Methods for registration of 3D medical images with robotic systems

Abstract

This invention discloses a method for registering three-dimensional medical images with a robotic system, including measures specifically addressing metallic interference to ensure accurate identification of the registration ball's position even in the presence of non-target metallic interference. This not only improves the recognition accuracy of the registration ball but also guarantees the reliability and accuracy of the registration process, enabling the robotic system to complete coordinate system registration more quickly and accurately before surgery. This is crucial for robot-assisted surgery requiring high-precision navigation, improving work efficiency during surgical preparation and contributing to increased surgical safety and success rates.

Description

Technical Field

[0001] This invention relates to the field of electromechanical equipment, and more particularly to a method for registering three-dimensional medical images with a robotic system. Background Technology

[0002] In recent years, 3D C-arm scanners have been increasingly applied in surgical settings such as dentistry and orthopedics. Compared to traditional CT scanners, 3D C-arm scanners are smaller, easier to transport, and more suitable for surgical scenarios. Compared to traditional CT images, cone-beam computed tomography (CBCT) images acquired by 3D C-arm scanners have higher spatial resolution, improving navigation accuracy during robotic surgery; simultaneously, their faster acquisition speed reduces radiation dose to patients. Before robotic surgery, registering the robot system coordinates with the real-world coordinate system is an essential step. Coordinate registration often employs a registration plate marker matching method, where markers are a set of small metal spheres embedded in the registration plate. During CBCT image acquisition, the robotic arm holds the registration plate, simultaneously acquiring the registration plate and the target surgical area into the CBCT image. After identifying the distribution of the metal spheres in the CBCT image, a coordinate transformation is performed between the registration plate and the calibrated distribution template to obtain the registration matrix.

[0003] Currently, metal ball recognition mainly uses thresholding and connected component filtering methods. Thresholding is a semi-automatic method that uses a fixed threshold to segment the ball and then directly calculates the center-to-center match. If the match is incorrect, the image is manually cropped to remove interference information. This method is inefficient and has difficulty handling situations where interference information overlaps with the metal ball.

[0004] In summary, a method for registering 3D medical images with robotic systems is needed to address the shortcomings of existing technologies. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for registering three-dimensional medical images with a robotic system, aiming to solve the aforementioned problems.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for registering three-dimensional medical images with a robotic system, comprising the following steps:

[0007] Step S1: Data preparation, simultaneously acquiring the registration plate and the target surgical area into the 3D image;

[0008] Step S2: Read the 3D image, perform interpolation resampling, and read the pre-calibrated center position of the metal sphere template and the diameter of the actual sphere;

[0009] Step S3: Target layer determination. The initial screening threshold is obtained by sorting the maximum values ​​of each slice layer using the maximum inter-class variance method, and the target layer for segmentation calculation is determined.

[0010] Step S4: Image segmentation threshold calculation. The segmentation threshold is calculated for the target layer using the maximum inter-class variance method. The 3D image is then segmented using the segmentation threshold.

[0011] Step S5: Image segmentation. Normalize the segmented 3D image, set an area threshold based on the maximum cross-sectional area, and obtain connected components smaller than the area threshold.

[0012] Step S6: Cluster calculation of sphere center. The number of clusters is determined by the clustering method, the cluster centers are initialized, and after clustering, the classes that meet the mention ratio and sphericity standards are selected to obtain the metal sphere identification result.

[0013] Step S7: Calculate the distance between all pairs of sphere centers that were actually identified, construct feature vectors, and match the feature vectors of the identified sphere centers with the corresponding template sphere centers;

[0014] Step S8: Solve the transformation matrix between the two coordinate systems using the singular value decomposition method to complete the coordinate system registration.

[0015] Optionally, the initial screening threshold is calculated in step S3 in the following way:

[0016] Step A1: Calculate the maximum value of voxels in each slice layer of the 3D image, sort the maximum values ​​of each layer, and form a list of maximum values;

[0017] Step A2: Normalize the voxel values ​​in the maximum value list to the range [0, 255];

[0018] Step A3: Set t1 to any integer in the range [0, 255], count the number N of elements less than or equal to t1 in the maximum value list, calculate the average of N elements meanN, and count the number M of elements greater than t1 in the maximum value list and the average of these M elements meanM.

