Implant space attitude estimation method for CT image

Abstract

The invention discloses a CT image-oriented implant space attitude estimation method, and belongs to the technical field of biomedical engineering and computer science. The method comprises the following steps: extracting a target three-dimensional model based on a CBCT image; the CBCT image is obtained by performing cone beam scanning on a target part where an implant is placed during or after an operation; obtaining an implant theoretical model pre-constructed for the implant; the mass center of the target three-dimensional model and the mass center of the implant theoretical model are calculated respectively, and the mass center of the target three-dimensional model and the mass center of the implant theoretical model are aligned; registering the target three-dimensional model and the implant theoretical model of which the mass centers are aligned by utilizing a main shaft rotating mode; and evaluating the registration precision based on multi-plane projection and similarity measurement, and outputting spatial attitude parameters of the implant. According to the method, the registration speed can be increased, the calculation complexity in the registration process is reduced, and the space registration precision can be further improved through a two-stage optimization strategy of centroid alignment and main shaft rotation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of biomedical engineering and computer science, and particularly relates to a CT image-oriented implant spatial posture estimation method. BACKGROUND

[0002] In medical operations of endosseous implants, such as oral implantation and orthopedic implant placement, it is crucial to accurately determine the spatial posture of the implant in the body. With the development of computer technology and medical imaging technology, in related technologies, CT images are segmented by threshold value to extract implant information, and then the spatial posture of the implant is estimated.

[0003] However, when the CT image has noise, artifacts, or the contrast between the implant and the surrounding tissue is low, the segmentation result will be inaccurate, which will affect the estimation accuracy of the spatial posture of the implant.

[0004] Therefore, there is an urgent need to provide a more accurate implant spatial posture estimation method. SUMMARY

[0005] The present application provides a CT image-oriented implant spatial posture estimation method, which can solve the problem of poor estimation accuracy in related technologies. The technical solution is as follows: On the one hand, a CT image-oriented implant spatial posture estimation method is provided, which comprises: extracting a target three-dimensional model based on a CBCT image; the CBCT image is obtained by cone beam scanning of a target site where an implant is placed during or after surgery; obtaining an implant theoretical model constructed in advance for the implant; calculating the centroids of the target three-dimensional model and the implant theoretical model respectively, and aligning the centroids of the target three-dimensional model and the implant theoretical model; aligning the target three-dimensional model and the implant theoretical model after centroid alignment by using the main shaft rotation method; evaluating the registration accuracy based on multi-plane projection and similarity measurement, and outputting the spatial posture parameters of the implant.

[0006] On the other hand, a CT image-oriented implant spatial posture estimation device is provided, which comprises: an extraction unit configured to extract a target three-dimensional model based on a CBCT image; the CBCT image is obtained by cone beam scanning of a target site where an implant is placed during or after surgery; an acquisition unit configured to obtain an implant theoretical model constructed in advance for the implant; an alignment unit configured to calculate the center of mass of the target three-dimensional model and the implant theoretical model respectively, and to align the center of mass of the target three-dimensional model and the implant theoretical model; a registration unit configured to register the target three-dimensional model and the implant theoretical model in the center of mass alignment by using a principal axis rotation method; an evaluation unit configured to evaluate the registration accuracy based on multi-plane projection and similarity measurement, and to output the spatial pose parameters of the implant.

[0007] In another aspect, a computer device is provided, which comprises a memory and a processor, the memory is configured to store a computer program, and the processor is configured to execute the computer program stored in the memory to implement the steps of the CT image-oriented implant spatial pose estimation method.

[0008] In another aspect, a computer readable storage medium is provided, which stores a computer program, and the computer program is executed by a processor to implement the steps of the CT image-oriented implant spatial pose estimation method.

[0009] In another aspect, a computer program product is provided, which comprises a computer program, and the computer program is executed by a processor to implement the steps of the CT image-oriented implant spatial pose estimation method.

[0010] The technical solution provided by the present application can bring at least the following beneficial effects: The present application can improve the registration speed and reduce the calculation complexity in the registration process by obtaining the target three-dimensional model and the implant theoretical model, calculating the center of mass of the two models, and implementing the registration by the center of mass alignment operation and the principal axis rotation method. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor.

