Method and terminal for generating a lower limb projection image

Abstract

The application discloses a lower limb projection image generation method and a terminal. The method comprises the following steps: acquiring a full lower limb three-dimensional model, and processing the full lower limb three-dimensional model to obtain a single lower limb long bone model and a single lower limb three-dimensional model; extracting a surface data point set of the single lower limb long bone model, and extracting a projection reference point of the single lower limb long bone model from the surface data point set based on a skeletal anatomical feature; determining a standard projection direction of the single lower limb three-dimensional model according to the projection reference point, and projecting the single lower limb three-dimensional model based on the standard projection direction to obtain a single lower limb projection image. The application can automatically calculate a projection direction, avoids subjectivity of manual adjustment, ensures accuracy of a projection image, and improves imaging efficiency of the projection image.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical imaging, in particular to a lower limb projection image generation method and terminal. BACKGROUND

[0002] In the field of medical imaging, full lower limb imaging technology mainly includes X-ray two-dimensional imaging and three-dimensional imaging (such as computed tomography CT, magnetic resonance imaging MRI, etc.). X-ray two-dimensional imaging requires shooting of the frontal and lateral films, wherein the frontal film requires the patella to be centered on the femoral condyle, but the positioning accuracy depends on the experience of the technician, and errors are easily caused when the patella is dislocated or the patient's activity is limited. In addition, the X-ray two-dimensional imaging lateral projection has the problem of overlapping of the left and right lower limb images, affecting the clarity. Although three-dimensional imaging can generate two-dimensional projections through maximum intensity projection (MIP) or average intensity projection (AIP), it needs to manually adjust the projection direction of the bilateral lower limbs, which is tedious and easy to introduce subjective errors, affecting the imaging efficiency and image accuracy. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a lower limb projection image generation method and terminal which can ensure the accuracy of the projection image and improve the imaging efficiency of the projection image.

[0004] To solve the above technical problems, the technical scheme adopted by the present application is:

[0005] A lower limb projection image generation method, comprising:

[0006] acquiring a full lower limb three-dimensional model and processing the full lower limb three-dimensional model to obtain a unilateral lower limb long bone model and a unilateral lower limb three-dimensional model;

[0007] extracting a surface data point set of the unilateral lower limb long bone model, and extracting a projection reference point of the unilateral lower limb long bone model from the surface data point set based on skeletal anatomical features;

[0008] determining a standard projection direction of the unilateral lower limb three-dimensional model according to the projection reference point, and projecting the unilateral lower limb three-dimensional model based on the standard projection direction to obtain a unilateral lower limb projection image.

[0009] To solve the above technical problems, another technical scheme adopted by the present application is:

[0010] A lower limb projection image generation terminal, comprising a memory, a processor, and a computer program stored on the memory and running on the processor, wherein the processor implements each step of the lower limb projection image generation method described above when executing the computer program.

[0011] The application has the beneficial effects that: through the segmentation processing of the full lower limb three-dimensional model, the whole model is decomposed into a single lower limb model, avoiding the problem of left and right lower limb image overlap. The surface data point set of the single lower limb long bone model is extracted, and the projection reference point is automatically positioned based on the skeletal anatomical features, replacing the traditional manual positioning to determine the projection direction. According to the projection reference point, the standard projection direction is calculated, which can eliminate subjective judgment errors and ensure that the projection direction is consistent with the spatial relationship of the skeletal anatomical structure, thereby generating a standardized lower limb projection image. The application provides independent data basis for subsequent single projection through segmentation processing; the extraction of the surface data point set combined with the anatomical feature screening reference point ensures that the projection direction is aligned with the skeletal physiological structure; at the same time, the projection direction is automatically calculated based on the projection reference point, avoiding the inefficiency and subjectivity of manual adjustment, and finally realizing the automatic generation of a standardized projection image without positioning intervention. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 A flowchart of a lower limb projection image generation method of the application;

[0013] Figure 2 A front view of the femoral condyle posterior point of the application;

[0014] Figure 3 A top view of the femoral condyle posterior point of the application;

[0015] Figure 4 An en face projection image of the application;

[0016] Figure 5 A lateral projection image of the application;

[0017] Figure 6 A structure schematic diagram of a lower limb projection image generation terminal of the application;

[0018] REFERENCE NUMERALS:

[0019] 100, a lower limb projection image generation terminal; 101, a memory; 102, a processor. DETAILED DESCRIPTION

[0020] To explain the technical content, the achieved purposes and effects of the application in detail, the following will be explained in combination with the embodiments and the drawings.

[0021] In the prior art, when the full lower limb projection image is obtained by using the X-ray two-dimensional imaging technology, the frontal view and the lateral view are usually needed. The key quality standard of the frontal view is to ensure that the patella is in the middle of the femoral condyle, and this standard is highly dependent on the position of the patient's lower limb. However, the accuracy of the position is limited by the experience of the technician, and often needs to be adjusted repeatedly, which increases the operation complexity. Moreover, for the case of patellar dislocation, forcibly placing the patella in the middle position may cause the lower limb to rotate outward or inward, which affects the accuracy of the measurement result. For severe patients with limited activity, it is difficult to adjust their lower limbs to the standard position, which further increases the difficulty of shooting. In addition, the X-ray two-dimensional imaging also has the problem of overlapping of the left and right lower limb images. Especially in the lateral projection, the tissues of the two lower limbs in the normal standing position are easily mixed, which reduces the image clarity and affects the diagnosis effect. Although the three-dimensional imaging technology does not need to distinguish the projection direction of the left and right lower limbs, it needs to manually adjust the projection direction of the two lower limbs respectively to ensure that it is not affected by the position and unified. This process not only increases the operation workload and affects the imaging efficiency, but also introduces subjective errors due to the dependence of the projection direction on subjective judgment, which affects the accuracy of the projection image.

