Fracture reduction dynamic planning system based on three-dimensional reconstruction technology

Abstract

The invention relates to the field of computer-aided foreign science, in particular to a fracture reduction dynamic planning system based on a three-dimensional reconstruction technology, and the system comprises the steps: obtaining a reference position and a reference path which are closest to a target position and have the most similar skeleton posture from all candidate paths except a target path; the similarity between the position lower constraint element attitude before the target position on the target path and the position lower constraint element attitude before the reference position on the reference path is recorded as a path switching index of the target position; marking the target position of which the path switching index is greater than a preset threshold value as a path switching point; and obtaining an optimal path and a standby reset path based on the path switching point. According to the method, when reduction errors occur in the fracture reduction process, a dynamic switching mechanism based on multiple standby reduction paths can be established, error tolerance is achieved through path switching index optimization, and the situation that fracture reduction cannot be completed according to expectation when reduction errors exist when a single reduction path is used is avoided.

Description

Technical Field

[0001] The present invention relates to the field of computer-assisted surgery, and in particular to a fracture reduction dynamic planning system based on three-dimensional reconstruction technology. Background Art

[0002] The fracture reduction process is typically accomplished using surgical navigation techniques, which rely on a preoperatively planned fracture reduction path. Existing fracture reduction paths are typically accomplished using computer-assisted planning techniques, such as dynamically planning the fracture reduction path by reconstructing a musculoskeletal model. Existing techniques for dynamic fracture reduction path planning, such as the one described in patent publication number CN116370073A, focus on using optimization simulation to obtain an optimal path, ensuring optimal fracture reduction in certain aspects.

[0003] However, due to the complexity of human tissue, during the fracture reduction process, reduction errors may occur when the fracture reduction is performed according to the optimal fracture reduction path designed before the operation. The reduction error refers to the position and posture of the tibia during the reduction process not changing according to the fracture reduction path, but deviating from the designed fracture reduction path. This results in the subsequent fracture reduction process having to rely on the doctor's experience to complete, or the fracture reduction result significantly deviates from expectations, and even when the fracture reduction is continued according to the fracture reduction path designed before the operation, it will obviously cause secondary damage to the patient.

[0004] For example, in pelvic fractures, the average reduction error of the intelligent reduction robot system is 3.41mm, and the postoperative residual displacement is 4.61mm, indicating that even with the use of automation technology, there are still non-negligible deviations when only one reduction path is used; for another example, Lei Jingtao, Wang Yang, Cheng Liya from Shanghai University and Hu Lei and Wang Tianmiao from Beijing University of Aeronautics and Astronautics published an article entitled "Reduction Robot Safety Strategy Based on Reduction Path Envelope Error and Improved Artificial Force Field Method" in the Journal of Mechanical Engineering, Issue 1, 2020. It is recorded that if the bone block offset exceeds the envelope error (usually set to 2mm) during the reduction of long bone fractures, it is necessary to rely on the doctor's experience to adjust.

[0005] In summary, the existing technology only uses one reduction path, and when a reduction error occurs, the fracture reduction may not be completed smoothly as expected. Summary of the Invention

[0006] In order to solve the problem that fracture reduction may not be completed smoothly as expected when only one reduction path is used and a reduction error occurs, the fracture reduction dynamic planning system based on three-dimensional reconstruction technology provided by the present invention not only obtains an optimal path, but also allocates multiple backup reduction paths to the optimal path. When a reduction error occurs in a specific fracture reduction process, a dynamic switching mechanism based on multiple backup reduction paths can be established, and error tolerance can be achieved through the optimization of path switching indicators.

[0007] The fracture reduction dynamic planning system based on three-dimensional reconstruction technology of the present invention adopts the following technical solutions:

[0008] One embodiment of the present invention provides a fracture reduction dynamic planning system based on three-dimensional reconstruction technology, which includes the following modules:

[0009] A 3D reconstruction module is used to 3D reconstruct a musculoskeletal model, which includes at least a skeleton and a muscle model. In the musculoskeletal model, fractured bones are recorded as daughter bones and mother bones. Bones other than daughter bones and muscles are recorded as constraint elements that interfere with the fracture reduction process.

[0010] A path acquisition module is used to obtain candidate paths on the musculoskeletal model that do not collide with constraint elements when the child bone moves to the parent bone during fracture reduction;

[0011] Each candidate path includes the skeletal posture of each position of the child bone during the movement process; the posture of the constraint element of each position of the child bone on each candidate path is simulated;

[0012] A path optimization module is used to record any candidate path as a target path and any position on the target path as a target position;

[0013] On all candidate paths other than the target path, obtain the position closest to the target position and with the most similar skeletal posture, record it as the reference position of the target position, and record the candidate path where the reference position is located as the reference path of the target path;

[0014] The similarity between the posture of the constraint element at the position before the target position on the target path and the posture of the constraint element at the position before the reference position on the reference path is recorded as the path switching index of the target position; the target position with the path switching index greater than the preset threshold th is recorded as the path switching point, and the candidate path where the reference position of the path switching point is located is recorded as the switching path of the target path;

[0015] The target path with the largest sum of the path switching indices of all path switching points and the most path switching points on the switching path is obtained and recorded as the optimal path. All switching paths of the optimal path are used as backup reduction paths when reduction errors occur in fracture reduction.

