CN109064472B - Fitting method and device for fitting plane of three-dimensional space model of vertebra - Google Patents

Abstract

The invention relates to a fitting method and device for fitting a three-dimensional space model of a vertebral bone, and is characterized in that the fitting device at least includes an identification module and a fitting module, wherein the identification module takes the triangular piece on the three-dimensional space model of the spine as the The seed points are grown and the adjacent planes with the maximum number of iterations are grown by the breadth-first traversal method to obtain a triangular patch set. The fitting module uses matrix multiplication and calculation matrix singularity based on the normal vector and vertex parameters of the triangular patch identified by the recognition module. The value method computes the equation of two opposing fit planes to obtain two opposing fit surfaces. The spinal correction device of the present invention has universality as a device for intercepting a local model from the whole, and can automatically obtain the spatial position and size of the bounding box to adapt to the vertebrae of different shapes. Compared with the prior art, the present invention has the advantages of simple operation, high automation and high accuracy.

Description

Fitting method and device for fitting plane of three-dimensional space model of vertebra

The present invention is a divisional application whose application number is 201710195660.3, the application date is March 28, 2017, the application type is invention, and the application name is a spinal correction device.

technical field

The invention relates to the technical field of medical images, in particular to a method and device for fitting a three-dimensional space model of a vertebrae to a plane.

Background technique

The chiropractic technique is a technique dedicated to repositioning the deviated and subluxated spine and adjusting the vertebral joints, similar to the pulling method and bone setting of traditional Chinese medicine. Chiropractic technology has systematic theoretical guidance and unique force-generating skills, and is a scientific and efficient treatment technology. Chiropractic technology has been incorporated into an important adjuvant treatment for intractable diseases, and has received very good results. Chiropractic technology has an important adjuvant treatment effect on diabetes, rheumatoid, cardiovascular and cerebrovascular diseases, etc. Most of today's scoliosis correction techniques use curved rod fixation correction. Before the operation, the doctor calculates and predicts the shape of the curved rod to be used in the operation according to the patient's spine CT image (including lateral view and tomographic scan) combined with his own experience, so as to achieve the purpose of using the curved rod to correct the spine.

In the process of generating the curved rod, the doctor can only use the two-dimensional CT image to calculate some spine parameters (such as the cobb angle, lumbar base angle) and determine the spatial relative position of the spine and the overall shape of the spine through the three views of the CT image. These states and the doctor's own experience predict the shape of the spine after correction. It can be seen that traditional chiropractic techniques require experienced clinicians to make good predictions, but relying only on the experience of clinicians and lack of scientific and systematic methods are prone to prediction errors, resulting in The curved rod is modified, which will inevitably increase the operation time, increase the patient's blood loss, and increase the operation risk. Therefore, it has become an urgent technical problem to provide a device specially designed for chiropractic technology to assist doctors to perform corrective surgery well.

To this end, a Chinese patent (publication number CN102968791A) discloses an interactive method and system for displaying three-dimensional medical image graphics. The interactive method of the patent includes the following steps: A. In the scene of the 3D medical image/graphic display, the range of interactive processing is selected by controlling the bounding box; B. The range delimited by the bounding box is applied to the 3D medical image/ In the processing of the graphic display, the corresponding partial display result is obtained; C. The partial display result is output to the segmentation algorithm, and the corresponding segmentation processing is performed. The interactive system of this patent includes: a selection module for selecting a range of interactive processing by controlling a bounding box in a scene of 3D medical image/graphic display; a bounding box processing module for applying the range bounded by the bounding box to In the processing of the three-dimensional medical image/graphic display, the corresponding partial display result is obtained; the execution module is used for outputting the partial display result to the segmentation algorithm, and executes the corresponding segmentation processing.

The method and system provided by this patent can realize the partial display of the region of interest/space of interest in the display mode, which is beneficial to the doctor's observation and diagnosis, but the interaction method and interaction system provided by this patent need to be It takes a lot of time, and it is difficult to quickly and accurately select objects that need to be interacted with. Therefore, there is an urgent need to provide a device that can accurately and automatically obtain the spatial position and size of a bounding box.

During the examination process of the parent case of the present invention, the first Office Action only pointed out the formal problems, and the closest prior art was not retrieved. Accordingly, the present invention has outstanding substantive features and significant advancements.

