A multi-localized image registration method based on body surface fixation and a multi-localized image drift detection method

Abstract

A method for multiple localized image registration based on body surface fixation and a method for detecting multiple localized image drift based on body surface fixation. The method for multiple localized image registration based on body surface fixation screens out irrelevant and erroneous point clouds step by step so that they do not participate in the registration operation, thereby accelerating the registration speed and avoiding mismatching; and screens out the correct point clouds as much as possible for registration step by step to ensure the registration accuracy. The method for detecting multiple localized image drift based on body surface fixation screens out erroneous grids step by step through the body surface model and the three-dimensional reconstruction model after operation so that they do not participate in the registration operation, thereby accelerating the registration speed and avoiding mismatching; and screens out the correct grids as much as possible step by step to accurately detect the drift result of the bone image after operation.

Description

Technical Field

[0001] The present invention relates to the fields of computer science and visualization recognition and simulation technology in medical digital surgery, and in particular to a body surface fixed modeling and multiple localized image registration method, and a body surface fixed multiple localized image drift detection method. Background Art

[0002] Intraoperative image drift refers to any change in the relative position of all pixels in a three-dimensional image. Changes in the human body's surface morphology and internal structure can cause changes in all pixels in the three-dimensional model and point cloud image. The safety, accuracy, and robustness of surgical navigation systems are very important for the prognosis of the subject. The three-dimensional spatial mapping relationship between the pre-operative image (virtual space) inspection scene image scan (real space) determines the success of the operation. The pre-operative CT scan, in-operative image registration, and navigation surgery application accuracy parameters of the surgical navigation system are particularly important. For example, spinal manipulation can cause changes in vertebral position and morphology, and key anatomical landmarks are difficult to assist in positioning. The spatial mapping misalignment of the instrument can also mislead the operator's judgment and lead to incorrect insertion of the implant.

[0003] The factors that cause image drift are: 1. Distortion of CT scan images before operation will cause image drift during operation; 2. Image drift during operation caused by operation scene scanning. For example, when the relative positions of the reference frame and the target vertebra are close, the possibility of accidental contact between the reference frame and the instrument increases, and this contact may affect the operation and the field of view of the image-guided instrument, or reduce the accuracy of the navigation system; 3. Registration error of the navigation map. The image registration of the navigation system is to enter the virtual navigation map into the world coordinate system and perform matrix conversion with the scene during operation to achieve coordinate system mapping; whether it is based on markers, textures, anatomical landmarks, screws or surface form image registration technology, its navigation accuracy has certain errors. 4. Tracking error of the instrument. For example, the optical navigation system nail placement instrument still has an error of about 0.6 mm after calibration. The error level depends on the spatial distance between the pointer, the reference frame and the optical tracking device. The reference frame is accidentally moved, the light source is projected and reflected, the nail placement instrument loses its signal for a long time, and the spatial mapping of the nail placement instrument is misplaced, which can also cause image drift during operation and even navigation operation failure. 5. Actual operation will cause image drift, which is caused by the elasticity and support of soft tissue and the rigidity of bone tissue. Soft tissue has a certain elasticity. When the puncture instrument reaches the lesion site, the surface of the tissue at the lesion site may be deformed or hollow, and the tissue morphology may not be able to rebound to its original state, causing drift. When puncturing and nailing, if the hand drill speed, cutting angle, cutting area, drill bit material and bone hardness are not well matched, the drill bit cutting force is likely to be insufficient and image drift will occur. Image drift caused by actual operation is unavoidable. Image drift before and during operation will affect the operation accuracy and success of the operation.

[0004] At present, there are some solutions for preventing image drift in surgical navigation systems. For example, in the field of orthopedic surgery, Tianji Orthopedic Robot (TINAVI, China) uses a restraint to limit trunk displacement, and Medtronic Mazor Robotic (medtronic, inc, USA) uses pins to fix the ilium, spinous process, femur, tibia and other parts, causing additional trauma to the patient's bone surface. In the field of neurosurgery, Sinnovation and Remebot use head pins combined with Leksell frames to fix the skull to prevent movement and displacement of the head, but brain tissue drift has not been completely resolved and it also causes additional head trauma.

[0005] Therefore, in view of the shortcomings of the prior art, it is necessary to provide a multi-localized image registration method based on body surface fixation and a multi-localized image drift detection method based on body surface fixation to solve the shortcomings of the prior art. Summary of the invention

[0006] The first object of the present invention is to avoid the shortcomings of the prior art and provide a multi-region image registration method based on body surface fixation. The method provides a set of body surface fixation methods, and at the same time, establishes a multi-region image registration method based on body surface fixation to perform multiple registrations on the body surface fixed model and point cloud and delete irrelevant or deformed point clouds, thereby preventing incorrect registration and improving registration accuracy.

[0007] The above-mentioned purpose of the present invention is achieved by the following technical measures:

[0008] A multi-region image registration method based on body surface fixation is provided, wherein the body surface of an object before operation is fixed to obtain a solidified object, a three-dimensional reconstruction model of the solidified object is obtained, and an operation planning is performed on the three-dimensional reconstruction model to obtain a planning model;

[0009] Then, the body surface point cloud and the environment point cloud of the solidified object are collected, and the body surface point cloud and the environment point cloud are multiple-registered with the planning model and irrelevant or deformed point clouds are deleted multiple times to obtain a body surface point cloud matrix;

[0010] Finally, the body surface point cloud matrix and the planning model are pasted, and then the body surface point cloud matrix and the planning model are moved to the virtual operation area to complete the registration.

[0011] The multi-region image registration method based on body surface fixation of the present invention is performed by the following steps:

[0012] S1, fixing the torso of the object before operation to obtain the solidified object; collecting CT data of the solidified object by a spiral CT device, reconstructing the CT data to obtain a three-dimensional reconstruction model, and performing operation planning on the three-dimensional reconstruction model to obtain a planning model;

[0013] S2, collecting the surface point cloud and environment point cloud of the solidified object;

[0014] S3, importing the planning model, the body surface point cloud and the environment point cloud obtained in S2 into the navigation software;

[0015] S4, moving the planning model closer to the body surface point cloud and the environment point cloud;

[0016] S5, coarsely aligning the body surface point cloud and the environment point cloud with the planning model to obtain the corresponding RMS value θ1;

[0017] S6, measuring the distance between the body surface point cloud and the environment point cloud and the planning model, marking the point cloud with a distance greater than α1 in the body surface point cloud and the environment point cloud, and there is θ1≤α1≤15θ1, and entering S7;

