Image three-dimensional reconstruction method and system for orthopedic postoperative monitoring

Abstract

The application discloses an image three-dimensional reconstruction method and system for orthopedic postoperative monitoring, and relates to the technical field of image processing. The method comprises the following steps: positioning a user's operation area pixel mask, and collecting reference three-dimensional image data; performing candidate base point analysis on the reference three-dimensional image data to obtain a candidate base point set; performing stability scoring on the candidate base point set, identifying an anchor base point set greater than a preset stability scoring threshold based on the stability scoring result, and selecting at least one reconstruction anchor base point; acquiring at least one follow-up three-dimensional image data, performing three-dimensional reconstruction on the at least one follow-up three-dimensional image data, and obtaining a three-dimensional image reconstruction result. The technical problems of difficult dynamic change analysis of image three-dimensional reconstruction, insufficient feature point stability and data registration accuracy in the prior art are solved, and the technical effects of multi-time three-dimensional image high-precision registration, improved consistency of three-dimensional reconstruction results at different follow-up time points and three-dimensional reconstruction accuracy are achieved.

Description

Technical Field

[0001] This application relates to the field of image processing technology, specifically to a method and system for three-dimensional image reconstruction for postoperative monitoring in orthopedics. Background Technology

[0002] Orthopedic surgeries involving fracture reduction, joint replacement, and spinal correction rely heavily on precise monitoring of the surgical area for postoperative rehabilitation and complication prevention. Three-dimensional reconstruction of medical images is widely used for postoperative evaluation. By spatially reconstructing tomographic data from CT and MRI, the three-dimensional structure of bones, implants, and surrounding soft tissues can be visually displayed. However, it has several limitations in long-term follow-up and dynamic monitoring. First, postoperative image comparison relies on global registration of the overall three-dimensional model. However, due to differences in patient position, soft tissue deformation, and scanning conditions at different time points, registration errors can easily accumulate, affecting monitoring accuracy. Second, relying on bone surface features for reconstruction and alignment is unstable for patients with osteoporosis or significant changes in bone surface morphology during fracture healing. Furthermore, implants may exhibit metallic artifacts in the images, interfering with the identification of key feature points and further reducing the accuracy of three-dimensional reconstruction.

[0003] Therefore, current technologies face challenges such as difficulty in analyzing dynamic changes during 3D image reconstruction, insufficient stability of feature points, and inadequate data registration accuracy. Summary of the Invention

[0004] This application provides a method and system for three-dimensional image reconstruction for postoperative monitoring in orthopedics, which solves the technical problems of difficulty in analyzing dynamic changes in three-dimensional image reconstruction, insufficient stability of feature points and insufficient data registration accuracy in the prior art. It achieves the technical effects of high-precision registration of multi-temporal three-dimensional images, improving the consistency of three-dimensional reconstruction results at different follow-up time points and improving the accuracy of three-dimensional reconstruction.

[0005] This application provides a method for three-dimensional image reconstruction for postoperative monitoring in orthopedic surgery. The method includes: locating a pixel mask of the user's surgical area and acquiring baseline three-dimensional image data of the pixel mask; performing candidate base point analysis on the baseline three-dimensional image data to obtain a set of candidate base points, wherein the candidate base points include geometric feature points on the implant, skeletal anatomical feature points, feature points within a stable area of ​​the soft tissue surface, and preset feature points on the user's anatomical structure mesh; performing a stability score on the set of candidate base points, identifying a set of anchor base points with a stability score greater than a preset stability score threshold from the set of candidate base points based on the stability score result, and selecting at least one reconstructed anchor base point; acquiring at least one follow-up three-dimensional image data, and performing three-dimensional reconstruction on the at least one reconstructed anchor base point to obtain the three-dimensional image reconstruction result of the pixel mask of the surgical area.

