Method for precisely positioning and navigating neck

Through CT image segmentation and real-time X-ray image registration, the problem of displacement correction of neck tumors in radiation therapy is solved, the precise positioning and navigation of neck tumors is achieved, and the accuracy and safety of treatment are improved.

CN118903706APending Publication Date: 2024-11-08JIAXING RES INST ZHEJIANG UNIV +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411037226.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-31
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

The existing radiation therapy technology has insufficient accuracy in the positioning and navigation of neck tumors, especially because the displacement caused by the composite motion pattern of the cervical spine is difficult to accurately correct, resulting in insufficient concentration of radiation dose, affecting the treatment effect.

Method used

Through image segmentation and extraction of preoperative CT images, cervical vertebrae and non-cervical vertebrae tissues are distinguished, and real-time X-ray images are used for registration to estimate the spatial transformation parameters of cervical vertebrae, and these parameters are used to initially estimate the displacement parameters of non-rigid body tissues. Iterative optimization to obtain the accurate displacement of non-rigid body tissue in the neck to achieve three-dimensional structural registration of the neck.

Benefits of technology

The use of sparse angles and low-dose X-ray imaging achieves precise positioning and navigation of neck tumors in radiation therapy, improves the accuracy of treatment, and reduces the radiation dose received by patients.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118903706A_ABST
    Figure CN118903706A_ABST
Patent Text Reader

Abstract

The invention discloses a method for precisely positioning and navigating a neck, and provides a technical problem that precise positioning and navigating of a neck tumor in radiotherapy are realized by using a composite model of rigid body motion and non-rigid body motion and adopting sparse-angle and low-dose X-ray imaging. The existing method of the composite motion model in the three-dimensional reconstruction of the human body motion mainly aims at estimating the rigid body and non-rigid body characteristics of the point cloud on the surface of the human body based on a visual camera such as Kinect and further estimating the change of the posture of the human body and the displacement of the motion. In the field of radiotherapy related to the method, displacement inside three-dimensional human tissue, not only the surface characteristics of the human body, needs to be estimated, and the method for positioning rigid body and non-rigid body composite tissue inside the human body by adopting a low-dose and sparse-angle X-ray image is an unsolved technical difficulty.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of radiotherapy, and in particular to a method for accurate positioning and navigation of a neck. Background Art

[0002] With the rapid development of medical technology, radiotherapy has become a very important means of treating cancer and some other diseases. Radiotherapy uses high-energy rays or other forms of radiation to kill or control the growth of abnormal cells while protecting the surrounding normal tissues to the greatest extent possible. To improve the accuracy of treatment, the radiotherapy system needs to accurately radiate the radiation dose to the tumor tissue while minimizing the damage to normal tissues.

[0003] Precise positioning and navigation are key technologies to ensure continuous and accurate radiation of tumors during radiotherapy. Positioning and navigation can use medical images (such as CT, MRI or PET scans) to determine the location, volume and shape of the tumor, and compare it with the treatment plan to achieve the positioning and real-time tracking of the treatment target.

[0004] The existing positioning and guiding methods are mainly divided into the following types, and there are corresponding problems:

[0005] 1.1.1. Fixtures

[0006] A head frame is a fixture used in gamma knife radiotherapy equipment for the head. Patients need to wear a specially designed head frame to ensure head stability. The head frame is usually fastened to the head with screws or clips to achieve precise positioning.

[0007] The main disadvantages of head frames include: 1) Fastening to the skull with screws will cause pain and discomfort to the patient, and cause damage to the skull, increasing the risk of infection. 2) Gamma Knife head frames are mainly used for intracranial treatment, so their application range is limited. For patients who need treatment of other parts (such as the neck), the head frame cannot provide proper fixation and positioning.

[0008] 1.1.2. CT and MRI scans

[0009] Before radiation therapy, doctors usually use CT and MRI scans to determine the anatomical information of the tumor and other organs to mark the tumor and key structures, determine the direction and dose distribution of the radiation beam, and develop a treatment plan.