[0018] S7, in the marked body point cloud and environment point cloud, delete the point cloud whose distance from the planning model is outside the range of ±α2, so as to remove the redundant scene point cloud, and obtain the first screening body point cloud, and there exists θ1≤α2≤10θ1;

[0019] S8, performing a first fine registration of the first screened body surface point cloud in S6 with the planning model to obtain a corresponding RMS value θ2, and proceeding to S9;

[0020] S9, measuring the distance between the first screened body surface point cloud and the planning model, deleting the point cloud whose distance from the planning model is outside the range of ±α3 in the first screened body surface point cloud, and α3=θ2, thereby removing the collapsed or expanded deformed point cloud, obtaining the second screened body surface point cloud, and entering S10;

[0021] S10, performing a second fine registration of the second screened-out body surface point cloud obtained in S9 with the planning model to obtain a corresponding RMS value θ3, and proceeding to S11;

[0022] S11, measuring the distance between the second screened body surface point cloud and the planning model, deleting the body surface point cloud whose distance from the planning model is outside the range of ±α4, and α4=θ3, thereby removing the collapsed or expanded deformed point cloud again, obtaining the third screened body surface point cloud, and entering S12;

[0023] S12, performing a third fine registration of the third-removed body surface point cloud obtained in S11 with the planning model, obtaining the registered third-removed body surface point cloud and corresponding coordinates, defining them as a body surface point cloud matrix, and proceeding to S13;

[0024] S13, pasting the body surface point cloud matrix into the planning model, and using an inverse matrix to move the body surface point cloud matrix and the planning model as a whole to a virtual operation area.

[0025] Preferably, the above S1 is performed by the following steps:

[0026] S1.1. Fixing the object before operation by a polymer polyurethane bandage to obtain a solidified object before operation;

[0027] S1.2. Collect CT data of a solidified object on a spiral CT device, perform three-dimensional reconstruction on the CT data to obtain a three-dimensional reconstruction model, and perform operation planning on the three-dimensional reconstruction model to obtain the planning model; wherein the planning model is provided with a CT body surface model, a CT bone model and a planning channel, and the three-dimensional reconstruction model is provided with a pre-operation CT body surface model and a pre-operation CT bone model.

[0028] Preferably, the fixing process is performed by the following steps:

[0029] a1. Place the subject's torso on a flat board before the operation;

[0030] a2. Use a moistened high-molecular polyurethane bandage to cover the body surface and both ends of the joints of the subject, then shape the high-molecular polyurethane bandage and squeeze the gap between the subject and the fixed plate so that the high-molecular polyurethane bandage fits the operation site;

[0031] a3. Spraying physiological saline onto the polymer polyurethane bandage to plasticize the polymer polyurethane bandage;

[0032] a4. Fix the end of the polymer polyurethane bandage on the flat plate by means of staples to obtain a solidified object before operation.

[0033] Preferably, the above S2 specifically involves collecting the surface point cloud and the environment point cloud of the solidified object through a 3D dynamic structured light camera.

[0034] Preferably, in the above S4, the planning model of S3 is brought closer to the body surface point cloud and the environment point cloud through a rotation-displacement function.

[0035] Preferably, in the above S5, a plurality of groups of common registration points are selected respectively in the planning model and the body surface point cloud and the environment point cloud, and then the planning model and the body surface point cloud are aligned through a coarse registration function.

[0036] The present invention provides a multi-localized image registration method based on body surface fixation. First, the body surface of the object before operation is fixed to obtain a solidified object, and then a three-dimensional reconstruction model of the solidified object is obtained, and the operation planning is performed on the three-dimensional reconstruction model to obtain a planning model; then, the body surface point cloud and the environmental point cloud of the solidified object are collected, and the body surface point cloud and the environmental point cloud are multiple-registered with the planning model and irrelevant or deformed point clouds are deleted multiple times to obtain a body surface point cloud matrix; finally, the body surface point cloud matrix is ​​pasted with the planning model, and then the body surface point cloud matrix and the planning model are moved to the virtual operation area to complete the registration. The multi-localized image registration method screens out irrelevant and erroneous point clouds step by step so that they do not participate in the registration operation, so as to speed up the registration speed and avoid mismatching; the correct point clouds are screened step by step to be retained as much as possible for registration to ensure the registration accuracy.

[0037] The second object of the present invention is to avoid the shortcomings of the prior art and provide a method for detecting multiple localized image drift based on body surface fixation. The method for detecting multiple localized image drift based on body surface fixation registers the body surface, and then performs multiple registrations and deletes irrelevant or deformed point clouds, thereby preventing misregistration and improving registration accuracy.

[0038] The above-mentioned purpose of the present invention is achieved by the following technical measures:

[0039] A method for detecting drift of multiple localized images based on body surface fixation is provided. First, CT data of a solidified object after operation is collected, and the CT data is three-dimensionally reconstructed to obtain a post-operation model, wherein the post-operation model includes a post-operation body surface model and a post-operation bone model.

[0040] Then, the three-dimensional reconstructed model is multiple-registered with the post-operation model and irrelevant or deformed grids are deleted multiple times to obtain a post-operation body surface grid matrix;

[0041] Finally, the post-operation body surface grid matrix is ​​pasted to the post-operation skeleton model, and the post-operation skeleton model is moved to the pre-operation CT skeleton model in the three-dimensional reconstruction model to obtain a skeleton image drift result.

[0042] The method for detecting drift of multiple localized images based on body surface fixation of the present invention is performed by the following steps:

[0043] A1, collecting CT data of the solidified object after the operation, and performing three-dimensional reconstruction on the CT data to obtain the post-operation model;

[0044] A2, importing the post-operation model and the three-dimensional reconstruction model into the navigation software, performing rough registration and alignment between the post-operation model and the pre-operation CT body surface model in the three-dimensional reconstruction model, obtaining the corresponding RMS value δ1, and proceeding to A3;

[0045] A3, measuring the distance between the post-operation body surface model and the pre-operation CT body surface model, marking the grids in the post-operation body surface model that are greater than the distance β1, and if δ1≤β1≤15δ1 exists, proceed to A4;

[0046] A4. In the marked grids of the post-operation body surface model, the grids whose distances from the pre-operation CT body surface model are outside the range of ±β2 are deleted to obtain the first screening post-operation body surface model, and δ1≤β2≤10δ1 exists;

[0047] A5, performing a first fine registration of the body surface model after the first screening operation with the CT body surface model before the operation to obtain a corresponding RMS value δ2, and proceeding to A6;

[0048] A6. measuring the distance between the model after the first screening operation and the CT body surface model before the operation, and deleting the grids of the body surface model after the first screening operation whose distances from the CT body surface model before the operation are outside the range of ±β3, and β3=δ2 exists, thereby removing the collapsed or expanded deformed grids, and obtaining the body surface model after the second screening operation;

