A body surface registration method and system based on tumor target area distance weight

Abstract

The present invention relates to the technical field of radiotherapy, and more specifically, to a body surface registration method and system based on tumor target area distance weights. A weight model is introduced in the process of searching a real-time point cloud to obtain a point pair set, so that the mean square error of the distance of corresponding points in the point pair is more focused on the error of corresponding points around the isocenter. The weight model is introduced in the calculation process of the transformation matrix, so that the calculated transformation matrix can better reflect the rigid transformation of the body surface point cloud around the target area. The weight of the area closer to the isocenter is larger, and the weight of the area farther from the isocenter is smaller, so that the deformation point cloud in the area farther from the isocenter has less influence on the registration result, that is, the real-time points and reference points in the area near the isocenter tend to coincide, and are not affected by the point cloud far from the isocenter.

Description

Technical Field

[0001] The present invention relates to the technical field of radiotherapy, and more specifically, to a body surface registration method and system based on tumor target area distance weight. Background Art

[0002] In traditional radiotherapy, patient positioning relies on static positioning markers (such as tattoo marks) and CT or MRI image data obtained before treatment. However, slight patient movement during treatment, including movement caused by breathing, changes in bladder filling status, and natural movement of organs, may cause the tumor position to shift, thus affecting the accuracy of treatment. Surface-Guided Radiation Therapy (SGRT) uses a high-precision optical camera system to capture three-dimensional images of the patient's body surface in real time. By comparing images before and after treatment, it can monitor and immediately correct slight movements of the patient's body surface during treatment, ensuring that the tumor is always within the precise irradiation range. This intelligent treatment method can overcome the accuracy defects of traditional radiotherapy. However, current SGRT products mostly use rigid registration algorithms to achieve registration of the entire body surface. This method performs well when treating tumors in rigid body surface areas such as the head. The registration error of the body surface can directly reflect the displacement of the tumor position. However, when treating soft tissue areas such as the chest and abdomen that are easily affected by respiratory movement, respiratory movement will cause dynamic changes in the body surface morphology, and the rigid registration algorithm cannot accurately capture the subtle displacements of the body surface areas adjacent to the tumor, and thus cannot accurately reflect the actual position changes of the tumor, which will have an adverse effect on the accuracy and treatment effect of radiotherapy. Summary of the invention

[0003] The purpose of the present invention is to overcome the deficiency of the rigid registration algorithm in the prior art that it cannot accurately reflect the actual position change of the tumor, and to provide a body surface registration method and system based on the tumor target area distance weight, which can accurately capture the subtle displacement of the body surface area adjacent to the tumor, accurately reflect the actual position change of the tumor, improve the accuracy of radiotherapy, reduce unnecessary radiation to healthy tissues, and improve the treatment effect and the quality of life of patients.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A body surface registration method based on tumor target area distance weight is provided, comprising the following steps:

[0006] Step S100. Acquire the real-time point cloud and the reference point cloud, and perform coarse registration to obtain a first transformation matrix; wherein the real-time point cloud and the reference point cloud are located in the same coordinate system;

[0007] Step S200. According to the first transformation matrix in step S100, the real-time point cloud is subjected to a first matrix transformation to obtain a transformed real-time point cloud;

[0008] Step S300. Traverse each real-time point in the real-time point cloud transformed in step S200 , at the reference point of the reference point cloud The nearest point is found by searching. If the distance between the real-time point and the nearest point found is within the threshold range, the real-time point and the reference point are matched as a point pair, and multiple groups of point pairs constitute a point pair set.

[0009] Step S400. Calculate the overlap fitness and mean square error inlier_rmse of the real-time point and the reference point in the point pair set of step S300, and determine whether the overlap fitness is less than the set overlap threshold and whether the mean square error inlier_rmse is less than the set mean square error mean;

[0010] Step S500. If the overlap fitness is less than the set overlap threshold and the mean square error inlier_rmse is less than the set mean square error mean, then stop the iteration to complete the registration;

[0011] If the overlap fitness is not less than the set overlap threshold, or the mean square error inlier_rmse is not less than the set mean square error mean, then the update matrix is ​​calculated according to the point pair set described in step S300, and the update matrix is ​​used to update the first transformation matrix in step S100 to obtain the second transformation matrix, and the transformed real-time point cloud is subjected to a second matrix transformation according to the second transformation matrix, and steps S200 to S500 are repeated;

[0012] The update matrix is ​​obtained as follows:

[0013]

[0014] In the formula, is the Jacobian matrix weighted by the gradient of the tumor target distance, is the residual vector weighted by the gradient of the distance to the tumor target.