[0010] However, preoperative CT and MRI scans can only help doctors outline the patient's treatment target area before radiotherapy, but cannot complete real-time position monitoring and displacement correction of the patient during treatment.

[0011] 1.1.3. Surface-Guided Radiation Therapy (SGRT)

[0012] Surface-guided radiation therapy (SGRT) systems use visual cameras or other imaging devices to capture features of the patient's body surface. Before treatment begins, the SGRT system captures a reference image of the patient's surface as a benchmark. During treatment, the SGRT system monitors the patient's surface features in real time, compares the current surface position with the reference image, and provides feedback, guides the beam, or pauses treatment when deviations are detected.

[0013] However, SGRT can only provide location information on the surface of the human body, and the patient's surface features do not always accurately reflect the location of tumors and organs inside the human body. Therefore, the accuracy of SGRT positioning is affected by the inconsistency between the displacement of the human body surface and internal organs.

[0014] 1.1.4. Image-Guided Radiation Therapy (IGRT)

[0015] Image-guided radiation therapy (IGRT) uses medical imaging technologies, such as X-ray imaging, CT imaging, and magnetic resonance imaging, to monitor the patient's body tissue structure in real time during radiotherapy and guide the direction of the radiation beam.

[0016] Among them, X-ray imaging can reduce the radiation dose generated by CT scans while obtaining two-dimensional images of the patient's internal anatomical structure. Due to its low dose, flexible design and easy embedding into radiotherapy equipment, positioning and navigation based on X-ray imaging have been applied in some radiotherapy equipment. However, compared with three-dimensional CT or MRI imaging, two-dimensional X-ray images have less data. Therefore, how to use X-ray images with less data to accurately locate three-dimensional anatomical tissues in the human body is a core issue facing the precision radiotherapy industry.

[0017] During image-guided radiotherapy, there are usually two modes for the registration of human tissues: rigid registration and non-rigid registration. Rigid registration is applicable to situations where the relative position remains unchanged, but there may be rotation or translation, such as head movement. Rigid registration performs a rigid transformation of translation and rotation on the image, during which the shape and structure of the object remain unchanged. In contrast, non-rigid registration uses individualized displacement correction parameters for the movement of each point in space, and is applicable to situations where the morphology of organs changes, such as deformation and swelling of internal organs such as the heart. However, its calculation method is relatively complex, and there are many parameters to be calculated. A large amount of data is required to accurately calculate the inconsistent displacements generated at different positions.

[0018] In addition to the two registration modes mentioned above, the movement of the spine is considered to be a composite mode. Due to the structure of the spine and the role of the intervertebral joints, the movement between the vertebrae mainly includes bending, extension, scoliosis, compression, and opening. Therefore, although the movement of individual vertebrae can be considered as rigid body movement, the relative movement between vertebrae makes the spine as a whole produce a non-rigid body movement form.

[0019] During radiotherapy, patients usually lie supine on the treatment bed. In this state, the patient's thoracic and lumbar vertebrae are close to the treatment bed and their positions are relatively fixed, while the cervical spine is easily displaced due to the patient's head movement. Therefore, the displacement of the cervical spine is a displacement pattern that is significantly different from the movement of the head and the movement of internal organs. For the precise radiotherapy program for the treatment of cervical tumors, the positioning and guidance of the neck has unique technical characteristics different from other organs. A report pointed out that if the neck displacement during radiotherapy is not corrected, 40.8% of the dose will be radiated beyond 5 mm from the target. Therefore, accurate positioning of the neck plays a vital role in precise radiotherapy. Summary of the invention

[0020] The present invention aims at solving the problems existing in the prior art and provides a method for accurate positioning and navigation of the neck. The present invention is achieved through the following technical solutions:

[0021] The present invention discloses a method for accurately positioning and navigating a neck, comprising:

[0022] 1) Perform image segmentation and extraction on the preoperative CT images to obtain cervical vertebrae bone tissue and non-cervical vertebrae bone tissue; thereby distinguishing the rigid tissue of the neck, cervical vertebrae bone tissue, from non-rigid tissue. The cervical vertebrae bone tissue is the rigid tissue, and the non-cervical vertebrae bone tissue is the non-rigid tissue;

[0023] 2) Take real-time single-angle or multi-angle X-ray images of the neck tissue to obtain X-ray projection images of the neck tissue structure at each angle;

[0024] 3) The cervical vertebrae bone tissue extracted from the CT image is registered with the real-time X-ray projection image of the cervical tissue structure at each angle to determine the displacement of the cervical vertebrae and obtain the spatial transformation parameters of the cervical vertebrae;

[0025] 4) Based on the spatial transformation parameters of the cervical vertebrae, the displacement parameters of the non-rigid tissues of the neck are preliminarily estimated;

[0026] 5) Iterate the displacement parameters of the non-rigid tissue according to the preoperative CT image and the real-time X-ray projection image of the neck tissue structure at each angle to obtain the accurate displacement of each voxel of the neck non-rigid tissue in the CT image.

[0027] As a further improvement, in step 1) described in the present invention, image segmentation adopts a method based on the CT value of each pixel in the image, and distinguishes cervical vertebrae bone tissue from non-cervical vertebrae bone tissue by setting a threshold. The tissue represented by pixels with CT values ​​higher than the threshold is cervical vertebrae bone tissue, and the tissue represented by pixels with CT values ​​lower than the threshold is non-cervical vertebrae bone tissue.

[0028] As a further improvement, in step 1) of the present invention, image segmentation is achieved by a K-means algorithm based on computer vision, and the algorithm specifically divides human tissue into different clusters, and then calculates the average CT value of the area of ​​each cluster. When the average CT value exceeds the demarcation threshold (such as 250) between non-cervical vertebrae bone tissue and cervical vertebrae bone tissue, the cluster is set as cervical vertebrae bone tissue, thereby completing the automatic segmentation and extraction of cervical vertebrae bones in the three-dimensional CT image.

[0029] As a further improvement, in step 1) described in the present invention, image segmentation is achieved through a deep learning-based model, and an artificial neural network is trained by annotating existing data, thereby completing automatic segmentation and extraction of cervical vertebrae in three-dimensional CT images.

[0030] As a further improvement, step 3) of the present invention is specifically to forward project the cervical vertebrae bone tissue extracted from the CT image in step 1) to simulate the two-dimensional X-ray image taken in step 2), the forward projection is completed by computer simulation, and a simulated structure in which the light source emitting X-rays and the detector receiving X-rays in the system are located at the same position is used to calculate the cervical vertebrae bone tissue in the three-dimensional CT image at the angle and distance of the X-ray image. The two-dimensional image generated by forward projection; then, the obtained two-dimensional image is compared with the X-ray image taken in real time in step 2) to determine the displacement (T) of the cervical vertebrae bone tissue during the actual treatment of the patient and the position of the patient during the preoperative CT scan. x ,T y ,T z ,R x ,R y ,R z ), T represents the translation of the cervical vertebra in the x, y, and z directions, and R represents the rotation of the cervical vertebra with the x, y, and z directions as the axes. Since the cervical vertebra is a rigid body tissue, these six translation and rotation parameters can fully describe the displacement of the cervical vertebra.

[0031] As a further improvement, step 4) of the present invention preliminarily estimates the displacement parameters of the non-rigid tissue of the neck as follows: first, use the displacement parameters of the rigid tissue to preliminarily estimate its displacement direction and displacement amplitude. For each non-rigid pixel, its displacement direction and amplitude are the weighted average values ​​of the surrounding cervical vertebrae bone tissue.

[0032] As a further improvement, the image of each pixel of the surrounding cervical vertebrae bone tissue to the non-rigid pixel is inversely proportional to the distance between the cervical vertebrae bone pixel and the non-rigid pixel. The farther the distance, the smaller the impact.