[0049] A7, performing a second fine registration on the body surface model after the second screening operation and the CT body surface model before the operation to obtain a corresponding RMS value δ3, and proceeding to A8;

[0050] A8, measuring the distance between the body surface model after the second screening operation and the three-dimensional reconstructed model, and deleting the grids whose distances from the CT body surface model before the operation are outside the range of ±β4 in the grids of the model after the second screening operation, and β4=δ3, thereby removing the collapsed or expanded deformed grids, obtaining the model after the third screening operation, and proceeding to A9;

[0051] A9, performing a third fine registration of the third screened post-operation body surface model obtained in A8 with the pre-operation CT body surface model, obtaining the registered third screened post-operation body surface model and the corresponding coordinates defined as a post-operation body surface grid matrix, and proceeding to A10;

[0052] A10. Paste the post-operation body surface grid matrix into the post-operation skeleton model, and use the cis matrix to move the post-operation skeleton model to the pre-operation CT skeleton model to obtain a skeleton image drift result.

[0053] Preferably, the above A2 specifically involves importing the post-operation model and the three-dimensional reconstructed model into the navigation system software, selecting multiple groups of common alignment points in the three-dimensional reconstructed model and the post-operation model respectively, and then aligning the three-dimensional reconstructed model and the post-operation model through a coarse alignment function.

[0054] The invention discloses a method for detecting drift of a multi-localized image based on body surface fixation. The CT data of a solidified object after operation is first collected. The CT data is three-dimensionally reconstructed to obtain a model after operation, wherein the model after operation includes a body surface model after operation and a bone model after operation. Then, the three-dimensional reconstructed model in the above-mentioned multi-localized image registration method is multiple-registered with the model after operation and irrelevant or deformed grids are deleted multiple times to obtain a body surface grid matrix after operation. Finally, the body surface grid matrix after operation is pasted to the bone model after operation, and the bone model after operation is moved to the CT bone model before operation to obtain a bone image drift result. The method for detecting drift of a multi-localized image based on body surface fixation screens out wrong grids step by step through the body surface model after operation and the three-dimensional reconstructed model so that the grids do not participate in the registration operation, so as to speed up the registration speed and avoid mismatching. The correct grids are screened step by step to be retained as much as possible, thereby accurately detecting the drift result of the bone image after operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] The present invention is further described with reference to the accompanying drawings, but the contents in the accompanying drawings do not constitute any limitation to the present invention.

[0056] Figure 1The flowchart is a multi-region image registration method based on body surface fixation.

[0057] Figure 2 The flowchart of a method for detecting drift of multiple localized images based on body surface fixation is shown in FIG.

[0058] Figure 3 This is a picture of the cured pig in Example 3.

[0059] Figure 4 for Figure 3 Another angle picture of .

[0060] Figure 5 Planning model diagram for solidifying pigs.

[0061] Figure 6 This is a point cloud diagram of the surface point cloud of the solidified object and the scene point cloud diagram of the environment in Example 3.

[0062] Figure 7 This is a schematic diagram of bringing the planning model closer to the body surface point cloud in S4 of Example 3.

[0063] Figure 8 This is the registration point diagram in S5 of Example 3.

[0064] Fig. 9 This is the image after rough registration in S5 of Example 3.

[0065] Fig.10 This is the marked figure in S6 of Example 3.

[0066] Fig.11 This is the first screening of the body surface point cloud map in S7 of Example 3.

[0067] Fig.12 This is the image after the first fine registration in S8 of Example 3.

[0068] Fig.13 This is the second screening body surface point cloud map in S9 of Example 3.

[0069] Fig.14 This is the image after the second fine registration in S10 of Example 3.

[0070] Fig.15 This is the third screened-out body surface point cloud in S11 of Example 3.

[0071] Fig.16 This is the image after the third fine registration in S12 of Example 3.

[0072] Fig.17 This is a diagram after moving to the virtual operation area in S13 of Example 3.

[0073] Fig.18This is the post-operation model of Example 4.

[0074] Fig.19 This is the image after rough registration in A2 of Example 4.

[0075] Fig. 20 This is the figure marked after A3 in Example 4.

[0076] Fig.21 This is the body surface model after the first screening operation in A4 of Example 4.

[0077] Fig. 22 This is the bone image drift result diagram in A10 of Example 4. DETAILED DESCRIPTION

[0078] The technical solution of the present invention is further described in conjunction with the following embodiments.

[0079] Example 1

[0080] A multi-region image registration method based on body surface fixation, such as Figure 1 First, the surface of the object before the operation is fixed to obtain a solidified object, then a three-dimensional reconstruction model of the solidified object is obtained, and the operation planning is performed on the three-dimensional reconstruction model to obtain a planning model;

[0081] Then, the body surface point cloud and the environment point cloud of the solidified object are collected, and the body surface point cloud and the environment point cloud are multiple-registered with the planning model and irrelevant or deformed point clouds are deleted multiple times to obtain a body surface point cloud matrix;

[0082] Finally, the body surface point cloud matrix and the planning model are pasted, and then the body surface point cloud matrix and the planning model are moved to the virtual operation area to complete the registration.

[0083] It should be noted that, firstly, the present invention performs fixation processing through the surface of the object, and does not need to fix the ilium, spinous process, femur and tibia, etc., and will not cause trauma to the object. Moreover, after the fixation operation, the surface of the object does not produce external force deformation, and the texture of the polymer polyurethane bandage also helps to improve the accuracy of image registration.

[0084] In actual situations, the object's bone image cannot be detected during the object processing process. The present invention uses a 3D dynamic structured light camera to collect the point cloud of the solidified object's body surface image and the environmental point cloud of the solidified object's environment, and then aligns the body surface point cloud and the environmental point cloud with the planning model.

[0085] The purpose of the multiple registration of the present invention is to filter out irrelevant and erroneous point clouds step by step, and at the same time, keep the correct point clouds as much as possible for subsequent registration to ensure the registration accuracy. Since the amount of point cloud data is very large, the present invention can reduce the amount of data processing and speed up the registration by repeatedly deleting irrelevant or deformed point clouds.

[0086] It should be noted that the deformed point cloud is caused by dehydration or edema of the object after solidification. When dehydrated, the body surface shrinks, causing the point cloud to collapse; when edematous, the body surface expands, causing the point cloud to expand.

