Bone registration method and system, surgical robot system, and storage medium

Abstract

A bone registration method, system, surgical robot system and storage medium for a surgical navigation system are provided, including: obtaining a source point set SPS using preoperative bone image data; generating a target point set TPS using point data collected from the actual bone surface; selecting 4 pairs of points from SPS and TPS for initial registration to obtain a transformation T1; using the transformation T1 and SPS to convert TPS into a three-dimensional model space that overlaps with SPS, and iteratively modifying the point L collected from the actual bone surface i , improving the T1 transformation error, and obtaining an accurate transformation matrix and registration error. The iterative modification is carried out by randomly selecting 4 points within a sphere centered at this point L i to obtain the transformation T1 between each random point and the corresponding point M i on the 3D virtual model and the new T1 transformation error. If the error is improved, the new point L i obtained based on these 4 points is used as the new sphere center to obtain new random points and repeat this process.

Description

Technical Field

[0001] The present invention relates to a method for improving the accuracy of bone registration, and more particularly, to a bone registration method and system, a storage medium, and a surgical robot system in orthopedic surgery. Background Art

[0002] Computer-assisted surgical navigation systems are widely used in various surgical operations, aiming to improve the accuracy and precision of surgeries. As a technology existing in various fields of medicine, navigation surgery requires an optical tracking system (such as a locator) to perform preoperative and intraoperative registration (also known as alignment) to achieve the tracking of surgical conditions.

[0003] Among them, registration is a very important process. Through it, the transformation between the preoperative plane coordinate system and the operating room (intraoperative) coordinate system can be found in the software, so that the position and orientation of the surgical instrument relative to the three-dimensional model of the intraoperative area can be visualized. The quality of the registration method will affect the surgical navigation accuracy in real time.

[0004] In the marker point registration method currently adopted by surgical navigation systems, the marker points include: bone implant screw marker points, anatomical landmark points, and marker points pasted on the skin surface. In addition, in current surgeries based on orthopedic surgical robot systems, a contact sampling method using a handheld probe is mostly used to collect and measure the position information of the bone surface for registration with the preoperative images of the patient. Although the marker point registration method is faster in alignment, however, due to factors such as the additional trauma and pain caused to the patient, human operation errors, recognition difficulty, light occlusion, marker point dropping, and the need for additional disinfection, in this method, it is generally inclined to limit the number of points to be collected as much as possible, or the system needs to guide the surgeon through display to collect the points for alignment. This not only easily affects the accuracy but also the operability. Summary of the Invention

[0005] In view of the above problems, the object of the present invention is to propose a new algorithm that allows for easy and accurate registration and alignment between the preoperative planning coordinate system in the software and the real bone coordinate system of the patient.

[0006] Specifically, according to one aspect of the present invention, there is provided a bone registration method for determining the transformation relationship between a preoperative coordinate system and an intraoperative coordinate system, which is characterized by including the following steps: using the preoperative bone image data, establishing a preoperative coordinate system of a 3D virtual model, and obtaining a source point set SPS from the surface of the 3D virtual model; using the point data collected from the actual bone surface during the operation, establishing an intraoperative coordinate system, and generating a target point set TPS; selecting 4 groups of corresponding point pairs from the source point set SPS and the target point set TPS to perform an initial registration between the source point set SPS and the target point set TPS, and obtaining a first transformation T1; using the first transformation T1 and the source point set SPS, converting the target point set TPS into a three-dimensional model space that overlaps with the three-dimensional model space of the source point set SPS, and iteratively modifying the point L i collected from the actual bone surface to improve the first transformation error until the error is less than a preset threshold or the maximum number of iterations is reached, thereby obtaining an accurate transformation matrix and a registration error. Among them, for the iterative modification of the point L i , it is carried out by randomly selecting 4 points within a sphere centered on this point L i , and obtaining the first transformation T1 and a new first transformation error between each random point and the corresponding point M i on the 3D virtual model, where i is a natural number.