[0033] As a further improvement, step 5) described in the present invention is specifically: based on the currently estimated displacement parameters of the cervical vertebrae bone tissue and the displacement parameters of the non-cervical vertebrae bone tissue, the three-dimensional structure reflecting the current human neck tissue is estimated, and through forward projection, a two-dimensional projection image with the same shooting angle and position as the X-ray imaging is simulated and generated, and compared with the two-dimensional X-ray image taken in real time, and the comparison degree between the forward projection and the X-ray image is optimized by continuously fine-tuning the displacement parameters of the non-rigid tissue. When the residual of the alignment of the forward projection and the X-ray image is minimized, it is considered that the displacement parameters of the non-rigid tissue at this time represent the displacement parameters of the non-cervical vertebrae bone tissue in the human tissue, thereby completing the overall alignment process of the human neck.

[0034] The beneficial effects of the present invention are as follows:

[0035] The present invention proposes to use a composite model of rigid motion + non-rigid motion to achieve the technical difficulty of accurate positioning and navigation of neck tumors in radiotherapy by adopting sparse angle and low-dose X-ray imaging. The existing composite motion model method in the three-dimensional reconstruction of human motion mainly focuses on estimating the rigid and non-rigid features of the human surface point cloud based on visual cameras such as Kinect, and then estimating the changes in human posture and displacement of motion. In the field of radiotherapy involved in this method, it is necessary to estimate the displacement inside the three-dimensional human tissue, not just the surface features of the human body.

[0036] In the field of radiotherapy, existing neck registration methods use methods based on the joint positioning of kiloelectron voltage X-rays and megaelectron voltage X-rays, or use three-dimensional cone beam CT (CBCT) imaging to determine the displacement changes of the internal structure of the neck. Both methods focus on collecting more image data to improve the accuracy of positioning, but at the same time increase the radiation dose received by the patient. Therefore, the method of using low-dose, sparse-angle X-ray images to locate the rigid + non-rigid composite tissue inside the human body is an unresolved technical difficulty. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a conventional flow chart of radiation therapy in the prior art;

[0038] Figure 2 It is a method flow chart of the scheme of the present invention;

[0039] Figure 3It is a flow chart of the present invention scheme that specifically explains how to estimate the registration parameters of non-rigid tissues through the registration parameters of rigid tissues. DETAILED DESCRIPTION

[0040] The present invention discloses a method for accurately positioning and navigating a neck, comprising:

[0041] 1) Perform image segmentation and extraction on the preoperative CT images to obtain cervical vertebrae bone tissue and non-cervical vertebrae bone tissue; the image segmentation method can be based on the CT value of each pixel in the image. Since the material density of cervical vertebrae bone is higher, its CT value is higher, while the density of non-cervical vertebrae bone tissue (such as muscle, ligament, etc.) is lower than that of bone, and its CT value is lower. Therefore, a threshold (such as 250) can be set to distinguish cervical vertebrae bone tissue from non-cervical vertebrae bone tissue. The tissue represented by the pixel with a CT value higher than the threshold is cervical vertebrae bone tissue, and the tissue represented by the pixel with a CT value lower than the threshold is non-cervical vertebrae bone tissue. In addition, the image segmentation method can also be implemented by a model based on deep learning. The existing deep learning model (such as U-Net) can train the artificial neural network by annotating the existing data, thereby completing the automatic segmentation and extraction of cervical vertebrae bone in three-dimensional CT images, and distinguishing the rigid tissue of the neck, cervical vertebrae bone, from non-rigid tissue. Cervical vertebrae bone tissue is rigid tissue, and non-cervical vertebrae bone tissue (such as muscle, ligament, etc.) is non-rigid tissue.

[0042] 2) Real-time single-angle, dual-angle or multi-angle X-ray imaging of the neck tissue is performed to obtain X-ray projection images of the neck tissue structure at different angles. In this step, the X-ray light source and the flat-panel detector are placed on both sides of the human tissue to take pictures of the human tissue.