[0087] The present invention is based on a multi-region image registration method with body surface fixation, and the registration is performed by the following steps:

[0088] S1, fixing the torso of the object before operation to obtain the solidified object; collecting CT data of the solidified object by a spiral CT device, reconstructing the CT data to obtain a three-dimensional reconstruction model, and performing operation planning on the three-dimensional reconstruction model to obtain a planning model;

[0089] S2, collecting the surface point cloud and environment point cloud of the solidified object;

[0090] S3, importing the planning model, the body surface point cloud and the environment point cloud obtained in S2 into the navigation software, specifically, using the capture function to import the body surface point cloud and the scene point cloud obtained in S2 into the navigation software;

[0091] S4, moving the planning model closer to the body surface point cloud and the environment point cloud;

[0092] S5, coarsely aligning the body surface point cloud and the environment point cloud with the planning model to obtain the corresponding RMS value θ1;

[0093] S6, measuring the distance between the body surface point cloud and the environment point cloud and the planning model, marking the point cloud whose distance is greater than α1 in the body surface point cloud and the environment point cloud, and θ1≤α1≤15θ1, and entering S7;

[0094] S7, in the marked body point cloud and environment point cloud, delete the point cloud whose distance from the planning model is outside the range of ±α2, so as to remove the redundant scene point cloud, and obtain the first screening body point cloud, and there exists θ1≤α2≤10θ1;

[0095] S8, performing a first fine registration of the first screened body surface point cloud in S6 with the planning model to obtain a corresponding RMS value θ2, and proceeding to S9;

[0096] S9, measuring the distance between the first screened body surface point cloud and the planning model, deleting the point cloud whose distance from the planning model is outside the range of ±α3 in the first screened body surface point cloud, and α3=θ2, thereby removing the collapsed or expanded deformed point cloud, obtaining the second screened body surface point cloud, and entering S10;

[0097] S10, performing a second fine registration of the second screened-out body surface point cloud obtained in S9 with the planning model to obtain a corresponding RMS value θ3, and proceeding to S11;

[0098] S11, measuring the distance between the second screened body surface point cloud and the planning model, deleting the body surface point cloud whose distance from the planning model is outside the range of ±α4, and α4=θ3, thereby removing the collapsed or expanded deformed point cloud again, obtaining the third screened body surface point cloud, and entering S12;

[0099] S12, performing a third fine registration of the third-removed body surface point cloud obtained in S11 with the planning model, obtaining the registered third-removed body surface point cloud and corresponding coordinates, defining them as a body surface point cloud matrix, and proceeding to S13;

[0100] S13, pasting the body surface point cloud matrix into the planning model, and using an inverse matrix to move the body surface point cloud matrix and the planning model as a whole to a virtual operation area.

[0101] It should be noted that S5 is a coarse alignment, and the range of its RMS value θ1 is relatively large. Therefore, setting α1 and α2 within the above range can retain the correct point cloud as much as possible. S8 belongs to fine alignment. At this time, the distance between the body surface point cloud and the planning model is very small. Therefore, by deleting the point cloud outside the range of ±θ2 in S9, most of the deformed point clouds can be deleted, thereby improving the accuracy of subsequent alignment. Similarly, S10 is the second fine alignment. This time, the distance between the body surface point cloud and the planning model is even smaller. Therefore, by deleting the point cloud outside the range of ±θ3 in S11, most of the deformed point clouds can be deleted, thereby improving the accuracy of subsequent alignment. Therefore, the present invention can improve the accuracy of alignment by multiple step-by-step screening of irrelevant and erroneous point clouds.

[0102] S1 is performed by the following steps:

[0103] S1.1. Fixing the object before operation by a polymer polyurethane bandage to obtain a solidified object before operation;

[0104] S1.2. Collect CT data of a solidified object on a spiral CT device, perform three-dimensional reconstruction on the CT data to obtain a three-dimensional reconstruction model, and perform operation planning on the three-dimensional reconstruction model to obtain the planning model; wherein the planning model is provided with a CT body surface model, a CT bone model and a planning channel, and the three-dimensional reconstruction model is provided with a pre-operation CT body surface model and a pre-operation CT bone model.

[0105] The fixing process of the present invention is carried out by the following steps:

[0106] a1. Place the subject's torso on a flat board before the operation;

[0107] a2. Use a moistened high-molecular polyurethane bandage to cover the body surface and both ends of the joints of the subject, then shape the high-molecular polyurethane bandage and squeeze the gap between the subject and the fixed plate so that the high-molecular polyurethane bandage fits the operation site;

[0108] a3. Spraying physiological saline onto the polymer polyurethane bandage to plasticize the polymer polyurethane bandage;

[0109] a4. Fix the end of the polymer polyurethane bandage on the flat plate by means of staples to obtain a solidified object before operation.

[0110] S2 of the present invention specifically collects the surface point cloud and the environmental point cloud of the solidified object through a 3D dynamic structured light camera. In S4, the planning model of S3 is moved closer to the surface point cloud and the environmental point cloud through a rotation-displacement function. In S5, multiple groups of common registration points are selected from the planning model and the surface point cloud and the environmental point cloud, respectively, specifically 4 to 8 groups, and then the planning model and the surface point cloud are aligned through a coarse registration function, and then the distance between the surface point cloud and the environmental point cloud and the planning model is measured.

[0111] It should be noted that experiments have shown that when the number of registration points exceeds 8 groups, the amount of calculation will be large and the RMS value θ1 will be large; when the number of registration points is less than 3 groups, although the amount of calculation is small, the RMS value θ1 will also be large. When the number of registration points is between 4 and 8 groups, the RMS value θ1 is relatively small to ensure the registration accuracy while reducing the amount of calculation.

[0112] It should be noted that the surface point cloud and environmental point cloud of the solidified object of the present invention, as well as the multiple registration and multiple deletion of irrelevant or deformed point clouds, are built by building an interface in the Cloudcompare software, and the interface is connected to the MPSizectorS SDK. Specifically, the surface point cloud and environmental point cloud of the solidified object are obtained through the 3D dynamic structured light camera of the MPSizectorS SDK, and the alignment, rough registration, precise registration, ranging, copying, and pasting of the planning model and the surface point cloud and environmental point cloud are all performed in the Cloudcompare software. The present invention can also be directly performed in the Xianhe Medical Navigation Puncture Software RCCSV1.0.

[0113] The reason why the multi-localized image registration method of the present invention can have registration accuracy and effectively prevent image drift before operation is:

[0114] 1. The present invention can improve the position consistency between the body surface point cloud and the CT model through image registration collected by a 3D dynamic structured light camera. Because the object needs to undergo a spiral CT scan and needs to be transported and transferred after the CT scan, it is difficult to maintain the position consistency of the object. The present invention uses a 3D dynamic structured light camera to collect point cloud scans of the scene before operation, and performs image registration on the point cloud and the planning model, which can moderately reduce the requirement for consistency between the object's current position and the CT scan position. In addition, the present invention can prevent the object's body surface from large-scale displacement by solidifying the object before operation, and keep the body surface in a fixed position on a fixed base plate. Therefore, the solidification process meets the assumption of object position consistency.