[0012] Figure 1 is a CT image-oriented implant spatial pose estimation method flow chart provided by an embodiment of the present application; Figure 2 is a CT image-oriented implant spatial pose estimation device structure diagram provided by an embodiment of the present application; Figure 3 is a hardware architecture diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0013] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in connection with the drawings of the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0014] The specific implementation of the present application concept will be described below.

[0015] Reference is made to Figure 1 The present application provides a CT image-oriented implant spatial posture estimation method, which comprises the following steps: Step 100: obtaining a target three-dimensional model and an implant theoretical model; the target three-dimensional model is obtained by cone beam scanning of a target part after implantation; the implant theoretical model is a three-dimensional geometric model constructed in advance for the implant; Step 102: calculating the center of mass of the target three-dimensional model and the implant theoretical model, and aligning the center of mass of the target three-dimensional model and the implant theoretical model; Step 104: using a principal axis rotation method to register the target three-dimensional model and the implant theoretical model after center of mass alignment; Step 106: evaluating the registration accuracy based on multi-plane projection and similarity measurement, and outputting the spatial posture parameters of the implant.

[0016] In the embodiments of the present application, the target three-dimensional model and the implant theoretical model are obtained, the center of mass of the two is calculated, and the registration is realized through the center of mass alignment operation and the principal axis rotation method. This not only improves the registration speed and reduces the calculation complexity in the registration process, but also further improves the spatial registration accuracy through the two-stage optimization strategy of center of mass alignment and principal axis rotation.

[0017] The execution of each step shown in the following description Figure 1 is described.

[0018] First, for step 100, the target three-dimensional model is extracted based on the CBCT image.

[0019] During the surgical procedure of implanting an implant into the target site, it is necessary to continuously estimate the spatial orientation of the implant to determine whether the implantation has been completed. After the implant has been implanted into the target site, it is also necessary to estimate the spatial orientation of the implant to observe its postoperative status. Therefore, this CBCT image is obtained by cone-beam scanning of the target site where the implant was placed during or after the procedure.

[0020] In this embodiment of the invention, all CT slice data after cone-beam scanning are stored in a designated file within a designated directory. All CT slice data can be loaded by traversing the designated files within that directory. This designated file can be a DICOM format file.

[0021] During storage and transmission, CT slice data may lose its natural spatial order due to scanning order, storage logic, and other reasons. Cone-beam CT requires two-dimensional slices to be stacked correctly according to spatial position in order to form a continuous and ordered three-dimensional voxel stack (i.e., CBCT image). If not sorted, the disordered slice order will cause misalignment and tortuosity in the subsequent three-dimensional reconstructed image, distorting the three-dimensional morphology of anatomical structures (such as teeth and jawbone), and seriously affecting the accuracy of medical analysis (such as dental arch morphology judgment, implant planning, etc.). Therefore, in this embodiment of the invention, after loading all CT slice data, the specified file can be spatially sorted according to the slice position information. The sorting basis is: sort_key(dcmi)=ImagePositionPatient[2], to obtain a continuous and ordered image voxel stack. The continuous and ordered image voxel stack is the CBCT image.

[0022] ImagePositionPatient is a core tag in the DICOM standard, used to characterize the image position in the patient coordinate system. It is a three-dimensional coordinate (x, y, z) that describes the spatial position of the CT slice in the patient's body. ImagePositionPatient[2] is used to characterize the third component of the three-dimensional coordinate (that is, the z component, which usually corresponds to the vertical direction of the human body, i.e., the z-axis direction). The sorting logic mainly uses the value of the z component in the ImagePositionPatient tag of each DICOM file as the sorting index to sort all CT slices according to the coordinate value in the z-axis direction, so that the slices are arranged sequentially in space along the z-axis direction (such as from bottom to top or from top to bottom), and finally form a continuous and ordered image voxel stack.

[0023] It should be noted that the location information of the slice refers to the spatial coordinates of the CT slice itself in the patient coordinate system, which is used to describe the spatial positioning of the slice.