[0043] 3) Align the cervical vertebrae bone tissue extracted from the CT image with the real-time X-ray projection image of the cervical tissue structure at each angle to determine the displacement of the cervical vertebrae and obtain the spatial transformation parameters of the cervical vertebrae; First, forward project the cervical vertebrae extracted from the CT image in step 1) to simulate the two-dimensional X-ray image of the cervical vertebrae bone tissue taken in step 2). The forward projection is completed by computer simulation, and the method is to use a simulated structure in which the light source emitting X-rays and the detector receiving X-rays in the system are located in the same position, calculate the angle and distance of the X-ray image of the cervical vertebrae bone tissue in the three-dimensional CT image in the actual radiotherapy system, and generate a two-dimensional image through forward projection. Subsequently, this two-dimensional projection is compared with the real-time X-ray image taken in step 2) to determine the displacement (T) of the cervical vertebrae bone tissue during the patient's actual treatment process compared to the patient's position during the preoperative CT scan. x ,T y ,T z ,R x ,R y ,Rz ). Among them, T represents the translation distance of the cervical vertebra in the x, y, and z directions, and R represents the rotation angle of the cervical vertebra with the x, y, and z directions as axes. Since the cervical vertebra is a rigid body tissue, these six translation and rotation parameters can fully describe the displacement of the cervical vertebra.

[0044] 4) Based on the spatial transformation parameters of the cervical vertebrae, the displacement parameters of the non-rigid tissues in the neck are preliminarily estimated. Since the non-cervical vertebrae bone tissue is a non-rigid structure, its displacement cannot be described by translation and rotation. Instead, the displacement trajectory of each pixel representing the non-cervical vertebrae bone tissue needs to be estimated separately. For the displacement trajectory of each pixel, the method used is to first use the displacement parameters of the rigid tissue near the non-rigid tissue to preliminarily estimate the displacement direction and displacement amplitude of the non-rigid tissue. The specific method is as follows: Figure 3 As shown. For each non-rigid pixel, its displacement direction and amplitude are the weighted average of the displacement of the surrounding cervical vertebrae bone tissue. Specifically, the image of each pixel of the surrounding cervical vertebrae bone tissue to the non-rigid pixel is inversely proportional to the distance between the cervical vertebrae bone pixel and the non-rigid pixel, that is, the farther the distance, the smaller the impact.

[0045] 5) Calculate the displacement parameters of the non-rigid tissue according to the preoperative CT image and the real-time X-ray projection image of the cervical tissue structure at each angle to obtain the accurate displacement of each voxel of the cervical non-rigid tissue in the CT image. The displacement of the non-rigid tissue preliminarily estimated in step 4) needs to be further accurately calculated for its accuracy. The specific method is to estimate the three-dimensional structure reflecting the current human neck tissue according to the displacement parameters of the currently estimated cervical vertebrae bone tissue and the displacement parameters of the non-cervical vertebrae bone tissue, and then simulate and generate a two-dimensional projection image with the same shooting angle and position as the X-ray imaging by forward projection, and compare it with the real-time two-dimensional X-ray image. The displacement parameters of the non-rigid tissue are continuously adjusted to optimize the comparison degree of the forward projection and the X-ray image. When the residual error of the registration of the forward projection and the X-ray image is the smallest, it can be considered that the displacement parameters of the non-rigid tissue at this time represent the displacement parameters of the non-cervical vertebrae bone tissue in the human tissue, thereby completing the overall registration process of the human neck.

[0046] The technical solution of the present invention is further described below by combining the accompanying drawings of the specification through specific embodiments:

[0047] Figure 1This is a conventional flow chart of radiotherapy with existing technology; usually, patients need to undergo CT and MRI imaging scans before treatment to outline human tissues and develop treatment plans. At the beginning of the treatment process, medical staff position the patient on the treatment bed and place the treatment target as close as possible to the treatment area of ​​the radiotherapy system. Subsequently, the physician uses the image guidance system to accurately move the treatment target into the treatment area by moving the treatment bed, and starts the radiotherapy process. During the treatment process, in order to ensure the effectiveness and safety of the treatment, the system needs to monitor the position of the treatment target in real time, and adjust the angle of the treatment bed and the treatment beam according to its displacement.