[0115] 2. The joint morphology and surrounding tissues enable the joint to maintain the anatomical posture. In addition to driving the joint movement, the muscles and ligaments attached around the joint also have certain tension characteristics to stabilize and limit the joint movement. In addition, the joint morphology is a simple mosaic structure, which can ensure that the joint maintains a certain direction and angle to form a local joint stable state. Therefore, the present invention fixes the two ends of the joint through a polymer polyurethane bandage to stabilize the joint.

[0116] 3. The solidification treatment of the object before operation can prevent large-scale deformation of the polymer mesh and skin, and keep the bones in the correct direction and position. The premise for the object's body surface to not have image drift is that the pre-operation CT solid surface model, the body surface point cloud and the post-operation CT body surface model do not have large-scale skin deformation. In the navigation system, the pre-operation CT body surface model is image-aligned with the body surface point cloud and the pre-operation CT body surface model. The distance level between the pre-operation CT model and the body surface point cloud after image alignment is observed successively. The distance test data and thermal map of the two are used to determine whether the pre-operation body surface model and the body surface point cloud of the object have large-scale image deformation. In this way, it is determined whether there is image drift. In addition, the body solidification treatment can prevent the object's body surface from external force deformation, and the polymer polyurethane bandage texture also helps to improve the accuracy of image registration.

[0117] The multi-localized image registration method screens out irrelevant and erroneous point clouds step by step so that they do not participate in the registration operation, thereby speeding up the registration speed and avoiding mismatching; and screens out the correct point clouds as much as possible for registration step by step to ensure the registration accuracy.

[0118] Example 2

[0119] A multi-localized image drift detection method based on body surface fixation, such as Figure 2 As shown, firstly, CT data of the solidified object after operation is collected, and the CT data is three-dimensionally reconstructed to obtain a post-operation model, wherein the post-operation model includes a post-operation body surface model and a post-operation bone model;

[0120] Then, the three-dimensional reconstructed model in the multi-region image registration method of Example 1 or Example 2 is multiple-registered with the post-operation model and irrelevant or deformed grids are deleted multiple times to obtain a post-operation body surface grid matrix;

[0121] Finally, the post-operation body surface grid matrix is ​​pasted to the post-operation skeleton model, and the post-operation skeleton model is moved to the pre-operation CT skeleton model in the three-dimensional reconstruction model to obtain a skeleton image drift result.

[0122] Since the bones may be slightly displaced due to external forces and other factors, the bones can elastically retract to their original positions when external forces and other factors are removed according to the elastic characteristics of soft tissue, so that the relative position relationship between the two remains relatively stable during the operation. The present invention uses the object's pre-operation CT body surface model to align the post-operation body surface model, and then performs a matrix operation to restore the pre-operation CT bone model to the corresponding position in the pre-operation CT body surface. Since the current pre-operation CT bone and the post-operation bone are in similar positions, it can be determined whether there is a positional offset (image drift) between the pre-operation bone and the post-operation bone. Finally, the distance level test function is used to detect the degree of bone offset.

[0123] It should be noted that the multi-localized image drift detection method based on body surface fixation of this embodiment is based on the multi-localized image registration method of embodiment 1, and after the solidified object is operated according to the operation plan, the solidified object after operation is obtained, and then CT data is collected for the solidified object after operation, and the post-operation model is obtained to perform multi-localized image drift detection. The present invention performs multiple registrations and multiple deletions of the body surface grid in the post-operation body surface model with the body surface of the three-dimensional reconstructed model, and then copies the post-operation body surface grid matrix after the final registration with the post-operation skeleton model, so that the post-operation skeleton model overlaps with the pre-operation skeleton model, and the skeleton drift is judged by the degree of overlap between the post-operation skeleton model and the pre-operation skeleton model.

[0124] It should also be noted that, since the solidified object causes body surface and bone drift after operation, the bone drift results before and after the operation can be detected through the multi-localized image drift detection method based on body surface fixation of the present invention.

[0125] The deformed mesh of the present invention is caused by dehydration or edema of the object after solidification or operation. When dehydrated, the body surface shrinks, causing the mesh to collapse; when edematous, the body surface expands, causing the mesh to expand.

[0126] The present invention is based on the body surface fixed multiple localized image drift detection method, which is detected by the following steps:

[0127] A1, collecting CT data of the solidified object after the operation, and performing three-dimensional reconstruction on the CT data to obtain the post-operation model;

[0128] A2, importing the post-operation model and the three-dimensional reconstruction model into the navigation software, performing rough registration and alignment between the post-operation model and the pre-operation CT body surface model in the three-dimensional reconstruction model, obtaining the corresponding RMS value δ1, and proceeding to A3;

[0129] A3, measuring the distance between the post-operation body surface model and the pre-operation CT body surface model, marking the grids in the post-operation body surface model that are greater than the distance β1, and if δ1≤β1≤15δ1 exists, proceed to A4;

[0130] A4. In the marked grids of the post-operation body surface model, the grids whose distances from the pre-operation CT bone model are outside the range of ±β2 are deleted to obtain the first screening post-operation body surface model, and δ1≤β2≤10δ1 exists;

[0131] A5, performing a first fine registration of the body surface model after the first screening operation with the CT body surface model before the operation to obtain a corresponding RMS value δ2, and proceeding to A6;

[0132] A6, measuring the distance between the model after the first screening operation and the CT skeleton model before the operation, and deleting the grids of the body surface model after the first screening operation whose distances from the CT skeleton model before the operation are outside the range of ±β3, and β3=δ2 exists, thereby removing the collapsed or expanded deformed grids, and obtaining the body surface model after the second screening operation;

[0133] A7, performing a second fine registration on the body surface model after the second screening operation and the CT bone model before the operation to obtain a corresponding RMS value δ3, and proceeding to A8;

[0134] A8, measuring the distance between the body surface model after the second screening operation and the CT skeleton model before the operation, and deleting the grids whose distances from the CT skeleton model before the operation are outside the range of ±β4 in the grid of the model after the second screening operation, and β4=δ3, thereby removing the collapsed or expanded deformed grids, obtaining the model after the third screening operation, and proceeding to A9;

[0135] A9, performing a third fine registration of the third screened post-operation body surface model obtained in A8 with the pre-operation CT bone model, obtaining the registered third screened post-operation body surface model and the corresponding coordinates defined as a post-operation body surface grid matrix, and proceeding to A10;

[0136] A10. Paste the post-operation body surface grid matrix into the post-operation skeleton model, and use the cis matrix to move the post-operation skeleton model to the pre-operation CT skeleton model to obtain a skeleton image drift result.