[0015] Preferably, the first transformation matrix in step S100 and the second transformation matrix in step S500 are both 4x4 matrices, expressed as :

[0016]

[0017] in, is a 3x3 three-dimensional rotation matrix, is the three-dimensional translation matrix;

[0018] The first matrix transformation in step S200 and the second matrix transformation in step S500 are both rotation and translation transformations.

[0019] Preferably, in step S500, the second transformation matrix is ​​obtained by the following formula:

[0020] .

[0021] Preferably, the Jacobian matrix based on the tumor target area distance gradient weighting is obtained as follows:

[0022]

[0023] The residual vector weighted by the gradient of the tumor target distance is obtained as follows:

[0024]

[0025] In the formula, For the The Jacobian vector for the point pair, For the The residuals for the point pairs, For the The transpose vector of the Jacobian vector for the pair of points, is the weight model, is the number of point pairs in the point pair set.

[0026] Preferably, The Jacobian vector of the point pair, The residuals for the point pairs are obtained as follows:

[0027] No. The first three rows of the Jacobian vectors for the point pairs = Point to point Middle and real time point Corresponding reference point The normal vector Directional projection;

[0028] No. The last three rows of the Jacobian vectors for the point pairs = Point to point Middle and real time point Corresponding reference point The normal vector ;

[0029] No. Residuals for pairs of points =No. Point to point Mid-real-time point The corresponding reference point The residual vector of Normal vector The projected length in the direction.

[0030] Preferably, the and stated Through the LDLT decomposition principle, a six-dimensional transformation vector is obtained; according to the matrix transformation principle, the six-dimensional transformation vector is converted into a 4x4 transformation matrix, and the 4x4 transformation matrix is ​​the update matrix.

[0031] Preferably, in step S500, the fitness of the point pair is calculated according to the following formula:

[0032]

[0033] In the formula, is the number of pairs in the pair set, is the number of real-time point cloud points.

[0034] Preferably, in step S500, the mean square error inlier_rmse of the point pair distance is calculated as follows:

[0035]

[0036]

[0037] in, is the number of pairs in the pair set, and are the 3D points of the i-th pair of points located on the real-time point cloud and the reference point cloud respectively; represents the three-dimensional coordinates of the real-time point in the i-th pair of points, represents the three-dimensional coordinates of the reference point in the i-th pair of points, is the weight model.

[0038] Preferably, the weight model Obtained by the following steps:

[0039] Use the kdtree algorithm to search for the point closest to the isocenter on the reference point cloud and the closest distance value miu, and set the standard deviation The value of

[0040] Searched point pairs Real-time point cloud , calculate its distance to the isocenter dist;

[0041] The weight model Set to:

[0042]

[0043] in, represents the Euclidean distance from the reference point to the isocenter in the i-th point pair, The introduction of the weight model makes the area closer to the isocenter have a larger weight, and the area farther from the isocenter has a smaller weight, so that the deformation point cloud in the area farther from the isocenter has less influence on the registration result, that is, the real-time points and reference points in the area near the isocenter tend to coincide, and are not affected by the point cloud far from the isocenter.

[0044] The present invention also provides a body surface registration system based on tumor target area distance weight, comprising:

[0045] Data acquisition module: used to obtain real-time point cloud and reference point cloud;

[0046] Rough registration module: used for receiving the point cloud data transmitted by the data acquisition module, and performing rough registration to obtain the first transformation matrix;

[0047] Transformation module: Based on the first transformation matrix obtained by the coarse registration module, the real-time point cloud is subjected to the first matrix transformation to obtain the transformed real-time point cloud;

[0048] Traversal module: used to traverse each real-time point in the real-time point cloud transformed by the transformation module, find the nearest reference point, and match the points whose distances between the real-time point and the reference point are within the threshold range as point pairs, and build a point pair set based on the point pairs;

[0049] Calculation module: used to calculate the overlap fitness and mean square error inlier_rmse of the real-time point and the reference point in the point pair set; and used to determine whether the overlap fitness is less than the set overlap threshold, and whether the mean square error inlier_rmse is less than the set mean square error mean;