[0016] Preferably, the method of obtaining a position closest to the target position and having the most similar skeletal posture on all candidate paths outside the target path, and recording the position as a reference position of the target position, comprises the following specific steps:

[0017] For any candidate path among all candidate paths other than the target path, obtain the Euclidean distance a between any position on the candidate path and the target position, and the similarity b between the skeletal posture of any position and the skeletal posture of the target position; define the first index of any position on the candidate path as: ba;

[0018] A position with the largest first index is obtained on any candidate path and recorded as a candidate position; and a candidate position with the largest first index is selected from all candidate paths other than the target path as a reference position of the target position.

[0019] Preferably, the similarity between the posture of the constraint element at the position before the target position on the target path and the posture of the constraint element at the position before the reference position on the reference path is recorded as the path switching index of the target position, and the specific steps include the following:

[0020] Each constraint element under each position on each candidate path corresponds to a posture vector;

[0021] The similarity between the posture vector sequence composed of the posture vectors of each constraint element at all positions before the target position on the target path and the posture vector sequence composed of the posture vectors of each constraint element at all positions before the reference position on the reference path is recorded as the posture change similarity of each constraint element; the average of the posture change similarities of all constraint elements is recorded as the path switching index of the target position.

[0022] Preferably, the step of obtaining the target path with the largest sum of the path switching indicators of all path switching points and the most path switching points on the switching path is recorded as the optimal path, and includes the following specific steps:

[0023] Obtain the sum of the path switching indices of all path switching points on each target path, and record it as the first feature of each target path; for each switching path of each target path, record the number of path switching points on each switching path, and record the average number of path switching points on all switching paths of each target path as the second feature; record the sum of the first feature and the second feature of each target path as the preferred index of each target path, and take the target path with the largest preferred index as the optimal path.

[0024] Preferably, the specific steps for obtaining the similarity of the skeleton postures are as follows:

[0025] The rotation angles included in the skeletal posture constitute a vector. The rotation angles are the rotation angles of the skeleton around each coordinate axis in the three-dimensional space where the musculoskeletal model is located. The cosine similarity of the vectors corresponding to any two skeletal postures is used as the similarity of the skeletal postures.

[0026] Preferably, the specific steps of obtaining the posture vector are as follows:

[0027] When using bones other than each child bone as constraint elements, the position of the bone and the rotation angle of the bone around each coordinate axis in the three-dimensional space where the musculoskeletal model is located are used as the posture vector;

[0028] When each muscle is used as a constraint element, the stretched length and volume of each muscle, as well as the position of each muscle, are used as the pose vector.

[0029] Preferably, the specific steps for obtaining the similarity of the posture changes are as follows:

[0030] Perform PCA dimensionality reduction on all posture vectors in each posture vector sequence, reduce each posture vector to one dimension, and record the sequence composed of the dimensionality reduction results of all posture vectors in each posture vector sequence as the reduced dimensionality sequence;

[0031] The DTW distance of the reduced dimension sequence of any two posture vector sequences is recorded as y, and exp(-y) is recorded as the posture change similarity, where exp() represents an exponential function with a natural constant as the base.

[0032] Preferably, the method of using all the switching paths of the optimal path as backup reduction paths when a reduction error occurs in fracture reduction includes the following specific steps:

[0033] When the optimal path is used to perform fracture reduction, the position of the sub-bone and the bone posture detected in real time are recorded as the real-time position and real-time posture respectively; the position where the reduction error occurs is obtained based on the difference between the real-time position and real-time posture and the position and bone posture on the optimal path, and is recorded as the reduction error position;

[0034] Among all the path switching points on the optimal path, obtain the path switching point closest to the reduction error position; when the distance between the reduction error position and the nearest path switching point is less than the third preset threshold th3, move the sub-bone from the reduction error position to the reference position of the nearest path switching point, and then reduce the fracture according to the alternative reduction path where the reference position is located.

[0035] Preferably, the step of obtaining the position where the reset error occurs based on the difference between the real-time position and the real-time posture and the position and the skeletal posture on the optimal path, which is recorded as the reset error position, comprises the following specific steps:

[0036] The position closest to the real-time position is obtained on the optimal path, and the distance between this position and the real-time position is recorded as ra. The similarity rb between the skeletal posture of this position on the optimal path and the real-time posture is obtained. When rb is less than the first preset threshold th1 and ra is greater than the second preset threshold th2, it is determined that there is a reset error, and this position is used as the reset error position.

[0037] Preferably, when the distance between the reduction error position and the nearest path switching point is greater than or equal to th3, the value of the preset threshold th is reduced multiple times. After each reduction of the preset threshold th, the target position on the optimal path where the path switching index is greater than the preset threshold th is recorded as the path switching point again; until the path switching point on the optimal path that is closest to the reduction error position has a distance from the reduction error position that is not greater than or equal to th3, the value of the preset threshold th is stopped from being reduced; then, after the sub-bone is moved from the reduction error position to the reference position of the nearest path switching point, the fracture is reduced according to the alternative reduction path where the reference position is located.

