A registration method, system, electronic device and storage medium for a fractured femur

Abstract

This application discloses a registration method, system, electronic device, and storage medium for femoral fractures. The method includes converting femoral CT volume data into a three-dimensional coordinate system to generate a volume data model; adjusting pose parameters to generate a target DRR image that coincides with the anatomical structure of the X-ray image; identifying anatomical key points jointly recognized by both and obtaining their corresponding two-dimensional coordinates; and then solving for the three-dimensional coordinates of these key points. Subsequently, a target function is constructed by combining the three-dimensional coordinates of the key points and the two-dimensional coordinates of the X-ray image; the optimal pose parameters of the model are solved with the function minimization as the objective; and finally, an optimal DRR image that precisely coincides with the X-ray image is generated based on these parameters. This method achieves precise 2D / 3D registration of femoral fractures using X-ray and CT scans, improving the robustness and accuracy of the registration, and providing a reliable three-dimensional pose reference for the diagnosis and treatment of femoral fractures.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to a registration method, system, electronic device and storage medium for fractured femurs. Background Technology

[0002] The diagnosis and treatment of femoral fractures require the combination of preoperative CT (Computed Tomography) and intraoperative / postoperative X-ray images. 2D / 3D registration of X-rays and CT scans restores the three-dimensional pose of the femur, serving the entire treatment process, including preoperative planning, intraoperative reduction assessment, and postoperative follow-up. Existing X-ray-CT 2D / 3D registration methods for femoral fractures are mainly divided into three categories: based on intensity similarity, based on anatomical features, and with artificial assistance. However, for the complex subject of femoral fractures, the core problem with existing technologies is the lack of a geometrically unified, closed-loop, and interference-resistant complete registration implementation scheme. Specifically, the registration process is highly sensitive to the initial pose, lacks sufficient resistance to fracture fragment occlusion / metal artifact interference, exhibits poor geometric consistency at each stage, and is prone to mismatch at key points from multiple perspectives. Furthermore, these methods only cover a single registration stage without a complete engineered process, resulting in low registration accuracy and poor robustness in femoral fracture scenarios, making it difficult to meet the needs of practical clinical applications. Summary of the Invention

[0003] To address the shortcomings mentioned above, this application provides a registration method, system, electronic device, and storage medium for fractured femurs.

[0004] Firstly, to achieve the above objectives, this application provides a registration method for fractured femurs, comprising: Acquire CT volume data and corresponding X-ray images of the femur, transform the voxel coordinates of each voxel in the CT volume data to a three-dimensional coordinate system, generate a femur volume data model, adjust the pose parameters of the volume data model, and generate a target DRR image that coincides with the femur anatomical structure in the X-ray image. Identify the common femoral anatomical key points in the target DRR image and the corresponding X-ray image, and obtain the two-dimensional coordinates of each anatomical key point in the target DRR image and the two-dimensional coordinates in the X-ray image respectively; Based on the two-dimensional coordinates of each anatomical key point in the target DRR image, determine the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system. Based on the three-dimensional coordinates of each anatomical key point and the two-dimensional coordinates in the X-ray image, an objective function is constructed, and the optimal pose parameters of the femoral body data model are obtained by minimizing the objective function. Based on the optimal pose parameters, an optimal DRR image that coincides with the femoral anatomy in the X-ray image is generated.

[0005] Preferably, the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system are determined based on the two-dimensional coordinates of each anatomical key point in the target DRR image, including: Based on the femoral body data model, determine the camera parameter matrix for generating the target DRR image; Based on the pose parameters of the volume data model, determine the rotation and translation parameters of the volume data model; Based on the camera parameter matrix, rotation parameter matrix, and translation parameter matrix, determine the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system.

[0006] Preferably, the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system are determined based on the camera parameter matrix, rotation parameter matrix, and translation parameter matrix, including: Based on the camera parameter matrix, rotation parameter matrix, and translation parameter matrix, construct the projection matrix. The formula for the projection matrix is ​​shown below: ; In the formula: The projection matrix; This is the camera parameter matrix; Here is the rotation parameter matrix; This is the translation parameter matrix; Based on the projection matrix, determine the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system.