[0022] To solve the above problems, the present application provides a method for generating a lower limb projection image, comprising:

[0023] Obtaining a full lower limb three-dimensional model and processing the full lower limb three-dimensional model to obtain a single lower limb long bone model and a single lower limb three-dimensional model;

[0024] Extracting a surface data point set of the single lower limb long bone model, and extracting a projection reference point of the single lower limb long bone model from the surface data point set based on the skeletal anatomical features;

[0025] Determining a standard projection direction of the single lower limb three-dimensional model according to the projection reference point, and projecting the single lower limb three-dimensional model based on the standard projection direction to obtain a single lower limb projection image.

[0026] From the above description, the beneficial effects of the present application are that by the segmentation processing of the full lower limb three-dimensional model, the whole model is decomposed into a single lower limb model, avoiding the problem of left and right lower limb image overlap. By extracting the surface data point set of the single lower limb long bone model, and automatically positioning the projection reference point based on the skeletal anatomical features, the traditional manual positioning method for determining the projection direction is replaced. According to the standard projection direction calculated based on the projection reference point, the subjective judgment error can be eliminated, and the spatial relationship between the projection direction and the skeletal anatomical structure is ensured to be consistent, so as to generate a standardized lower limb projection image. The present application provides independent data basis for subsequent single projection through segmentation processing; the extraction of the surface data point set combined with the screening of the reference point based on the anatomical features ensures that the projection direction is aligned with the skeletal physiological structure; at the same time, the projection direction is automatically calculated based on the projection reference point, avoiding the inefficiency and subjectivity of manual adjustment, and finally realizing the automatic generation of a standardized projection image without positioning intervention.

[0027] Further, the single lower limb long bone model includes a single femur model;

[0028] The projection reference point includes a femoral lateral condyle posterior point and a femoral medial condyle posterior point;

[0029] The extraction of the projection reference point of the single lower limb long bone model from the surface data point set based on the skeletal anatomical features includes:

[0030] The femoral lateral condyle posterior point and the femoral medial condyle posterior point are extracted from the single femur model according to the spatial distribution characteristics of the surface data point set and the skeletal anatomical features.

[0031] From the above description, the femur is used as the anatomical characteristics of the largest long bone of the lower limb, and the distal condyle part has significant morphological characteristics and stable position, which provides reliable anatomical basis for the selection of the projection reference point. The projection reference point is specifically the femoral lateral condyle posterior point and the femoral medial condyle posterior point, and these two bone markers have clear anatomical positioning significance and can form a stable spatial connection relationship in three-dimensional space. By extracting feature points according to the spatial distribution characteristics of the surface data point set, the subjectivity of traditional manual identification can be effectively overcome, and the anatomical marker points are automatically identified by using the spatial topological relationship of the three-dimensional model surface points. Among them, the feature extraction based on the spatial distribution characteristics of the surface data point set can adapt to the difference of different individual skeletal morphology, automatically identify the key anatomical points by mathematical method, avoid the error caused by manual intervention, and provide accurate reference point data for the automatic calculation of the subsequent projection direction.

[0032] Further, the extraction of the femoral lateral condyle posterior point and the femoral medial condyle posterior point from the single femur model according to the spatial distribution characteristics of the surface data point set and the skeletal anatomical features includes:

[0033] According to a first spatial distribution feature of the surface data point set in a human height direction, surface data points located at a distal end of the single-side femur model are marked as a knee joint data point set;

[0034] According to a second spatial distribution feature of the knee joint data point set in a human front-rear direction, a knee joint data point closest to a rear of the human body is marked as a first femoral posterior condyle point;

[0035] Distances between the first femoral posterior condyle point and each knee joint data point in the knee joint data point set are respectively calculated, and a knee joint data point with a distance greater than a preset threshold is marked as a candidate femoral posterior condyle point set;

[0036] According to the second spatial distribution feature of the candidate femoral posterior condyle point set, a candidate femoral posterior condyle point closest to the rear of the human body is marked as a second femoral posterior condyle point;

[0037] A femoral lateral posterior condyle point and a femoral medial posterior condyle point are determined according to a third spatial distribution feature of the first femoral posterior condyle point and the second femoral posterior condyle point in a human left-right direction.

[0038] As can be seen from the above description, precise positioning of anatomical landmark points is achieved through multi-dimensional spatial feature analysis. First, the knee joint region data points are screened based on the spatial distribution feature in the human height direction, effectively excluding the interference of non-target data points; then the first femoral posterior condyle point closest to the rear is locked by using the spatial distribution feature in the human front-rear direction, establishing an initial positioning reference; the candidate point set is screened by using the preset distance threshold to exclude adjacent points on the same condyle; the second femoral posterior condyle point is determined from the candidate set based on the human front-rear direction feature again, ensuring the anatomical distance between the two posterior condyle points; finally, the medial and lateral posterior condyle points are accurately distinguished by combining the spatial distribution feature in the left-right direction. This phased and multi-dimensional feature screening mechanism overcomes the subjectivity and error accumulation problems of traditional manual positioning through hierarchical progressive analysis of spatial distribution features, providing accurate anatomical reference for automatic determination of the subsequent projection direction.