[0038] The beneficial effects of the technical solution of the present invention are:

[0039] The optimal path obtained by the present invention can ensure that when a reduction error occurs, on the one hand, it can be reliably switched to an alternative fracture reduction path. The reliability refers to: the optimal path as a whole has a larger path switching index, and the path switching index is obtained from the position of the sub-bone and the skeleton posture and the posture change of the constraint element, ensuring that the sum of the path switching indexes of all path switching points is maximized, so that even if reduction errors occur in multiple places, there is a chance to switch to the alternative reduction path, or the risk of switching to the alternative reduction path can be avoided as much as possible when switching; on the other hand, if a reduction error still exists after switching to the alternative reduction path, it is still possible to switch to other alternative reduction paths.

[0040] Compared with the situation where only one reduction path is used and the fracture reduction may not be completed smoothly as expected when a reduction error occurs, the optimal path obtained by the present invention has multiple reliable backup reduction paths, which can ensure the smooth completion of the fracture reduction process and achieve error tolerance. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] Figure 1A diagram showing the framework of a dynamic planning system for fracture reduction based on three-dimensional reconstruction technology provided by one embodiment of the present invention;

[0043] Figure 2 This is a flow chart of a fracture reduction path switching method in a fracture reduction dynamic planning system based on three-dimensional reconstruction technology provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0044] To further illustrate the technical means and effects employed by the present invention to achieve the intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail the specific implementation, structure, features, and effects of the dynamic planning system for fracture reduction based on three-dimensional reconstruction technology proposed by the present invention. In the following description, different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0045] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0046] The specific scheme of the fracture reduction dynamic planning system based on three-dimensional reconstruction technology provided by the present invention is described in detail below with reference to the accompanying drawings.

[0047] Example 1:

[0048] See also Figure 1 , which shows a framework structure diagram of a fracture reduction dynamic planning system based on three-dimensional reconstruction technology provided by an embodiment of the present invention. The system includes: a three-dimensional reconstruction module, a path acquisition module and a path optimization module.

[0049] Specifically, the three-dimensional reconstruction module is used to three-dimensionally reconstruct the musculoskeletal model. Specifically, based on the CTA image, the bone and skin models are reconstructed in three dimensions using the Marching-Cubes algorithm, and the nerve distribution is detected by preoperative MRI (magnetic resonance imaging); CTA-MRI image fusion and registration technology are used to achieve three-dimensional geometric reconstruction of important tissues such as bones, muscles, blood vessels, and nerves. This process is a well-known technology. For example, the technology is disclosed in a lower limb bone high-energy injury reduction navigation method described in the patent publication number CN116370073A. The musculoskeletal model of this embodiment includes: bone, muscle, blood vessel and nerve models.

[0050] In addition, the fracture reduction process follows the principle of "the child seeks the mother," that is, the distal fracture end (the daughter bone) is moved to meet the proximal fracture end (the mother bone). Bones other than the daughter bone, as well as models of muscles, blood vessels, and nerves, are recorded as constraint elements. These constraint elements may block or interfere with the movement of the daughter bone toward the mother bone. For example, the bones, blood vessels, and muscles between the daughter bone and the mother bone may block or interfere with the movement of the daughter bone toward the mother bone. For example, the muscles attached to the daughter bone may also interfere with this process. When the daughter bone collides with these bones or blood vessels during movement, or stretches these muscles, it will leave sequelae to the fracture reduction or interfere with the postoperative recovery process.

[0051] The path acquisition module is used to obtain a candidate path on the musculoskeletal model that does not collide with the constraint elements when the child bone moves to the parent bone during the fracture reduction process.

[0052] As an example, the method for obtaining candidate paths includes:

[0053] Professional medical staff in the orthopedic field mark the movement path of the child bone when it moves towards the parent bone on the musculoskeletal model, delete the movement path that collides with the constraint elements or is determined to be inappropriate by medical staff, and record the remaining movement paths as candidate paths.

[0054] As another example, the method for obtaining candidate paths includes:

[0055] Patent publication number CN116370073A describes a method for repositioning and navigating high-energy lower limb skeletal injuries. This method constructs a repositioning path expression based on collision detection of bones, blood vessels, and nerves. Based on this, a series of basic repositioning paths are solved, and the muscle forces (i.e., the net force exerted on the muscles during deformation) are simulated and calculated. In this embodiment, the repositioning path in the series of basic repositioning paths where the muscle force is no greater than the maximum tolerance of each muscle is recorded as a candidate path.