[0006] In a possible implementation, the image 3D reconstruction method for postoperative orthopedic monitoring further performs the following processing: acquiring stability training samples, which include multiple stability factors under different postoperative orthopedic samples, wherein the multiple stability factors include the temporal variance of the base point at different time points, the average confidence level of the data detection, the structural rigidity score of the base point location, and the spatial distance between the base point and the lesion; performing logistic regression training based on the stability training samples to obtain multiple learning weights corresponding to the multiple stability factors; fitting a stability scoring model using the multiple learning weights and the multiple stability factors; and using the stability scoring model to score the stability of the candidate base point set.

[0007] In a possible implementation, the three-dimensional image reconstruction method for postoperative monitoring in orthopedic surgery further performs the following processing: randomly selecting initial reconstruction anchor points from the set of anchor points; defining a multi-objective optimization function, the multi-objective function including a first optimization objective and a second optimization objective, the first optimization objective being to maximize the stability of the initial reconstruction anchor points, and the second optimization objective being that the spatial distribution uniformity of the initial reconstruction anchor points satisfies a preset uniformity threshold; and using the multi-objective optimization function to perform a greedy algorithm to optimize the initial reconstruction anchor points in the set of anchor points to obtain at least one reconstruction anchor point.

[0008] In a possible implementation, the image three-dimensional reconstruction method for postoperative monitoring in orthopedics further performs the following processing: optimizing the initial reconstruction anchor points in the anchor point set using the multi-objective optimization function; wherein the candidate anchor point set has a preset priority sequence, which is arranged in descending order as geometric feature points on the implant, skeletal anatomical feature points, feature points in the stable area of ​​the soft tissue surface, and preset feature points on the user's anatomical structure mesh.

[0009] In a possible implementation, the image three-dimensional reconstruction method for postoperative monitoring in orthopedics further performs the following processing: detecting the reconstruction marker status data of the anchoring base point set, including the number of times the markers are used for reconstruction and the marker frequency; replacing the anchoring base point set with invalid markers based on the reconstruction marker status data, wherein if the markers are not marked for a preset number of consecutive times or the marker frequency is less than a preset threshold, the current anchoring base point is determined to be invalid.

[0010] In a possible implementation, the image three-dimensional reconstruction method for postoperative monitoring in orthopedics further performs the following processing: if the at least one reconstruction anchor point returns multiple points, the at least one reconstruction anchor point is input into a rigid transformation solving algorithm to perform prior sparse three-dimensional reconstruction on the at least one follow-up three-dimensional image data to obtain a sparse three-dimensional image reconstruction result; the sparse three-dimensional image reconstruction result is then subjected to posterior registration three-dimensional reconstruction to obtain a registered three-dimensional image reconstruction result.

[0011] In a possible implementation, the image three-dimensional reconstruction method for postoperative monitoring in orthopedics further performs the following processing: obtaining a reference three-dimensional image reconstruction result based on the reference three-dimensional image data of the pixel mask of the surgical area; fusing the three-dimensional image reconstruction result with the reference three-dimensional image reconstruction result to output a fused three-dimensional image reconstruction result; performing deformation analysis on the fused three-dimensional image reconstruction result to output deformation features, including average displacement, peak displacement, angle change, local volume change, and deformation rate; calculating a risk index of the at least one follow-up three-dimensional image data based on the deformation features, and sending a risk warning message to the user based on the risk index.

[0012] This application also provides a three-dimensional image reconstruction system for postoperative monitoring in orthopedics. The system includes: a three-dimensional image data acquisition module for locating the user's surgical area pixel mask and acquiring baseline three-dimensional image data of the surgical area pixel mask; a candidate base point analysis module for performing candidate base point analysis on the baseline three-dimensional image data to obtain a set of candidate base points, wherein the candidate base points include geometric feature points on the implant, skeletal anatomical feature points, feature points within the stable area of ​​the soft tissue surface, and preset feature points on the user's anatomical structure mesh; an anchoring base point set identification module for performing stability scoring on the candidate base point set, identifying anchoring base point sets greater than a preset stability scoring threshold from the candidate base point set based on the stability scoring results, and selecting at least one reconstructed anchoring base point; and a three-dimensional reconstruction module for acquiring at least one follow-up three-dimensional image data, performing three-dimensional reconstruction on the at least one reconstructed anchoring base point, and obtaining the three-dimensional image reconstruction result of the surgical area pixel mask.