SUMMARY OF THE INVENTION

Aiming at the problems in the prior art that when the vertebra is partially segmented, the bounding box is manually adjusted to adapt to different shapes of vertebrae, which is time-consuming and labor-intensive, and it is difficult to quickly and accurately obtain the spatial position and size of the bounding box. A spinal correction device for obtaining the spatial position and size of a bounding box, especially a device for partially segmenting a three-dimensional space model of a medical spine to obtain a single vertebra. The method adopted by the spine correction device provided by the present invention is mainly based on the three-dimensional spine model generated in the Visual Toolkit tool. A space bounding box is generated in the space area, the six faces of the bounding box are set as cut planes, and the vertebrae to be divided are stored in the bounding box. inside the box. The bounding box of the present invention can also freely move the position and adjust the size and angle, thus realizing the interception of a single vertebra with different morphological features. Further, in order to make the positioning of the bounding box more accurate and fast, the present invention provides a method of firstly performing plane recognition on the upper and lower planes of the vertebrae to be intercepted on the three-dimensional spine model to obtain parameters such as normal vectors and spatial positions of the upper and lower planes, and then using these parameters. parameters, carry out a realistic calculation process, and finally can automatically generate a more ideal space bounding box, the user only needs to fine-tune to cut out an ideal single vertebra from the overall spine model. Further, in order to make the interception process more convenient and fast, the bounding box processing module adopts the left and right two parts for contrast interception, the left half is placed with the overall 3D spine model and the bounding box, and the right half is placed with the partial model and bounding box intercepted by the bounding box. The bounding boxes on both sides are fully synchronized, which makes the capture process more convenient.

According to a preferred embodiment, the method for cutting a single vertebra by a chiropractic device includes the following process: adopting plane growth and plane fitting based on the normal vector of the surface model of the three-dimensional space model of the spine to determine the shape of the upper and lower planes of the bounding box; Translate the upper and lower fitting planes by a certain distance along their respective normal vectors to include the entire vertebra to be cut; determine the center of the bounding box according to the line connecting the center points of the two fitting planes. The normal vector with the smaller angle with the vertical direction determines the normal vector of the upper and lower planes of the bounding box, thereby determining the shape and size of the bounding box. The left and right parts are compared for interception, and the bounding box is completely interactive, which makes the interception process more convenient and intuitive; after the local area is intercepted, the local model is immediately denoised, the maximum connected domain of the model in space is saved, and small blocks are eliminated. of impurities.

According to a preferred embodiment, a fitting device for fitting a three-dimensional space model of a vertebra to a plane includes at least an identification module and a fitting module. Wherein, the identification module uses the triangular patch on the three-dimensional space model of the spine as a seed point to grow, and uses the breadth-first traversal method to grow adjacent planes with the maximum number of iterations to obtain a triangular patch set, and the fitting module is based on The normal vector and vertex parameters of the triangular pieces identified by the identification module are calculated by the method of matrix multiplication and calculation of matrix singular values to calculate the equations of two relative fitting planes to obtain two opposite fitting planes.

According to a preferred embodiment, the spinal correction device at least includes an identification module, a fitting module, a bounding box processing module and a correction module. The identification module uses the triangular patch on the three-dimensional space model of the spine as a seed point to grow, and uses the breadth-first traversal method to grow the adjacent plane with the maximum number of iterations to obtain a triangular patch set. The fitting module calculates the fitting plane equation by using the method of matrix multiplication and calculating the singular value of the matrix based on the normal vectors and vertex parameters of all the triangular pieces in the triangular piece set identified by the identification module. The bounding box processing module determines the center of the bounding box, the distance between the two opposite faces, and the normal vector of the two opposite faces based on the center points and normal vectors of the two opposite fitting planes fitted by the fitting module. The spatial position and size of the bounding box are determined based on the determined center of the bounding box, the distance between the two opposite faces, and the normal vectors of the two opposite faces. The correction module generates a bending parameter of the curved rod based on the vertebral bone parameters intercepted by the bounding box determined by the bounding box processing module, so as to realize spine correction through the curved rod.

According to a preferred embodiment, the vertebral correction device generates the bending parameters of the curved rod, so that the doctor can customize the curved rod to be corrected according to the patient's spinal pathology. Preferably, the correction module generates the bending scheme of the curved rod based on the vertebral bone parameters intercepted by the bounding box determined by the bounding box processing module. More preferably, the orthodontic device obtains the position and orientation of the curved rod screw based on the vertebral parameters intercepted by the bounding box determined by the bounding box processing module, and converts the screw position and orientation using one or more geometry-based algorithms. into a series of bending instructions. Preferably, the bending algorithm for the bent rod is to acquire and digitize points in space, analyze the points and calculate the bending instructions and the length of the bent rod required to bend the bent rod with a mechanical bending device. The invention can be applied to the spatial model reference of scoliosis surgery and the pre-generation of curved rods in medicine, so as to improve the manufacturing accuracy of the curved rods, reduce the blood volume of patients during the operation, reduce the labor intensity of doctors, reduce the operation time and reduce the operation risk. , which is of great significance in clinical application.

According to a preferred embodiment, the identification module includes at least a selection unit, a label array establishment unit, a first storage unit and a growth unit. The selecting unit is used for selecting an ID of a triangular slice from the surface of the three-dimensional space model of the spine. The marker array establishment unit is used to create a marker array after the selection unit selects the triangular slice, and the marker array is used to mark the usage of the triangular slice and the size of the marker array is the size of the surface of the spine three-dimensional space model to be intercepted. Total number of triangles. The first storage unit is used for storing the data of the three-dimensional space model of the spine, the ID linked list of the triangular slices to be compared, and/or the triangular slice ID linked list identified by the identification module. The growing unit uses the triangular patch normal vector selected by the selecting unit as a reference and uses a breadth-first traversal method to grow adjacent planes with a maximum number of iterations to obtain the triangular patch set.