[0039] Further, the standard projection direction includes a lateral projection direction and an en face projection direction;

[0040] Determining the standard projection direction of the single-side lower limb three-dimensional model according to the projection reference point includes:

[0041] Determining the lateral projection direction of the single-side lower limb three-dimensional model according to a direction of a line segment between the femoral lateral posterior condyle point and the femoral medial posterior condyle point;

[0042] Determining the en face projection direction of the single-side lower limb three-dimensional model according to the lateral projection direction and a preset plumb line direction.

[0043] As can be known from the above description, the lateral projection direction is determined according to the direction of the line between the posterior point of the lateral femoral condyle and the posterior point of the medial femoral condyle, the anatomical symmetry feature of the human femoral condyle is used as a space reference to ensure that the projection direction is aligned with the real anatomical axis of the human body. The orthogonal relationship between the preset vertical line direction and the lateral projection direction is used to determine the frontal projection direction, the objective physical reference of the gravity direction is combined with the anatomical reference to construct an orthogonal projection coordinate system that meets the medical image standard. This composite projection direction determination method based on the spatial relationship of the anatomical landmark points and the physical reference avoids the subjective error of manual adjustment and ensures the standardization and repeatability of the projection direction in the three-dimensional space.

[0044] Further, the single lower limb projection image includes a lateral projection image and a frontal projection image;

[0045] The single lower limb three-dimensional model includes a single lower limb voxel model;

[0046] Projecting the single lower limb three-dimensional model based on the standard projection direction to obtain a single lower limb projection image includes:

[0047] Respectively calculating the average density value of the voxel unit of the single lower limb voxel model in the lateral projection direction and the frontal projection direction;

[0048] Generating a lateral projection image and a frontal projection image according to the average density values of the lateral projection direction and the frontal projection direction, respectively.

[0049] As can be known from the above description, the single lower limb three-dimensional model is a voxel model, which can directly utilize the three-dimensional spatial distribution characteristics of the voxel data to provide structured data support for subsequent projection calculation. By respectively calculating the average density values of the voxel units in the lateral and frontal standard projection directions, a standard projection parameter system based on anatomical features is established, so that the projection direction does not need to be adjusted manually. According to the average density values of different projection directions, the corresponding projection images are generated, which not only retains the anatomical structure information of the three-dimensional model, but also effectively eliminates the voxel data noise through density averaging processing, ensuring that the generated two-dimensional projection images meet the standardization requirements of clinical diagnosis and have good image quality. The cooperative calculation of the lateral projection direction and the frontal projection direction realizes the automatic matching of multi-angle projection parameters, solving the problem of workload in the traditional method of processing different projection directions respectively.

[0050] Further, processing the full lower limb three-dimensional model to obtain a single lower limb long bone model and a single lower limb three-dimensional model includes:

[0051] Performing voxel segmentation on the full lower limb three-dimensional model to obtain a full lower limb long bone voxel model;

[0052] extracting isosurface patches of the full lower limb long bone voxel model, and constructing a full lower limb long bone mesh model according to the isosurface patches;

[0053] constructing connected domains based on mesh data points in the full lower limb long bone mesh model;

[0054] segmenting the full lower limb long bone mesh model into a left single lower limb long bone model for the left side of the human body and a right single lower limb long bone model for the right side of the human body according to center positions of the connected domains;

[0055] calculating a minimum bounding box of the single lower limb long bone model, and extracting a single lower limb three-dimensional model corresponding to the single lower limb long bone model from the full lower limb three-dimensional model according to the minimum bounding box.

[0056] As can be seen from the above description, the full lower limb three-dimensional model is voxel segmented to accurately obtain a full lower limb long bone voxel model; the isosurface patches are constructed to generate a mesh model, which provides an accurate geometric structure basis for subsequent connected domain analysis; the left and right sides are segmented based on the center positions of the connected domains, and the anatomical symmetry feature is used to achieve accurate side segmentation; finally, the minimum bounding box is matched with the original three-dimensional model to ensure the spatial correspondence between the single lower limb three-dimensional model and the long bone model, and to avoid segmentation errors. The application of the pre-trained model improves the segmentation speed, the combination of the mesh model and the connected domain enhances the segmentation robustness, and the geometric feature matching of the minimum bounding box ensures the anatomical accuracy of the segmentation result.

[0057] Further, the surface data point set of the single lower limb long bone model includes:

[0058] calculating the number of voxels of each connected domain according to the single lower limb long bone model and the full lower limb long bone voxel model;

[0059] labeling the connected domain with the number of voxels greater than a preset voxel threshold as an effective connected domain;

[0060] labeling the connected domain with the number of voxels less than or equal to the preset voxel threshold as noise;

[0061] extracting all mesh data points in the effective connected domain to obtain a surface data point set.

[0062] From the above description, first, the number of voxels contained in each connected domain is calculated by combining the unilateral lower limb long bone model and the whole lower limb long bone voxel model, and the effectiveness of each connected domain is objectively evaluated by a quantitative index. Secondly, the preset voxel threshold is used for binary classification, and the region with the voxel scale meeting the standard is marked as an effective connected domain, and the noise region with sparse voxels is excluded. Finally, only the grid data points in the effective connected domain are extracted to form a surface data point set, ensuring that the geometric feature data relied on by the subsequent projection reference point extraction are all from the effective anatomical structure region, avoiding the interference of noise points on the positioning of key feature points.

[0063] Further, determining the lateral projection direction of the unilateral lower limb three-dimensional model according to the direction of the line segment between the lateral femoral condyle posterior point and the medial femoral condyle posterior point comprises:

[0064] A standard vector is calculated from the medial femoral condyle posterior point to the lateral femoral condyle posterior point, and the standard vector is normalized to obtain a standard unit vector;

[0065] The standard unit vector is determined as the lateral projection direction of the unilateral lower limb three-dimensional model.