[0007] Preferably, if the first transformation error does not meet the improvement condition, then select 4 more random points within the sphere, recalculate the first transformation T1 and its error until the error decreases, update the center of the sphere, and repeat the iterative modification process; if the first transformation error meets the improvement condition, use another new point L i+1 as the center of the new sphere to obtain new random points and repeat the iterative modification process.

[0008] Preferably, the radius of the sphere is 2 mm.

[0009] Preferably, the natural number i = 1, 2,... 30.

[0010] Preferably, if the above registration process reaches the maximum number of iterations and does not converge to the established error threshold, then use the generated first transformation T1 as the initial transformation of the rigid registration algorithm.

[0011] Preferably, let the input be: source point set SPS, target point set TPS, four points PM n selected from the 3D virtual model, four points PB n selected from the actual femur, the maximum number of iterations m, the general error threshold G e , and let the target output of the algorithm be: registration error R e , the second transformation T2. This method further includes the following steps: setting an initial error E; based on PBn Find the corresponding sphere center C for the point set n ; Start a loop for i = 1, 2, …, m natural numbers: Among them, based on PB n point set, PM n point set to obtain the transformation matrix T. According to the point sets SPS, TPS and the transformation matrix T, the registration error E is obtained i , when the error E i is less than the initial error E, perform an assignment operation for error convergence: E = E i , T1 = T, (Equation 1). Thus, according to the value of the initial error E, reduce the number of points in the point set SPS, and obtain a new corresponding point set PB within the range of the sphere center C n n , when the initial error E is less than the general error threshold G e , interrupt the loop. By aligning the virtual point set and the physical point set, obtain the first transformation T1. Thus, according to the source point set SPS, the target point set TPS and the first transformation T1, use ICP iteration to obtain the registration error R e and the second transformation T2.

[0012] According to another aspect of the present invention, there is provided a surgical robot system, including a surgical navigation system having an optical tracking system and a computer. By the computer executing the steps of the above-mentioned bone registration method, register the intraoperative coordinate system with the preoperative coordinate system, and guide or control the surgical robot to perform surgery based on the coordinates of the actual bone.

[0013] According to one aspect of the present invention, there is provided a storage medium, which is a computer-readable storage medium storing a computer program, characterized in that the computer program is executed to implement the steps of the above-mentioned bone registration method.

[0014] According to one aspect of the present invention, there is provided a bone registration system for intraoperative navigation, used to determine the transformation relationship between the preoperative coordinate system and the intraoperative coordinate system, characterized in that it includes: a source point set acquisition unit, which uses preoperative bone image data to establish a preoperative coordinate system of a 3D virtual model, and obtains a source point set SPS from the surface of the 3D virtual model; a target point set generation unit, which uses point data collected from the surface of the actual bone during surgery to establish an intraoperative coordinate system and generate a target point set TPS; an initial registration unit, which performs initial registration between the source point set SPS and the target point set TPS by selecting 4 groups of corresponding point pairs from the source point set SPS and the target point set TPS, and obtains the first transformation T1; a precise registration unit, which uses the first transformation T1 and the source point set SPS to convert the target point set TPS into a three-dimensional model space that overlaps with the three-dimensional model space of the source point set SPS, so as to obtain a precise transformation matrix and a registration error. The initial registration unit further includes an iterative modification unit, which, by the point L collected from the surface of the actual bone​i Iterative modification is performed to improve the first transformation error until the error is less than a pre-set threshold or the maximum number of iterations is reached. For the point L i the iterative modification is carried out by randomly selecting 4 points within a sphere centered at this point L i and calculating the first transformation T1 between each random point and the corresponding point M on the 3D virtual model i as well as the new T1 transformation error. Here, i is a natural number.

[0015] Preferably, if the first transformation error does not meet the improvement condition, 4 more random points are selected within the sphere, the first transformation T1 and its error are recalculated until the error decreases, the center of the sphere is updated, and the iterative modification process is repeated; if the first transformation error meets the improvement condition, another new point L i+1 is used as the center of the new sphere to obtain new random points and the iterative modification process is repeated.