[0048] Figure 2 This is a flow chart of the method of the present invention; a schematic diagram of the method in the overall process of radiotherapy. The method mainly focuses on segmenting and extracting tissue from cervical CT and performing projection registration with real-time generated X-ray images, thereby first determining the motion parameters of the rigid body, and then estimating the nonlinear displacement parameters of the non-rigid structure based on the prior information of the displacement transformation parameters of the rigid body.

[0049] This technical solution is divided into the following steps: 1) Segment and extract the preoperative CT image; 2) Align the cervical vertebrae extracted from the CT image with the real-time two-dimensional X-ray image to determine the displacement of the cervical vertebrae; 3) Preliminarily estimate the displacement parameters of the non-rigid tissue of the neck based on the spatial transformation parameters of the cervical vertebrae; 4) Iterate the displacement parameters of the non-rigid tissue based on the preoperative CT image and the real-time two-dimensional X-ray image to obtain the accurate displacement of each voxel of the non-rigid tissue of the neck in the CT image. The overall technical solution is as follows:

[0050] As shown in the figure above, the solution proposed by this method is 1) first estimate the six-dimensional transformation parameters (T x ,T y ,T z ,R x ,R y ,R z ) ; 2) The displacement motion field of each non-rigid voxel i is a function of the displacement parameters of all rigid tissues around it (F). The influence weight of different rigid tissues on each voxel is related to the relative distance between the main voxel and the rigid tissue. The specific technical solution is as follows:

[0051] 1) Image segmentation and extraction of CT images

[0052] First, a 3D CT image of the human body is collected on a medical CT scanner as a preoperative CT image. Based on the differentiated CT values ​​(Hounsfield Unit) presented by different human tissues in the CT image and the classification method in computer vision technology or the feature segmentation method based on the deep learning neural network architecture, the cervical vertebrae, ligaments and other tissues in the neck tissue are extracted. The specific implementation method is as follows:

[0053] For example, assuming that the CT value of cervical vertebrae is usually greater than 250, then pixels with CT values ​​greater than 250 in the CT image can be first extracted as cervical vertebrae tissue, and pixels with CT values ​​less than 250 can be taken as non-rigid tissue. Next, the erosion-dilation method in computer vision is used to denoise pixels with CT values ​​greater than 250 to remove scattered high CT value pixels in the CT image.

[0054] In addition, the K-means algorithm in computer vision can also be used to divide human tissue into different clusters. The number of clusters in the K-means algorithm can vary with the number of cervical vertebrae in the field of view. Then, the average CT value of each cluster area is calculated. When the average CT value exceeds the boundary threshold between soft tissue and cervical vertebrae (such as 250), the cluster is set as cervical vertebrae.

[0055] In addition, a deep learning-based model can also be used to segment the cervical vertebrae. The U-Net architecture can segment the 3D image into regions using a pre-trained human tissue segmentation model to determine the tissue category of each region, thereby extracting the cervical vertebrae and other tissues.

[0056] 2) Align the 2D X-ray image of the cervical vertebra with the 3D CT image to calculate the displacement between the standard position before surgery and the real-time position during surgery

[0057] The cervical vertebra extracted from the X-ray image is spatially registered with the two-dimensional forward projection of the preoperative CT image of the phantom. The best fitting position of the two-dimensional forward projection of the CT image and the X-ray image is obtained by adjusting the translation and rotation parameters of the CT image in the spatial coordinate system, thereby obtaining the relative displacement of the phantom during the intraoperative and preoperative CT image scanning, and determining the displacement parameters of the six degrees of freedom of each cervical vertebra (T x ,T y ,T z ,R x ,R y ,R z ).