[0066] From the above description, first, a standard vector is generated based on the direction from the medial femoral condyle posterior point to the lateral femoral condyle posterior point. This vector directly reflects the physiological structure characteristics of the human femoral condyle and can accurately represent the horizontal reference direction required for lateral projection. By normalizing the standard vector, the influence of the size difference of the lower limbs of different patients on the projection direction is eliminated, ensuring the consistency of the projection direction calculation among different individuals. Finally, the normalized standard unit vector is directly used as the lateral projection direction, realizing fully automatic direction determination based on anatomical feature points, avoiding the operation errors caused by manual intervention, and significantly improving the efficiency of generating projection images.

[0067] Further, determining the lateral projection direction of the unilateral lower limb three-dimensional model according to the direction of the line segment between the lateral femoral condyle posterior point and the medial femoral condyle posterior point comprises:

[0068] A vertical line vector corresponding to the preset vertical line direction is obtained;

[0069] A target vector perpendicular to the standard unit vector and the vertical line vector is calculated, and the target vector is normalized to obtain a target unit vector;

[0070] The target unit vector is determined as the lateral projection direction of the unilateral lower limb three-dimensional model.

[0071] As can be known from the above description, obtaining the plumb line vector converts the human anatomy standard body position into a mathematical vector reference, and establishes the correlation between the projection direction and the gravity direction; calculating the target vector perpendicular to the standard unit vector and the plumb line vector ensures the spatial right-angle coordinate system formed by the orthogonal characteristics of the vector, eliminates the directional deviation; normalizing the target vector generates the unit vector, and ensures the standardization of the projection direction parameter; finally, the target unit vector is directly used as the frontal projection direction, which replaces the artificial subjective judgment through the certainty of vector operation, and realizes the automatic calculation of the projection direction parameter. The scheme converts the medical image standard positioning requirement into a calculable geometric constraint condition through the mathematical method of spatial vector orthogonal decomposition, and generates the frontal projection direction meeting the clinical standard without manual intervention.

[0072] The application further provides a lower limb projection image generation terminal, comprising a memory, a processor and a computer program stored in the memory and running on the processor, and each step in the lower limb projection image generation method is realized when the processor executes the computer program.

[0073] As can be known from the above description, the application has the beneficial effects that: through the segmentation processing of the full lower limb three-dimensional model, the whole model is decomposed into a single lower limb model, avoiding the problem of overlapping of left and right lower limb images. The surface data point set of the single lower limb long bone model is extracted, and the projection reference point is automatically positioned based on the skeletal anatomical features, replacing the traditional projection direction determination mode relying on artificial positioning. The standard projection direction is calculated according to the projection reference point, which can eliminate the subjective judgment error and ensure that the projection direction is consistent with the spatial relationship of the skeletal anatomical structure, so as to generate a standardized lower limb projection image. The application provides independent data basis for subsequent single projection through segmentation processing; the extraction of the surface data point set in combination with the anatomical feature screening reference point ensures that the projection direction is aligned with the skeletal physiological structure; and the projection direction is automatically calculated based on the projection reference point, avoiding the inefficiency and subjectivity of artificial adjustment, and finally realizing the automatic generation of the standardized projection image without positioning intervention.

[0074] Embodiments of the application provide a lower limb projection image generation method and terminal suitable for the medical imaging field, which can automatically determine the projection direction, avoid the subjectivity of artificial adjustment, thereby ensuring the accuracy of the projection image and improving the imaging efficiency of the projection image. The following will be described through specific embodiments:

[0075] Please refer to Figures 1 to 5 An embodiment of the application is:

[0076] As Figure 1 shown, a lower limb projection image generation method comprises steps S1-S3.

[0077] S1, acquire a full lower limb three-dimensional model, and process the full lower limb three-dimensional model to obtain a single lower limb long bone model and a single lower limb three-dimensional model.

[0078] The full lower limb three-dimensional model refers to a three-dimensional data model containing bilateral lower limb bones and soft tissues reconstructed through CT or MRI scanning, and can be stored in the form of a voxel grid or a curved surface grid. The full lower limb three-dimensional model provides a complete anatomical structure data basis for segmentation processing. The single lower limb long bone model refers to a three-dimensional data model containing only one side of the lower limb bones. The single lower limb three-dimensional model refers to a three-dimensional data model containing only one side of the lower limb bones and soft tissues.

[0079] S2, extract a surface data point set of the single lower limb long bone model, and extract a projection reference point of the single lower limb long bone model from the surface data point set based on skeletal anatomical features.

[0080] The surface data point set is a three-dimensional coordinate set of the surface of the single lower limb long bone model, and can be obtained by an isosurface extraction algorithm. The skeletal anatomical features refer to typical markers or characteristics of the skeleton in terms of morphology, structure, position, and relationship with other tissues. The single lower limb long bone model includes a single femur model. The human lower limb long bones mainly include the femur and the tibia. The femur is the largest long bone in the lower limb and has the most prominent morphological features. The condyle part has a stable geometric structure in the three-dimensional space. The projection reference point is a key marker point selected based on the skeletal anatomical features. In this embodiment, the projection reference point includes the lateral condyle back point of the femur and the medial condyle back point of the femur, which provides a geometric reference for the projection direction.

[0081] Specifically, in step S2, the projection reference point of the single lower limb long bone model is extracted from the surface data point set based on the skeletal anatomical features, including step S250.

[0082] S250, extracting the lateral condyle back point of the femur and the medial condyle back point of the femur from the single femur model according to the spatial distribution characteristics of the surface data point set and the skeletal anatomical features.