[0016] According to the present invention, the number of points to be collected can be further restricted. Restricting the number of points taken from the traditional 40 to 30 can ensure the registration accuracy, and it is not necessary for the system to guide the surgeon through display to collect the points for registration. Thus, not only the accuracy is improved, but also the operability is enhanced. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Exemplarily, the iterative process of updating points is shown in a two-dimensional diagram.

[0018] Figure 2 Exemplarily, the reduction of points during the iterative registration process and the first registration process is shown.

[0019] Figure 3 The flowchart of the registration algorithm according to an embodiment is shown. DETAILED DESCRIPTION

[0020] The exemplary embodiments of the present invention will be described in detail below with reference to the drawings. The exemplary embodiments described below and shown in the drawings are intended to teach the principles of the present invention so that those skilled in the art can implement and use the present invention in several different environments and for several different applications. Therefore, the protection scope of the present invention is defined by the appended claims, and the exemplary embodiments are not intended to, and should not be considered as, a restrictive description of the protection scope of the present invention.

[0021] <Overview>

[0022] Through research, the present inventor has proposed a new method for bone registration in, for example, knee joint transplantation surgery.

[0023] According to this new registration algorithm, it is no longer necessary for the system to specify that the surgeon collects points on the physical bone (also known as the entity, actual bone) that strictly correspond to those on the three-dimensional bone model. Among them, the estimated T1 of the first transformation is used to align the correspondence through singular value decomposition SVD.

[0024] Initially, for example, 4 points M previously selected on the surface of the three-dimensional bone model i and 4 corresponding points L on the patient's leg i are used to calculate the transformation T1. This transformation will be iteratively improved later to reduce its error. Using the transformation T1 and the point set of the three-dimensional bone model, the bone surface TPS is converted into the three-dimensional model space, and the transformation error is calculated point-to-point between the remaining, for example, 30 points selected on the bone surface.

[0025] In this algorithm, the points obtained on the three-dimensional virtual model of the bone are regarded as exact points, rather than the corresponding points on the patient's leg (actual bone) that may have errors. This error range is caused by not selecting points exactly the same as those on the three-dimensional virtual model, and it mainly increases at three important points, namely the hip joint center and the external and internal malleoli. These points are the most difficult to locate corresponding points in the three-dimensional virtual model because when rotating the leg, the hip center has to be determined indirectly through certain algorithms, and the ankle joints are determined above the skin covering them.

[0026] To improve the T1 transformation error, the L i points (preferably 30 points in this embodiment) will be iteratively modified until the error is less than a pre-set threshold or the maximum number of iterations is reached.

[0027] Each modification of the L i points is carried out by randomly selecting 4 points within a sphere with a radius of, for example, 2 mm and a center at L i .

[0028] When these 4 points are modified, T1 will be recalculated using Mi and the new Li points, and the new transformation error will also be calculated. If the error is improved (corresponding to meeting the improvement condition), the new Li points will be used as the center of a new sphere with a radius of, for example, 2 mm to obtain new random points and repeat the process.

[0029] If the error is not improved (corresponding to not meeting the improvement condition), 4 more random points will be selected within the same sphere, T1 and its error will be recalculated until the error decreases, the center of the sphere will be updated, and this process will be repeated.

[0030] Since the SPS set is much larger than the TPS set, for each iteration, if the error decreases sufficiently, the number of points in the SPS can be reduced. The reduction strategy is based on selecting the nearest point to each TPS point within a distance threshold greater than the T1 error.

[0031] <Example>

[0032] In intraoperative navigation, surgical instruments tracked by, for example, a locator need to be displayed in real time on a three-dimensional anatomical structure reconstructed preoperatively. Therefore, it is necessary to register the patient with the three-dimensional anatomical structure reconstructed in the image space.

[0033] By calculating the transformation relationship between the coordinate system of the positioning system (intraoperative coordinate system) and the three-dimensional medical image coordinate system (preoperative coordinate system) preoperatively, the actual intraoperative position of the patient and the three-dimensional anatomical structure are accurately registered, so that the three-dimensional model seen by the doctor on the display device can truly reflect the distance and positional relationship of the surgical instrument relative to the target bone, such as the lesion.