[0056] In other examples, the method for obtaining candidate paths includes:

[0057] The movement path of the sub-bone is planned based on the method of artificial force field. The method adopts a virtual force field to guide the object (i.e., sub-bone) to avoid obstacles (i.e., constraint elements). For example, a repulsive force field is created between the sub-bone and other bones or muscles; specifically, the product of the shortest distance between the sub-bone and other bones (or muscles) and a first preset parameter (e.g., 1) can be used as the repulsive force field. For another example, a muscle-bone force field is created between the sub-bone and the muscles connected thereto; specifically, the product of the stretching amount of each muscle and a second preset parameter (e.g., 0.5) is used as the first muscle-bone force field, and the product of the torsion of the muscle and a third preset parameter (e.g., 0.5) is used as the second muscle-bone force field. The resultant force field of the repulsive force field, the first muscle-bone force field, and the second muscle-bone force field is regarded as a virtual force field. Path planning is performed based on the virtual force field using the method of artificial force field.

[0058] Path planning methods using artificial force fields and virtual force fields are well known and will not be described in detail in this embodiment. In this embodiment, by setting the first, second, and third preset parameters to different values, the artificial force field method outputs several movement paths, which are recorded as candidate paths.

[0059] In some other examples, the methods for obtaining candidate paths include:

[0060] The candidate paths obtained from all the above examples are taken as a set, and a number (for example, 20) candidate paths are screened from the set using an expert scoring method as the candidate paths obtained in this example.

[0061] So far, several candidate paths are obtained.

[0062] The candidate paths described in this embodiment refer to all positions of the child bone during its movement toward the parent bone (e.g., one position is sampled every 1mm). Each position is three-dimensional in the musculoskeletal model. Each child bone corresponds to a skeletal pose at each position, representing the posture of the child bone at that position. In this embodiment, the skeletal pose is the rotation angle of the child bone around each coordinate axis in the three-dimensional space of the musculoskeletal model (referred to as the rotation angle of the child bone).

[0063] In summary, the candidate path describes the translation and rotation at each position during the movement of the child bone.

[0064] Furthermore, the constraint element posture of each child bone at each position on each candidate path is simulated; specifically:

[0065] When the child bone moves on each candidate path, it will directly or indirectly affect the posture of the constrained element.

[0066] When the child bone moves to each position (that is, when it moves from the previous position to the current position), the posture of the constraint element at each position (that is, the posture of the constraint element at each position) is simulated.

[0067] As an example, when each other bone (except the child bone) is used as a constraint element, the position of each other bone (specifically the center position of the bone model) and the rotation angle of each other bone are used as the constraint element posture at each position.

[0068] As an example, when each muscle is used as a constraint element, the muscle force of each muscle (that is, the resultant force on the muscle during deformation) and the position of each muscle model are used as the constraint element posture at each position.

[0069] In other embodiments, the stretched length or volume of each muscle and the position of each muscle model may be used as the constraint element pose at each position.

[0070] For example, when using blood vessels or nerves as constraint elements, the musculoskeletal model must be divided into several segments, for example, one centimeter per segment. Each segment is then treated as a constraint element. The length, position, and rotation angle of each segment are used as the constraint element pose at each position.

[0071] In other embodiments, the extrusion deformation (eg, the deformation of the average diameter or volume) of each segment of the blood vessel or nerve model may be used as the constraint element pose at each position.

[0072] As an example, a specific method for simulating the posture of the constrained element at each position (i.e., the posture of the constrained element at each position) includes:

[0073] The Hill three-element model (contraction unit, parallel elastic unit, and serial elastic unit) is used to describe the muscle dynamics characteristics. In platforms such as Unity, the Rigidbody component and the Character Joint joint system are used to simulate the posture of the constraint element at each position.

[0074] In other embodiments, the Ziva VFX soft tissue dynamics engine may be used to simulate the pose of the constrained element at each position.

[0075] In some other embodiments, a finite element analysis method may be used to simulate the posture of the constraint element at each position.

[0076] The above method is a well-known technology and will not be described in detail in this embodiment.

[0077] The path optimization module is used to select a candidate path from all candidate paths as the optimal fracture reduction path (abbreviated as the optimal path).

[0078] The issue that this module considers is that during the fracture reduction process, whether it is manual fracture reduction or fracture reduction using robot surgical navigation, due to the complexity of human tissue, there may be reduction errors when reduction is performed according to the fracture reduction path designed before the operation. The reduction error refers to the fact that the position and posture of the sub-bone do not change according to the fracture reduction path during the sub-bone reduction process, but deviate from the designed fracture reduction path; this causes the subsequent fracture reduction process to rely on the doctor's experience to complete, or the fracture reduction result deviates significantly from expectations, and even when the reduction continues according to the fracture reduction path designed before the operation, it will obviously cause secondary damage to the patient.

[0079] This embodiment takes the above-mentioned problems into consideration when designing or planning the optimal path, so that even when there is a reduction error during the reduction process, there is still an available backup reduction path, thus avoiding as much as possible the above-mentioned problems existing in always relying on a single fracture reduction path.

[0080] Specifically:

[0081] (1) Any candidate path is recorded as the target path, and any position on the target path is recorded as the target position; on all candidate paths other than the target path, the position closest to the target position and with the most similar skeletal posture is obtained and recorded as the reference position of the target position, and the candidate path where the reference position is located is recorded as the reference path of the target path.