[0007] Preferably, the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system are determined based on the projection matrix, including: Based on the projection matrix, the formula for the linear equation is constructed as follows: ; In the formula: Three-dimensional coordinates; The formula is shown below: ; In the formula: Two-dimensional coordinates of key anatomical points from at least two perspectives; By solving the linear equations using singular value decomposition, the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system are obtained.

[0008] Preferably, the objective function is constructed based on the three-dimensional coordinates of each anatomical key point and the two-dimensional coordinates in the X-ray image, including: Determine the predicted coordinates of each anatomical key point by projecting its three-dimensional coordinates onto the X-ray image; An objective function is constructed based on the error between the predicted coordinates and the two-dimensional coordinates of the corresponding anatomical key points in the X-ray image.

[0009] Preferably, key anatomical points of the femur that are commonly identified in the target DRR image and the corresponding X-ray image are determined, including: CT body data includes femoral segmentation labels, which are then resampled into the voxel network of the femoral body data model. The femoral body data model projects femoral segmentation labels onto the target DRR image and X-ray image, and determines the common femoral anatomical key points in the target DRR image and the corresponding X-ray image based on the femoral segmentation labels.

[0010] Preferably, after determining the common femoral anatomical key points identified in the target DRR image and the corresponding X-ray image, the method further includes: The key anatomical points of each femur are sorted according to a preset rule to form a fixed key point sequence. The three-dimensional coordinates and optimal pose parameters are then processed based on the fixed key point sequence. The preset rule is to sort the key points in ascending order by the numerical suffix of their names.

[0011] Secondly, this application also provides a registration system for fractured femurs, comprising: The acquisition module is used to acquire CT scan data and corresponding X-ray images of the femur; The processing module is used to transform the voxel coordinates of each voxel in the CT volume data to a three-dimensional coordinate system, generate a femoral body data model, adjust the pose parameters of the volume data model, and generate a target DRR image that coincides with the femoral anatomical structure in the X-ray image; identify the common femoral anatomical key points in the target DRR image and the corresponding X-ray image, and obtain the two-dimensional coordinates of each anatomical key point in the target DRR image and the two-dimensional coordinates in the X-ray image respectively; determine the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system based on the two-dimensional coordinates of each anatomical key point in the target DRR image; construct an objective function based on the three-dimensional coordinates of each anatomical key point and the two-dimensional coordinates in the X-ray image, and obtain the optimal pose parameters of the femoral body data model by minimizing the objective function. The execution module is used to generate an optimal DRR image that coincides with the femoral anatomy in the X-ray image based on the optimal pose parameters.

[0012] Thirdly, this application also provides an electronic device, including at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the program is executed by the processing unit, the processing unit performs the above-described method.

[0013] Fourthly, this application also provides a storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the above-described method.

[0014] Compared with the prior art, the beneficial effects of this application are as follows: This application generates a femoral body data model from CT body data, adjusts the femoral body data model to generate a target DRR image that overlaps with the femoral anatomical structure in the X-ray image, and then identifies the femoral anatomical key points that can be jointly identified in the DRR image and the X-ray image. Based on the femoral anatomical key points, the optimal pose parameters are obtained, realizing the accurate correspondence between the anatomical features of X-ray and CT images. This reduces the difficulty of identifying and corresponding anatomical key points in scenarios with fracture fragments and cortical bone fractures, and improves the problem of poor feature matching effect in the femoral fracture scenario. Attached Figure Description

[0015] Figure 1 This is a flowchart of the registration method for fractured femurs in this application. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] This application provides a registration method for fractured femurs, including: Acquire CT volume data and corresponding X-ray images of the femur, transform the voxel coordinates of each voxel in the CT volume data to a three-dimensional coordinate system, generate a femur volume data model, adjust the pose parameters of the volume data model, and generate a target DRR image that coincides with the femur anatomical structure in the X-ray image. Specifically, the camera parameters for generating the target DRR image from the femoral body data model include the source-to-detector distance, output image side length, and pixel size, which remain constant. The pose parameters are six degrees of freedom, including rotation parameters around the X, Y, and Z axes of the three-dimensional coordinate system, and translation parameters along the X, Y, and Z axes. These settings ensure geometric consistency in the generated DRR image, avoiding registration deviations caused by fluctuations in imaging parameters. The six-degree-of-freedom pose parameters can be flexibly adjusted in all dimensions, achieving precise alignment of the target DRR image with the femoral anatomy in the X-ray image.