[0034] The method and device for establishing the preoperative coordinate system and the intraoperative coordinate system are not the focus of the present invention and will not be elaborated here. Existing means can be used to achieve this.

[0035] Next, in combination with Figures 1-3 , the registration algorithm for realizing the registration between two point sets according to the present invention will be described in detail by way of examples as follows.

[0036] Specifically, the registration method described in this embodiment includes the following steps:

[0037] <Step S201: Output of point sets>

[0038] In the registration method, on the one hand, the three-dimensional medical image of the patient is input preoperatively, and the surface contour point set data (corresponding to the source point set) is output; on the other hand, for example, in the operating room (intraoperatively), the patient surface point set information (corresponding to the target point set) is output.

[0039] That is, the image data obtained by photographing the patient preoperatively, such as CT, MRI, X-ray, etc., is used as the anatomical structure input data to establish a three-dimensional (3D) virtual model and convert it into a point set, that is, the source point set SPS (Source Point Set), and during the operation, the surface point set data of the bone of interest is obtained through an appropriate device such as a probe, that is, the target point set TPS (Target Point Set).

[0040] Among them, SPS is the point obtained from the surface of the bone 3D virtual model, which represents the area of concern during the bone surgery process. TPS is the point obtained from the actual bone surface of the patient. For example, a total of 30 points can be extracted from the femoral surface, or a total of 30 points can be extracted from the tibial surface.

[0041] Preferably, TPS should be collected sparsely to cover as much bone area as possible.

[0042] <Step S202: Registration of point sets>

[0043] Common point set registration algorithms are generally divided into initial registration (the first stage) and precise registration (the second stage). Initial registration makes an approximation between two point sets, and precise registration finds the best possible alignment between two point sets.

[0044] <Step S2021: Initial Registration>

[0045] To perform initial registration, 4 groups of corresponding point pairs are selected from two point sets. For example, 4 points are taken from the 3D virtual model of the femur and 4 points are taken from the corresponding actual femur. It is also possible to take 4 points from the 3D virtual model of the tibia and 4 points from the corresponding actual femur. Here, 4 points are sufficient to support the rigid transformation of the three-dimensional space required for initial registration.

[0046] The anatomical points collected on the femur are, for example: the center of the hip joint, the center of the femoral knee joint, the lateral epicondyle, and the medial epicondyle. The anatomical points collected on the tibia are, for example: the center of the tibial knee joint, the tibial tuberosity, the lateral malleolus, and the medial malleolus. These anatomical points can be easily and precisely located by a surgeon, for example, using a probe.

[0047] Different from common surgical navigation systems such as MAKO, etc., this algorithm does not require the system to specify that the surgeon collects the corresponding points on the physical bone and on the three-dimensional bone model.

[0048] Specifically, between the 4 previously selected points M on the surface of the three-dimensional bone model i and the 4 corresponding points L on the patient's leg i an initial transformation T1 is calculated. Here, i = 1, 2, 3, 4.

[0049] In this way, in this embodiment, a relatively rough registration can be performed when the transformation between the two point sets SPS and TPS is completely unknown, so that the two point sets SPS and TPS are basically aligned in space to provide a better initial transformation value for precise registration.

[0050] Thus, the bone surface TPS can also be converted into the three-dimensional model space using the initial transformation T1 and the point set SPS of the three-dimensional bone model.

[0051] That is, as Figure 2 illustrated, on the three-dimensional model space of the point set SPS represented by each white area (or white patch), the three-dimensional model space of the point set TPS represented by each small gray dot is overlapped and shown.

[0052] Next, through precise registration, this transformation T1 will be iteratively improved to reduce its error. Among them, the transformation error will be calculated point by point for the remaining multiple points, for example, 30 points, selected on the bone surface.

[0053] <Step S2022: Precise Registration>

[0054] To further improve the overall registration accuracy and reduce errors, precise registration is required.

[0055] As mentioned above, since the premise assumption is that the points obtained on the 3D virtual model of the bone are regarded as accurate points, while the points L obtained on the patient's leg with a probe or the like i may have errors, which may cause errors in the T1 transformation.