[0019] Step A4: Traverse the range [0,255] to find t1 that maximizes Var, and obtain the initial screening threshold.

[0020] Optionally, the target layer for segmentation calculation is determined in step S3 by the following method:

[0021] Step B1: Select layers with a voxel maximum value greater than the initial screening threshold as the initial screening layers, forming the initial screening layer group L, and count the number of layers N. L Based on the number and diameter of the metal spheres, the theoretical number of distribution layers Lt is calculated.

[0022] Step B2: Calculate the mean L and standard deviation stdL of the list of maximum values ​​of the initial screening layer group L;

[0023] Step B3: Calculate the ratio of meanL to standard deviation and compare it with the metal threshold. If the ratio is greater than the metal threshold, take the layer containing the median of the maximum voxel values ​​in the initial screening layer group L as the target layer for subsequent calculation. Otherwise, take the Xth layer from smallest to largest as the target layer for segmentation calculation.

[0024] Optionally, the segmentation threshold in step S4 is used to segment the 3D image in the following way:

[0025] The segmentation threshold Ts is calculated for the target layer using the Otsu's method. The entire 3D image is then segmented using the segmentation threshold Ts. Voxel values ​​less than Ts are set to 0, while voxel values ​​greater than or equal to Ts remain unchanged.

[0026] Optionally, the image segmentation in step S5 is performed in the following manner:

[0027] Step S51: Normalize the 3D image using the global maximum and minimum values;

[0028] Step S52: Calculate the theoretical maximum cross-sectional area Sa of the metal sphere in the two-dimensional image based on the diameter of the metal sphere and the physical pixel spacing of the image;

[0029] Step S53: Set an area threshold St, traverse each slice layer of the 3D image, calculate the two-dimensional connected components, retain the connected components with an area smaller than the area threshold, and obtain the segmented 3D image.

[0030] Optionally, the cluster center is initialized in step S6 in the following way:

[0031] Step S61: Calculate the distance threshold Td and randomly select the first cluster center C0;

[0032] Step S62: Calculate the distance from all other points to C0, and select the point with the largest distance from C0 as the center of the second cluster C1;

[0033] Step S63: Calculate the distance from the remaining points to the second cluster center, and select the point with the largest distance as the next cluster center Cx;

[0034] Step S64: Repeat step S63 until the distance from all points to the nearest cluster center is less than the distance threshold Td.

[0035] Optionally, the center of the sphere is calculated in the following way:

[0036] The theoretical sphere volume in the image space is calculated, and volume and sphericity are filtered. The sphericity of each type is sorted, and the type with the largest number of sphericities is selected as the preset number of metal spheres. The metal sphere recognition result is obtained, and the center of each type is the recognition center.

[0037] Optionally, in step S7, the feature vectors of the identified sphere center and the corresponding template sphere center are matched and identified in the following way:

[0038] Step S71: Calculate the distance matrix based on the coordinates of each sphere's center;

[0039] Step S72: Calculate the minimum value in the distance matrix other than zero, which is the minimum distance between the sphere centers;

[0040] Step S73: Reorder the elements in the feature vector of the m-th ball in ascending order;

[0041] Step S74: Calculate the distance matrix and feature vectors of each ball for both the identification ball center and the template ball center. Calculate the distance between two feature vectors. Match the identification ball center and the template ball center corresponding to the two feature vectors with the smallest distance to obtain ball center pairs.

[0042] Step S75: Repeat step S74 to match the identified ball center with the template ball center one by one.

[0043] A three-dimensional medical image and robot system registration system, employing the aforementioned three-dimensional medical image and robot system registration method, includes a data acquisition module, a preprocessing module, an initial screening threshold calculation module, a target layer determination module, an image segmentation module, a normalization and connected component extraction module, a cluster analysis module, a feature vector matching module, and a coordinate system registration module;

[0044] The data acquisition module is responsible for simultaneously acquiring three-dimensional image data of the registration plate and the target surgical area;

[0045] The preprocessing module is used to perform interpolation and resampling to make the image data suitable for further analysis;

[0046] The initial screening threshold calculation module is used to obtain the initial screening threshold from the maximum value sorting of each slice layer by the maximum inter-class variance method, so as to determine which layers are the target layers for segmentation calculation.