[0083] In step S2, the surface data point set is a three-dimensional coordinate set of the surface of the single femur model. The spatial distribution characteristics of the surface data point set refer to the geometric distribution law of the three-dimensional model surface vertices in the spatial coordinate system. The vertex positions behind the medial and lateral condyles can be accurately located by analyzing the distribution density and curvature change of the surface data point set on the coronal plane, the sagittal plane, and the transverse plane. The coronal plane refers to the plane that divides the front-back direction of the human body (front-back), the sagittal plane refers to the plane that divides the left-right direction of the human body (left-right), and the transverse plane refers to the plane that divides the horizontal direction of the human body (up-down). Figure 2 and Figure 3As shown, the femoral condyle posterior point refers to the protruding point of the distal end of the femur (femoral condyle) on the last side (the most backward); the femoral lateral condyle posterior point refers to the bony landmark point of the lateral femoral condyle most protruding backward on the sagittal plane, located on the lateral side of the knee joint; the femoral medial condyle posterior point refers to the bony landmark point of the medial femoral condyle most protruding backward on the sagittal plane, located on the medial side of the knee joint.

[0084] In an optional embodiment, step S250 includes steps S2501-S2505.

[0085] S2501, according to the first spatial distribution characteristics of the surface data point set in the human body height direction, the surface data points located at the distal end of the unilateral femur model are marked as the knee joint data point set.

[0086] Wherein, the distal end of the unilateral femur model refers to the end away from the center of the human body, i.e. the lower end of the femur. The first spatial distribution characteristics refer to the coordinate point distribution along the human body height axis direction, which can be realized by comparing the coordinate values in the human body height axis direction to quickly locate the knee joint region.

[0087] S2502, according to the second spatial distribution characteristics of the knee joint data point set in the human body front-back direction, the knee joint data point closest to the back of the human body is marked as the first femoral condyle posterior point.

[0088] Wherein, the second spatial distribution characteristics refer to the coordinate point distribution along the human body front-back direction, which can be used to determine the initial position of the condyle posterior point by comparing the coordinate values in the human body front-back direction.

[0089] S2503, the distance between the first femoral condyle posterior point and each knee joint data point in the knee joint data point set is calculated respectively, and the knee joint data point with a distance greater than a preset threshold is marked as a candidate femoral condyle posterior point set.

[0090] Wherein, the preset threshold refers to the interval range of the medial and lateral femoral condyles in anatomy. By calculating the distance between the first femoral condyle posterior point and all knee joint data points, the data points within the interval range of the medial and lateral femoral condyles are selected (i.e. excluding adjacent points on the same condyle), so as to ensure that the two femoral condyle posterior points come from different condyles.

[0091] S2504, according to the second spatial distribution characteristics of the candidate femoral condyle posterior point set, the candidate femoral condyle posterior point closest to the back of the human body is marked as the second femoral condyle posterior point.

[0092] S2505, according to the third spatial distribution characteristics of the first femoral condyle posterior point and the second femoral condyle posterior point in the human body left-right direction, the femoral lateral condyle posterior point and the femoral medial condyle posterior point are determined.

[0093] The third spatial distribution feature refers to the distribution of coordinate points along the left-right direction of the human body, and can be specifically distinguished by comparing the coordinate values of the left-right direction of the human body to distinguish the medial and lateral posterior points of the femoral condyle.

[0094] In a specific application scenario, taking the right femur model in the standing position as an example, the three-dimensional coordinate system of the right femur model is that the front-back direction of the human body is the Y axis, the height direction of the human body is the Z axis, and the left-right direction of the human body is the X axis. The positive direction of the Y axis is the front of the human body, the positive direction of the Z axis is the upper side of the human body, and the positive direction of the X axis is the left side of the human body. The coordinates of the i-th data point in the surface data point set D0 are [x i ,y i ,z i ]. At this time, the specific process of step S250 includes steps a-e.

[0095] Step a, obtaining the maximum value of the Z axis in the surface data point set D0 as the maximum height of the right femur model, i.e., the maximum height H = max z i ; obtaining the minimum value of the Z axis in the surface data point set as the minimum height of the right femur model, i.e., the minimum height h = min z i ; wherein [x i ,y i ,z i ] . The surface data points whose Z axis coordinate values satisfy the condition z i <(h+(H-h) / 5) are marked as knee joint data points, and a knee joint data point set D1 is obtained. The condition z i <(h+(H-h) / 5) means that the Z axis coordinate value is lower than 1 / 5 of the total height of the femur (H-h) calculated from the lowest point (h) upwards.

[0096] Step b, marking the knee joint data point with the minimum Y axis coordinate value in the knee joint data point set D1 as the first posterior point of the femoral condyle P1. That is, if the Y axis coordinate value of the k-th data point is equal to argmin y k , wherein [x k ,y k ,z k ] , and argmin y represents finding the value that makes the Y axis coordinate value minimum, then the k-th data point is the first posterior point of the femoral condyle P1.

[0097] Step c, calculating the distance d from the j-th knee joint data point in the knee joint data point set D1 to the first posterior point of the femoral condyle P1, wherein [x j ,y j ,z j ] , and D jWhen the preset threshold value D' is met, the jth knee joint data point belongs to the candidate posterior femoral condyle point set D2.

[0098] Step d, mark the candidate posterior femoral condyle point with the minimum Y-axis coordinate value in the candidate posterior femoral condyle point set D2 as the second posterior femoral condyle point P2. That is, if the Y-axis coordinate value of the mth data point is equal to argmin y m , wherein [x m , y m , z m ] , the mth data point is the second posterior femoral condyle point P2.