[0056] Therefore, to improve the T1 transformation error, the following algorithm will be used to iteratively modify the L i points until the error is less than a pre-set threshold or the maximum number of iterations is reached.

[0057] Here, i = 5, 6, 7, 8…, 34, L i i.e., corresponding to the remaining 30 points selected on the bone surface except for L1 to L4.

[0058] Among them, for the modification of the point L i is carried out by randomly selecting 4 points within a sphere with a radius of 2 mm and a center of L i . Here, the radius is not limited to 2 mm and can be appropriately selected according to the accuracy requirements.

[0059] As Figure 1 shown, first, within a sphere with a center of the point L i and a radius of 2 mm, randomly select 4 points Ra, Rb, Rc, Rd, and perform the transformation T1 between each random point Ra, Rb, Rc, Rd and the corresponding point M i (i = 5, 6, 7, 8…, 34) on the surface of the 3D virtual model.

[0060] Correspondingly, the new T1 transformation error (refer to Figure 3 the registration error E in Step S6 i ) will also be calculated.

[0061] If the error is improved, the new L i point (as Figure 1 shown L i+1 ) will be used as the center of a new sphere with a radius of 2 mm to obtain new random points and repeat the process.

[0062] If the error is further improved, the new L i point (as Figure 1 shown L i+2 ) will be used as the center of a sub-new sphere with a radius of 2 mm to obtain sub-new random points and repeat the process.

[0063] If the error is not improved, four more random points will be selected within the same sphere (for example, at least one point is different from Figure 1 the points Ra, Rb, Rc, Rd shown in), and T1 and its error will be recalculated until the error decreases. The center of the sphere will be updated and this process will be repeated.

[0064] Here, the so-called "new L i point" (or "sub-new") is another point reselected when the error improvement effect can be obtained based on the four random points Ra, Rb, Rc, Rd. As Figure 1 shown, after iterative modification of the L i point, iterative modification will be performed successively for other L i+1 points, L i+2 points, etc. However, Figure 1 the L i point, L i+1 point, L i+2 point, Ra, Rb, Rc, Rd points are all illustratively shown and do not strictly correspond to or limit their selection order and orientation. For example, although it is shown that the L i+1 point is within the sphere with the center at the L i point, and the L i+2 point is outside the sphere with the center at the L i+1 point, the relative positions of the L i point, L i+1 point and the subsequent L i+2 points are not necessarily as shown in the figure.

[0065] Since the SPS set is much larger than the TPS set (as the area of the white region in Figure 2 is significantly larger than the area of each small dot (corresponding to the TPS point set)), therefore, for each iteration, if the error decreases sufficiently, the area of the SPS (see Figure 2 a) of, Figure 2 b) of) or the number of points (see Figure 2 c) of, Figure 2 d) of) can be reduced or decreased.

[0066] For example, in the cases where the number of iterations shown successively in Figure 2 a), b), c), d) are 1, 2, 20, 50 times respectively, the area of the white region (corresponding to the SPS point set) around each small dot (corresponding to the TPS point set) gradually shrinks. Here, the shrinking strategy is based on selecting the nearest point of each TPS point within the distance threshold range greater than the T1 transformation error.

[0067] Thus, the best possible alignment method is gradually achieved between the SPS and TPS point sets, reflecting the gradual precise registration of the patient's actual intraoperative position and the preoperative three-dimensional anatomical structure.

[0068] If this first registration process reaches the maximum number of iterations and does not converge to the established error threshold, the resulting transformation T1 will be used as the initial transformation for a rigid registration algorithm such as the Iterative Closest Point (ICP) or some of its variants.

[0069] The matrix transformation T2 output by this algorithm will be the sought-after transformation.

[0070] <Registration algorithm>

[0071] Regarding the above registration algorithm, refer to Figure 3 for a more detailed description of the flowchart.

[0072] Let the inputs to the algorithm be: the source point set SPS, the target point set TPS, four points PM selected from the 3D virtual model n , four points PB selected from the actual femur n , the maximum number of iterations m, the general error threshold G e , where both n and m are natural numbers.