[0047] The target layer determination module is used to determine the target layer that actually needs to be segmented based on the layer whose voxel maximum value is greater than the initial screening threshold.

[0048] Optionally, the image segmentation module is used to calculate a segmentation threshold for the target layer using the maximum inter-class variance method, and to segment the three-dimensional image using this threshold.

[0049] The normalization and connected component extraction module is used to normalize the segmented 3D image and set an area threshold based on the maximum cross-sectional area to obtain connected components smaller than the area threshold.

[0050] The cluster analysis module is used to determine the number of clusters and initialize cluster centers through clustering methods, filter out the classes that meet the mention ratio and sphericity criteria, and finally obtain the metal ball recognition results;

[0051] The feature vector matching module is used to calculate the distance between all pairs of sphere centers that are actually identified, construct feature vectors, and match the feature vectors of the identified sphere centers with the corresponding template sphere centers.

[0052] The coordinate system registration module is used to solve the transformation matrix between two coordinate systems using the singular value decomposition method, thereby completing the coordinate system registration.

[0053] The beneficial effects of this invention are:

[0054] 1. The targeted measures for handling metallic interference in this invention ensure accurate identification of the registration ball's position even in the presence of non-target metallic interference. This not only improves the identification accuracy of the registration ball but also guarantees the reliability and accuracy of the registration process.

[0055] 2. In this invention, the robot system can complete coordinate system registration more quickly and accurately before surgery, which is crucial for robot-assisted surgery that requires high-precision navigation, improves the efficiency of the surgical preparation stage, and also helps to improve the safety and success rate of the surgery.

[0056] 3. In this invention, the maximum inter-class variance method is used to calculate the initial screening threshold and the segmentation threshold. This method can effectively separate the target metal ball from complex CBCT images, reduce the possibility of misjudgment, and reduce the computational complexity by transforming the threshold problem in three-dimensional space through two two-dimensional threshold calculations. It does not require manual specification, increases the automation and adaptability of the method, and makes it more suitable for metal ball recognition in different situations. Attached Figure Description

[0057] Figure 1 This is a flowchart of one method of the present invention.

[0058] Figure 2 This is a schematic diagram of a primary image screening layer according to the present invention.

[0059] Figure 3 This is a schematic diagram of a binarized image after image segmentation according to the present invention.

[0060] Figure 4 This is a schematic diagram of the clustering analysis results of the present invention.

[0061] Figure 5 This is a schematic diagram of a template sphere center matching result according to the present invention.

[0062] Figure 6 This is a schematic diagram of a matching result according to the present invention. Detailed Implementation

[0063] To more clearly illustrate the technical solutions in the embodiments of the invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] like Figures 1 to 6 As shown, a method for registering three-dimensional medical images with a robotic system includes the following:

[0065] Design a registration plate containing more than three metal spheres. The diameter and spacing of the spheres are known, the pairwise spacing between spheres is unique, and the center position of each sphere is precisely calibrated. Considering that metal artifacts in CBCT images can affect the accuracy of sphere center detection, titanium spheres, which have less artifacts, were chosen. However, titanium spheres have lower response values ​​than commonly used metals such as steel. Therefore, a more robust metal interference recognition mechanism is needed when other metal interferences are present in CBCT images.

[0066] Data preparation

[0067] CBCT images are acquired, and a robotic arm is used to hold the registration plate, simultaneously capturing all metal balls and the target surgical area into the CBCT images.

[0068] Read CBCT 3D images and interpolate and resample the 3D images based on the physical pixel spacing of the 2D images to make the physical spacing of the 3D image slices the same as that of the 2D images.

[0069] Read the center position of each metal ball template and the actual diameter of the ball.

[0070] Threshold calculation

[0071] To reduce computational load, the threshold calculation in three-dimensional space is transformed into two two-dimensional threshold calculations.

[0072] Target layer determination

[0073] Calculate the maximum voxel value in each slice layer of the 3D image, sort the maximum values ​​of each layer, and use the Otsu's method to calculate the initial screening threshold Tf from the list of maximum values. The specific steps for calculating the initial screening threshold Tf are as follows:

[0074] ① Normalize the list of maximum values ​​to the range of 0-255;

[0075] ② Let t1 be any integer in the range of 0-255. Count the number N of elements less than or equal to t1 in the maximum value list and the average of these N elements, meanN. Count the number M of elements greater than t1 in the maximum value list and the average of these M elements, meanM.