[0050] Execution module: used to execute step S500 according to the calculation result of the calculation module.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] The body surface registration method and system based on tumor target area distance weight of the present invention perform a first matrix transformation on the real-time point cloud through a first transformation matrix, perform a second matrix transformation on the real-time point cloud after the first transformation through a second transformation matrix, and perform multiple matrix transformations until the overlap of all point pairs in the point pair set and the mean square error of the point pairs are less than the set value, thereby completing the registration of the real-time point and the reference point; the present invention introduces a weight model in the process of searching the real-time point cloud to obtain the point pair set, so that the mean square error of the distance of the corresponding points in the point pair is more focused on the error of the corresponding points around the distance isocenter point, and introduces a weight model in the calculation process of the transformation matrix, so that the calculated matrix can better reflect the rigid transformation of the body surface point cloud around the target area, the area closer to the distance isocenter point has a larger weight, and the area farther from the distance isocenter point has a smaller weight, so that the deformation point cloud in the area farther from the isocenter point has a smaller influence on the registration result, that is, the real-time point and the reference point in the area near the isocenter point tend to coincide, and are not affected by the point cloud far from the isocenter point. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 The flowchart of the body surface registration method based on tumor target area distance weight;

[0054] Figure 2 This is the effect diagram of the registration based on the rigid registration algorithm;

[0055] Figure 3 This is the registration effect diagram of the body surface registration method based on the tumor target area distance weight; DETAILED DESCRIPTION

[0056] The present invention will be further described below in conjunction with specific implementation modes.

[0057] Embodiment 1

[0058] like Figure 1 The first embodiment of the body surface registration method based on tumor target area distance weight of the present invention is shown, comprising the following steps:

[0059] Step S100. Acquire the real-time point cloud and the reference point cloud, and perform coarse registration to obtain a first transformation matrix; wherein the real-time point cloud and the reference point cloud are located in the same coordinate system;

[0060] Step S200. According to the first transformation matrix in step S100, the real-time point cloud is subjected to a first matrix transformation to obtain a transformed real-time point cloud;

[0061] Step S300. Traverse each real-time point in the real-time point cloud transformed in step S200 , at the reference point of the reference point cloud The nearest point is found by searching. If the distance between the real-time point and the nearest point searched is within the threshold range, the real-time point and the reference point are matched as a point pair, and multiple groups of point pairs constitute a point pair set. In this embodiment, the threshold is set according to different treatment sites and deformation sizes, and is generally set to 1.5 to 3 times the voxel size.

[0062] Step S400. Calculate the overlap fitness and mean square error inlier_rmse of the real-time point and the reference point in the point pair set of step S300, and determine whether the overlap fitness is less than the set overlap threshold and whether the mean square error inlier_rmse is less than the set mean square error mean;

[0063] Step S500. If the overlap fitness is less than the set overlap threshold and the mean square error inlier_rmse is less than the set mean square error mean, then stop the iteration to complete the registration;

[0064] If the overlap fitness is not less than the set overlap threshold, or the mean square error inlier_rmse is not less than the set mean square error, then the update matrix is ​​calculated according to the point pair set described in step S300, and the update matrix is ​​used to update the first transformation matrix in step S100 to obtain the second transformation matrix, and the transformed real-time point cloud is subjected to a second matrix transformation according to the second transformation matrix, and steps S200 to S500 are repeated. Among them, the overlap threshold and the mean square error are set according to the actual application. In this embodiment, the overlap threshold is set to 0.95 and the mean square error is set to 0.05.

[0065] In step S100, the first transformation matrix and the second transformation matrix in step S500 are both 4x4 matrices, expressed as :

[0066]

[0067] in, is a 3x3 three-dimensional rotation matrix, is the three-dimensional translation matrix;

[0068] In step S200, the first matrix transformation and in step S500, the second matrix transformation are both rotation and translation transformations.

[0069] In step S500, the second transformation matrix is ​​obtained by the following formula:

[0070] .

[0071] The update matrix is ​​obtained as follows:

[0072]

[0073] In the formula, is the Jacobian matrix weighted by the gradient of the tumor target distance, is the residual vector weighted by the gradient of the distance to the tumor target.