[0013] This application proposes a method and system for three-dimensional image reconstruction for postoperative monitoring in orthopedic surgery. The method involves locating the pixel mask of the user's surgical area and acquiring baseline three-dimensional image data. Candidate base point analysis is performed on the baseline three-dimensional image data to obtain a set of candidate base points. A stability score is applied to the candidate base point set, and anchor base points with scores exceeding a preset stability threshold are identified. At least one reconstructed anchor base point is selected. At least one follow-up three-dimensional image is acquired, and three-dimensional reconstruction is performed on this data to obtain the three-dimensional image reconstruction result. This method solves the technical problems of difficulty in analyzing dynamic changes in three-dimensional image reconstruction, insufficient feature point stability, and inadequate data registration accuracy in existing technologies. It achieves high-precision registration of multi-temporal three-dimensional images, improves the consistency of three-dimensional reconstruction results at different follow-up time points, and enhances the accuracy of three-dimensional reconstruction. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Flowcharts are used in this application to illustrate the operations performed by the system according to the embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0015] Figure 1 This is a schematic diagram of the three-dimensional image reconstruction method for postoperative monitoring in orthopedics provided in an embodiment of this application.

[0016] Figure 2 This is a schematic diagram of candidate base points obtained in the three-dimensional image reconstruction method for postoperative monitoring of orthopedic surgery provided in the embodiments of this application.

[0017] Figure 3 This is a schematic diagram of the reconstruction anchoring points selected in the three-dimensional image reconstruction method for postoperative monitoring in orthopedics provided in the embodiments of this application.

[0018] Figure 4 This is a schematic diagram of the structure of a three-dimensional image reconstruction system for postoperative monitoring in orthopedics, provided in an embodiment of this application.

[0019] Figure labeling: 3D image data acquisition module 10, candidate base point analysis module 20, anchor base point set identification module 30, 3D reconstruction module 40. Detailed Implementation

[0020] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of this application are given below.

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application will be provided in conjunction with the accompanying drawings. The described embodiments should not be considered as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0022] In the following description, references to "some embodiments" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, 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 application belongs. The terminology used herein is for the purpose of describing embodiments of this application only.

[0023] This application provides a three-dimensional image reconstruction method for postoperative monitoring in orthopedics, such as... Figure 1 As shown, the method includes:

[0024] Step S100: Locate the pixel mask of the user's surgical area and acquire the reference three-dimensional image data of the pixel mask of the surgical area.

[0025] Preferably, the surgical area refers to the anatomical location where the patient undergoes orthopedic surgery, such as the acetabular region in hip replacement, a specific vertebral segment in spinal surgery, or a bone segment for internal fixation of a fracture. The pixel mask refers to a binary image region of the same size as the original image data, marking which pixels belong to the surgical area and which do not. The pixel set of the surgical area is extracted from CT / MRI / DICOM images through image segmentation to generate the corresponding mask layer, i.e., the surgical area pixel mask. Tomographic data is acquired using medical imaging equipment. Based on the surgical area pixel mask, high-quality three-dimensional image data containing only the surgical area is collected as a zero-point reference for follow-up comparison, i.e., baseline three-dimensional image data is obtained.

[0026] Step S200: Perform candidate base point analysis on the reference three-dimensional image data to obtain a set of candidate base points. The candidate base points include geometric feature points on the implant, skeletal anatomical feature points, feature points in the stable area of ​​the soft tissue surface, and preset feature points on the user's anatomical structure mesh.