[0076] ③ Calculate the inter-class variance, Var = (N(meanN-mean)) 2 +M(meanM-mean) 2 ) / 256, where mean is the average of the list of maximum values;

[0077] ④ Traverse the range 0-255 to find t1 that maximizes var, which is the initial screening threshold Tf.

[0078] Layers with voxel values ​​greater than the initial screening threshold are retained as the initial screening layer group L, and the number of layers NL is counted. Additionally, the theoretical distribution number Lt is calculated based on the number and diameter of the metal spheres. Due to the registration plate design and imaging rules, no two target metal spheres will appear in the same layer. The specific calculation steps for determining the target layer for segmentation are as follows:

[0079] ① Calculate the mean L and standard deviation stdL of the list of maximum values ​​of L in the initial screening layer group;

[0080] ② Calculate the ratio of meanL to standard deviation and compare it with the metal threshold. If the ratio is greater than the metal threshold, the layer containing the median of the maximum voxel values ​​in the initial screening layer group L is taken as the target layer for subsequent calculation. Otherwise, the Xth layer from smallest to largest is taken as the target layer for segmentation calculation, X = 0.6×Lt.

[0081] Calculate image segmentation threshold

[0082] The segmentation threshold Ts is calculated for the target layer using the Otsu's method, following the same steps as described above. The segmentation threshold Ts is then applied to the entire 3D image; voxel values ​​less than Ts are set to 0, while voxel values ​​greater than or equal to Ts remain unchanged.

[0083] Image segmentation

[0084] The 3D image is normalized using the global maximum and minimum values ​​of the 3D image.

[0085] The theoretical maximum cross-sectional area Sa of the metal sphere in the 2D image is calculated based on the diameter of the metal sphere and the physical pixel spacing of the image. An area threshold St = 1.5Sa is set. Each slice layer of the 3D image is traversed, and the 2D connected components are calculated. Connected components with an area smaller than the area threshold St are retained to obtain the segmented 3D image.

[0086] K-means clustering to calculate the center of the sphere

[0087] K-means clustering is an unsupervised machine learning method. Its basic idea is to iteratively find a classification scheme for K (hyperparameter) clusters that minimizes the loss function corresponding to the clustering result. The loss function is generally defined as the sum of squared errors between each sample and its cluster center. When the initial cluster centers are optimal, K-means clustering is computationally much more efficient than 3D connected component extraction. Traditional K-means clustering requires inputting the target number of clusters K first. This paper proposes an improved K-means clustering method that can automatically determine the number of clusters K. This is possible because, in the current problem, the target metal sphere has high pixel density per sphere and large distances between spheres, while interfering metal objects also have similar characteristics.

[0088] The specific steps for cluster center initialization are as follows:

[0089] ① Calculate the distance threshold ThresDist = 3 × (known sphere diameter D / spacing), and randomly select the first cluster center C0;

[0090] ② Calculate the distance from all other points to C0, and select the point with the largest distance from C0 as the second cluster center C1;

[0091] ③ Calculate the distance from all other points to the nearest cluster center, and take the point with the largest distance as the next cluster center Cx;

[0092] ④ Repeat step ③ until the distance from all points to the nearest cluster center is less than the distance threshold ThresDist.

[0093] After initialization, initial cluster centers are obtained automatically based on the known sphere diameters. Using the number of initial cluster centers M as the hyperparameter K, K-means clustering is performed, resulting in M ​​predefined clusters. Since distance judgment has already been performed during initialization, the clustering process converges quickly.

[0094] When interfering metallic objects are present, the clustering results may include classes of non-target metallic spheres, i.e., M is greater than the preset number of metallic spheres N. In this case, further filtering is needed to identify the metallic sphere classes. The specific filtering steps are as follows:

[0095] ① Calculate the theoretical sphere volume of the image space: V = 4π / 3 * (0.5 * D / spacing) 3 D is the diameter of the known sphere;

[0096] ② Volume filtering: The number of points in a class is the class volume Vc. Calculate the volume ratio of each class Rv = Vc / V, and retain classes with Rv in the interval (0.5, 1.3).