[0058] Specifically, the current position is assumed to be the initial estimated position, and the CT image is forward projected at the current position to generate a two-dimensional projection image. Next, the position difference between the projection image and the real-time X-ray image is calculated, and the three-dimensional spatial position parameters (T x ,T y ,T z ,R x ,R y ,R z ) to translate the CT image (T x ,T y ,T z ) or rotation (R x ,R y ,R z ) to generate two-dimensional projection images at different positions. By continuously adjusting the three-dimensional space position parameters (T x ,T y ,T z ,R x ,R y ,R z ) to optimize the matching degree between the projection image and the X-ray image. When the matching degree between the projection image and the X-ray image reaches the best, that is, when the difference between the projection image and the X-ray image is the smallest, or when the loss function describing the difference between the projection image and the X-ray image is the smallest, the three-dimensional spatial position parameter (T x ,T y ,T z ,R x ,R y ,R z ) is the displacement parameter between the preoperative planned CT image and the real-time position during surgery.

[0059] 3) Estimation of the displacement parameters of the non-rigid tissue of the neck based on the displacement parameters of the cervical vertebrae

[0060] For each voxel in a 3D CT image, the parameters of its coordinate transformation can be estimated from the parameters of the rigid body transformation of the adjacent vertebrae and a nonlinear model.

[0061] Specifically, assume that there are n cervical vertebrae in the CT image, and the spatial displacement parameter of each cervical vertebra has been calculated as (T x 1 ,T y 1 ,T z 1 ,R x 1 ,R y 1 ,R z 1 ) to (T xn ,T y n ,T z n ,R x n ,R y n ,R z n ), then the displacement motion field of pixel i where the non-rigid tissue is located in the 3D CT image can be represented by the following function:

[0062]

[0063] Where F represents the motion field of pixel i, which is described by two parameters: direction and displacement. These two parameters are the rigid body displacement model (T x ,T y ,T z ,R x ,R y ,R z ) is a weighted average of the spatial displacement transformation of pixel i. The image weight of each cervical vertebra k on pixel i is related to the distance between the cervical vertebra k and pixel i, and is in inverse proportion, that is, the greater the distance, the lower the influence weight of the cervical vertebra k on pixel i. The specific displacement calculation method of pixel i is as follows:

[0064]

[0065] Among them, Q represents the current position of non-rigid tissue pixel i, P represents the position of non-rigid tissue pixel i after displacement, ω represents the influence weight of cervical vertebrae on pixels, and M represents the influence weight of translation and rotation (T x ,T y ,T z ,R x ,R y ,R z ) represents the spatial transformation matrix. In this way, the vector QP represents the displacement direction and amplitude of pixel i.

[0066] After the displacement vector of each non-rigid tissue pixel i is calculated by the above method, its vector parameters can be fine-tuned within a range to optimize the matching degree between the two-dimensional DRR projection generated by the displaced CT image and the real-time generated X-ray image, thereby realizing the overall displacement estimation of the CT image.

[0067] The above description is not a limitation of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several changes, modifications, additions or substitutions can be made without departing from the essential scope of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for accurate positioning and navigation of the neck, It is characterized in that include: 1) performing image segmentation and extraction on the preoperative CT image to obtain cervical vertebrae bone tissue and non-cervical vertebrae bone tissue, wherein the cervical vertebrae bone tissue is a rigid tissue, and the non-cervical vertebrae bone tissue is a non-rigid tissue; 2) Take real-time single-angle or multi-angle X-ray images of the neck tissue to obtain X-ray projection images of the neck tissue structure at each angle; 3) The cervical vertebrae bone tissue extracted from the CT image is registered with the real-time X-ray projection image of the cervical tissue structure at each angle to determine the displacement of the cervical vertebrae and obtain the spatial transformation parameters of the cervical vertebrae; 4) Based on the spatial transformation parameters of the cervical vertebrae, the displacement parameters of the non-rigid tissues of the neck are preliminarily estimated; 5) Iterate the displacement parameters of the non-rigid tissue according to the preoperative CT image and the real-time X-ray projection image of the neck tissue structure at each angle to obtain the accurate displacement of each voxel of the neck non-rigid tissue in the CT image.