[0027] Preferred, such as Figure 2 As shown, candidate base point analysis is performed on the baseline 3D image data. This involves identifying feature points in the baseline 3D image data that are likely to remain stable during follow-up and can be used for registration, resulting in multiple candidate base points, forming a candidate base point set. These candidate base points include geometric feature points on the implant (implant_geo), skeletal anatomical feature points (bone_anatomy), feature points within stable soft tissue surface regions (soft_tissue), and pre-set feature points on the user's anatomical structure mesh (mesh_preset). Specifically, geometric feature points on the implant refer to clearly identifiable geometric shape features on the surface of orthopedic implants such as screws, plates, and joint prostheses, such as the center point of the screw head, etc. The features include: corner points at the body edge and the center of the screw hole on the steel plate; skeletal anatomical feature points, which are anatomical landmarks naturally present on the bone surface in the image, such as the apex of the intercondylar crest of the knee joint, the anterior superior iliac spine and bony tubercle of the pelvis, and bony ridges, which serve as reference points for surgical navigation or image registration; feature points in the stable region of the soft tissue surface, which are feature points extracted from areas of the skin, tendons, and other soft tissue surfaces that do not significantly deform at different follow-up times, such as the center point of a surgical scar or the intersection of specific folds on the skin; and feature points preset on the user's anatomical structure mesh, which are reference points predefined on the three-dimensional anatomical mesh model, such as specific node numbers in the three-dimensional bone mesh model or mesh intersections evenly distributed on the anatomical surface, which can be used to increase the density of feature points during auxiliary registration.

[0028] Step S300: Perform stability scoring on the candidate base point set, identify anchor base points with a value greater than a preset stability score threshold from the candidate base point set based on the stability score results, and select at least one reconstructed anchor base point.

[0029] Step S300 further includes step S310, obtaining stability training samples, wherein the stability training samples include multiple stability factors under different orthopedic postoperative samples, wherein the multiple stability factors include the time variance of the base point at different time points, the average confidence level of data detection, the structural rigidity score of the base point location, and the spatial distance between the base point and the lesion; step S320, performing logistic regression training based on the stability training samples to obtain multiple learning weights corresponding to the multiple stability factors, fitting a stability scoring model using the multiple learning weights and the multiple stability factors, and using the stability scoring model to score the stability of the candidate base point set.

[0030] Preferably, stability training samples are obtained from multiple orthopedic postoperative imaging follow-up data. Each sample is labeled with whether each feature point is stable at different time points. The stability training samples include multiple stability factors under different orthopedic postoperative samples, that is, the stability training samples cover different types of surgeries such as fracture fixation, joint replacement, and spinal internal fixation, to ensure the model's generalization ability. Among them, multiple stability factors include the time variance of the base point at different time points, the average confidence score of the data detection, the structural rigidity score of the base point's location, and the spatial distance between the base point and the lesion. Specifically, the time variance of the base point at different time points refers to the calculation of... The variance of the spatial location change of this base point during multiple follow-ups is considered; the smaller the variance, the higher the stability. The average confidence score of the data detection refers to the average confidence score of the output of the image processing or feature point detection algorithm. Points that are stable and easily identifiable have a high confidence score. The structural rigidity score of the base point's location refers to the physical rigidity of the anatomical structure where the base point is located. For example, cortical bone has high rigidity, while soft tissue has low rigidity. This can be assessed through material properties or local CT density. The spatial distance between the base point and the lesion refers to the distance between the base point and the surgical lesion or implant. Points that are too close to the lesion may move with changes in the lesion and are not stable enough.

[0031] Preferably, logistic regression training is performed based on stability training samples. This involves using multiple stability factors from the training samples as input and known "stable / unstable" labels as output to train a logistic regression model. The model learns multiple learning weights corresponding to the stability factors, i.e., it learns the importance of each stability factor. Different stability factors are assigned different weights to indicate their impact on stability. For example, "temporal variance" may have a high weight, while "distance from lesion" has a lower weight. Then, a stability scoring model is fitted using multiple learning weights and multiple stability factors. Finally, the stability scoring model is used to score the stability of the candidate baseline set. This involves outputting the weights of each factor for each candidate baseline and performing a weighted sum to obtain a stability score. A higher score indicates that the baseline has higher reliability in follow-up reconstruction, thereby ensuring that the most reliable anchor points are selected.