[0082] Among them, the reference position on the reference path is closest to the target position on the target path, and has the same or similar skeletal posture; then when the sub-bone is fractured along the target path, the position of the sub-bone and its skeletal posture are monitored in real time. If the sub-bone does not move toward the target position, or the skeletal posture of the sub-bone monitored in real time is significantly different from the skeletal posture of the target position, it means that the sub-bone has deviated from the target path at the target position. At this time, the reference path may be used as a backup reduction path for fracture reduction, for reference in the subsequent fracture reduction process.

[0083] As an example, on all candidate paths outside the target path, the position closest to the target position and having the most similar skeletal posture is obtained and recorded as the reference position of the target position. The methods include:

[0084] For any candidate path among all candidate paths other than the target path, obtain the Euclidean distance a between any position on the candidate path and the target position, and the similarity b between the skeleton posture of any position and the skeleton posture of the target position.

[0085] The first index of any position on the candidate path is defined as: ba; the larger the first index is, the closer the position is to the target position and the more similar the bone posture is.

[0086] A position with the maximum first index on any candidate path is obtained and recorded as the candidate position.

[0087] Similarly, for all candidate paths other than the target path, each candidate path corresponds to a candidate position. Among these candidate positions, the candidate position with the largest first index is again selected as the reference position of the target position.

[0088] In some embodiments, b×w1-a×w2 may be used as the first indicator, where w1 and w2 are preset attention coefficients. Some embodiments are described using w1 and w2 as 0.1 and 1, respectively, as examples. In other embodiments, w1 and w2 may be set to other values.

[0089] As an example, the method for calculating the similarity of skeletal postures includes:

[0090] Since the skeletal posture contains multiple values (such as the rotation angle around each coordinate axis in three-dimensional space), these values constitute a vector, and the cosine similarity of the vectors corresponding to any two skeletal postures is used as the similarity of the skeletal postures.

[0091] In other embodiments, the Euclidean distance x between vectors corresponding to any two skeletal postures and exp(-x) may be used as the similarity of the skeletal postures; exp() represents an exponential function with a natural constant as the base.

[0092] (2) The similarity between the posture of the constraint element at the position before the target position on the target path and the posture of the constraint element at the position before the reference position on the reference path is recorded as the path switching index of the target position; the target position where the path switching index is greater than the preset threshold is recorded as the path switching point, and the candidate path where the reference position of the path switching point is located is recorded as the switching path of the target path.

[0093] This embodiment further considers that the reference position on the reference path and the target position on the target path merely describe the differences or similarities between the positions of the sub-bone and the skeletal posture. The reference path obtained in this case cannot necessarily be used as a preferred backup path for the target path when there is a reset error. The path switching index in this embodiment further describes the similarities between the postures of the constraint elements when the sub-bone moves according to the target path and the reference path, respectively. The path switching point and switching path obtained based on the path switching index can more reliably serve as a preferred backup path for the target path when there is a reset error. The larger the path switching index, the more it indicates that the reference path not only has similar sub-bone postures at the reference and target positions as the target path, but also has similar effects and interferences on the constraint elements. In this case, the reference path can be used more effectively as a backup path for the target path when there is a reset error, or even as a substitute for the target path.

[0094] As an example, the similarity between the pose of the constraint element at the position before the target position on the target path and the pose of the constraint element at the position before the reference position on the reference path is recorded as the path switching index of the target position, including:

[0095] First, it's important to note that for any position on any candidate path, the pose of the constraint element at that position describes the poses of all constraint elements simulated when the child bone moves along that candidate path. Each constraint element's pose also contains multiple values (see the path acquisition module for details), which together form the constraint element's pose vector. In other words, each constraint element at each position on each candidate path corresponds to a pose vector.

[0096] For all positions before the target position on the target path (including the target position), the posture vectors of each constraint element at all positions constitute a posture vector sequence, which reflects the posture change of each constraint element before the target position on the target path.

[0097] For all positions before the reference position on the reference path (including the reference position), the posture vectors of each constraint element at all positions also constitute a posture vector sequence, which reflects the posture change of each constraint element before the reference position on the reference path.

[0098] For the same constraint element, the similarity between the pose vector sequence obtained for that constraint element on the target path and the pose vector sequence obtained for that constraint element on the reference path is recorded as the pose change similarity for each constraint element. The average of the pose change similarities for all constraint elements is recorded as the path switching index for the target location.

[0099] As an example, a method for obtaining the similarity of two posture vector sequences includes:

[0100] First of all, it should be noted that the lengths of the two posture vector sequences are not necessarily the same.

[0101] PCA dimensionality reduction is performed on all posture vectors in each posture vector sequence, and each posture vector is reduced to one dimension. The sequence composed of the dimensionality reduction results of all posture vectors in each posture vector sequence is recorded as the reduced dimensionality sequence.

[0102] The DTW distance of the reduced dimension sequence of the two posture vector sequences is recorded as y, and exp(-y) is recorded as the similarity of the two posture vector sequences.

[0103] The DTW distance is obtained by a DTW algorithm, which is a well-known technique and is not specifically limited in this embodiment.