[0024] In this embodiment of the invention, after acquiring the CBCT image, the target 3D model can be extracted in the following manner: S11: Extract pixel arrays from CBCT images and reconstruct 3D volumetric data based on the pixel arrays; wherein the data dimensions of the 3D volumetric data are: ; S12: Convert the reconstructed 3D volume data from the anatomical LPS coordinate system to the medical imaging standard RAS coordinate system; the conversion formula is: in, The coordinates are converted to the standard medical imaging coordinate system; flip() is the flip operation function, used to flip the input vector or array on the specified coordinate axis (such as mirroring or exchanging directions) to achieve orientation alignment between coordinate systems; V is the coordinates in the anatomical coordinate system before conversion; axis = (0,1) is used to specify the coordinate axis dimension to be flipped. axis = (0,1) indicates that the flip is performed on the 0th and 1st axes (axis index starts from 0). This is because the LPS coordinate system and the RAS coordinate system are defined oppositely in the "left / right (L / R)" and "front / back (P / A)" directions, and the coordinate system conversion needs to be completed by flipping these two axes.

[0025] S13: Convert the 3D volume data, which has been transformed to the medical imaging standard coordinate system, into the SimpleITK image format and normalize its pixel intensity; S14: The image is subjected to binarization masking based on the global threshold segmentation algorithm to obtain a binarization mask; S15: Optimize the connectivity of the binarization mask through morphological closing operations to obtain a binarized 3D model of the target.

[0026] The normalization method in step S13 is as follows: Where I(x,y,z) is the normalized value of the 3D data at spatial coordinates (x, y, z); V(x,y,z) is the original value of the 3D volume data at spatial coordinates (x, y, z) before normalization (such as CT value, gray value, etc.); μ is the mean of the 3D volume data before normalization, that is, the average value of all voxel original values ​​V(x,y,z), used to reflect the central tendency of the data; σ is the standard deviation of the 3D volume data before normalization, that is, the degree of dispersion of all voxel original values ​​V(x,y,z) relative to the mean σ, used to measure the dispersion of the data.

[0027] The global threshold segmentation algorithm in step S13 can be implemented using the following segmentation formula: Where M(x,y,z) is the three-dimensional binary mask obtained after global thresholding, which is 1 (target region) or 0 (background region), and T threshold To set pixel values.

[0028] The resulting 3D model is an NIfTI format file, with the same spatial resolution as the original data.

[0029] In this embodiment of the invention, by standardizing the coordinates of the specified file, it can be ensured that the three-dimensional volume data conforms to the medical image analysis standard, providing a unified spatial basis for subsequent registration.

[0030] Then, for step 102, obtain the theoretical model of the implant that was pre-constructed for the implant.

[0031] In this embodiment of the invention, the theoretical model of the implant can be obtained by cone-beam scanning of the implant before surgery, or it can be constructed according to the size of the implant.

[0032] Taking the implant as a cylinder as an example, the method for constructing the theoretical model of the implant is as follows: S21: Define the dimensional parameters of the theoretical cylinder, including diameter, height, and spatial resolution; The dimension parameter can be: diameter 2r, height h, and spatial resolution size / 900, where size is the dimension parameter of the three-dimensional matrix in the length, width, and height dimensions when constructing the theoretical cylinder.

[0033] Furthermore, in order to ensure that the spatial resolution of the implant theoretical model and the target 3D model are the same, the spatial resolution data of the target 3D model can be read in advance. When constructing the theoretical cylinder, the parameter size is used to match the spatial dimension scale of the target 3D model. By constructing a 3D matrix of size (size, size, size), the voxel mesh scale of the theoretical cylinder can be matched with the spatial dimension features of the target 3D model, thereby ensuring that the number of voxels per unit physical length is consistent, which ultimately results in the same spatial resolution.

[0034] S22: Construct a theoretical model of a cylinder in three-dimensional space with central symmetry; The mathematical expression for the cylinder theoretical model is: Among them, c x c y Let h be the coordinates of the center of the circle along the x and y axes. start h end The initial and final heights of the theoretical cylinder along the z-axis are given.

[0035] The shape constraints of the theoretical cylinder model can be achieved through the mathematical expression of this cylinder theoretical model.