[0097] ③ Sphericity screening: Take the distance dc between the farthest point from the class center and the class center, calculate the volume Vdc of the sphere with radius dc, let the sphericity of the class be Sc=Vc / Vdc, and keep the classes with Sc greater than 0.5;

[0098] ④ Sort the various types of Sc, select the type with the largest number of Sc equal to the preset number of metal balls, which is the metal ball recognition result, and the center of each type is the center of the recognized ball.

[0099] Sphere center matching + calculation of coordinate system transformation

[0100] Calculate the distance between the center of each pair of spheres actually identified, and record the minimum distance. Divide all distances by the minimum distance; at this point, each identified sphere center has a feature vector. Perform the same calculation on the center of the metal sphere template. Identified sphere centers and template sphere centers with matching feature vectors are then matched. The specific calculation steps are as follows:

[0101] ① When the number of balls is n, calculate the distance matrix based on the coordinates of each ball's center.

[0102] , where d1n is the distance between the 1st ball and the nth ball (such as the Euclidean distance), it can be seen that the matrix is ​​a symmetric matrix and all values ​​on the diagonal are 0;

[0103] ② Calculate the minimum value (excluding 0) in the distance matrix Adist, which is the minimum inter-sphere distance distmin. ;

[0104] ②The m-th ball (m≤n) has the following eigenvector: Reorder the elements in the feature vector Vm so that they are arranged in ascending order;

[0105] ③ Calculate the distance matrix and eigenvectors of both the identification sphere center and the template sphere center. Take the eigenvector Vrp of a certain identification sphere center and the eigenvector Vtq of a certain template sphere center, and calculate the Euclidean distance between the two eigenvectors. The two feature vectors with the smallest distance correspond to the recognition ball center and the template ball center, which are the matched ball center pairs.

[0106] ⑤ Repeat step ④ to ensure that the center of the identified ball matches the center of the template ball one by one.

[0107] Once the center of the sphere is correctly matched with the center of the template sphere, the singular value decomposition (SVD) method can be used to solve for the transformation matrix RT between the two coordinate systems, thus completing the coordinate system registration.

[0108] This invention incorporates targeted measures to handle metallic interference in each calculation step, ensuring accurate identification of the registration ball's position even in the presence of non-target metallic interference. This not only improves the identification accuracy of the registration ball but also guarantees the reliability and accuracy of the registration process. The invention also introduces the K-means clustering method from machine learning to calculate the ball centers. This method has higher computational efficiency compared to traditional three-dimensional connected component extraction, especially when the initial cluster centers are optimal, allowing the clustering process to converge quickly, thereby significantly shortening the computation time and improving the overall system efficiency.

[0109] The ability to quickly and accurately complete coordinate system registration before surgery is crucial for robot-assisted surgery requiring high-precision navigation. This improves efficiency during surgical preparation and also contributes to increased surgical safety and success rates. The method employs maximum inter-class variance (MOV) to calculate the initial screening and segmentation thresholds. This approach effectively separates the target metal sphere from complex CBCT images, reducing the possibility of misjudgment. Furthermore, by transforming the two-dimensional threshold calculations into a three-dimensional threshold problem, computational complexity is reduced. The improved K-means clustering algorithm automatically determines the optimal number of clusters K without manual specification, increasing the method's automation and adaptability, making it better suited for metal sphere identification under different conditions.

[0110] A three-dimensional medical image and robot system registration system, employing the aforementioned three-dimensional medical image and robot system registration method, includes a data acquisition module, a preprocessing module, an initial screening threshold calculation module, a target layer determination module, an image segmentation module, a normalization and connected component extraction module, a cluster analysis module, a feature vector matching module, and a coordinate system registration module;

[0111] The data acquisition module is responsible for simultaneously acquiring three-dimensional image data of the registration plate and the target surgical area;

[0112] The preprocessing module is used to perform interpolation and resampling to make the image data suitable for further analysis;

[0113] The initial screening threshold calculation module is used to obtain the initial screening threshold from the maximum value sorting of each slice layer by the maximum inter-class variance method, so as to determine which layers are the target layers for segmentation calculation.

[0114] The target layer determination module is used to determine the target layer that actually needs to be segmented based on the layer whose voxel maximum value is greater than the initial screening threshold.

[0115] Optionally, the image segmentation module is used to calculate a segmentation threshold for the target layer using the maximum inter-class variance method, and to segment the three-dimensional image using this threshold.