[0104] Furthermore, the target position whose path switching index is greater than a preset threshold th is recorded as a path switching point, and the candidate path where the reference position of the path switching point is located is recorded as a switching path of the target path.

[0105] This embodiment is described by taking th=0.3 as an example. In other embodiments, it may be set to other values, which is not specifically limited in this embodiment.

[0106] At this point, several path switching points can be obtained on any candidate path.

[0107] (3) Path switching points are obtained on the target path, and a path switching index is obtained for each path switching point. For all target paths, the sum of the path switching indexes of all path switching points on each target path is obtained and recorded as the first feature of each target path. The larger the first feature, the more reliable it is to switch to an alternative fracture reduction path from multiple locations when there is a reduction error on the target path.

[0108] On the other hand, each switching path for each target path also has a number of path switching points. The number of path switching points is recorded, and the mean of the number of path switching points for all switching paths for each target path is recorded as the second feature, representing the average number of path switching points for each target path's switching path. A larger second feature indicates a greater number of path switching points on the target path's switching path.

[0109] The sum of the first and second characteristics of each target path is recorded as the preferred index of each target path, and the target path with the largest preferred index is used as the fracture reduction path (hereinafter referred to as the optimal path). All switching paths of the optimal path are used as backup reduction paths.

[0110] The optimal path obtained above can ensure that when a reduction error occurs, on the one hand, it can be reliably switched to an alternative fracture reduction path. The reliability refers to: the optimal path as a whole has a larger path switching index, which is obtained from the position of the sub-bone and the skeletal posture and the posture change of the constraint element, so that the above-mentioned first feature and the second feature have the maximum value at the same time, so that even if reduction errors occur in multiple places, there is a chance to switch to the alternative reduction path, or the risk of switching to the alternative reduction path can be avoided as much as possible when switching, such as when switching, because the sub-bone position needs to be moved significantly or the skeletal posture of the sub-bone needs to be significantly changed, or when the posture of the interfering element under the alternative reduction path cannot be adapted, there is a risk that the fracture reduction cannot be completed smoothly and is not conducive to subsequent rehabilitation. On the other hand, if there is still a reduction error after switching to the alternative reduction path, it is still necessary to switch to other alternative reduction paths as much as possible.

[0111] Compared with the situation where only one reduction path is used and fracture reduction may not be completed smoothly as expected when reduction errors occur, the optimal path obtained in this embodiment has multiple reliable backup reduction paths, which can ensure the smooth completion of the fracture reduction process.

[0112] Example 2:

[0113] In the above-mentioned first embodiment, the fracture reduction path (ie, the optimal path) and several alternative paths of the optimal path are obtained by using the three-dimensional reconstruction technology before fracture reduction. This process is completed before the fracture reduction surgery.

[0114] Next, the fracture is reduced according to the optimal path, which is a well-known process; for example, manual reduction according to the optimal path; another example is using a surgical navigation system robot to fix the bone and automatically reduce it according to the optimal path; another example is a combination of manual reduction and automatic reduction.

[0115] This embodiment provides a method for obtaining reduction error in fracture reduction, including:

[0116] By marking the surface of the sub-bone with an infrared optical tracker or an electromagnetic sensor, the position of the sub-bone and the skeleton posture are obtained in real time, which are recorded as the real-time position and real-time posture respectively. For example, a binocular vision method is used to track the position and posture (i.e., rotation angle) of multiple reflective ball markers fixed on the skeleton. The position closest to the real-time position is obtained on the optimal path, and the distance between this position and the real-time position is recorded as ra. The similarity rb of the skeleton posture at this position on the optimal path and the real-time posture is obtained. When rb is less than a first preset threshold th1 and ra is greater than a second preset threshold th2, it is indicated that the real-time position and real-time posture have obviously deviated from the optimal path. At this time, it is determined that a reset error occurs at this position on the optimal path.

[0117] This embodiment is described by taking th1=0.2 and th2=3 as an example. In other embodiments, th1 and th2 may be set to other values, which are not specifically limited in this embodiment.

[0118] It should be noted that the unit of the Euclidean distance between the positions of bones in all embodiments of the present invention is millimeter.

[0119] This embodiment provides a method, such as Figure 2 As shown, it is used to switch to the backup reset path when a reset error occurs.

[0120] First, it should be noted that, as can be seen from the first embodiment, there are several path switching points on the optimal path, each path switching point corresponds to a reference position, and the reference position is on the backup reset path.

[0121] When the optimal path obtained above is used to reduce the fracture, the position where the reduction error occurs is recorded as the reduction error position. Among all the path switching points on the optimal path, the path switching point closest to the reduction error position is obtained. When the Euclidean distance between the nearest path switching point and the reduction error position is less than the third preset threshold th3, the sub-bone is directly moved from the reduction error position to the reference position of the nearest path switching point. The movement trajectory is obtained by linear interpolation of the reduction error position (and the bone posture at this position) and the reference position of the nearest path switching point (and the bone at this position). After moving to the reference position of the nearest path switching point, the fracture is subsequently reduced according to the alternative reduction path where the reference position is located (the alternative reduction path at this time is regarded as the optimal path).