According to a preferred embodiment, the growing unit obtains the triangular piece set in the following manner: the growing unit starts from the triangular piece selected by the selection unit, and sequentially marks and judges the usage of the triangular piece to be selected. When the triangle to be selected has been used, the growing unit discards the comparison between the triangle to be selected and the seed point triangle; when the triangle to be selected is not used, the growth unit calculates the triangle to be selected. The sum of the absolute values of the component differences between the patch normal vector and the triangular patch normal vector of the seed point.

According to a preferred embodiment, when the growing unit calculates that the sum of the absolute values of the differences of the components of the normal vector of the triangle to be selected and the normal vector of the triangle of the seed point is not greater than 0.5, the growing unit will The linked list of the IDs of the triangle pieces to be selected is sent to the first storage unit for storage. When the growing unit calculates that the sum of the absolute values of the differences between the normal vector of the triangular patch to be selected and the normal vector of the triangular patch of the seed point is greater than 0.5, the growing unit accesses the next triangular patch to be selected. Compared with the prior art judging the angle between the two vectors to generate a new plane, the present invention has the advantages of simplicity, high efficiency and time saving by judging the sum of the absolute values of the differences of the components of the two vectors.

According to a preferred embodiment, the growing unit takes the triangular patch stored in the first storage unit as the object to be compared, and uses the difference between the normal vector of the triangular patch to be selected and each component of the normal vector of the triangular patch to be compared. The linked list of the to-be-selected triangle slice IDs whose absolute value sum is not greater than 0.5 is sent to the first storage unit for storage and used as the object to be compared in the next cycle, and this cycle is repeated until all triangle slices are traversed, and After the loop is over, the growing unit retrieves the linked list of triangle IDs stored in the first storage unit and colors the corresponding triangles with a color different from that of the unselected triangles to generate an approximately planar hyper flat.

According to a preferred embodiment, the fitting module includes at least a first calculation unit, a second storage unit, a second calculation unit, a third calculation unit and a verification unit. Wherein, the first computing unit averages the normal vectors of all the triangular patches in the triangular patch set identified by the identification module to obtain the average normal vector of the triangular patches and stores the average normal vector to the second storage unit. The second storage unit is configured to store the average normal vector and vertex parameters of all the triangular patches in the triangular patch set identified by the identification module. The second calculating unit calculates the average values of the triangle vertices in the X, Y and Z axes respectively based on the vertex parameters stored in the second storage unit. The third calculation unit adopts matrix multiplication and calculates matrix singular values based on the average normal vector calculated by the first calculation unit and the average value of the triangular patch vertices in the X, Y, and Z axes calculated by the second calculation unit method to calculate the fitted plane equation. The verification unit calculates the normal vector of the fitted plane based on the fitted plane equation and compares it with the average normal vector calculated by the first calculation unit, and recalculates the normal vector when the difference between the two reaches a set threshold. Describe the fitted plane equation. Preferably, the fitting module uses matrix multiplication and the method of calculating the singular value of the matrix to calculate the upper and lower two vertebrae of the single vertebra to be intercepted based on the normal vectors and vertex parameters of all the triangular pieces in the triangular piece set identified by the recognition module. The fitted plane equation for the plane.

The average normal vector and the vertex average value calculated by the first calculation unit and the second calculation unit of the present invention are used as the reference vector and the reference value for calculating the center of the fitting plane, and are not directly used for the fitting plane, thereby reducing the difference between the fitting plane and the fitting plane. The deviation of the ideal plane makes the fitted plane more accurate.

According to a preferred embodiment, the bounding box processing module determines the center of the bounding box according to the line connecting the center points of the two opposite surfaces fitted by the fitting module, and uses the length of the center line as the bounding box. The distance between the two opposite faces of the box, the normal vector of the two opposite faces of the bounding box is determined according to the smaller angle between the normal vector of the two opposite faces and the vertical direction, and the bounding box processing module is based on the The center of the bounding box, the distance between the two opposite faces of the bounding box, and the normal vector of the two opposite faces of the bounding box are used to determine the spatial position and size of the bounding box.

According to a preferred embodiment, the bounding box processing module includes at least a bounding box determination unit, an adjustment unit and a post-processing unit. Wherein, the bounding box determining unit determines the spatial position of the bounding box based on the center point and normal vector parameters of the two opposite surfaces fitted by the fitting module. The adjusting unit rotates the spatial position of the bounding box determined by the bounding box determining unit so that the bounding box can fit the surface of the three-dimensional space model of the spine. After the post-processing unit intercepts the vertebrae based on the bounding box rotated by the adjustment unit, the local model intercepted in the bounding box is denoised to save the maximum connected domain of the local model in space.

