputted based on the floating image in the CT image and the body surface depth image for training to obtain a trained optical body surface internal and external correlation model; wherein the optical body surface internal and external correlation model is composed of a UNet network and a Kolmogorov-Arnold network, specifically including the following steps: The UNet network is an encoder and decoder structure; the input body surface depth image is downsampled by the encoder to extract multi-scale spatial features; Through the decoder jump connection, the multi-scale spatial features extracted in the encoding process are fused and upsampled to the original resolution to obtain a feature representation containing spatial information of different scales.

4. A tumor location monitoring method according to claim 3, characterized in that: The Kolmogorov-Arnold network is combined with the UNet network, and the Kolmogorov-Arnold network is embedded in the bottleneck part of the encoder and decoder paths of the UNet network to perform nonlinear feature extraction.

5. A tumor location monitoring method according to claim 1, characterized in that: The step of obtaining a body surface depth image of a target patient, inputting the body surface depth image of the target patient into a trained optical body surface internal and external correlation model for processing, and outputting a tumor position corresponding to the body surface depth image of the target patient specifically includes the following steps: Generate a 3D deformation field based on the trained optical body surface internal and external correlation model, where the 3D deformation field is the dynamic deformation of the target patient at the current moment; By applying the 3D deformation field to the target patient's body surface depth image through the spatial transformation module, a 3D image at the current moment and the corresponding 3D tumor position are obtained.

6. A tumor location monitoring method according to claim 1, characterized in that: The three-dimensional image is used as a gold standard to obtain a loss function, which is used to train an optical body surface internal and external correlation model.

7. A system for monitoring tumor location according to claim 1, characterized in that: include: Preprocessing module: acquiring a CT image of a patient, performing data enhancement on the CT image, and obtaining a three-dimensional image and a body surface depth image; Training module: inputting the optical body surface internal and external correlation model for training based on the floating image in the CT image and the body surface depth image, so as to obtain a trained optical body surface internal and external correlation model; wherein the optical body surface internal and external correlation model is composed of a UNet network and a Kolmogorov-Arnold network; Test module: obtain the target patient's body surface depth image, input the target patient's body surface depth image into the trained optical body surface internal and external correlation model for processing, and output the 3D tumor position corresponding to the target patient's body surface depth image.

8. A device, characterized in that The device includes a processor and a memory coupled to the processor, wherein the memory stores program instructions for implementing a tumor position monitoring method as described in any one of claims 1-6; and the processor is used to execute the program instructions stored in the memory to implement tumor position monitoring.

9. A storage medium, characterized in that: Program instructions executable by a processor are stored, and the program instructions are used to execute a tumor position monitoring method according to any one of claims 1 to 6.

Images