[0122] This embodiment is described by taking th3=0.5×th2 as an example. In other embodiments, 0.5 here can be replaced by other values, preferably a value less than 1 and greater than or equal to 0, which is not specifically limited in this embodiment.

[0123] Specifically, when the Euclidean distance between the nearest path switching point and the reset error position is greater than or equal to the preset threshold th3, it indicates that there is no path switching point near the reset error position. At this time, the preset threshold th in the first embodiment is reduced (for example, by 10% of the initial value, where the initial value of th is the value set for th in the first embodiment), and then the path switching point on the optimal path is obtained again (that is, the target position on the optimal path where the path switching index is greater than the preset threshold th is again recorded as the path switching point). Then, it is again determined whether the Euclidean distance between the nearest path switching point and the reset error position is greater than or equal to the preset threshold th3. If not, switching to the backup reset path is performed according to the above method. If so, the preset threshold th in the first embodiment is reduced again (for example, by 10% of the initial value), and so on, until the Euclidean distance between the nearest path switching point and the reset error position is no greater than or equal to the preset threshold th3.

[0124] In particular, when the reduced preset threshold th is less than 0.1, the preset threshold th is no longer reduced. At this time, the backup reduction path is no longer switched to. Instead, the sub-bone is directly moved from the reduction error position to the nearest position on the optimal path, and then the fracture reduction is continued according to the optimal path obtained in Example 1.

[0125] When only one optimal path is used for fracture reduction (that is, when a single path is used in the traditional method for fracture reduction), the success rate is 70% in 100 simulation experiments, that is, 30% of the simulation experiments need to rely on the doctor's experience to complete the fracture reduction, and the expected fracture reduction effect is not achieved in these 30% of simulation experiments. In contrast, in the 100 simulation experiments of this embodiment, the success rate is 92%, and the expected fracture reduction effect is achieved in 92% of the simulation experiments. It can be seen that this embodiment greatly improves the success rate of fracture reduction by dynamically switching between the optimal path and the backup reduction path. The values of th and the number of path switching points in the 100 simulation experiments of this embodiment are shown in Table 1. The average length of the optimal path in the 100 simulation experiments is 18 mm.

[0126] Table 1: Comparison table of th values and number of path switching points:

[0127] The value of th 0.3 0.27 0.24 0.21 0.18 0.15 0.12 Number of path switching points 7.1 8.7 9.4 12.7 13.2 13.6 15.5 Number of simulation experiments 28 17 15 10 13 11 6

[0128] For example, the second column of Table 1 shows that in 28 simulations, when th was ultimately set to 0.3, an average of 7.1 path switching points were achieved on each of the optimal path and all alternative reduction paths. Table 1 shows that fracture reduction was achieved when th = 0.3 in the majority of simulations.

[0129] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A dynamic planning system for fracture reduction based on three-dimensional reconstruction technology, characterized by: The system includes the following modules: A 3D reconstruction module is used to 3D reconstruct a musculoskeletal model, which includes at least a skeleton and a muscle model. In the musculoskeletal model, fractured bones are recorded as daughter bones and mother bones. Bones other than daughter bones and muscles are recorded as constraint elements that interfere with the fracture reduction process. The path acquisition module is used to obtain candidate paths on the musculoskeletal model that do not collide with constraint elements when the child bone moves to the parent bone during fracture reduction. Each candidate path includes the skeletal posture of each position of the child bone during the movement process; and simulates the posture of the constraint elements at each position of the child bone on each candidate path. The path optimization module is used to record any candidate path as the target path and any position on the target path as the target position; on all candidate paths other than the target path, the position closest to the target position and with the most similar skeletal posture is obtained and recorded as the reference position of the target position, and the candidate path where the reference position is located is recorded as the reference path of the target path; The similarity between the posture of the constraint element at the position before the target position on the target path and the posture of the constraint element at the position before the reference position on the reference path is recorded as the path switching index of the target position; the target position with the path switching index greater than the preset threshold th is recorded as the path switching point, and the candidate path where the reference position of the path switching point is located is recorded as the switching path of the target path; The target path with the largest sum of the path switching indices of all path switching points and the most path switching points on the switching path is obtained and recorded as the optimal path. All switching paths of the optimal path are used as backup reduction paths when reduction errors occur in fracture reduction.

2. The fracture reduction dynamic planning system based on three-dimensional reconstruction technology according to claim 1, characterized in that: The method of obtaining the position closest to the target position and having the most similar skeletal posture on all candidate paths outside the target path is recorded as the reference position of the target position, including the following specific steps: For any candidate path among all candidate paths other than the target path, obtain the Euclidean distance a between any position on the candidate path and the target position, and the similarity b between the skeletal posture of any position and the skeletal posture of the target position; define the first index of any position on the candidate path as: ba; A position with the largest first index is obtained on any candidate path and recorded as a candidate position; and a candidate position with the largest first index is selected from all candidate paths other than the target path as a reference position of the target position.