[0018] Furthermore, the acquired femoral CT volume data includes femoral segmentation labels. These labels are resampled into the voxel network of the femoral body data model using nearest-neighbor interpolation to generate enhanced CT volume data for DRR rendering. At least two enhancement methods are available: one is a highlighting mode, which keeps the CT values ​​of the extrafemoral region unchanged and only increases the voxel values ​​of the femoral region by a preset increment ΔHU; the other is a femoral-only mode, which sets the voxel values ​​of the extrafemoral region to a preset background value. These settings highlight the femoral structure, suppress background interference, and address the problem of low structural recognition caused by cortical fracture and fragment occlusion in fracture scenarios. The generated enhanced CT volume data can render DRR images with clearer contours, improving the accuracy of subsequent key point recognition.

[0019] For example, a set of views is employed, including an orthogonal view and views with small-angle perturbations applied in the yaw and pitch directions. This could include, for instance, seven views such as orthogonal, yaw (±5 degrees), pitch (±5 degrees), and combinations thereof. Each view corresponds to a set of six-DOF pose parameters, preferably represented using Euler angles plus translation. For each view, ray integration or sampling is performed on the CT volume data based on the pose parameters to obtain the corresponding DRR image. This setup overcomes the limitations of single-view projection information, covers the multi-projection angle features of the femur, and resolves feature loss caused by fragment occlusion under single-view conditions.

[0020] In some embodiments, the X-ray image to be registered is converted into a grayscale image, and the grayscale image is uniformly mapped onto an S×S square canvas to generate a grayscale image of uniform fixed size. During the mapping process, anatomical key points in the grayscale image are mapped accordingly. The pose parameters of the femoral body data model are adjusted so that the generated target DRR image coincides with the femoral anatomical structure in the X-ray image. Through the above settings, the synchronous mapping of key points ensures the geometric consistency of coordinates, and the pose adjustment achieves the coincidence of DRR with the femoral structure in the X-ray, solving the problem of missing two-dimensional feature matching benchmarks in fracture scenarios and improving the accuracy of subsequent key point recognition and matching.

[0021] Furthermore, the femoral label is projected as a two-dimensional mask in the current pose and displayed on the X-ray as a red semi-transparent overlay or outline overlay to improve the recognizability of the fractured femoral boundary and key structures. Through these settings, the fracture boundary, fracture ends, and key structures are highlighted, addressing the problem of low femoral structure recognition in X-ray images. This assists the surgeon in quickly aligning and marking key points, reducing operational difficulty and improving efficiency.

[0022] Identify the common femoral anatomical key points in the target DRR image and the corresponding X-ray image, and obtain the two-dimensional coordinates of each anatomical key point in the target DRR image and the two-dimensional coordinates in the X-ray image respectively; In some embodiments, key anatomical points of the femur are sorted according to a preset rule to form a fixed key point sequence. The three-dimensional coordinates and optimal pose parameters are then processed based on this fixed key point sequence. The preset rule is to sort the key points in ascending order by the numerical suffix of their names. From each viewpoint, the two-dimensional coordinates of each anatomical key point in the target DRR image and the two-dimensional coordinates in the X-ray image are determined sequentially according to the fixed key point sequence. Through this setup, the correspondence between key points in multiple views and processing stages is solidified, completely resolving the mismatch problem of key points in multiple views caused by differences in manual annotation habits and changes in data traversal order, thus improving the stability and repeatability of the registration results.

[0023] Based on the two-dimensional coordinates of each anatomical key point in the target DRR image, determine the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system. In some embodiments, the camera parameter matrix for generating the target DRR image is determined based on the femoral body data model; Based on the pose parameters of the volume data model, determine the rotation and translation parameters of the volume data model; Based on the camera parameter matrix, rotation parameter matrix, and translation parameter matrix, construct the projection matrix. The formula for the projection matrix is ​​shown below: ; In the formula: The projection matrix; This is the camera parameter matrix; Here is the rotation parameter matrix; This is the translation parameter matrix; Based on the projection matrix, the formula for the linear equation is constructed as follows: ; In the formula: Three-dimensional coordinates; The formula is shown below: ; In the formula: Two-dimensional coordinates of key anatomical points from at least two perspectives; By solving the linear equations using singular value decomposition, the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system are obtained.