[0074] The Jacobian matrix based on the tumor target area distance gradient weighting is obtained as follows:

[0075]

[0076] The residual vector weighted by the gradient of the tumor target distance is obtained as follows:

[0077]

[0078] In the formula, For the The Jacobian vector for the point pair, For the The residuals for the point pairs, For the The transpose vector of the Jacobian vector for the pair of points, is the weight model, is the number of pairs in the pair set.

[0079] No. The Jacobian vector of the point pair, The residuals for the point pairs are obtained as follows:

[0080] No. The first three rows of the Jacobian vectors for the point pairs = Point to point Middle and real time point Corresponding reference point The normal vector The projection of the direction; that is, the product of the three-dimensional coordinate vector of the corresponding point belonging to the real-time point cloud and the three-dimensional normal vector of the corresponding point belonging to the reference point cloud;

[0081] No. The last three rows of the Jacobian vectors for the point pairs = Point to point Middle and real time point Corresponding reference point The normal vector ; That is, the three-dimensional normal vector of the point in the corresponding point that belongs to the reference point cloud;

[0082] No. Residuals for pairs of points =No. Point to point Mid-real-time point The corresponding reference point The residual vector of Normal vector The projected length in the direction.

[0083] Said and stated Through the LDLT decomposition principle, a six-dimensional transformation vector is obtained; according to the matrix transformation principle, the six-dimensional transformation vector is converted into a 4x4 transformation matrix, and the 4x4 transformation matrix is ​​the update matrix.

[0084] In step S500, the fitness of the point pair and the mean square error inlier_rmse of the point pair distance are calculated as follows:

[0085]

[0086] In the formula, is the number of point pairs, is the number of real-time point cloud points;

[0087]

[0088]

[0089] in, is the number of corresponding point pairs, and are the 3D points of the i-th pair of points located on the real-time point cloud and the reference point cloud respectively; represents the three-dimensional coordinates of the real-time point in the i-th pair of points, represents the three-dimensional coordinates of the reference point in the i-th pair of points, is the weight model.

[0090] The weight model Obtained by the following steps:

[0091] Use the kdtree algorithm to search for the point closest to the isocenter on the reference point cloud and the closest distance value miu, and set the standard deviation The value of

[0092] The weight model Set to:

[0093]

[0094] in, represents the Euclidean distance from the reference point to the isocenter in the i-th point pair, The introduction of the weight model makes the area closer to the isocenter have a larger weight, and the area farther from the isocenter has a smaller weight, so that the deformation point cloud in the area farther from the isocenter has less influence on the registration result, that is, the real-time points and reference points in the area near the isocenter tend to coincide, and are not affected by the point cloud far from the isocenter.

[0095] After the above steps, a weight model is introduced in the process of searching the real-time point cloud to obtain a point pair set, so that the mean square error of the distance of the corresponding points in the point pair is more focused on the error of the corresponding points around the isocenter. The weight model is introduced in the calculation process of the transformation matrix, so that the calculated matrix can better reflect the rigid transformation of the surface point cloud around the target area. The area closer to the isocenter has a larger weight, and the area farther from the isocenter has a smaller weight, so that the deformation point cloud in the area farther from the isocenter has less influence on the registration result, that is, the real-time points and reference points in the area near the isocenter tend to coincide, and are not affected by the point cloud far from the isocenter.

[0096] During the treatment of breast tumors, respiratory motion will cause dynamic changes in the body surface morphology. Due to respiratory motion, the position of the tumor will also change, and the optical body surface is to detect the body surface motion as much as possible to monitor whether the tumor target area is shifted. The results of the existing rigid registration method are as follows: Figure 2 The body surface registration method based on tumor target area distance weight in this embodiment is used for registration during breast tumor treatment. The registration result is shown in Figure 3 shown. Figure 2 and Figure 3 In the figure, the center points of the three coordinate axes coincide with the isocenter of the breast tumor target area.

[0097] When using the rigid registration algorithm for registration, the real-time point cloud and the reference point cloud tend to overlap as much as possible, which cannot accurately capture the subtle displacement of the surface area adjacent to the tumor, and thus cannot accurately reflect the actual position change of the tumor, which has an adverse effect on the accuracy and treatment effect of radiotherapy. Figure 2 As shown; Figure 2 In the figure, the lighter point cloud is the reference point cloud, and the darker point cloud is the real-time point cloud.