[0116] The normalization and connected component extraction module is used to normalize the segmented 3D image and set an area threshold based on the maximum cross-sectional area to obtain connected components smaller than the area threshold.

[0117] The cluster analysis module is used to determine the number of clusters and initialize cluster centers through clustering methods, filter out the classes that meet the mention ratio and sphericity criteria, and finally obtain the metal ball recognition results;

[0118] The feature vector matching module is used to calculate the distance between all pairs of sphere centers that are actually identified, construct feature vectors, and match the feature vectors of the identified sphere centers with the corresponding template sphere centers.

[0119] The coordinate system registration module is used to solve the transformation matrix between two coordinate systems using the singular value decomposition method, thereby completing the coordinate system registration.

[0120] The embodiments described above are merely preferred solutions of this application and are not intended to limit this application in any way. Other variations and modifications are possible without departing from the technical solutions described in the claims.

Claims

1. A method for registering three-dimensional medical images with a robotic system, characterized in that, Includes the following steps: Step S1: Data preparation, simultaneously acquiring the registration plate and the target surgical area into the 3D image; Step S2: Read the 3D image, perform interpolation resampling, and read the pre-calibrated center position of the metal sphere template and the diameter of the actual sphere; Step S3: Target layer determination. The initial screening threshold is obtained by sorting the maximum values ​​of each slice layer using the maximum inter-class variance method, and the target layer for segmentation calculation is determined. The target layer for segmentation calculation is determined as follows: Step B1: Select layers with a voxel maximum value greater than the initial screening threshold as the initial screening layers, forming the initial screening layer group L, and count the number of layers N. L Based on the number and diameter of the metal spheres, the theoretical number of distribution layers Lt is calculated. Step B2: Calculate the mean L and standard deviation stdL of the list of maximum values ​​of the initial screening layer group L; Step B3: Calculate the ratio of meanL to standard deviation and compare it with the metal threshold. If the ratio is greater than the metal threshold, take the layer containing the median of the maximum voxel values ​​in the initial screening layer group L as the target layer for subsequent calculation. Otherwise, take the Xth layer from smallest to largest as the target layer for segmentation calculation, X=0.6×Lt. Step S4: Image segmentation threshold calculation. The segmentation threshold is calculated for the target layer using the maximum inter-class variance method. The 3D image is then segmented using the segmentation threshold. Step S5: Image segmentation. Normalize the segmented 3D image, set an area threshold based on the maximum cross-sectional area, and obtain connected components smaller than the area threshold. Step S6: Cluster calculation of sphere center. The number of clusters is determined by the clustering method, the cluster centers are initialized, and after clustering, the classes that meet the mention ratio and sphericity standards are selected to obtain the metal sphere identification results. Step S7: Calculate the distance between all pairs of sphere centers that were actually identified, construct feature vectors, and match the feature vectors of the identified sphere centers with the corresponding template sphere centers; Step S8: Solve the transformation matrix between the two coordinate systems using the singular value decomposition method to complete the coordinate system registration.

2. The method for registering three-dimensional medical images with a robotic system according to claim 1, characterized in that, The initial screening threshold is calculated in step S3 using the following method: Step A1: Calculate the maximum value of voxels in each slice layer of the 3D image, sort the maximum values ​​of each layer, and form a list of maximum values; Step A2: Normalize the voxel values ​​in the maximum value list to the range [0, 255]; Step A3: Set t1 to any integer in the range [0, 255], count the number N of elements less than or equal to t1 in the maximum value list, calculate the average of N elements meanN, and count the number M of elements greater than t1 in the maximum value list and the average of these M elements meanM. Step A4: Calculate the between-class variance Var, using the following formula: Var=(N(meanN-mean) 2 +M(meanM-mean) 2 ) / 256, In the formula, mean is the average of the list of maximum values; Step A5: Traverse the range [0,255] to find t1 that maximizes Var, and obtain the initial screening threshold.

3. The method for registering three-dimensional medical images with a robotic system according to claim 1, characterized in that, In step S4, the segmentation threshold is used to segment the 3D image, and this is done in the following way: The segmentation threshold Ts is calculated for the target layer using the Otsu's method. The entire 3D image is then segmented using the segmentation threshold Ts. Voxel values ​​less than Ts are set to 0, while voxel values ​​greater than or equal to Ts remain unchanged.