[0099] Step e, when the X-axis coordinate value of the first posterior femoral condyle point P1 is greater than the X-axis coordinate value of the second posterior femoral condyle point P2, that is, x k > x m , the first posterior femoral condyle point P1 is the posterior medial femoral condyle point, and the second posterior femoral condyle point P2 is the posterior lateral femoral condyle point; when the X-axis coordinate value of the first posterior femoral condyle point P1 is less than the X-axis coordinate value of the second posterior femoral condyle point P2, that is, x k < x m , the first posterior femoral condyle point P1 is the posterior lateral femoral condyle point, and the second posterior femoral condyle point P2 is the posterior medial femoral condyle point.

[0100] S3, determining a standard projection direction of the unilateral lower limb three-dimensional model according to the projection reference point, and projecting the unilateral lower limb three-dimensional model based on the standard projection direction to obtain a unilateral lower limb projection image.

[0101] The standard projection direction includes a lateral projection direction and an anterior projection direction. The lateral projection direction is a projection axis direction perpendicular to the sagittal plane, and is used to eliminate the projection deviation caused by the rotation of the lower limb. The anterior projection direction is a projection axis direction perpendicular to the coronal plane, and ensures that the projection direction conforms to the standing posture of the human body.

[0102] Specifically, in step S3, determining the standard projection direction of the unilateral lower limb three-dimensional model according to the projection reference point includes steps S310-S320.

[0103] S310, determining the lateral projection direction of the unilateral lower limb three-dimensional model according to the direction of the line segment between the posterior lateral femoral condyle point and the posterior medial femoral condyle point.

[0104] In an optional embodiment, step S310 includes: calculating a standard vector pointing from the posterior medial femoral condyle point to the posterior lateral femoral condyle point, normalizing the standard vector to obtain a standard unit vector, and determining the standard unit vector as the lateral projection direction of the unilateral lower limb three-dimensional model.

[0105] In a specific application scenario, taking the calculation of the lateral projection direction of the right lower limb three-dimensional model as an example, the coordinates of the lateral femoral condyle posterior point are [x1, y1, z1], and the coordinates of the medial femoral condyle posterior point are [x2, y2, z2], and the lateral projection direction is a standard unit vector :

[0106] .

[0107] S320, determining the frontal projection direction of the unilateral lower limb three-dimensional model according to the lateral projection direction and the preset plumb line direction.

[0108] The plumb line direction refers to the vector expression of the gravity direction in the three-dimensional space, and can be specifically represented by a vertical axis vector in the three-dimensional coordinate system as a physical reference for the calculation of the frontal projection direction.

[0109] In an optional implementation, step S320 includes: obtaining a plumb line vector corresponding to the preset plumb line direction, calculating a target vector perpendicular to the standard unit vector and the plumb line vector, normalizing the target vector to obtain a target unit vector, and determining the target unit vector as the frontal projection direction of the unilateral lower limb three-dimensional model.

[0110] In a specific application scenario, taking the calculation of the frontal projection direction of the right lower limb three-dimensional model as an example, the plumb line vector corresponding to the plumb line direction is , the standard unit vector is , and the frontal projection direction is the target unit vector :

[0111] .

[0112] The unilateral lower limb projection image includes a lateral projection image and a frontal projection image; and the unilateral lower limb three-dimensional model includes a unilateral lower limb voxel model, which is a unilateral lower limb model established based on a voxel unit.

[0113] In an optional implementation, in step S3, the projection of the unilateral lower limb three-dimensional model based on the standard projection direction to obtain the unilateral lower limb projection image includes steps S330-S340.

[0114] S330, respectively calculating the average density values of the voxel units of the unilateral lower limb voxel model in the lateral projection direction and the frontal projection direction.

[0115] S340, respectively generating the lateral projection image and the frontal projection image according to the average density values of the lateral projection direction and the frontal projection direction.

[0116] In a specific application scenario, the projection image is specifically represented as ; wherein, u, v represents the pixel coordinate of the projection image, AIP u, v represents the pixel value of the projection image at the pixel coordinate u, v G x, y, z represents the density value of the voxel unit of the single lower limb voxel model at the coordinate x, y, z Mean represents the average density value of the density value of the voxel unit along the standard projection direction. It can be seen that each pixel coordinate u, v on the projection plane only corresponds to a ray (i.e. projection line) which is parallel to the standard projection direction and passes through u, v , which extends in the three-dimensional space and penetrates the entire single lower limb voxel model. The voxel units in the single lower limb voxel model are spatially sampled on the ray, and the continuous ray is discretized into a series of sampling points (x, y, z), so as to realize the average density value calculation.

[0117] In another optional embodiment, the projection image is generated by calculating the maximum density value or the minimum density value of the voxel unit of the single lower limb voxel model under the standard projection direction.

[0118] As shown in Figure 4 and Figure 5 , the right lower limb orthostatic projection image and the lateral projection image can be obtained based on the above-mentioned generation method of the lower limb projection image.

[0119] In summary, the present application generates the single lower limb projection image through the full lower limb three-dimensional model, and automatically calculates the standard projection direction of each of the two lower limbs, so as to realize the full-automatic projection, which is more efficient than the manual adjustment of the projection direction. In addition, the present application determines the standard projection direction by identifying the posterior point of the medial femoral condyle and the posterior point of the lateral femoral condyle, which is not affected by the patella and is suitable for serious patients with limited activity of the lower limb, so that the projection image quality is not affected by the patient's positioning, and the problem that the full lower limb orthostatic image is not suitable for serious patients due to the high requirement for positioning is solved. In addition, the present application can automatically determine the single lower limb model, and only project the single lower limb according to the single lower limb model, so as to solve the problem of the overlapping of the left and right lower limb images of the full lower limb lateral image.