[0036] S23: Use the OpenCV library to generate a two-dimensional circular template and stack it along the axis; The core algorithm in this step is: `cv2.circle(circle,(c,c),r,255,-1)`. Specifically, this core algorithm is used to draw a circle with a fill of 255 (white) in the two-dimensional image matrix of `circle`, with the center at coordinates (c, c) (both x-axis and y-axis coordinates are c, achieving central symmetry) and a radius of `r`. The parameter `-1` indicates that the circle is filled with a solid fill (instead of just drawing the outline), ultimately generating a two-dimensional circular template (so that two-dimensional graphics can be stacked later to form a three-dimensional cylinder).

[0037] Under the shape constraints of the theoretical cylinder model, two-dimensional circular templates with circular constraints in the x and y planes generated in this step are used. These two-dimensional circular templates are then stacked along the z-axis to construct a complete theoretical cylinder model.

[0038] S24: Generate a cylindrical theoretical model with the same spatial resolution as the target 3D model.

[0039] The following explanations will focus on steps 104, "calculate the centroids of the target 3D model and the implant theoretical model respectively, and align the centroids of the target 3D model and the implant theoretical model", and 106, "register the centroid-aligned target 3D model and implant theoretical model using the principal axis rotation method".

[0040] The centroids of the target 3D model and the implant theoretical model can be calculated using the following formula: Among them, c target Let c be the centroid of the target 3D model. source Let M be the centroid of the implant's theoretical model, N be the total number of points in the target 3D model involved in the centroid calculation, and M be the total number of points in the implant's theoretical model involved in the centroid calculation. The coordinates of the i-th point in the target 3D model are divided into... Let be the coordinates of the i-th point in the theoretical model of the implant.

[0041] In one embodiment of the present invention, the centroids of the target 3D model and the implant theoretical model can be aligned by the following translation transformation: Where T is the displacement required to translate the centroid of the implant's theoretical model to the centroid of the target 3D model, and c target Let c be the centroid of the target 3D model. source This is the centroid of the theoretical model of the implant.

[0042] In one embodiment of the present invention, when registering the target 3D model and the implant theoretical model after centroid alignment using the principal axis rotation method, the process may specifically include: calculating the principal axis directions of the target 3D model and the implant theoretical model based on covariance matrix decomposition, constructing a rotation matrix according to the principal axis directions, adjusting the posture of the implant theoretical model through affine transformation, and using inverse rotation to compensate for the offset, thereby achieving spatial registration of the target 3D model and the implant theoretical model.

[0043] The formula for calculating the main axis direction is: Where G represents the covariance matrix, and N is the total number of points involved in the calculation. Let i be the coordinates of the i-th point on the target 3D model. Let be the coordinates of the i-th point on the theoretical model of the implant; The orientation of the implant's theoretical model is adjusted using the following affine transformation: Where R is the rotation matrix, and U and V are the orthogonal matrices obtained by singular value decomposition (SVD). SVD() is the singular value decomposition operation, which can extract the main direction information of the matrix and provide a basis for constructing the rotation matrix. The offset is compensated by the following reverse rotation: Among them, R inv R is the inverse rotation matrix. -1 Let R be the inverse of the rotation matrix. These are the final coordinates of points on the implant's theoretical model after inverse rotation compensation. The intermediate coordinates of the points on the theoretical model of the implant after transformation by the rotation matrix R.

[0044] Because the initial position alignment of existing registration is based on geometric center registration, such as taking the minimum or maximum midpoint of the coordinate range, but the geometric center does not take into account the mass distribution of the model. In medical images, the voxel density and effective pixel weights of the target region are not uniform, for example, the density of hard tissue in teeth is higher than that of the surrounding soft tissue. Therefore, using the geometric center for registration will lead to a large initial position deviation, and subsequent registration will converge slowly or even fail.

[0045] Based on this, in this embodiment of the invention, centroid alignment is performed based on the mass weighting center, so that the centroid can better reflect the center of the model's mass distribution, rather than simply the midpoint of the coordinate range.

[0046] The centroid alignment method in this invention is better suited to the non-uniformity of medical models. Medical 3D models (such as teeth and implants) have a non-uniform mass (density) distribution, and the centroid is more representative of the core anatomical / design location of the model than the geometric center. For example, the centroid of a tooth will be closer to the concentrated area of ​​the hard tissue of the tooth, rather than a simple coordinate midpoint.