[0073] In addition, let the target outputs of the algorithm be: the registration error R e , the matrix transformation T2 (corresponding to the second transformation or the second transformation matrix).

[0074] When starting the algorithm (step S1), first set the initial registration error E (step S2) (also known as the initial error).

[0075] Then, based on the PB n point set, find the corresponding center of the sphere C n (step S3).

[0076] For example Figure 1 as shown, first find the center of the sphere L i point, then find the center of the sphere L i+1 point when the error is improved, and then the subsequent L i+2 point.

[0077] Subsequently, start the loop for i = 1, 2,..., m times (step S4).

[0078] In this loop, based on the PB n point set and the PM n point set, obtain the transformation matrix T (step S5). According to the point sets SPS, TPS, and the transformation matrix T, use the iterative ICP(SPS, TPS, T) to obtain the registration error E i (step S6).

[0079] Subsequently, the process proceeds to step S7 to determine the error E iIs it less than the initial error E? When the judgment result is "yes", the process proceeds to step S8 for the assignment operation for error convergence:

[0080] E = E i , T1 = T, (Equation 1).

[0081] Thus, according to the value of the error E, the number of points in the point set SPS can be reduced (step S9). And within the sphere center C n (L i points, L i+1 points and subsequent L i+2 points, etc.), a new corresponding point set PB n (step S10) is obtained.

[0082] Subsequently, or when the judgment result of step S7 is "no", the process proceeds to step S11 to determine whether the initial error E is less than the general error threshold G e。 When the judgment result is "no", the process returns to step S4 for the next cycle; when the judgment result is "yes", the loop is interrupted, and the process proceeds to step S12. By aligning the virtual point set and the physical point set, the general error threshold G e is returned and the transformation matrix T1 (corresponding to the first transformation or the first transformation matrix) is obtained.

[0083] Thus, based on the source point set SPS, the target point set TPS, and the transformation matrix T1, the registration error R e and the transformation matrix T2 can be obtained by ICP iteration (step S13). Then the alignment algorithm ends (step S14).

[0084] <Specific technical effects>

[0085] In this way, in this embodiment, first, a relatively rough registration can be performed when the transformation between the two point sets SPS and TPS is completely unknown, so that the distances between the two point sets SPS and TPS are basically aligned in space, providing a better initial transformation value for fine registration.

[0086] Then, based on the result of the rough registration, using a preset iterative algorithm, the intraoperative point set data is accurately registered with the virtual point set data based on the preoperative medical images. For the second registration stage, the sought transformation T2 can be decomposed into rotation and translation. Because at this stage, since the point sets have been made relatively close by T1.

[0087] Thus, this algorithm allows registration between the preoperative planning coordinate system in the software and the patient's real bone coordinate system.

[0088] Preferably, this algorithm is designed for computer - assisted knee transplantation surgery.

[0089] As described above, by introducing random points based on the center of the sphere (point L i ), compared with the traditional case of selecting 40 fixed points (point L i ), only about 30 fixed points (point L i ) need to be selected to ensure higher accuracy through the loop approximation algorithm.

[0090] Moreover, as random points within the sphere, they can be randomly selected within a certain range based on the four selected points PM n , PB n during the iterative process.

[0091] <Computer device>

[0092] According to another aspect of the present application, a computer device is provided, which includes a processor and a memory. The memory stores a computer program. When the computer program is executed by the processor, the processor is caused to execute the steps of the above method. This also includes the case involving a remote computer. For example, the remote computer can be connected to the user computer through any type of network - including a local area network or a wide area network - or can be connected to an external computer.

[0093] <Storage medium>

[0094] According to another aspect of the present application, a non - volatile computer - readable storage medium is provided, on which computer - readable instructions are stored. When the instructions are executed by a processor, the processor is caused to execute the steps of the above method. For example, it can be any tangible medium that contains or stores a program, and the program can be used by a computer device or a surgical robot system.