According to a preferred embodiment, the bounding box determining unit determines the spatial position of the bounding box in the following manner: the bounding box determining unit is based on calculating the difference between the center points of the two opposite surfaces fitted by the fitting module. The average value is obtained to obtain the center of the bounding box, and the bounding box determination unit extends the bounding box in the X and Y axis directions according to the left and right preset thresholds of the center, respectively, and the bounding box in the Z axis direction according to the predetermined threshold. The center extends up and down by a preset threshold to determine the spatial position of the bounding box.

According to a preferred embodiment, the adjustment unit adjusts the spatial position of the bounding box in the following manner: the adjustment unit calculates the space vector with the smallest angle with the Z axis and determines the angle of the angle between the space vector and the Z axis value, the adjustment unit calculates the common vertical line between the space vector and the Z axis, and the adjustment unit makes the bounding box take the common vertical line as the rotation axis, and use the angle between the space vector and the Z axis Rotate for the rotation angle.

The spinal correction device of the present invention, as a device for intercepting a local model from the whole, has versatility, and can automatically obtain the spatial position and size of the bounding box to adapt to different shapes of vertebrae. In order to better adapt to different shapes of vertebrae, The bounding box of the present invention is also capable of automatic rotation. Compared with the prior art, the present invention has the advantages of simple operation, high automation and high accuracy.

Description of drawings

Fig. 1 is the module schematic diagram of a preferred embodiment of the chiropractic device of the present invention;

FIG. 2 is a schematic diagram of the effect of a preferred embodiment of the bounding box determined by the present invention; and

FIG. 3 is a schematic diagram of the effect of another preferred embodiment of the bounding box determined by the present invention.

List of reference signs

10: Recognition module 101: Select unit

102: mark array establishment unit 103: first storage unit

104: Growth Unit 20: Fitting Module

201: first computing unit 202: second storage unit

203: The second computing unit 204: The third computing unit

205: Verification unit 30: Bounding box processing module

301: Bounding box determination unit 302: Adjustment unit

303: Post-processing unit 40: Correction module

Detailed ways

The following describes in detail with reference to the accompanying drawings and embodiments.

Aiming at the embarrassing situation that only two-dimensional CT images can be used for simple measurement and estimation in the current hospital for scoliosis correction, the present invention provides a spinal correction device, through which the spinal bone model generated by the three-dimensional modeling technology can be analyzed. Perform local segmentation and use the segmented local model to simulate and correct the simulation. Specifically, by using the MarchingCubes method of the Visual Toolkit (VTK) tool for 3D reconstruction of CT tomographic images, a 3D image of the human skeleton is restored, and then the reconstructed 3D model is segmented and processed to finally obtain a 3D model of a single vertebra. And save them separately for subsequent measurement and adjustment. Preferably, by encapsulating different processing classes, VTK enables users to conveniently generate 3D models using CT tomographic images and MRI images, including volume rendering and surface rendering. More preferably, the present invention uses a surface rendering method to perform three-dimensional modeling on a set of CT tomographic scans to generate a realistic three-dimensional surface model. The 3D surface model contains a series of information required by the 3D model, such as points, lines, surfaces, normal vectors, etc. With this information, a series of operations can be performed on the 3D model to achieve functions such as surgical simulation.

Further, the present invention provides a device for obtaining a single vertebra by partial segmentation using a three-dimensional space model of a medical spine. The device uses a spatial bounding box and manually adjusts the size and spatial shape of the bounding box to locally intercept the entire spine. The device can also perform plane recognition on the upper and lower surfaces of the vertebrae on the three-dimensional space model of the spine, and determine the position and shape of the space bounding box by using various parameters of the recognized plane (such as the normal vector and the length in each direction), and then fine-tune to determine the position and shape of the space bounding box. A single vertebra was cut. Label the cut-out vertebrae and store the necessary vertebrae parameters, so as to facilitate the subsequent vertebra adjustment and correction simulation work.

The terms involved in the present invention are explained as follows.

_Breadth -first traversal method: The breadth-first traversal method is a traversal strategy for connected graphs. The breadth-first traversal method is in the order of layers, and after all nodes on a certain layer are searched, the next layer is searched. It includes three steps: (1) Starting from a certain vertex V ₀ in the graph and visiting this vertex. (2) Starting from V ₀ , visit each unvisited adjacent point W ₁ , W ₂ ...... W _K of V ₀ , and then visit each unvisited adjacent point from W ₁ , W ₂ ...... W _K accordingly . (3) Repeat step (2) until all vertices are visited.

Bounding Box: The bounding box is defined as the smallest hexahedron containing the object with sides parallel to the coordinate axes. The bounding box can be freely moved and rotated in the spatial coordinate system, and can be freely resized. The bounding box has seven balls that control the position of each face and the overall position. When the mouse is clicked on the face instead of the ball, the center ball is used as the center of rotation for free rotation.