[0137] The present invention A2 specifically imports the post-operation model and the three-dimensional reconstructed model into the navigation system software, selects multiple groups of common registration points in the pre-operation CT surface model and the post-operation model respectively, and then aligns the pre-operation CT surface model and the post-operation model through a coarse registration function.

[0138] It should be noted that the bone image drift result of the present invention can be directly observed through the overlapping pictures of two models, and can also be obtained by calculating the measured distance between the bone model after the operation and the CT bone model before the operation.

[0139] The reason why the method for detecting image drift based on multiple localized images fixed on the body surface can accurately detect image drift is as follows:

[0140] 1. The solidification treatment of the object before the operation can prevent the polymer mesh and skin from large-scale deformation, so that the bones are still in the correct direction and position. The premise for the object's body surface to not have image drift is that the pre-operation CT solid surface model, the surface point cloud and the post-operation CT surface model do not have large-scale skin deformation. In the navigation system, the pre-operation and post-operation CT surface models are image registered, and the distance level between the pre-operation CT model and the post-operation model after the image registration is observed successively. The distance test data and thermal map of the two are used to judge whether there is large-scale image deformation in the morphology of the pre-operation surface model and the post-operation surface model of the object. In this way, the image drift after the operation is judged.

[0141] 2. The body surface and bones have anatomical position consistency. Since the bones can be slightly displaced due to external forces and other factors, the bones can elastically retract to their original positions when external forces and other factors are removed according to the elastic characteristics of soft tissue, so that the relative position relationship between the two remains relatively stable during the operation. The pre-operative body surface model of the object is used to align the post-operative body surface model, and then the matrix operation is performed to restore the pre-operation bone model to the corresponding position in the pre-operation body surface. Since the positions of the pre-operation bone and the post-operation bone are similar, it can be determined whether there is a positional offset (image drift) between the pre-operation bone and the post-operation bone; finally, the distance level test function is used to verify the degree of bone offset.

[0142] This multi-localized image drift detection method based on body surface fixation gradually screens out erroneous grids through the post-operation body surface model and the three-dimensional reconstruction model, so that they do not participate in the registration operation, thereby speeding up the registration speed and avoiding mismatching. The correct grids are retained as much as possible through gradual screening, thereby accurately detecting the drift results of the bone image after the operation.

[0143] Example 3

[0144] A multi-region image registration method based on body surface fixation as in Example 1, the object of this example is a pig. The fixation process of the pig is performed by the following steps:

[0145] a1. Place the pig trunk and limbs on a wooden board before operation;

[0146] a2. Fold the polymer polyurethane bandage in half to form a 1 cm folded bandage, then place the polymer polyurethane bandage in water for about 3 seconds, remove it and squeeze out the water to obtain a moistened polymer polyurethane bandage; then cover the moistened polymer polyurethane bandage on both ends of the pig's joints and the body surface, while exposing the surgical field, adjusting the position and range of the surgical field, placing the pig in the placement area in the center of the wooden board, and then pressurizing and fixing the pig with the polymer polyurethane bandage, then shaping the polymer polyurethane bandage and squeezing the gap between the object and the fixed plate, so that the polymer polyurethane bandage fits the operation site;

[0147] a3. Spraying physiological saline onto the polymer polyurethane bandage to plasticize the polymer polyurethane bandage;

[0148] a4. After the polymer polyurethane bandage is completely plasticized, the end of the polymer polyurethane bandage is fixed to the flat plate by means of a staple to obtain a solidified pig before operation. Figure 3 and Figure 4 .

[0149] The multiple localized image registration method of this embodiment is specifically performed by the following steps:

[0150] S1, collecting CT data of a solidified pig through a spiral CT device, reconstructing the CT data to obtain a three-dimensional reconstruction model, and performing operation planning on the three-dimensional reconstruction model to obtain a planning model, such as Figure 5 ;

[0151] S2. Open MPSizectorS SDK and collect the surface point cloud and environment point cloud of the solidified pig, such as Figure 6 , where the scene point cloud includes cured pigs, equipment, other facilities, and operators.

[0152] S3, importing the planning model, the body surface point cloud and the environment point cloud obtained in S2 into the navigation software, wherein the body surface point cloud and the environment point cloud are imported into the navigation software through the capture function;

[0153] S4, the planning model is moved closer to the body surface point cloud and the environment point cloud through the rotation-displacement function, such as Figure 7 ;

[0154] S5, the planning model and the body surface point cloud and the environment point cloud respectively select 7 groups of common registration points, such as A0-A6 in the body surface point cloud and the environment point cloud on the left side of the figure, and R0-R6 in the planning model on the right side, and then align R0-R6 of the planning model with A0-A6 of the body surface point cloud and the environment point cloud through the coarse registration function, as shown in FIG. Figure 8 and Fig. 9 , and the corresponding RMS value θ1 is obtained. The RMS value θ1 is basically between 1mm and 3mm.

[0155] S6. Measure the distance between the body surface point cloud and the planning model, mark the point cloud with a distance greater than α1 in the body surface point cloud and the environment point cloud, and θ1≤α1≤15θ1 exists. In this embodiment, α1 is set to 10 mm, and the color of the point cloud with a distance greater than 10 mm is set to red, such as Fig.10 , enter S7;

[0156] S7. In the marked body surface point cloud and environment point cloud, delete the point cloud whose distance from the planning model is outside the range of ±α2, so as to remove the redundant scene point cloud and obtain the first screening body surface point cloud. In this embodiment, α2 is set to 5mm, and the point cloud outside ±5mm is deleted. Fig.11 , there exists θ1≤α2≤10θ1;

[0157] S8, performing a first fine registration of the first screened body surface point cloud of S6 with the planning model, such as Fig.12 , and obtain the corresponding RMS value θ2, where the RMS value θ2 is 0.78464 mm, and enter S9;

[0158] S9, measuring the distance between the first screening body surface point cloud and the planning model, and deleting the point cloud whose distance to the planning model is outside the range of ±0.78464 in the first screening body surface point cloud, such as Fig.13 , thereby removing the collapsed or expanded deformed point cloud, obtaining the second screened surface point cloud, and entering S10;

[0159] S10, performing a second fine registration of the second screened-out body surface point cloud obtained in S9 with the planning model, such as Fig.14 , and obtain the corresponding RMS value θ3, where the RMS value θ3 is 0.226051 mm, and enter S11;

[0160] S11, measuring the distance between the second screened body surface point cloud and the planning model, in the second screened body surface point cloud, such as Fig.15 , deleting the body surface point cloud whose distance from the planning model is outside the range of ±0.226051, thereby removing the collapsed or expanded deformed point cloud again, obtaining the third screened body surface point cloud, and entering S10;

[0161] S12, performing a third fine registration of the third screened-out body surface point cloud obtained in S11 with the planning model, such as Fig.16 At this time, the RMS value is 0.0700926 mm, and the third screened body surface point cloud and the corresponding coordinates after registration are defined as the body surface point cloud matrix, and enter S13;

[0162] S13, pasting the body surface point cloud matrix into the planning model, and using the inverse matrix to move the body surface point cloud matrix and the planning model as a whole to the virtual operation area, such as Fig.17 .