[0032] Preferred, such as Figure 3As shown, based on the stability score results, multiple candidate base points with a value greater than the preset stability score threshold (e.g., 0.8) are identified from the candidate base point set and combined to form an anchor base point set. Points that are not stable enough are filtered out to ensure that subsequent registration is based on high-quality base points. Finally, through multi-objective optimization and a greedy algorithm, at least one reconstruction anchor base point is selected from the anchor base point set for use in the 3D registration and reconstruction of the follow-up images. For example, the total number of candidate base points in the candidate base point set is 150, and the number of identified anchor base points is 6. The anchor base point information is as follows: Anchor base point 1 coordinates (21.0, 20.0, 19.0), stability 0.662; Anchor base point 2 coordinates (28.0, 13.0, 10.0), stability 0.664; Anchor base point 3 coordinates (31.0, 12.0, 11.0), stability 0.670; Anchor base point 4 coordinates (33.0, 12.0, 15.0), stability 0.670; Anchor base point 5 coordinates (37.0, 24.0, 20.0), stability 0.670; Anchor base point 6 coordinates (37.0, 25.0, 19.0), stability 0.670; The average stability score of the 6 anchor base points is 0.668, and the registration error (root mean square error) is 1.629.

[0033] Furthermore, step S300 also includes step S330, randomly selecting initial reconstruction anchor points from the set of anchor points; step S340, defining a multi-objective optimization function, the multi-objective function including a first optimization objective and a second optimization objective, the first optimization objective being to maximize the stability of the initial reconstruction anchor points, and the second optimization objective being that the spatial distribution uniformity of the initial reconstruction anchor points satisfies a preset uniformity threshold; step S350, using the multi-objective optimization function to perform a greedy algorithm to optimize the initial reconstruction anchor points in the set of anchor points, obtaining at least one reconstruction anchor point.

[0034] Preferably, multiple base points are randomly selected from the set of anchor base points as initial reconstruction anchor base points. Maximizing the stability of the initial reconstruction anchor base points is the first optimization objective. Satisfying a preset uniformity threshold in the spatial distribution of the initial reconstruction anchor base points is the second optimization objective. This preset uniformity threshold is set based on historical data to prevent all base points from concentrating in a localized area, thereby improving the overall stability and coverage of the reconstruction. A multi-objective optimization function is constructed with the first and second optimization objectives as constraints. The spatial distribution uniformity is the average distance variance of the multiple initial reconstruction anchor base points; a smaller variance indicates a more uniform distribution. When the spatial distribution uniformity meets the preset uniformity threshold... The score for this item is 1; otherwise, it is penalized proportionally. The weight parameters for stability and spatial distribution uniformity are adjusted according to the importance of the task, for example, the weights are 0.7 and 0.3 respectively. Through a multi-objective optimization function, a greedy algorithm is used to optimize the initial reconstruction anchor points in the set of anchor points. That is, starting from the initial reconstruction anchor points, among the remaining unselected anchor points, the anchor point that will increase the multi-objective optimization function score the most after being added is found, and it is compared whether the preset uniformity threshold is still met after adding the anchor point. If it is met, the anchor point is added to the reconstruction anchor points. The optimization is repeated until the preset number is reached or the multi-objective optimization function score can no longer be improved. Finally, at least one reconstruction anchor point is determined, thereby improving the accuracy and consistency of 3D reconstruction.

[0035] Furthermore, step S350 also includes optimizing the initial reconstruction anchor points in the anchor point set using the multi-objective optimization function; wherein the candidate anchor point set has a preset priority sequence, and the priority sequence is arranged in descending order as geometric feature points on the implant, skeletal anatomical feature points, feature points in the stable area of ​​the soft tissue surface, and preset feature points on the user's anatomical structure mesh.