Example 1

Fig. 1 shows a schematic block diagram of a preferred embodiment of the chiropractic device of the present invention. As shown in FIG. 1 , the spinal correction device of the present invention at least includes an identification module 10 , a fitting module 20 , a bounding box processing module 30 and a correction module 40 . The identification module 10 uses the triangular patch on the three-dimensional space model of the spine as the seed point to grow, and uses the breadth-first traversal method to grow the adjacent plane with the maximum number of iterations to obtain the triangular patch set. The fitting module 20 calculates the equations of the two relative fitting planes based on the normal vectors and vertex parameters of the triangular pieces identified by the identifying module 10 by means of matrix multiplication and calculating matrix singular values to obtain two opposite fitting planes. Preferably, the two opposite surfaces are the upper and lower surfaces of the single vertebra to be cut. The bounding box processing module 30 determines the center of the bounding box, the distance between the two opposite faces, and the normal vector of the two opposite faces based on the center points and normal vectors of the two opposite fitting planes fitted by the fitting module 20, and based on the Determine the center of the bounding box, the distance between the two opposite faces, and the normal vector of the two opposite faces to determine the spatial position and size of the bounding box. The correction module 40 generates the bending parameters of the curved rod based on the vertebral bone parameters intercepted by the bounding box determined by the bounding box processing module 30 to realize spine correction through the curved rod. A schematic diagram of the effect of the bounding box determined in this embodiment is shown in FIG. 2 .

According to a preferred embodiment, each triangular piece of the spine three-dimensional space model has a normal vector. When a certain triangular piece is selected for regional growth, it can be compared with the normal vector of the adjacent triangular pieces, and the direction to the selected triangle The adjacent triangular slices whose space included angle does not exceed a certain small measure (determined according to the actual situation) are grown, and the spatial plane fitting can be realized by performing plane fitting with many triangular slices after the growth. Preferably, many triangular pieces are obtained by the following method: selecting a triangular piece on a plane by using a triangular piece selector, using the triangular piece as a seed point to grow, combining the normal vector of each triangle piece, and growing according to the seed The angle between the normal vector of the point triangle and the normal vector of other triangles, the breadth-first traversal method is used to select the adjacent triangles with the angle within a small range, and the adjacent planes with the maximum number of iterations are grown, so that we can get Larger range of flat triangles.

Referring to FIG. 1 again, the identification module 10 at least includes a selection unit 101 , a label array establishment unit 102 , a first storage unit 103 and a growth unit 104 . Preferably, the selecting unit 101 is used to select an ID of a triangular slice from the surface of the three-dimensional space model of the spine. Preferably, the marker array establishment unit 102 is configured to create a marker array after the selection unit 101 selects a triangular slice, the marker array is used to mark the usage of each triangle slice and the size of the marker array is the size of the surface of the spine three-dimensional space model to be intercepted. The total number of triangles. Preferably, after the selection unit 101 selects a triangle slice from the surface of the spine three-dimensional space model, if there is no marker array, a marker array is created by the marker array establishment unit 102, and the marker array is used to determine the triangle on the plane to be identified. whether the slice is used. More preferably, the size of the marker array is the total number of triangles in the plane where the selected triangle is located. Preferably, the first storage unit 103 is used for storing the data of the three-dimensional space model of the spine, the ID linked list of the triangular slices to be compared and/or the linked list of the IDs of the triangular slices finally selected. Preferably, the ID linked list of the triangles to be compared is the IDs of the triangles selected by the selection unit 101 from the surface of the spine three-dimensional space model, and the data stored in the first storage unit 103 is prepared for the breadth-first traversal of the growth unit 104 . Preferably, the growing unit 104 uses the normal vector of the triangular patch selected by the selecting unit 101 as a reference vector, and uses the breadth-first traversal method to grow adjacent planes with the maximum number of iterations to obtain a set of triangular patches that meet the requirements. Preferably, the breadth-first traversal method at least includes the following steps:

S1: The growing unit 104 firstly sets the number of cycles of breadth-first traversal. Preferably, the number of cycles is the total number of triangles in the plane where the triangles selected by the selection unit 101 are located.

S2: During the first cycle, the growth unit 104 first obtains the normal vector information of all triangular slices in the three-dimensional space model of the spine. Preferably, the normal vector information of all triangular pieces is obtained and stored in the three-dimensional data when the VTK constructs the three-dimensional surface model. More preferably, the first storage unit 103 stores the data of the three-dimensional space model of the spine, and the growth unit 104 can obtain the normal vector information of all triangular slices from the first storage unit 103 .

S3: The growing unit 104 starts to cycle from the triangular slices selected by the selection unit 101, and sequentially marks and judges the usage of the triangular slices to be compared. When the triangle to be selected has been used, the growing unit 104 discards the comparison of the triangle to be selected with the seed point triangle. When the triangular patch to be selected is not used, the growing unit 104 marks and compares the sum of the absolute values of the differences of the components of the normal vector of the triangular patch to be selected and the triangular patch of the seed point.

Preferably, when the growing unit 104 calculates that the sum of the absolute values of the difference between the normal vector of the triangle to be selected and the normal vector of the seed point triangle is not greater than 0.5, the growing unit 104 sends the linked list of the IDs of the triangles to be selected to The first storage unit 103 performs storage. The triangular patch to be selected stored in the first storage unit 103 is used as a triangular patch in the final result set. When the growing unit 104 calculates that the sum of the absolute values of the difference between the normal vector of the triangle to be selected and the normal vector of the seed point triangle is greater than 0.5, the growing unit 104 accesses the next triangle to be selected.