[0120] Another embodiment of the present application is:

[0121] A generation method of a lower limb projection image, which is different from the first embodiment in that the specific implementation of the surface data point set of the single lower limb long bone model extracted in steps S1 and S2 is limited.

[0122] In the present embodiment, the obtained full lower limb three-dimensional model is a CT three-dimensional model. ​​​​

[0123] Specifically, in step S1, the full lower limb three-dimensional model is processed to obtain a single lower limb long bone model and a single lower limb three-dimensional model, which includes steps S110-S150.

[0124] S110, voxel segmentation is performed on the full lower limb three-dimensional model to obtain a full lower limb long bone voxel model.

[0125] In an alternative embodiment, the full lower limb three-dimensional model is voxel segmented by a pre-trained deep learning segmentation model. In a specific application scenario, specifically, the deep learning segmentation model classifies each voxel unit in the full lower limb three-dimensional model according to the region it belongs to, and assigns a value of 0, 1 or 2 to the classification result of each voxel unit, where 1 represents the femur region, 2 represents the tibia region, and 0 represents other regions. Thus, the segmentation result is a full lower limb long bone voxel model containing a femur voxel model and a tibia voxel model.

[0126] In another alternative embodiment, the full lower limb three-dimensional model is voxel segmented by a threshold segmentation method to obtain a full lower limb long bone voxel model.

[0127] S120, the isosurface patches of the full lower limb long bone voxel model are extracted, and a full lower limb long bone mesh model is constructed according to the isosurface patches.

[0128] In a specific application scenario, assuming that the threshold value of the isosurface to be extracted is 0.5, a plurality of cubic voxels with a size of 2x2x2 are constructed with each voxel unit in the full lower limb long bone voxel model as the center, the voxel values at the vertices of the voxels are compared with the given threshold value, the voxels intersecting the isosurface are first found, then the intersection points of the isosurface and the edges of the voxels are found by interpolation method, and finally the intersection points are connected to form a triangle to constitute an isosurface patch. The set of triangles in all voxels constitutes a full lower limb long bone mesh model.

[0129] In an alternative embodiment, a model registration method can be used to construct the full lower limb long bone mesh model.

[0130] S130, a connected domain is constructed based on the grid data points in the full lower limb long bone mesh model.

[0131] In a specific application scenario, surface data points that have direct or indirect connection paths between each other are marked as the same connected domain, and finally a plurality of connected domains that are not connected to each other are obtained.

[0132] S140, the full lower limb long bone mesh model is segmented into a single lower limb long bone model on the left side of the human body and a single lower limb long bone model on the right side of the human body according to the center positions of the connected domains.

[0133] In an optional embodiment, step S140 specifically comprises: performing a binary classification on each voxel unit within the full lower limb long bone voxel model according to whether it is located inside the connected domain, and counting the number of voxel units located inside the connected domain, which is the volume of the connected domain. A volume threshold is set, and the connected domain with a volume greater than the volume threshold is identified as the connected domain of the lower limb long bone. At this time, the unilateral lower limb long bone model can be determined by comparing the coordinate values of the center positions of the connected domains in the left-right direction of the human body. In a specific application scenario, if the center position of the first connected domain is represented as (x c1 ,y c1 ,z c1 ), and the center position of the second connected domain is represented as (x c2 ,y c2 ,z c2 ), then when x c1 >x c2 , the first connected domain is the left lower limb long bone model, and the second connected domain is the right lower limb long bone model; when x c1 <x c2 , the first connected domain is the right lower limb long bone model, and the second connected domain is the left lower limb long bone model.

[0134] S150, calculate the minimum bounding box of the unilateral lower limb long bone model, and extract the unilateral lower limb three-dimensional model corresponding to the unilateral lower limb long bone model from the full lower limb three-dimensional model according to the minimum bounding box.

[0135] In a specific application scenario, the minimum bounding boxes of the left femur grid model and the left tibia grid model are calculated, and the left lower limb grid model is obtained after appropriately expanding the range, so that all the voxel units of the left lower limb bones and soft tissues are included in the left lower limb grid model. The right lower limb grid model is the same, and details are not repeated here.

[0136] Specifically, in step S2, the surface data point set of the unilateral lower limb long bone model comprises steps S210-S240.

[0137] S210, calculate the number of voxels of each connected domain according to the unilateral lower limb long bone model and the full lower limb long bone voxel model.

[0138] S220, mark the connected domain with a number of voxels greater than a preset voxel threshold as an effective connected domain.

[0139] S230, mark the connected domain with a number of voxels less than or equal to the preset voxel threshold as noise.

[0140] S240, extract all grid data points in the effective connected domain to obtain the surface data point set.

[0141] In an alternative embodiment, the method for identifying the effective connected domain in steps S210-S220 is the same as the method for identifying the connected domain of the long bone of the lower limb in step S140, which will not be described herein.

[0142] Please refer to Figure 6 Another embodiment of the present application is:

[0143] A terminal 100 for generating a projection image of a lower limb comprises a memory 101, a processor 102, and a computer program stored in the memory 101 and running on the processor 102, wherein the processor 102 implements each step of the method for generating a projection image of a lower limb according to the above-mentioned embodiments when executing the computer program.

[0144] The above-mentioned embodiments are merely examples of the present application, and do not limit the patent scope of the present application. Any equivalent transformation or direct or indirect application in the related technical field based on the content of the specification and drawings of the present application is also included in the patent protection scope of the present application.