[0098] When the surface registration method based on the tumor target area distance weight of the present embodiment is used for registration, the body surface of the breast that moves due to respiratory movement is accurately monitored, thereby indirectly monitoring the movement of the tumor target area. The real-time points and reference points in the area near the isocenter tend to coincide with each other and are not affected by the point cloud far from the isocenter. Therefore, the slight displacement of the body surface area adjacent to the tumor can be accurately captured, and the actual position change of the tumor can be accurately reflected, thereby improving the accuracy and treatment effect of radiotherapy. Figure 3 In the figure, the lighter point cloud is the reference point cloud, and the darker point cloud is the real-time point cloud.

[0099] The subtle displacement also reflects an overall rigid transformation of the tumor's adjacent surface. Specifically, each point of the chest and abdominal tumor adjacent surface will be displaced in the Z-axis direction with breathing, but the displacement of each point is inconsistent. Based on the registration method of this application, the displacement of each point in the entire adjacent area is calculated according to the different weights of each point, thereby calculating the changes in the entire body surface area to predict or replace the movement of the tumor target area, thereby improving the accuracy and safety of radiotherapy.

[0100] Embodiment 2

[0101] This embodiment is an embodiment of a body surface registration system based on tumor target area distance weights. The body surface registration system based on tumor target area distance weights includes:

[0102] Data acquisition module: used to obtain real-time point cloud and reference point cloud;

[0103] Rough registration module: used for receiving the point cloud data transmitted by the data acquisition module, and performing rough registration to obtain the first transformation matrix;

[0104] Transformation module: Based on the first transformation matrix obtained by the coarse registration module, the real-time point cloud is subjected to the first matrix transformation to obtain the transformed real-time point cloud;

[0105] Traversal module: used to traverse each real-time point in the real-time point cloud transformed by the transformation module, find the nearest reference point, and match the points whose distances between the real-time point and the reference point are within the threshold range as point pairs, and build a point pair set based on the point pairs;

[0106] Calculation module: used to calculate the overlap fitness and mean square error inlier_rmse of the real-time point and the reference point in the point pair set; and used to determine whether the overlap fitness is less than the set overlap threshold, and whether the mean square error inlier_rmse is less than the set mean square error mean;

[0107] Execution module: used to execute step S500 according to the calculation result of the calculation module.

[0108] The body surface registration system based on tumor target area distance weight of the present invention introduces a weight model in the process of searching the real-time point cloud to obtain a point pair set, so that the mean square error of the distance of corresponding points in the point pair is more focused on the error of corresponding points around the isocenter. The weight model is introduced in the calculation process of the transformation matrix, so that the calculated matrix can better reflect the rigid transformation of the body surface point cloud around the target area. The area closer to the isocenter has a larger weight, and the area farther from the isocenter has a smaller weight, so that the deformation point cloud in the area farther from the isocenter has less influence on the registration result, that is, the real-time points and reference points in the area near the isocenter tend to coincide, and are not affected by the point cloud far from the isocenter.

[0109] In the specific contents of the above-mentioned specific implementation methods, the various technical features can be combined in any non-contradictory manner. In order to make the description concise, not all possible combinations of the above-mentioned technical features are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0110] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A body surface registration method based on tumor target area distance weight, characterized in that: The following steps are involved: Step S100. Acquire the real-time point cloud and the reference point cloud, and perform coarse registration to obtain a first transformation matrix; wherein the real-time point cloud and the reference point cloud are located in the same coordinate system; Step S200. According to the first transformation matrix in step S100, the real-time point cloud is subjected to a first matrix transformation to obtain a transformed real-time point cloud; Step S300. Traverse each real-time point in the real-time point cloud transformed in step S200 , at the reference point of the reference point cloud The nearest point is found by searching. If the distance between the real-time point and the nearest point found is within the threshold range, the real-time point and the reference point are matched as a point pair, and multiple groups of point pairs constitute a point pair set. Step S400. Calculate the overlap fitness and mean square error inlier_rmse of the real-time point and the reference point in the point pair set of step S300, and determine whether the overlap fitness is less than the set overlap threshold and whether the mean square error inlier_rmse is less than the set mean square error mean; Step S500. If the overlap fitness is less than the set overlap threshold and the mean square error inlier_rmse is less than the set mean square error mean, then stop the iteration to complete the registration; If the overlap fitness is not less than the set overlap threshold, or the mean square error inlier_rmse is not less than the set mean square error mean, then the update matrix is ​​calculated according to the point pair set described in step S300, and the update matrix is ​​used to update the first transformation matrix in step S100 to obtain the second transformation matrix, and the transformed real-time point cloud is subjected to a second matrix transformation according to the second transformation matrix, and steps S200 to S500 are repeated; Wherein, the update matrix X is obtained as follows: In the formula, is the Jacobian matrix weighted by the gradient of the tumor target distance, is the residual vector weighted by the gradient of the distance to the tumor target area; The Jacobian matrix based on the tumor target area distance gradient weighting is obtained as follows: The residual vector weighted by the gradient of the tumor target distance is obtained as follows: In the formula, For the The Jacobian vector for the point pair, For the The residuals for the point pairs, For the The transpose vector of the Jacobian vector for the pair of points, is the weight model, is the number of point pairs in the point pair set; The weight model is obtained by the following steps: Use the kdtree algorithm to search for the point closest to the isocenter on the reference point cloud and the closest distance value miu, and set the standard deviation The value of The weight model Set to: in, represents the Euclidean distance from the reference point to the isocenter in the i-th point pair, is the distance variance.