[0024] The above settings enable accurate correspondence of the two-dimensional projection features of the fractured femur, solving the problems of no unified matching benchmark between DRR and X-ray and poor feature correspondence, and avoiding the decrease in registration accuracy caused by feature point selection deviation.

[0025] Based on the three-dimensional coordinates of each anatomical key point and the two-dimensional coordinates in the X-ray image, an objective function is constructed, and the optimal pose parameters of the femoral body data model are obtained by minimizing the objective function. In some embodiments, the predicted coordinates of each anatomical key point projected onto the X-ray image are determined; an objective function is constructed based on the error between the predicted coordinates and the corresponding anatomical key point's two-dimensional coordinates in the X-ray image. This error includes at least one of root mean square error and mean square error. Preferably, the root mean square error is used as the objective function.

[0026] Furthermore, with the objective function minimization as the goal, an adaptive gradient optimizer is used to iteratively update the pose parameters during the optimization process. To improve stability, gradient pruning is preferably performed, and the best loss solution throughout the process is retained; simultaneously, the loss value, pose parameters, and projection visualization results of each iteration are saved to a log file.

[0027] By using the above settings, the femoral body data model and the X-ray image are accurately aligned, effectively reducing pose deviation and solving the problems of insufficient registration accuracy and blind pose optimization.

[0028] Based on the optimal pose parameters, an optimal DRR image that coincides with the femoral anatomy in the X-ray image is generated.

[0029] This application also provides a registration system for fractured femurs, including: The acquisition module is used to acquire CT scan data and corresponding X-ray images of the femur; The processing module is used to transform the voxel coordinates of each voxel in the CT volume data to a three-dimensional coordinate system, generate a femoral body data model, adjust the pose parameters of the volume data model, and generate a target DRR image that coincides with the femoral anatomical structure in the X-ray image; identify the common femoral anatomical key points in the target DRR image and the corresponding X-ray image, and obtain the two-dimensional coordinates of each anatomical key point in the target DRR image and the two-dimensional coordinates in the X-ray image respectively; determine the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system based on the two-dimensional coordinates of each anatomical key point in the target DRR image; construct an objective function based on the three-dimensional coordinates of each anatomical key point and the two-dimensional coordinates in the X-ray image, and obtain the optimal pose parameters of the femoral body data model by minimizing the objective function. The execution module is used to generate an optimal DRR image that coincides with the femoral anatomy in the X-ray image based on the optimal pose parameters.

[0030] This application also provides an electronic device, including at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program, and when the program is executed by the processing unit, the processing unit performs the above-described method.

[0031] This application also provides a storage medium storing a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the above-described method.

[0032] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A registration method for fractured femurs, characterized in that, include: Acquire CT volume data of the femur and the corresponding X-ray image, transform the voxel coordinates of each voxel in the CT volume data to a three-dimensional coordinate system, generate a femoral volume data model, adjust the pose parameters of the volume data model, and generate a target DRR image that coincides with the femoral anatomical structure in the X-ray image. Identify the common femoral anatomical key points in the target DRR image and the corresponding X-ray image, and obtain the two-dimensional coordinates of each anatomical key point in the target DRR image and the two-dimensional coordinates in the X-ray image respectively; Based on the two-dimensional coordinates of each of the anatomical key points in the target DRR image, determine the three-dimensional coordinates of each of the anatomical key points in the three-dimensional coordinate system; Based on the three-dimensional coordinates of each anatomical key point and the two-dimensional coordinates in the X-ray image, an objective function is constructed, and the optimal pose parameters of the femoral body data model are obtained by minimizing the objective function. Based on the optimal pose parameters, an optimal DRR image that coincides with the femoral anatomy in the X-ray image is generated.

2. The registration method for fractured femurs according to claim 1, characterized in that, Determining the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system based on its two-dimensional coordinates in the target DRR image includes: Based on the femoral body data model, determine the camera parameter matrix for generating the target DRR image; Based on the pose parameters of the volume data model, determine the rotation and translation parameters of the volume data model; Based on the camera parameter matrix, the rotation parameter matrix, and the translation parameter matrix, the three-dimensional coordinates of each of the anatomical key points in the three-dimensional coordinate system are determined.