Claims

1. A method of generating a lower extremity projection image, characterized by, The method comprises: acquiring a full lower limb three-dimensional model, and processing the full lower limb three-dimensional model to obtain a single lower limb long bone model and a single lower limb three-dimensional model; extracting a surface data point set of the single lower limb long bone model, and extracting projection reference points of the single lower limb long bone model from the surface data point set based on skeletal anatomical features; the projection reference points comprise a femur lateral condyle posterior point and a femur medial condyle posterior point; determining a standard projection direction of the single lower limb three-dimensional model according to the projection reference points, and projecting the single lower limb three-dimensional model based on the standard projection direction to obtain a single lower limb projection image; the standard projection direction comprises a lateral projection direction and an anterior projection direction; determining the standard projection direction of the single lower limb three-dimensional model according to the projection reference points comprises: determining the lateral projection direction of the single lower limb three-dimensional model according to a direction of a line segment between the femur lateral condyle posterior point and the femur medial condyle posterior point; determining the anterior projection direction of the single lower limb three-dimensional model according to the lateral projection direction and a preset plumb line direction.

2. The method of generating a lower extremity projection image according to claim 1, wherein, the single lower limb long bone model comprises a single femur model; extracting the projection reference points of the single lower limb long bone model from the surface data point set based on skeletal anatomical features comprises: extracting the femur lateral condyle posterior point and the femur medial condyle posterior point from the single femur model according to spatial distribution characteristics and skeletal anatomical features of the surface data point set.

3. The method of generating a lower extremity projection image according to claim 2, wherein, extracting the femur lateral condyle posterior point and the femur medial condyle posterior point from the single femur model according to spatial distribution characteristics and skeletal anatomical features of the surface data point set comprises: according to a first spatial distribution characteristic of the surface data point set in a human body height direction, marking surface data points located at a distal end of the single femur model as a knee joint data point set; according to a second spatial distribution characteristic of the knee joint data point set in a human body front-rear direction, marking a knee joint data point closest to a rear side of the human body as a first femur condyle posterior point; respectively calculating distances between the first femur condyle posterior point and each knee joint data point in the knee joint data point set, and marking a knee joint data point with a distance greater than a preset threshold as a candidate femur condyle posterior point set; according to the second spatial distribution characteristic of the candidate femur condyle posterior point set, marking a candidate femur condyle posterior point closest to the rear side of the human body as a second femur condyle posterior point; determining the femur lateral condyle posterior point and the femur medial condyle posterior point according to a third spatial distribution characteristic of the first femur condyle posterior point and the second femur condyle posterior point in a human body left-right direction.

4. The method of generating a lower extremity projection image according to claim 1, wherein, the single lower limb projection image comprises a lateral projection image and an anterior projection image; the single lower limb three-dimensional model comprises a single lower limb voxel model; projecting the single lower limb three-dimensional model based on the standard projection direction to obtain a single lower limb projection image comprises: respectively calculating average density values of voxel units of the single lower limb voxel model in the lateral projection direction and the anterior projection direction; respectively generating a lateral projection image and an anterior projection image according to the average density values of the lateral projection direction and the anterior projection direction.

5. The method of generating a lower extremity projection image according to claim 1, wherein, processing the full lower limb three-dimensional model to obtain a single lower limb long bone model and a single lower limb three-dimensional model comprises: perform voxel segmentation on the full lower limb three-dimensional model to obtain a full lower limb long bone voxel model; extract an isosurface patch of the full lower limb long bone voxel model, and construct a full lower limb long bone grid model according to the isosurface patch; construct a connected domain based on grid data points in the full lower limb long bone grid model; segment the full lower limb long bone grid model into a left single lower limb long bone model of a human body and a right single lower limb long bone model of a human body according to a center position of the connected domain; calculate a minimum bounding box of the single lower limb long bone model, and extract a single lower limb three-dimensional model corresponding to the single lower limb long bone model from the full lower limb three-dimensional model according to the minimum bounding box.

6. The method of generating a lower extremity projection image according to claim 5, wherein, extracting a surface data point set of the single lower limb long bone model includes: calculating a voxel number of each connected domain according to the single lower limb long bone model and the full lower limb long bone voxel model; labeling a connected domain with a voxel number greater than a preset voxel threshold as a valid connected domain; labeling a connected domain with a voxel number less than or equal to the preset voxel threshold as noise; extracting all grid data points in the valid connected domain to obtain a surface data point set.

7. The method of generating a lower extremity projection image according to claim 1, wherein, determining a lateral projection direction of the single lower limb three-dimensional model according to a direction of a line segment between the femoral lateral condyle posterior point and the femoral medial condyle posterior point includes: calculating a standard vector pointing from the femoral medial condyle posterior point to the femoral lateral condyle posterior point, and normalizing the standard vector to obtain a standard unit vector; determining the standard unit vector as the lateral projection direction of the single lower limb three-dimensional model.

8. The method of generating a lower extremity projection image according to claim 7, wherein, determining an en face projection direction of the single lower limb three-dimensional model according to the lateral projection direction and a preset plumb line direction includes: obtaining a plumb line vector corresponding to the preset plumb line direction; calculating a target vector perpendicular to the standard unit vector and the plumb line vector, and normalizing the target vector to obtain a target unit vector; determining the target unit vector as the en face projection direction of the single lower limb three-dimensional model.

9. A lower extremity projection image generation terminal, characterized by, A computer program product includes a memory, a processor, and a computer program stored on the memory and running on the processor, and the processor implements each step in the generation method of the lower limb projection image according to any one of claims 1-8 when executing the computer program.

Images