2. The body surface registration method based on tumor target area distance weight according to claim 1, characterized in that: The first transformation matrix in step S100 and the second transformation matrix in step S500 are both 4x4 matrices, expressed as : in, is a 3x3 three-dimensional rotation matrix, is the three-dimensional translation matrix; In step S200, the first matrix transformation and in step S500, the second matrix transformation are both rotation and translation transformations.

3. The body surface registration method based on tumor target area distance weight according to claim 1, characterized in that: In step S500, the second transformation matrix is ​​obtained by the following formula: 。 4. The body surface registration method based on tumor target area distance weight according to claim 1, characterized in that: No. The Jacobian vector of the point pair, The residuals for the point pairs are obtained as follows: No. The first three rows of the Jacobian vectors for the point pairs = Point to point Middle and real time point Corresponding reference point The normal vector Directional projection; No. The last three rows of the Jacobian vectors for the point pairs = Point to point Middle and real time point Corresponding reference point The normal vector ; No. Residuals for pairs of points =No. Point to point Mid-real-time point The corresponding reference point The residual vector of Normal vector The projected length in the direction.

5. The body surface registration method based on tumor target area distance weight according to claim 1, characterized in that: Said and stated Through the LDLT decomposition principle, a six-dimensional transformation vector is obtained; according to the matrix transformation principle, the six-dimensional transformation vector is converted into a 4x4 transformation matrix, and the 4x4 transformation matrix is ​​the update matrix.

6. The body surface registration method based on tumor target area distance weight according to claim 1, characterized in that: In step S500, the fitness of the point pair is calculated as follows: In the formula, is the number of pairs in the pair set, is the number of real-time point cloud points.

7. The body surface registration method based on tumor target area distance weight according to claim 1, characterized in that: In step S500, the mean square error inlier_rmse of the point pair distance is calculated as follows: in, is the number of pairs in the pair set, and are the 3D points of the i-th pair of points located on the real-time point cloud and the reference point cloud respectively; represents the three-dimensional coordinates of the real-time point in the i-th pair of points, represents the three-dimensional coordinates of the reference point in the i-th pair of points, is the weight model.

8. A body surface registration system based on tumor target area distance weight, characterized in that: The system is used to implement the body surface registration method based on tumor target area distance weight according to any one of claims 1 to 7, comprising: Data acquisition module: used to obtain real-time point cloud and reference point cloud; Rough registration module: used for receiving the point cloud data transmitted by the data acquisition module, and performing rough registration to obtain the first transformation matrix; Transformation module: Based on the first transformation matrix obtained by the coarse registration module, the real-time point cloud is subjected to the first matrix transformation to obtain the transformed real-time point cloud; Traversal module: used to traverse each real-time point in the real-time point cloud transformed by the transformation module, find the nearest reference point, and match the points whose distances between the real-time point and the reference point are within the threshold range as point pairs, and build a point pair set based on the point pairs; Calculation module: used to calculate the overlap fitness and mean square error inlier_rmse of the real-time point and the reference point in the point pair set; and used to determine whether the overlap fitness is less than the set overlap threshold, and whether the mean square error inlier_rmse is less than the set mean square error mean; Execution module: used to execute step S500 according to the calculation result of the calculation module.

Images