3. The registration method for fractured femurs according to claim 2, characterized in that, Determining the three-dimensional coordinates of each of the anatomical key points in the three-dimensional coordinate system based on the camera parameter matrix, the rotation parameter matrix, and the translation parameter matrix includes: Based on the camera parameter matrix, the rotation parameter matrix, and the translation parameter matrix, a projection matrix is ​​constructed, and the formula for the projection matrix is ​​as follows: ； In the formula: The projection matrix; This is the camera parameter matrix; Here is the rotation parameter matrix; This is the translation parameter matrix; Based on the projection matrix, the three-dimensional coordinates of each of the anatomical key points in the three-dimensional coordinate system are determined.

4. The registration method for fractured femurs according to claim 3, characterized in that, Determining the three-dimensional coordinates of each of the anatomical key points in the three-dimensional coordinate system based on the projection matrix includes: Based on the projection matrix, a linear equation formula is constructed, as shown below: ； In the formula: Three-dimensional coordinates; The formula is shown below: ； In the formula: The two-dimensional coordinates of the key anatomical points from at least two perspectives; By solving the linear equations using singular value decomposition, the three-dimensional coordinates of each of the anatomical key points in the three-dimensional coordinate system are obtained.

5. The registration method for fractured femurs according to claim 1, characterized in that, The step of constructing the objective function based on the three-dimensional coordinates of each of the anatomical key points and the two-dimensional coordinates in the X-ray image includes: Determine the predicted coordinates of each of the anatomical key points by projecting their three-dimensional coordinates onto the X-ray image; The objective function is constructed based on the error between the predicted coordinates and the corresponding two-dimensional coordinates of the anatomical key points in the X-ray image.

6. The registration method for fractured femurs according to claim 1, characterized in that, The determination of the common femoral anatomical key points identified in the target DRR image and the corresponding X-ray image includes: The CT body data includes femoral segmentation labels, and the femoral segmentation labels are resampled into the voxel network of the femoral body data model. The femoral body data model projects the femoral segmentation label onto the target DRR image and the X-ray image, and determines the common femoral anatomical key points in the target DRR image and the corresponding X-ray image based on the femoral segmentation label.

7. The registration method for fractured femurs according to claim 6, characterized in that, After determining the common femoral anatomical key points identified in the target DRR image and the corresponding X-ray image, the method further includes: The femoral anatomical key points are sorted according to a preset rule to form a fixed key point sequence, and the three-dimensional coordinates and the optimal pose parameters are processed according to the fixed key point sequence. The preset rule is to sort the key points in ascending order by the numerical suffix of the key point name.

8. A registration system for fractured femurs, characterized in that, include: The acquisition module is used to acquire CT scan data and corresponding X-ray images of the femur; The processing module is used to transform the voxel coordinates of each voxel in the CT volume data to a three-dimensional coordinate system, generate a femoral body data model, adjust the pose parameters of the volume data model, and generate a target DRR image that coincides with the femoral anatomical structure in the X-ray image. Identify the common femoral anatomical key points in the target DRR image and the corresponding X-ray image, and obtain the two-dimensional coordinates of each anatomical key point in the target DRR image and the two-dimensional coordinates in the X-ray image respectively; determine the three-dimensional coordinates of each anatomical key point in the three-dimensional coordinate system based on the two-dimensional coordinates of each anatomical key point in the target DRR image. Based on the three-dimensional coordinates of each anatomical key point and the two-dimensional coordinates in the X-ray image, an objective function is constructed, and the optimal pose parameters of the femoral body data model are obtained by minimizing the objective function. The execution module is used to generate an optimal DRR image that coincides with the femoral anatomy in the X-ray image based on the optimal pose parameters.

9. An electronic device, characterized in that, The method includes at least one processing unit and at least one storage unit, wherein the storage unit stores a computer program that, when executed by the processing unit, causes the processing unit to perform the method described in any one of claims 1 to 7.

10. A storage medium, characterized in that, It stores a computer program executable by an electronic device, which, when run on the electronic device, causes the electronic device to perform the method described in any one of claims 1 to 7.

Images