S4: The growing unit 104 starts from the triangular slices to be selected that have been screened, finds its adjacent triangular slices, and stores the IDs of the adjacent triangular slices in the array to be compared in the first storage unit 103, as the next iteration in the cycle. To select the triangular piece to use, enter the next cycle until the number of cycles is used up.

Preferably, the growing unit 104 takes the selected triangular patch as the object to be compared, and selects the triangular patch whose sum of absolute values of the difference between the normal vector of the triangular patch to be selected and the normal vector of the triangular patch to be compared is not greater than 0.5 The ID is stored in the to-be-compared array of the first storage unit 103 to be used as the to-be-compared object in the next cycle, and the cycle is repeated until all triangles are traversed.

S5: After the loop ends, the growing unit 104 retrieves the triangle ID stored in the first storage unit 103 and colorizes the corresponding triangle with a color different from the unselected triangle to generate an approximate plane hyperplane. Preferably, the triangular pieces stored in the first storage unit 103 are approximately on the same plane.

After obtaining the triangle set that approximates the plane, the fitting module 20 uses the selected triangle pieces to perform plane fitting to generate a plane, and the generated plane serves as the basis for the subsequent positioning of the bounding box.

Continuing to refer to FIG. 1 , the fitting module 20 includes at least a first calculation unit 201 , a second storage unit 202 , a second calculation unit 203 , a third calculation unit 204 and a verification unit 205 . Preferably, the first calculation unit 201 averages the normal vectors of the triangular patches identified by the identification module 10 to obtain an average normal vector of the triangular patches and stores the averaged normal vectors in the second storage unit 202 . Preferably, the second storage unit 202 further stores the vertex parameters of the triangle pieces identified by the identification module 10 . Preferably, the second calculating unit 203 calculates the average values of the triangle vertices in the X, Y and Z axes respectively based on the vertex parameters stored in the second storage unit 202 . The average value is used as the reference value for the subsequent calculation of the center of the fitted plane, rather than directly used as the center of the fitted plane, so that the deviation between the fitted plane and the ideal plane can be reduced. Preferably, the third calculating unit 204 adopts the method of matrix multiplication and calculating the singular value of the matrix based on the average normal vector calculated by the first calculating unit 201 and the average value of the triangular patch vertices in the X, Y and Z axes calculated by the second calculating unit 203 Calculate the equation for the fitted plane. Preferably, the third calculation unit 204 specifies the plane coefficients of the fitted plane according to the singular vectors obtained from the calculated singular values of the matrix. Preferably, the verification unit 205 calculates the normal vector of the fitted plane based on the fitted plane equation and compares it with the average normal vector calculated by the first calculation unit 201, and recalculates the fitted plane when the deviation between the two reaches a set threshold. formula. More preferably, the projection direction of the normal vector of the fitted plane on the original average normal vector is consistent with the original average normal vector. Among them, the normal vector of the fitting plane points from the inside of the vertebral plane to the outside of the vertebral plane.

After fitting by the fitting module 20, an approximate plane fitting plane can be generated on the surface of the three-dimensional space model of the spine, and a plane floating on the surface of the three-dimensional space model of the spine can be generated by appropriately translating the fitted plane. After fitting the plane, the bounding box processing module 30 can further determine the spatial position and size of the bounding box. Preferably, the bounding box processing module 30 determines the center of the bounding box according to the line connecting the center points of the two opposite faces fitted by the fitting module 20, and takes the length of the center line as the distance between the two opposite faces of the bounding box, according to The normal vector of the two opposite faces of the bounding box is determined by the smaller angle between the normal vectors of the two opposite faces and the vertical direction. The bounding box processing module 30 determines the position and size of the bounding box based on the center of the bounding box, the distance between two opposite faces of the bounding box, and the normal vectors of the two opposite faces of the bounding box. Subsequent fine-tuning can easily and accurately cut out a single piece of vertebrae.

Continuing to refer to FIG. 1 , the bounding box processing module 30 at least includes a bounding box determination unit 301 , an adjustment unit 302 and a post-processing unit 303 . Preferably, the bounding box determining unit 301 determines the spatial position of the bounding box based on the center point and normal vector parameters of the two opposite surfaces fitted by the fitting module 20 . Preferably, the adjusting unit 302 rotates based on the spatial position of the bounding box determined by the bounding box determining unit 301 so that the bounding box can fit the surface of the three-dimensional space model of the spine. Preferably, after the post-processing unit 303 intercepts the vertebrae based on the bounding box rotated by the adjustment unit 302, the local model intercepted in the bounding box is denoised to save the maximum connected domain of the model in space. Preferably, the vtkPolyDataConnectivity Filter method in VTK is used to extract the maximum connected domain of the intercepted part.