[0163] The multi-localized image registration method of this embodiment gradually screens out irrelevant and erroneous point clouds, and the registration RMS value drops from the initial 1mm to 3mm to 0.0700926mm at the third fine registration. The lower the RMS value, the higher the registration accuracy, which proves that the multi-localized image registration method of the present invention has the advantage of high registration accuracy.

[0164] Example 4

[0165] After the operation on the immobilized pig in Example 3, the immobilized pig is further tested according to the multi-localized image drift detection method based on body surface fixation of the present invention, and the detection is specifically performed in the following steps:

[0166] A1. Collect the CT data of the cured pig after the operation, and perform 3D reconstruction on the CT data to obtain the model after the operation, such as Fig.18 ;

[0167] A2. Import the post-operation model and the 3D reconstructed model into the navigation system software, select multiple groups of common registration points in the 3D reconstructed model and the post-operation model respectively, and the registration points in this embodiment are 7 groups, such as R0-R6 in the left 3D reconstructed model and A0-A6 in the right post-operation model, and then align the 3D reconstructed model and the post-operation model through the coarse registration function, such as Fig.19 , get the corresponding RMS value δ1, enter A3, the RMS value δ1 is greater than 3mm;

[0168] A3, measuring the distance between the post-operation body surface model and the pre-operation CT body surface model, marking the grids in the post-operation body surface model that are larger than the distance β1, and if δ1≤β1≤15δ1, proceed to A4, specifically, in this embodiment, β1 is set to 10 mm, and the color of the grids that are larger than the distance 10 mm is set to red, such as Fig. 20 ,

[0169] A4. In the marked grids of the post-operation body surface model, the grids whose distances from the pre-operation CT skeleton model are outside the range of ±β2 are deleted. Specifically, β2 in this embodiment is set to 5 mm, and the post-operation body surface model is obtained for the first screening, such as Fig.21 ;

[0170] A5, performing a first fine registration of the body surface model after the first screening operation with the CT bone model before the operation, obtaining a corresponding RMS value δ2, wherein the RMS value δ2 is 0.821775 mm, and proceeding to A6;

[0171] A6. Measuring the distance between the model after the first screening operation and the CT skeleton model before the operation, and deleting the grids of the body surface model after the first screening operation whose distances from the CT skeleton model before the operation are outside the range of ±0.821775, thereby removing the collapsed or expanded deformed grids, and obtaining the body surface model after the second screening operation;

[0172] A7, performing a second fine registration on the body surface model after the second screening operation and the CT bone model before the operation to obtain a corresponding RMS value δ3, wherein the RMS value δ3 is 0.335722 mm, and proceeding to A8;

[0173] A8, measuring the distance between the body surface model after the second screening operation and the CT bone model before the operation, and deleting the grids of the model after the second screening operation whose distances to the CT bone model before the operation are outside the range of ±0.335722 mm, thereby removing the collapsed or expanded deformed grids, obtaining the model after the third screening operation, and proceeding to A9;

[0174] A9, performing a third fine registration of the third screened post-operation body surface model obtained in A8 with the pre-operation CT bone model, wherein the RMS value is 0.179005 mm, obtaining the registered third screened post-operation body surface model and the corresponding coordinates defined as a post-operation body surface grid matrix, and proceeding to A10;

[0175] A10, pasting the post-operation body surface grid matrix into the post-operation skeleton model, and using the cis matrix to move the post-operation skeleton model to the pre-operation CT skeleton model, to obtain a skeleton image drift result, such as Fig. 22 .

[0176] Through Fig. 22 The distance between the spine of the skeleton model after the operation and the spine of the skeleton model before the operation was calculated, and the skeleton image drift was found to be 1.10661mm.

[0177] This multi-localized image drift detection method based on body surface fixation gradually screens out erroneous grids through the post-operation body surface model and the three-dimensional reconstruction model, so that they do not participate in the registration operation, thereby speeding up the registration speed and avoiding mismatching. The correct grids are retained as much as possible through gradual screening, thereby accurately detecting the drift results of the bone image after the operation.

[0178] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the essence and scope of the technical solution of the present invention.

Claims

1. A multi-region image registration method based on body surface fixation, characterized in that: Firstly, the surface of the object before operation is fixed to obtain a solidified object, then a three-dimensional reconstruction model of the solidified object is obtained, and operation planning is performed on the three-dimensional reconstruction model to obtain a planning model; Then, the body surface point cloud and the environment point cloud of the solidified object are collected, and the body surface point cloud and the environment point cloud are multiple-registered with the planning model and irrelevant or deformed point clouds are deleted multiple times to obtain a body surface point cloud matrix; Finally, the body surface point cloud matrix and the planning model are pasted, and then the body surface point cloud matrix and the planning model are moved to the virtual operation area to complete the registration; The registration is performed by the following steps: S1, fixing the torso of the object before operation to obtain the solidified object; collecting CT data of the solidified object by a spiral CT device, reconstructing the CT data to obtain a three-dimensional reconstruction model, and performing operation planning on the three-dimensional reconstruction model to obtain a planning model; S2, collecting the surface point cloud and environment point cloud of the solidified object; S3, importing the planning model, the body surface point cloud and the environment point cloud obtained in S2 into the navigation software; S4, moving the planning model closer to the body surface point cloud and the environment point cloud; S5, coarsely aligning the body surface point cloud and the environment point cloud with the planning model to obtain the corresponding RMS value θ1; S6, measuring the distance between the body surface point cloud and the environment point cloud and the planning model, marking the point cloud with a distance greater than α1 in the body surface point cloud and the environment point cloud, and there is θ1≤α1≤15θ1, and entering S7; S7, in the marked body point cloud and environment point cloud, delete the point cloud whose distance from the planning model is outside the range of ±α2, so as to remove the redundant scene point cloud, and obtain the first screening body point cloud, and there exists θ1≤α2≤10θ1; S8, performing a first fine registration of the first screened body surface point cloud in S6 with the planning model to obtain a corresponding RMS value θ2, and proceeding to S9; S9, measuring the distance between the first screened body surface point cloud and the planning model, deleting the point cloud whose distance from the planning model is outside the range of ±α3 in the first screened body surface point cloud, and α3=θ2, thereby removing the collapsed or expanded deformed point cloud, obtaining the second screened body surface point cloud, and entering S10; S10, performing a second fine registration of the second screened-out body surface point cloud obtained in S9 with the planning model to obtain a corresponding RMS value θ3, and proceeding to S11; S11, measuring the distance between the second screened-out body surface point cloud and the planning model, deleting the body surface point cloud whose distance from the planning model is outside the range of ±α4, and α4=θ3, thereby removing the collapsed or expanded deformed point cloud again, obtaining the third screened-out body surface point cloud, and entering S12; S12, performing a third fine registration of the third-removed body surface point cloud obtained in S11 with the planning model, obtaining the registered third-removed body surface point cloud and corresponding coordinates, defining them as a body surface point cloud matrix, and proceeding to S13; S13, pasting the body surface point cloud matrix into the planning model, and using an inverse matrix to move the body surface point cloud matrix and the planning model as a whole to a virtual operation area.