4. The method for registering three-dimensional medical images with a robotic system according to claim 1, characterized in that, The image segmentation in step S5 is performed in the following manner: Step S51: Normalize the 3D image using the global maximum and minimum values; Step S52: Calculate the theoretical maximum cross-sectional area Sa of the metal sphere in the two-dimensional image based on the diameter of the metal sphere and the physical pixel spacing of the image; Step S53: Set an area threshold St, traverse each slice layer of the 3D image, calculate the two-dimensional connected components, retain the connected components with an area smaller than the area threshold, and obtain the segmented 3D image.

5. The method for registering three-dimensional medical images with a robotic system according to claim 1, characterized in that, In step S6, the cluster center is initialized in the following way: Step S61: Calculate the distance threshold Td and randomly select the first cluster center C0; Step S62: Calculate the distance from all other points to C0, and select the point with the largest distance from C0 as the center of the second cluster C1; Step S63: Calculate the distance from the remaining points to the second cluster center, and select the point with the largest distance as the next cluster center Cx; Step S64: Repeat step S63 until the distance from all points to the nearest cluster center is less than the distance threshold Td.

6. The method for registering three-dimensional medical images with a robotic system according to claim 5, characterized in that, The center of the sphere is calculated in the following way: The theoretical sphere volume in the image space is calculated, and volume and sphericity are filtered. The sphericity of each type is sorted, and the type with the largest number of sphericities is selected as the preset number of metal spheres. The metal sphere recognition result is obtained, and the center of each type is the recognition center.

7. The method for registering three-dimensional medical images with a robotic system according to claim 1, characterized in that, In step S7, the feature vectors of the identified sphere center and the corresponding template sphere center are matched and identified in the following way: Step S71: Calculate the distance matrix based on the coordinates of each sphere's center. ， In the formula, d 1n This represents the distance between the 1st ball and the nth ball. Step S72: Calculate the minimum value in the distance matrix other than zero, which is the minimum distance between the sphere centers; Step S73: Reorder the elements in the feature vector of the m-th ball in ascending order; Step S74: Calculate the distance matrix and feature vectors of each ball for both the identification ball center and the template ball center. Calculate the distance between two feature vectors. Match the identification ball center and the template ball center corresponding to the two feature vectors with the smallest distance to obtain ball center pairs. Step S75: Repeat step S74 to match the identified ball center with the template ball center one by one.

8. A three-dimensional medical image and robot system registration system, employing the three-dimensional medical image and robot system registration method as described in any one of claims 1-7, characterized in that, It includes a data acquisition module, a preprocessing module, a preliminary screening threshold calculation module, a target layer determination module, an image segmentation module, a normalization and connected component extraction module, a cluster analysis module, a feature vector matching module, and a coordinate system registration module; The data acquisition module is responsible for simultaneously acquiring three-dimensional image data of the registration plate and the target surgical area; The preprocessing module is used to perform interpolation and resampling to make the image data suitable for further analysis; The initial screening threshold calculation module is used to obtain the initial screening threshold from the maximum value sorting of each slice layer by the maximum inter-class variance method, so as to determine which layers are the target layers for segmentation calculation. The target layer determination module is used to determine the target layer that actually needs to be segmented based on the layer whose voxel maximum value is greater than the initial screening threshold.

9. The registration system for three-dimensional medical imaging and robotic systems according to claim 8, characterized in that, The image segmentation module is used to calculate the segmentation threshold for the target layer using the maximum inter-class variance method, and to segment the three-dimensional image using this threshold. The normalization and connected component extraction module is used to normalize the segmented 3D image and set an area threshold based on the maximum cross-sectional area to obtain connected components smaller than the area threshold. The cluster analysis module is used to determine the number of clusters and initialize cluster centers through clustering methods, filter out the classes that meet the mention ratio and sphericity criteria, and finally obtain the metal ball recognition results; The feature vector matching module is used to calculate the distance between all pairs of sphere centers that are actually identified, construct feature vectors, and match the feature vectors of the identified sphere centers with the corresponding template sphere centers. The coordinate system registration module is used to solve the transformation matrix between two coordinate systems using the singular value decomposition method, thereby completing the coordinate system registration.

Images