According to a preferred embodiment, the bounding box determination unit 301 determines the spatial position of the bounding box in the following manner: the bounding box determination unit 301 obtains the bounding box based on the average value of the center points of the two opposite surfaces fitted by the fitting module 20 box center. The bounding box determining unit 301 extends the bounding box in the X and Y axis directions to the left and right according to the preset thresholds respectively, and extends the bounding box in the Z axis direction respectively according to the center up and down the preset threshold value to determine the spatial position of the bounding box. Preferably, the distance that the bounding box extends left and right in the X and Y axis directions according to the center is greater than the distance that the bounding box extends up and down according to the center in the Z axis direction. The specific extension distance can be adjusted based on the actual situation. For example, the bounding box extends 30 to 60 units left and right in the X and Y axis directions, respectively, according to the center, and the bounding box extends 1 to 2 units up and down in the Z axis direction, respectively, according to the center.

According to a preferred embodiment, the adjustment unit 302 adjusts the spatial position of the bounding box in the following manner: the adjustment unit 302 calculates the space vector with the smallest angle with the Z axis and determines the angle value of the angle between the space vector and the Z axis, and the adjustment unit 302 calculates The space vector is the common perpendicular to the Z axis. Preferably, the adjustment unit 302 obtains the common perpendicular by calculating the three components of the common perpendicular and using the third-order determinant to obtain the cross product of the two vectors. The adjustment unit 302 rotates the bounding box with the common perpendicular as the rotation axis and the included angle between the space vector and the Z axis as the rotation angle.

Example 2

The spinal correction device provided in Embodiment 1 of the present invention is used as a device for cutting a partial model from an overall model. It is versatile and easy to adjust. However, for some complex and imprecise three-dimensional models, the device provided in Embodiment 1 is used for cutting. It seems that the steps are cumbersome, for this reason, Example 2 provides a method of performing multiple adjustments by a manual method for cutting.

According to a preferred embodiment, the spine correction device uses VTK to take a point on the surface of the constructed 3D model, and uses this point as the center to expand several units in the positive and negative directions of the X, Y, and Z directions to generate a VTKBox bounding box. The six faces of the bounding box are set as cutting planes, and the model inside the bounding box is taken as the interception result, so that partial interception can be realized. The bounding box can be enlarged, reduced, and rotated to accommodate different vertebrae. Preferably, in order to avoid the influence of the intercepted impurities on subsequent operations, the vertebral correction device extracts the largest connected body in the space as the final result of the interception of the spine region. A schematic diagram of the effect of the bounding box determined in this embodiment is shown in FIG. 3 .

According to a preferred embodiment, the chiropractic device divides the interception window into two parts, left and right, and the left half is used to display the entire three-dimensional spine space model. The spine correction device first selects a point on the surface of a certain vertebra to be intercepted on the three-dimensional spine space model on the left side of the interception window, and then generates a bounding box of a fixed size with the point as the center. At the same time, the same bounding box and the extracted local model are displayed in the interception window on the right. Both sides are completely synchronized, and the size and rotation angle of the bounding box can be adjusted manually. The vertebral correction device of the present invention divides the interception window into two parts, left and right, for comparison, so that the intercepted model is more accurate.

The accuracy of the single vertebrae obtained by the vertebral correction device used in Example 1 and Example 2 was compared in the following manner, and the comparison results are shown in Table 1 and Table 2.

The spine correction device of Embodiment 1: the spine correction device first identifies the upper and lower surfaces of the vertebrae to be intercepted, determines a center point and a normal vector based on the identified surface, and automatically generates a space bounding box with a reasonable spatial position and size, and fine-tunes it through mouse interaction. The location and size of the bounding box to extract a single vertebra.

The spine correction device of Embodiment 2: The spine correction device first selects a point on the surface of the three-dimensional space model of the spine, and defines a bounding box radius in the X, Y, and Z directions with the point as the center, thereby generating a parallel to the coordinate axis. Spatial bounding box, and then adjust the size and position of the bounding box through mouse interaction, so as to cut out a single vertebra.

First determine the center point of the ideal model, and then compare all the points of the actual initial interception models in Example 1 and Example 2 with the center point of the ideal model. Here, the criterion for judging the initial interception effect of a local model is: the closer the intercepted model is to the center point of the ideal model, the closer the model is to the ideal model, and the better it can represent the shape of the ideal model. The number of points whose distance from the center point is greater than a certain threshold is used as the standard. The smaller the number, the better the model can represent the shape characteristics of the ideal model. Because the larger the interval is, the more discrete the points of the model will be. Preferably, the threshold is determined according to the boundary value in the X direction of the ideal model, and the method is to first obtain the boundary values in the X, Y, and Z directions of the ideal model, and then take 1/1 of the distance value in the X direction. 2 as the criterion for judgment. The X-direction is chosen here because the spine slope in the X-direction is larger, which makes it easier to reflect the differences.