2. The method for accurate positioning and navigation of the neck according to claim 1, It is characterized in that In the step 1), the image segmentation adopts a method based on the CT value of each pixel in the image, and the cervical vertebrae bone tissue and non-cervical vertebrae bone tissue are distinguished by setting a threshold. The tissue represented by the pixels with CT values ​​higher than the threshold is the cervical vertebrae bone tissue, and the tissue represented by the pixels with CT values ​​lower than the threshold is the non-cervical vertebrae bone tissue.

3. The method for accurate positioning and navigation of the neck according to claim 1, It is characterized in that In the step 1), image segmentation is achieved through the K-means algorithm based on computer vision. The algorithm divides human tissue into different clusters, and then calculates the average CT value of the area of ​​each cluster. When the average CT value exceeds the boundary threshold between non-cervical vertebrae bone tissue and cervical vertebrae bone tissue, the cluster is set as cervical vertebrae bone tissue, thereby completing the automatic segmentation and extraction of cervical vertebrae in the three-dimensional CT image.

4. The method for accurate positioning and navigation of the neck according to claim 1, It is characterized in that In the step 1), image segmentation is achieved through a deep learning-based model, and an artificial neural network is trained by annotating existing data, thereby completing automatic segmentation and extraction of cervical vertebrae in three-dimensional CT images.

5. The method for accurate positioning and navigation of the neck according to claim 1, 2, 3 or 4, It is characterized in that The step 3) is specifically to forward project the cervical vertebrae bone tissue extracted from the CT image in step 1) to simulate the two-dimensional X-ray image taken in step 2), wherein the forward projection is completed by computer simulation, using a simulated structure in which the light source emitting X-rays and the detector receiving X-rays in the system are located at the same position, and calculating the two-dimensional image generated by forward projection at the angle and distance of the X-ray image in the three-dimensional CT image; then, the obtained two-dimensional image is compared with the X-ray image taken in real time in step 2) to determine the displacement (T) of the cervical vertebrae bone tissue during the actual treatment of the patient and the position of the patient during the preoperative CT scan. x ,T y ,T z ,R x ,R y ,R z ), T represents the translation of the cervical vertebrae in the x, y, and z directions, and R represents the rotation of the cervical vertebrae around the x, y, and z directions.

6. The method for accurate positioning and navigation of the neck according to claim 5, It is characterized in that The step 4) preliminarily estimates the displacement parameters of the non-rigid tissue of the neck as follows: first, the displacement direction and displacement amplitude are preliminarily estimated using the displacement parameters of the rigid tissue. For each non-rigid pixel, its displacement direction and amplitude are the weighted average values ​​of the surrounding cervical vertebrae bone tissue.

7. The method for accurate positioning and navigation of the neck according to claim 6, It is characterized in that The image of each pixel of the surrounding cervical vertebrae bone tissue to the non-rigid pixel is inversely proportional to the distance between the cervical vertebrae bone pixel and the non-rigid pixel. The farther the distance, the smaller the impact.

8. The method for accurate positioning and navigation of the neck according to claim 1 or 2 or 3 or 4 or 6 or 7, It is characterized in that The step 5) is specifically as follows: based on the currently estimated displacement parameters of the cervical vertebrae bone tissue and the displacement parameters of the non-cervical vertebrae bone tissue, the three-dimensional structure reflecting the current human neck tissue is estimated; by means of forward projection, a two-dimensional projection image with the same shooting angle and position as the X-ray imaging is simulated and generated, and compared with the two-dimensional X-ray image taken in real time; by continuously fine-tuning the displacement parameters of the non-rigid tissue, the comparison degree between the forward projection and the X-ray image is optimized; when the residual error of the registration of the forward projection and the X-ray image is minimized, it is considered that the displacement parameters of the non-rigid tissue at this time represent the displacement parameters of the non-cervical vertebrae bone tissue in the human tissue, thereby completing the overall registration process of the human neck.