[0047] The centroid alignment method in this embodiment of the invention can reduce the initial position deviation. Centroid alignment provides a more accurate initial position reference for subsequent registration (such as spindle rotation and fine registration), avoiding the registration iteration from not converging or getting trapped in local optima due to large initial position deviation, thus improving the stability of registration accuracy from the source.

[0048] The centroid alignment method of this invention is more universal, requiring no manual point selection or reliance on specific features. It can automatically calculate the centroid based on the inherent distribution of the model, adapting to target 3D models and implants of different shapes, and improving the automation and robustness of the registration process.

[0049] Existing registration rotation alignment methods, such as iterative optimization, rely on iterative convergence. If the initial rotation deviation is large, it is easy to get trapped in local optima and the calculation is time-consuming. Another example is manually specifying the direction or forcing the coordinate axis alignment, which cannot adapt to the inherent shape orientation of the model. For example, tilted implants or curved dental arches are not naturally parallel to the coordinate axes, resulting in extremely poor registration accuracy for complex shape models.

[0050] Based on this, in this embodiment of the invention, since the principal axis of the model is its most representative morphological direction and can reflect the main extension trend of the model's distribution in space, rotation alignment is performed by rotating the principal axis, utilizing the inherent geometric features of the model. For example, the dental arch has a natural arc-shaped principal axis, and the implant has a designed axial principal axis; by decomposing the covariance matrix to extract the principal axis, the natural morphological direction of the model can be accurately matched, rather than being forcibly aligned with the coordinate axes.

[0051] The spindle rotation alignment method of this invention does not require iterative optimization. It directly calculates the optimal rotation matrix through matrix decomposition (such as SVD), which avoids the risk of local optima and greatly improves registration efficiency (without repeated iterations). It can achieve rotation alignment in one go and improve the upper limit of accuracy.

[0052] The spindle rotation alignment method of this invention is robust to complex shapes. For models with irregular shapes, tilts, or angular deviations, such as pathological dental arches and personalized implants, spindle rotation can capture their natural extension direction. Compared with forced coordinate axis alignment or manually specifying the direction, the registration accuracy is significantly improved, which is more in line with the needs of actual medical scenarios.

[0053] Furthermore, in this embodiment of the invention, the combination of centroid alignment and spindle rotation can first minimize the positional deviation through centroid alignment, and then minimize the directional deviation through spindle rotation, thus forming a precise registration closed loop and fundamentally improving the overall registration accuracy.

[0054] Finally, for step 108, the registration accuracy is evaluated based on multi-plane projection and similarity metric, and the spatial pose parameters of the implant are output.

[0055] In one implementation, the registration accuracy can be evaluated as follows: For the registered target 3D model and the implant theoretical model, generate the corresponding three-view orthogonal projection images of the model; calculate the Dice similarity coefficient based on the orthogonal projection images of the target 3D model and the implant theoretical model on each view, and evaluate the registration overlap based on the calculation results, and determine the registration overlap as the registration accuracy.

[0056] After registration, both models are projected using the following projection formula: The formula for calculating the Dice similarity coefficient is: in, This refers to the effective pixel region of the implant in the target 3D model. This refers to the effective pixel region of the implant in the theoretical model of the implant.

[0057] Furthermore, a pre-set registration precision can be established. When the registration precision is higher than the set precision, the registration is considered complete, and the centroid coordinates and principal axis direction vectors of the registered model are output to represent the spatial orientation of the implant. The spatial orientation of the implant can be: in, These are the three-dimensional coordinates of the model's centroid after registration. These represent the principal axis direction vectors of the registered model on the x, y, and z axes, respectively. To determine the final centroid of the model after registration, This refers to the principal axis direction vector of the registered model.

[0058] In this embodiment of the invention, the parameter error of the implant's spatial orientation needs to meet the following requirements: , .

[0059] In this embodiment of the invention, the computational complexity of the registration algorithm is: O , where N is the number of voxels.