Therefore, according to the two values, the number of points that are greater than a certain threshold from the center point of the ideal model (called the number of failure points) and the percentage of valid points to the total number of points in the model (called the effective rate), the two values can be compared. The difference of the effect of each method, the smaller the number of failure points, the greater the efficiency of the model, the closer to the ideal interception model. Here, the ideal truncated model refers to a single vertebrae model obtained by extracting the largest spatially connected domain of the truncated model, which does not contain small impurities and belongs to the model that the user finally wants to obtain.

Table 1 The number of failure points and the effective rate of a single vertebra cut in Example 1 and Example 2

From Table 1, it can be seen that the effective rate of the single vertebra cut in Example 1 is 70%-93.1%, and the effective rate of the single vertebra cut in Example 2 is 54.1%-75.0%. Among them, the curvature of the vertebrae from T4 to L1 increased in turn, and the curvature of the vertebrae from L1 to L4 decreased in turn. It can also be seen from Table 1 that the difference between Example 1 and Example 2 is greater at the position of the vertebrae with larger curvature. It can be seen that the vertebral correction device of Example 1 is closer to the ideal model when cutting a single vertebra. , especially when the curvature of the vertebral bone is large, the use of the vertebral correction device of implementation 1 can better cut out a single vertebral bone solved by the ideal model.

Using the ratio of the number of intercepted model points in Example 1 and Example 2 to the number of ideal intercepted model points as a comparison standard, the results are shown in Table 2.

Table 2 The ratio of the number of intercepted model points to the ideal model in Example 1 and Example 2

spine number Example 1 Example 2 T4 1.12334 1.24389 T5 1.16403 1.22328 L1 1.12559 1.22543 L2 1.12554 1.34737 L3 1.09258 1.28872 L4 1.06319 1.15600

It can be seen from Table 2 that the ratio of the intercepted model to the ideal model in Example 1 is closer to 1.

To sum up, compared with Example 2, the single vertebrae obtained by the spinal correction device of Example 1 is more accurate and closer to the ideal model.

It should be noted that the above-mentioned specific embodiments are exemplary, and those skilled in the art can come up with various solutions inspired by the disclosure of the present invention, and these solutions also belong to the disclosure scope of the present invention and fall within the scope of the present invention. within the scope of protection of the invention. It should be understood by those skilled in the art that the description of the present invention and the accompanying drawings are illustrative rather than limiting to the claims. The protection scope of the present invention is defined by the claims and their equivalents.

Claims (4)

1. A fitting device for a three-dimensional space model fitting plane of a vertebra, wherein the fitting device at least comprises an identification module (10) and a fitting module (20), wherein, The identification module (10) uses the triangular slices on the three-dimensional space model of the spine as seed points to grow and uses the breadth-first traversal method to grow adjacent planes with the maximum number of iterations to obtain a triangular slice set, The fitting module (20) uses matrix multiplication and the method of calculating matrix singular values to calculate the equations of two relative fitting planes based on the normal vectors and vertex parameters of the triangular pieces identified by the identification module (10) to obtain two equations. The relative fitting surface, The identification module (10) includes at least a first storage unit (103), and the first storage unit (103) is used for storing the data of the three-dimensional space model of the spine, the ID linked list of the triangle slices to be compared and/or the identification The triangular piece ID linked list identified by the module (10); The fitting module (20) includes at least a first calculation unit (201) and a second storage unit (202), wherein, The first computing unit (201) averages the normal vectors of all the triangular patches in the triangular patch set identified by the identifying module (10) to obtain the average normal vector of the triangular patches and uses the The average normal vector is stored in the second storage unit (202); The second storage unit (202) is configured to store the average normal vector and the vertex parameters of all the triangle patches in the triangle patch set identified by the identification module (10). The second storage unit (202) is used for for storing the average normal vector and the vertex parameters of all triangular patches in the triangular patch set identified by the identification module (10); The fitting module (20) further includes a second calculation unit (203), which respectively calculates the triangle patch vertices based on the vertex parameters stored in the second storage unit (202) Average value in X, Y, Z axis; The fitting module (20) further includes a third calculation unit (204), the third calculation unit (204) is based on the average normal vector calculated by the first calculation unit (201) and the second calculation unit ( 203) Calculate the average value of the triangular piece vertices in the X, Y, and Z axes by using the method of matrix multiplication and calculating the singular value of the matrix to calculate the fitting plane equation. 2. The fitting device according to claim 1, characterized in that, the fitting device further comprises a bounding box processing module (30), wherein, The bounding box processing module (30) determines the center of the bounding box, the distance between the two opposite surfaces, and the The space position and size of the bounding box are determined based on the determined center of the bounding box, the distance between the two opposing faces, and the normal vectors of the two opposing faces. 3. The fitting device according to claim 2, wherein the fitting module (20) further comprises a verification unit (205), wherein the verification unit (205) calculates the The normal vector of the fitted plane is compared with the average normal vector calculated by the first calculation unit (201), and the fitted plane equation is recalculated when the deviation between the two reaches a set threshold. 4. The fitting device according to claim 3, wherein the average normal vector and the vertex average calculated by the first calculation unit (201) and the second calculation unit (203) are used as reference vectors and Calculate the reference value for the center of the fitted plane.

Images