[0036] Preferably, the candidate base point set has a preset priority sequence, representing the importance of different types of base points. The priority sequence is arranged from highest to lowest as follows: geometric feature points on the implant, skeletal anatomical feature points, feature points in the stable area of ​​the soft tissue surface, and preset feature points on the user's anatomical structure grid. Specifically, geometric feature points on the implant have the highest stability (the implant is fixed), and the metal structure is clearly visible in the image; skeletal landmarks in the skeletal anatomical feature points change little during the postoperative stabilization period; feature points in the stable area of ​​the soft tissue surface have moderate stability, and some are affected by position or deformation; preset feature points on the user's anatomical structure grid are auxiliary selection points and have the lowest natural stability. When the scores of the multi-objective optimization function are similar among different types of base points, the higher priority type of points will be retained first. In the greedy algorithm optimization process, the priority can be used as an additional weight for scoring, including adding extra points to different types of base points according to the priority, to ensure that the optimization process prioritizes implant feature points when the stability is similar, followed by skeletal feature points, then stable soft tissue points, and finally preset grid points.

[0037] Furthermore, step S300 also includes step S360, detecting the reconstruction marker status data of the anchoring base point set, including the number of markers used for reconstruction and the marker frequency; step S370, replacing the anchoring base point set with a failed one according to the reconstruction marker status data, wherein if the anchoring base point is not marked for a preset number of consecutive times or the marker frequency is less than a preset threshold, the current anchoring base point is determined to be failed.

[0038] Preferably, the reconstruction marker status data of the anchor point set is detected, that is, the usage record of each anchor point is recorded during the follow-up 3D image reconstruction process, including the number of times the marker is used for reconstruction and the marker frequency. The number of times the marker is used refers to the total number of times the anchor point has been successfully identified and used for registration in a number of follow-up reconstructions. For example, if the point is matched 4 times in 5 follow-ups, the number of times the marker is used is 4. The marker frequency is the ratio of the number of times the marker is used to measure the attendance rate of the anchor point in long-term follow-up. For example, in the above example, the marker frequency is 80%. This reflects whether the anchor points are truly stable and usable in long-term monitoring. Based on the reconstruction marker status data, the anchor point set is replaced with invalid markers. Specifically... If a base point cannot be detected and marked for a preset number of consecutive times (e.g., 3 times), the current anchoring base point is determined to be invalid. This may be due to soft tissue deformation causing mismatch, metal artifact occlusion, or changes in bone structure. Alternatively, if the marking frequency is less than a preset threshold (e.g., a marking frequency of ≥70%), the base point is considered unreliable in long-term follow-up and is also determined to be invalid. Then, a new high-stability base point is used to replace it. That is, the invalid base point is removed from the anchoring base point set, and a new base point that meets the requirements of stability and spatial distribution uniformity is selected from the candidate base point set to replace it. This ensures that the anchoring base point set is always composed of high-quality and highly available base points, thereby ensuring the reliability and consistency of the 3D image reconstruction.

[0039] Step S400: Obtain at least one follow-up three-dimensional image data, and perform three-dimensional reconstruction on the at least one follow-up three-dimensional image data based on the at least one reconstruction anchor point to obtain the three-dimensional image reconstruction result of the pixel mask of the surgical area.

[0040] Preferably, at least one follow-up 3D image data is acquired. This follow-up 3D image data refers to the surgical area pixel mask collected at different postoperative times (e.g., 3 months, 6 months, 1 year post-surgery). The corresponding medical 3D image data may be CT, MRI, 3D scan, etc., meaning only the image portion related to the surgical site is retained. Then, 3D reconstruction is performed on the at least one follow-up 3D image data based on at least one reconstruction anchor point. Specifically, anchor points at corresponding positions are found in the baseline 3D image data and the at least one follow-up 3D image data, and the two sets of anchor points are spatially registered and aligned. Then, the position of the reconstruction anchor point is detected in the at least one follow-up 3D image data. Based on the positional relationship between the reconstruction anchor point and the anchor point in the baseline image, a rigid or non-rigid transformation matrix is ​​calculated. Then, a spatial transformation is performed on the at least one follow-up 3D image data to precisely align it with the baseline 3D image data. Based on the alignment, a 3D model of the surgical area is reconstructed, i.e., the volume corresponding to the surgical area pixel mask. Finally, the registered 3D surgical area model, in the same coordinate system as the baseline image, is output, determining the 3D image reconstruction result of the surgical area pixel mask.