3. The fracture reduction dynamic planning system based on three-dimensional reconstruction technology according to claim 1, characterized in that: The similarity between the posture of the constraint element at the position before the target position on the target path and the posture of the constraint element at the position before the reference position on the reference path is recorded as the path switching index of the target position, and the specific steps include the following: Each constraint element under each position on each candidate path corresponds to a posture vector; The similarity between the posture vector sequence composed of the posture vectors of each constraint element at all positions before the target position on the target path and the posture vector sequence composed of the posture vectors of each constraint element at all positions before the reference position on the reference path is recorded as the posture change similarity of each constraint element; the average of the posture change similarities of all constraint elements is recorded as the path switching index of the target position.

4. The fracture reduction dynamic planning system based on three-dimensional reconstruction technology according to claim 1, characterized in that: The step of obtaining a target path having the largest sum of path switching indicators of all path switching points and the most path switching points on the switching path is recorded as the optimal path, and includes the following specific steps: Obtain the sum of the path switching indices of all path switching points on each target path, and record it as the first feature of each target path; For each switching path of each target path, record the number of path switching points on each switching path, and record the average number of path switching points on all switching paths of each target path as the second feature; The sum of the first feature and the second feature of each target path is recorded as the preferred index of each target path, and the target path with the largest preferred index is taken as the optimal path.

5. The fracture reduction dynamic planning system based on three-dimensional reconstruction technology according to claim 2, characterized in that: The specific steps for obtaining the similarity between the skeleton posture at any one position and the skeleton posture at the target position are as follows: The rotation angles included in the skeletal posture constitute a vector. The rotation angles are the rotation angles of the skeleton around each coordinate axis in the three-dimensional space where the musculoskeletal model is located. The cosine similarity of the vectors corresponding to any two skeletal postures is used as the similarity of the skeletal postures.

6. The fracture reduction dynamic planning system based on three-dimensional reconstruction technology according to claim 3, characterized in that: The specific steps for obtaining the posture vector are as follows: When using bones other than each child bone as constraint elements, the position of the bone and the rotation angle of the bone around each coordinate axis in the three-dimensional space where the musculoskeletal model is located are used as the posture vector; When each muscle is used as a constraint element, the stretched length and volume of each muscle, as well as the position of each muscle, are used as the pose vector.

7. The fracture reduction dynamic planning system based on three-dimensional reconstruction technology according to claim 3, characterized in that: The specific steps for obtaining the similarity of the posture changes are as follows: Perform PCA dimensionality reduction on all posture vectors in each posture vector sequence, reduce each posture vector to one dimension, and record the sequence composed of the dimensionality reduction results of all posture vectors in each posture vector sequence as the reduced dimensionality sequence; The DTW distance of the reduced-dimensional sequence of any two posture vector sequences is denoted as y, and exp(-y) is denoted as the posture change similarity, where exp() represents an exponential function with a natural constant as the base.

8. The fracture reduction dynamic planning system based on three-dimensional reconstruction technology according to claim 1, characterized in that: The specific steps of using all the switching paths of the optimal path as the backup reduction paths when a reduction error occurs in fracture reduction are as follows: When the optimal path is used to perform fracture reduction, the position of the sub-bone and the bone posture detected in real time are recorded as the real-time position and real-time posture respectively; the position where the reduction error occurs is obtained based on the difference between the real-time position and real-time posture and the position and bone posture on the optimal path, and is recorded as the reduction error position; Among all the path switching points on the optimal path, obtain the path switching point closest to the reduction error position; when the distance between the reduction error position and the nearest path switching point is less than the third preset threshold th3, move the sub-bone from the reduction error position to the reference position of the nearest path switching point, and then reduce the fracture according to the alternative reduction path where the reference position is located.

9. The fracture reduction dynamic planning system based on three-dimensional reconstruction technology according to claim 8, characterized in that: The method of obtaining the position where the reset error occurs based on the difference between the real-time position and real-time posture and the position and skeleton posture on the optimal path, which is recorded as the reset error position, includes the following specific steps: The position closest to the real-time position is obtained on the optimal path, and the distance between this position and the real-time position is recorded as ra. The similarity rb between the skeletal posture of this position on the optimal path and the real-time posture is obtained. When rb is less than the first preset threshold th1 and ra is greater than the second preset threshold th2, it is determined that there is a reset error, and this position is used as the reset error position.

10. The fracture reduction dynamic planning system based on three-dimensional reconstruction technology according to claim 8, characterized in that: When the distance between the reduction error position and the nearest path switching point is greater than or equal to th3, the value of the preset threshold th is reduced multiple times. After each reduction of the preset threshold th, the target position on the optimal path where the path switching index is greater than the preset threshold th is again recorded as the path switching point; until the path switching point on the optimal path that is closest to the reduction error position is no more than or equal to th3 away from the reduction error position, the reduction of the preset threshold th is stopped; then, after the sub-bone is moved from the reduction error position to the reference position of the nearest path switching point, the fracture is reduced according to the alternative reduction path where the reference position is located; where 0.1≤th≤0.3.

Images