[0060] Through a two-stage optimization strategy (centroid alignment + spindle rotation), registration efficiency and accuracy were significantly improved. The final output spatial attitude parameters can be directly integrated into the surgical navigation system, providing millimeter-level precision spatial reference for clinical operations.

[0061] Please refer to Figure 2 This invention provides an implant spatial pose estimation device for CT images, the device comprising: An extraction unit is used to extract a target three-dimensional model based on CBCT images; the CBCT images are obtained by cone-beam scanning of the target site where an implant is placed during or after surgery. The acquisition unit is used to acquire the theoretical model of the implant that has been pre-constructed for the implant. An alignment unit is used to calculate the centroids of the target 3D model and the implant theoretical model respectively, and to align the centroids of the target 3D model and the implant theoretical model. The registration unit is used to register the target 3D model and the implant theoretical model after centroid alignment by using a principal axis rotation method; The evaluation unit is used to evaluate the registration accuracy based on multi-plane projection and similarity metrics, and output the spatial pose parameters of the implant.

[0062] In one embodiment of the present invention, the spatial resolution of the implant theoretical model during the construction process is the same as that of the target three-dimensional model.

[0063] In one embodiment of the present invention, when the alignment unit performs centroid alignment of the target 3D model and the implant theoretical model, it specifically includes: aligning the centroids of the target 3D model and the implant theoretical model through the following translation transformation: Where T is the displacement required to translate the centroid of the implant's theoretical model to the centroid of the target 3D model, and c target Let c be the centroid of the target 3D model. source Let q be the centroid of the theoretical model of the implant. i Let be the coordinates of the i-th point in the theoretical model of the implant.

[0064] In one embodiment of the present invention, the registration unit is specifically used to calculate the principal axis directions of the target 3D model and the implant theoretical model based on covariance matrix decomposition, construct a rotation matrix according to the principal axis directions, adjust the posture of the implant theoretical model through affine transformation, and use inverse rotation to compensate for the offset, so as to achieve spatial registration of the target 3D model and the implant theoretical model.

[0065] In one embodiment of the present invention, the formula for calculating the main axis direction is: Where G represents the covariance matrix, and N is the total number of points involved in the calculation. Let i be the coordinates of the i-th point on the target 3D model. Let be the coordinates of the i-th point on the theoretical model of the implant; The orientation of the implant's theoretical model is adjusted using the following affine transformation: Where R is the rotation matrix, U and V are the orthogonal matrices obtained by singular value decomposition, and SVD() is the singular value decomposition operation; The offset is compensated by the following reverse rotation: Among them, R inv R is the inverse rotation matrix. -1 Let R be the inverse of the rotation matrix. These are the final coordinates of points on the implant's theoretical model after inverse rotation compensation. The intermediate coordinates of the points on the theoretical model of the implant after transformation by the rotation matrix R.

[0066] In one embodiment of the present invention, when the evaluation unit performs the evaluation of the registration accuracy based on multi-plane projection and similarity measurement, it specifically includes: generating three-view orthogonal projection images of the corresponding models for the registered target three-dimensional model and the implant theoretical model respectively; calculating the Dice similarity coefficient based on the orthogonal projection images of the target three-dimensional model and the implant theoretical model on each view, so as to evaluate the registration overlap based on the calculation results, and determining the registration overlap as the registration accuracy.

[0067] It should be noted that the implant spatial pose estimation device for CT images provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the implant spatial pose estimation device for CT images provided in the above embodiments and the implant spatial pose estimation method for CT images belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0068] Embodiments of this application also provide a computer device, please refer to... Figure 3 The computer device includes a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, the at least one instruction, at least one program, code set or instruction set being loaded and executed by the processor to implement the implant spatial pose estimation method for CT images provided in the above method embodiments.

[0069] Embodiments of this application also provide a computer-readable storage medium storing at least one instruction, at least one program, code set, or instruction set, wherein the at least one instruction, at least one program, code set, or instruction set is loaded and executed by a processor to implement the implant spatial pose estimation method for CT images provided in the above-described method embodiments.

[0070] Embodiments of this application also provide a computer program product comprising a computer program, wherein a processor of a computer device reads the computer program from a computer-readable storage medium, and the processor executes the computer program, causing the computer device to perform any of the implant spatial pose estimation methods for CT images described in the above embodiments.