[0041] Furthermore, step S400 also includes step S410, if the at least one reconstruction anchor point is returned as multiple, the at least one reconstruction anchor point is input into the rigid transformation solving algorithm to perform prior sparse 3D reconstruction on the at least one follow-up 3D image data to obtain sparse 3D image reconstruction results; step S420, the sparse 3D image reconstruction results are subjected to posterior registration 3D reconstruction to obtain registered 3D image reconstruction results.

[0042] Preferably, if at least one reconstructed anchor point is returned as multiple points, that is, multiple reconstructed anchor points are found in the follow-up image, the spatial transformation relationship between the reference 3D image data and the follow-up 3D image data is calculated based on the multiple anchor points. Then, at least one reconstructed anchor point is input into the rigid transformation solution algorithm for solution. The rigid transformation includes rotation and translation. That is, the 3D coordinates of the anchor point in the reference 3D image data and the 3D coordinates of the corresponding anchor point in the follow-up 3D image data are input, and the rotation matrix and translation vector are output as rigid transformation parameters to align the follow-up image to the reference image coordinate system.

[0043] Preferably, prior sparse 3D reconstruction is performed on at least one follow-up 3D image data. This involves first performing rapid 3D reconstruction on key regions of the follow-up 3D image data, where key regions include sparse point clouds or low-resolution voxels, resulting in a roughly aligned 3D result, i.e., the sparse 3D image reconstruction result. Then, posterior registration 3D reconstruction is performed on the sparse 3D image reconstruction result. Specifically, based on the prior rigid alignment, fine-grained spatial registration is performed, typically using non-rigid registration that allows local deformation. This not only matches anchor points but can also use whole-pixel information or surface features for matching, thus completing the posterior registration 3D reconstruction. This eliminates the small errors remaining after rigid alignment, ultimately outputting a registered high-precision 3D image model, i.e., the registered 3D image reconstruction result, ensuring the accuracy of the 3D reconstruction of images used for postoperative monitoring in orthopedics.

[0044] Furthermore, step S400 also includes step S430, obtaining a reference three-dimensional image reconstruction result based on the reference three-dimensional image data of the surgical area pixel mask; step S440, fusing the three-dimensional image reconstruction result with the reference three-dimensional image reconstruction result, and outputting a fused three-dimensional image reconstruction result; step S450, performing deformation analysis on the fused three-dimensional image reconstruction result, and outputting deformation features, including average displacement, peak displacement, angle change, local volume change, and deformation rate; step S460, calculating a risk index of the at least one follow-up three-dimensional image data based on the deformation features, and sending a risk warning message to the user based on the risk index.

[0045] Preferably, the baseline 3D image data of the pixel mask in the surgical area is reconstructed to obtain an analyzable baseline 3D model, which serves as the baseline 3D image reconstruction result. Then, the 3D image reconstruction result is fused with the baseline 3D image reconstruction result; that is, after spatial alignment, the two 3D models are superimposed, including mesh merging, point cloud registration, or voxel fusion in the same coordinate system, outputting the fused 3D image reconstruction result. Next, deformation quantitative analysis is performed on the fused 3D image reconstruction result, including quantitative analysis of average displacement, peak displacement, angle change, local volume change, and deformation rate. Specifically, in the fused and aligned model, vector difference operations are performed on corresponding points to determine the displacement field, and then various indicators are statistically analyzed to obtain the average displacement, peak displacement, angle change, local volume change, and deformation rate. The average displacement is the average displacement of all corresponding points of the follow-up model relative to the baseline model; the peak displacement is the value of the point or region with the largest displacement; the angle change refers to the rotational angle of the bone segment or implant; the local volume change refers to the volume increase or decrease of a specific region; and the deformation rate refers to the rate of deformation change per unit time.