[0095] <Computer software program>

[0096] According to this embodiment, the process described above with reference to the flowchart can be implemented as a computer software program. Each block in the flowchart can represent a module, a program segment, or a part of the code. For example, it can include a computer program carried on a computer - readable medium, and the computer program contains program code for executing the method shown in the flowchart. The program code can be executed entirely on the user computer, partially on the user computer, executed as an independent software package, partially on a remote computer, or entirely on a remote computer or server.

[0097] When the computer program is executed by a processor such as a central processing unit, the above - defined functions in the above method are executed.

[0098] <Bone registration system>

[0099] There is also provided a bone registration system for an intraoperative navigation system, including: devices such as a probe for obtaining intraoperative bone surface point set data of a patient; a computer for image input and point set output, connected to the probe and other devices, capable of performing corresponding data operations and processing; a display device for displaying the registered three-dimensional image, connected to the computer. The display device includes but is not limited to a monitor.

[0100] For example, the probe can also be replaced with a three-dimensional scanner for obtaining intraoperative bone surface point set data of a patient.

[0101] <Surgical robot system>

[0102] According to the surgical robot system of the present invention, it includes a surgical navigation system having an optical tracking system (such as a locator) and a computer. By the computer executing the steps of the above-mentioned bone registration method, the intraoperative coordinate system is registered with the preoperative coordinate system, and the surgical robot is guided or controlled to perform surgery with the coordinates of the actual bone.

[0103] The anatomical structures discussed above are not limited to the above-mentioned femur or tibia, and can also be other bones or multiple bones, or even the whole or part of the anatomical structure outside of them.

[0104] Although the present invention has been described with reference to various specific embodiments, it should be understood that modifications can be made within the spirit and scope of the described inventive concept. Therefore, it is intended that the present invention is not limited to the described embodiments, but will have the full scope defined by the language of the appended claims.

Claims

1. A bone registration method for determining the transformation relationship between a preoperative coordinate system and an intraoperative coordinate system, characterized in that It includes the following steps: Using the preoperative bone image data, establish a preoperative coordinate system for the 3D virtual model, and obtain the source point set SPS from the surface of the 3D virtual model; Using the point data collected from the actual bone surface during the operation, establish an intraoperative coordinate system, and generate the target point set TPS; Select 4 groups of corresponding point pairs from the source point set SPS and the target point set TPS, perform an initial registration between the source point set SPS and the target point set TPS, and obtain the first transformation T1; Use the first transformation T1 and the source point set SPS to convert the target point set TPS into a three-dimensional model space that overlaps with the three-dimensional model space of the source point set SPS; Among them, by iteratively modifying the points L collected from the actual bone surface i to improve the first transformation error until the error is less than a preset threshold or reaches the maximum number of iterations, so as to obtain the accurate transformation matrix and the registration error Among them, for the iterative modification of the point L i is performed by randomly selecting 4 points within a sphere centered at this point L i to obtain the first transformation T1 and the new first transformation error between each random point and the corresponding point M i on the 3D virtual model, where i is a natural number If the first transformation error does not meet the improvement condition, then select 4 random points within the sphere, recalculate the first transformation T1 and the first transformation error until the error decreases, update the center of the sphere, and repeat the iterative modification process; If the first transformation error satisfies the improvement condition, another new point L i+1 is used as the center of the new sphere to obtain a new random point and repeat the iterative modification process, If the maximum number of iterations is reached and no convergence to the established error threshold occurs, then use the generated first transformation T1 as the initial transformation of the rigid registration algorithm; Let the input be: Source Point Set SPS, Target Point Set TPS, four points PM selected from the 3D virtual model n , four points PB selected from the actual femur n , maximum number of iterations m, general error threshold G e Let the target output of this algorithm be: registration error R e , second transformation T2 This method further includes the following steps: Set the initial error E; Based on PB n Find the corresponding sphere center C for the point set n ; Start a loop of i = 1, 2, …, m natural numbers: Among them, based on the PB n point set and the PM n point set, a transformation matrix T is obtained. According to the point sets SPS, TPS and the transformation matrix T, the registration error E is obtained i , When the error E i is less than the initial error E, an assignment operation for error convergence is performed: E = E i , T1 = T, (Equation 1), Thus, according to the value of the initial error E, the number of points in the point set SPS is reduced, and a new corresponding point set PB is obtained within the range of the sphere center C n n ,​ When the initial error E is less than the general error threshold G e then interrupt the loop By aligning the virtual point set and the physical point set, obtain the first transformation T1; Therefore, according to the source point set SPS, the target point set TPS, and the first transformation T1, the registration error R is obtained by ICP iteration. e and the second transformation T2.