2. The multi-region image registration method based on body surface fixation according to claim 1, characterized in that: The S1 is performed by the following steps: S1.

1. Fixing the object before operation by a polymer polyurethane bandage to obtain a solidified object before operation; S1.

2. Collect CT data of a solidified object on a spiral CT device, perform three-dimensional reconstruction on the CT data to obtain a three-dimensional reconstruction model, and perform operation planning on the three-dimensional reconstruction model to obtain the planning model; wherein the planning model is provided with a CT body surface model, a CT bone model and a planning channel, and the three-dimensional reconstruction model is provided with a pre-operation CT body surface model and a pre-operation CT bone model.

3. The multi-region image registration method based on body surface fixation according to claim 2, characterized in that: The fixing process is performed by the following steps: a1. Place the subject's torso on a flat board before the operation; a2. Use a moistened high-molecular polyurethane bandage to cover the body surface and both ends of the joints of the subject, then shape the high-molecular polyurethane bandage and squeeze the gap between the subject and the fixed plate so that the high-molecular polyurethane bandage fits the operation site; a3. Spraying physiological saline onto the polymer polyurethane bandage to plasticize the polymer polyurethane bandage; a4. Fix the end of the polymer polyurethane bandage on the flat plate by means of staples to obtain a solidified object before operation.

4. The multi-region image registration method based on body surface fixation according to claim 3, characterized in that: The S2 specifically collects the surface point cloud and the environment point cloud of the solidified object through a 3D dynamic structured light camera.

5. The multi-region image registration method based on body surface fixation according to claim 4, characterized in that: In S4, the planning model of S3 is moved closer to the body surface point cloud and the environment point cloud through a rotation-displacement function.

6. The multi-region image registration method based on body surface fixation according to claim 5, characterized in that: In S5, a plurality of groups of common registration points are selected from the planning model, the body surface point cloud and the environment point cloud respectively, and then the planning model and the body surface point cloud are aligned through a coarse registration function.

7. A method for detecting drift of multiple localized images based on body surface fixation, characterized in that: Firstly, CT data of the solidified object after operation is collected, and the CT data is three-dimensionally reconstructed to obtain a post-operation model, wherein the post-operation model includes a post-operation body surface model and a post-operation bone model; Then, the three-dimensional reconstructed model in the multi-regional image registration method according to any one of claims 1 to 6 is multiple-registered with the post-operation model and irrelevant or deformed grids are deleted multiple times to obtain a post-operation body surface grid matrix; Finally, the post-operation body surface grid matrix is ​​pasted to the post-operation skeleton model, and the post-operation skeleton model is moved to the pre-operation CT skeleton model in the three-dimensional reconstruction model to obtain a skeleton image drift result; The detection is carried out by the following steps: A1, collecting CT data of the solidified object after the operation, and performing three-dimensional reconstruction on the CT data to obtain the post-operation model; A2, importing the post-operation model and the three-dimensional reconstruction model into the navigation software, performing rough registration and alignment between the post-operation model and the pre-operation CT body surface model in the three-dimensional reconstruction model, obtaining the corresponding RMS value δ1, and proceeding to A3; A3, measuring the distance between the post-operation body surface model and the pre-operation CT body surface model, marking the grids in the post-operation body surface model that are greater than the distance β1, and if δ1≤β1≤15δ1 exists, proceed to A4; A4. In the marked grids of the post-operation body surface model, the grids whose distances from the pre-operation CT body surface model are outside the range of ±β2 are deleted to obtain the first screening post-operation body surface model, and δ1≤β2≤10δ1 exists; A5, performing a first fine registration of the body surface model after the first screening operation with the CT body surface model before the operation to obtain a corresponding RMS value δ2, and proceeding to A6; A6. measuring the distance between the body surface model after the first screening operation and the CT body surface model before the operation, and deleting the grids of the body surface model after the first screening operation whose distances from the CT body surface model before the operation are outside the range of ±β3, and β3=δ2 exists, thereby removing the collapsed or expanded deformed grids, and obtaining the body surface model after the second screening operation; A7, performing a second fine registration on the body surface model after the second screening operation and the CT body surface model before the operation to obtain a corresponding RMS value δ3, and proceeding to A8; A8, measuring the distance between the body surface model after the second screening operation and the CT body surface model before the operation, and deleting the grids of the body surface model after the second screening operation whose distances to the CT body surface model before the operation are outside the range of ±β4, and β4=δ3 exists, thereby removing the collapsed or expanded deformed grids, obtaining the model after the third screening operation, and proceeding to A9; A9, performing a third fine registration of the third screened post-operation body surface model obtained in A8 with the pre-operation CT body surface model, obtaining the registered third screened post-operation body surface model and the corresponding coordinates defined as a post-operation body surface grid matrix, and proceeding to A10; A10. Paste the post-operation body surface grid matrix into the post-operation skeleton model, and use the cis matrix to move the post-operation skeleton model to the pre-operation CT skeleton model to obtain a skeleton image drift result.

8. The method for detecting drift of multiple localized images based on body surface fixation according to claim 7, characterized in that: The A2 specifically imports the post-operation model and the three-dimensional reconstructed model into the navigation system software, selects a plurality of groups of common registration points in the pre-operation CT body surface model and the post-operation model respectively, and then aligns the pre-operation CT body surface model and the post-operation model through a coarse registration function.

Images