[0071] For ease of description, the above systems or devices are described separately as various modules or units based on their functions. Of course, in implementing this application, the functions of each unit can be implemented in one or more software and / or hardware components.

[0072] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0073] Finally, it should be noted that in this document, relational terms such as first, second, third, and fourth are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0074] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for estimating the spatial pose of an implant based on CT images, characterized in that, The method includes: The target 3D model is extracted based on CBCT images; the CBCT images are obtained by cone-beam scanning of the target site where the implant is placed during or after surgery. Obtain a theoretical model of the implant that has been pre-constructed for the implant; Calculate the centroids of the target 3D model and the implant theoretical model respectively, and align the centroids of the target 3D model and the implant theoretical model. The target 3D model and the implant theoretical model are registered by using a principal axis rotation method after the centroids are aligned. The registration accuracy is evaluated based on multi-plane projection and similarity metrics, and the spatial orientation parameters of the implant are output.

2. The method according to claim 1, characterized in that, The spatial resolution of the implant theoretical model during the construction process is the same as that of the target three-dimensional model.

3. The method according to claim 1, characterized in that, Aligning the centroids of the target 3D model and the implant theoretical model includes: The centroids of the target 3D model and the implant theoretical model are aligned using the following translation transformation: Where T is the displacement required to translate the centroid of the implant's theoretical model to the centroid of the target 3D model, and c target Let c be the centroid of the target 3D model. source Let q be the centroid of the theoretical model of the implant. i Let be the coordinates of the i-th point in the theoretical model of the implant.

4. The method according to claim 3, characterized in that, The registration of the target 3D model and the implant theoretical model after centroid alignment using a principal axis rotation method includes: The principal axis directions of the target 3D model and the implant theoretical model are calculated based on the covariance matrix decomposition. A rotation matrix is ​​constructed based on the principal axis directions, and the orientation of the implant theoretical model is adjusted by affine transformation. Inverse rotation is used to compensate for the offset, thereby achieving spatial registration between the target 3D model and the implant theoretical model.

5. The method according to claim 4, characterized in that, The formula for calculating the direction of the main axis is: Where G represents the covariance matrix, and N is the total number of points involved in the calculation. Let be the coordinates of the i-th point on the target 3D model; The orientation of the implant's theoretical model is adjusted using the following affine transformation: Where R is the rotation matrix, U and V are the orthogonal matrices obtained by singular value decomposition, and SVD() is the singular value decomposition operation; The offset is compensated by the following reverse rotation: Among them, R inv R is the inverse rotation matrix. -1 Let R be the inverse of the rotation matrix. These are the final coordinates of points on the implant's theoretical model after inverse rotation compensation. The intermediate coordinates of the points on the theoretical model of the implant after transformation by the rotation matrix R.

6. The method according to any one of claims 1-5, characterized in that, The evaluation of the registration accuracy based on multi-plane projection and similarity metrics includes: For the registered target 3D model and implant theoretical model, orthogonal projection images of the corresponding models in three views are generated respectively. Based on the orthogonal projection images of the target 3D model and implant theoretical model in each view, the Dice similarity coefficient is calculated to evaluate the registration overlap based on the calculation results, and the registration overlap is determined as the registration accuracy.

7. A device for estimating the spatial pose of an implant based on CT images, characterized in that, The device includes: The extraction unit is used to extract a target three-dimensional model based on CBCT images; the CBCT images are obtained by cone-beam scanning of the target site where an implant is placed during or after surgery. The acquisition unit is used to acquire the theoretical model of the implant that has been pre-constructed for the implant. An alignment unit is used to calculate the centroids of the target 3D model and the implant theoretical model respectively, and to align the centroids of the target 3D model and the implant theoretical model. The registration unit is used to register the target 3D model and the implant theoretical model after centroid alignment by using a principal axis rotation method; The evaluation unit is used to evaluate the registration accuracy based on multi-plane projection and similarity metrics, and output the spatial pose parameters of the implant.

8. A computer device, characterized in that, The computer device includes a memory and a processor. The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to implement the steps of the method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the steps of the method described in any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1-6.

Images