2. The bone registration method according to claim 1, wherein The radius of the sphere is 2 mm.

3. The bone registration method according to claim 1, wherein The natural number i = 1, 2, …, 30.

4. A surgical robot system, including a surgical navigation system having an optical tracking system and a computer, which performs the steps of the bone registration method according to any one of claims 1 to 3 through the computer, registers the intraoperative coordinate system with the preoperative coordinate system, and guides or controls the surgical robot to perform surgery based on the coordinates of the actual bone.

5. A storage medium, which is a computer-readable storage medium storing a computer program, characterized in that, This computer program is executed to implement the steps of the bone registration method according to any one of claims 1 to 3.

6. A bone registration system for intraoperative navigation, which is used to determine the transformation relationship between the preoperative coordinate system and the intraoperative coordinate system, characterized in that, It includes: A source point set acquisition unit, which uses the preoperative bone image data to establish a preoperative coordinate system for the 3D virtual model, and obtains the source point set SPS from the surface of the 3D virtual model; A target point set generation unit, which uses the point data collected from the actual bone surface during the operation to establish an intraoperative coordinate system, and generates the target point set TPS; An initial registration unit, which performs an initial registration between the source point set SPS and the target point set TPS by selecting 4 groups of corresponding point pairs from the source point set SPS and the target point set TPS, and obtains the first transformation T1; An accurate registration unit, which uses the first transformation T1 and the source point set SPS to convert the target point set TPS into a three-dimensional model space that overlaps with the three-dimensional model space of the source point set SPS, thereby obtaining an accurate transformation matrix and a registration error; The initial registration unit further includes an iterative modification unit, which iteratively modifies the points L collected from the actual bone surface i to improve the first transformation error until the error is less than a preset threshold or the maximum number of iterations is reached. Among them, for the iterative modification of the point L i is carried out by randomly selecting 4 points within a sphere centered at this point L i and obtaining the first transformation T1 and the new first transformation error between each random point and the corresponding point M i on the 3D virtual model, where i is a natural number If the first transformation error does not meet the improvement condition, then select 4 random points within the sphere, recalculate the first transformation T1 and the first transformation error until the error decreases, update the center of the sphere, and repeat the iterative modification process; If the first transformation error satisfies the improvement condition, another new point L i+1 is used as the center of a new sphere to obtain a new random point and the iterative modification process is repeated. If the maximum number of iterations is reached and convergence to the established error threshold has not occurred, the resulting first transformation T1 is used as the initial transformation for the rigid registration algorithm. Let the input be: Source Point Set (SPS), Target Point Set (TPS), four points PM selected from the 3D virtual model n , four points PB selected from the actual femur n , maximum number of iterations m, general error threshold G e Let the target output of this algorithm be: Registration Error (R) e , second transformation T2 It further includes the following steps: Set an initial error E; Based on PB n Find the corresponding sphere center C for the point set n ; Start a loop for i = 1, 2, …, m natural numbers: Among them, based on the PB n point set and the PM n point set, a transformation matrix T is obtained. According to the point sets SPS, TPS, and the transformation matrix T, the registration error E is obtained i , When the error E i is less than the initial error E, an assignment operation for error convergence is performed: E = E i , T1 = T, (Equation 1), Thus, according to the value of the initial error E, the number of points in the point set SPS is reduced, and a new corresponding point set PB is obtained within the range of the sphere center C n n ,​ When the initial error E is less than the general error threshold G e Interrupt the loop Obtain the first transformation T1 by aligning the virtual point set and the physical point set. Thus, according to the source point set SPS, the target point set TPS, and the first transformation T1, the registration error R is obtained by ICP iteration. e and the second transformation T2.

Images