[0046] Preferably, a risk scoring model is trained using logistic regression or decision trees and a large amount of case data. Deformed features are input into the risk scoring model for calculation, and a risk score or risk level is output as a risk indicator for at least one follow-up 3D imaging data session. This indicates the likelihood of the patient's condition worsening or complications occurring compared to the baseline 3D imaging. Finally, risk alerts are sent to users based on the risk indicators, such as through automatic pop-ups in the hospital system to doctors, to facilitate timely adjustments to treatment or scheduling of follow-up examinations. Self-monitoring awareness is also enhanced through SMS messages and health management mini-programs on the patient's end. The risk alert information includes the risk level, key indicator values, and corresponding recommended measures.

[0047] In the above text, refer to Figure 1 A three-dimensional image reconstruction method for postoperative monitoring in orthopedics according to embodiments of the present invention is described in detail. Next, reference will be made to... Figure 2 A three-dimensional image reconstruction system for postoperative monitoring in orthopedics is described according to an embodiment of the present invention.

[0048] The image three-dimensional reconstruction system for postoperative monitoring in orthopedic surgery according to embodiments of the present invention addresses the technical problems in the prior art, such as difficulties in analyzing dynamic changes in three-dimensional image reconstruction, insufficient stability of feature points, and inadequate data registration accuracy. It achieves high-precision registration of multi-temporal three-dimensional images, improves the consistency of three-dimensional reconstruction results at different follow-up time points, and enhances the accuracy of three-dimensional reconstruction. Figure 2 As shown, the image three-dimensional reconstruction system for postoperative monitoring in orthopedics includes: a three-dimensional image data acquisition module 10, a candidate base point analysis module 20, an anchor base point set identification module 30, and a three-dimensional reconstruction module 40.

[0049] The three-dimensional image data acquisition module 10 is used to locate the user's surgical area pixel mask and acquire the baseline three-dimensional image data of the surgical area pixel mask; the candidate base point analysis module 20 is used to perform candidate base point analysis on the baseline three-dimensional image data to obtain a candidate base point set, the candidate base points including geometric feature points on the implant, skeletal anatomical feature points, feature points in the stable area of ​​the soft tissue surface, and preset feature points on the user's anatomical structure mesh; the anchoring base point set identification module 30 is used to perform stability scoring on the candidate base point set, and based on the stability scoring results, identify the anchoring base point set greater than the preset stability scoring threshold from the candidate base point set, and select at least one reconstructed anchoring base point; the three-dimensional reconstruction module 40 is used to acquire at least one follow-up three-dimensional image data, and perform three-dimensional reconstruction on the at least one reconstructed anchoring base point to obtain the three-dimensional image reconstruction result of the surgical area pixel mask.

[0050] The specific configuration of the anchoring base point set identification module 30 will be described in detail below. The anchoring base point set identification module 30 further includes: acquiring stability training samples, which include multiple stability factors under different orthopedic postoperative samples, wherein the multiple stability factors include the temporal variance of the base point at different time points, the average confidence level of the data detection, the structural rigidity score of the base point location, and the spatial distance between the base point and the lesion; performing logistic regression training based on the stability training samples to obtain multiple learning weights corresponding to the multiple stability factors; fitting a stability scoring model using the multiple learning weights and the multiple stability factors; and using the stability scoring model to score the stability of the candidate base point set.

[0051] The specific configuration of the anchoring base point set identification module 30 will be described in detail below. The anchoring base point set identification module 30 further includes: randomly selecting initial reconstruction anchoring base points from the anchoring base point set; defining a multi-objective optimization function, which includes a first optimization objective and a second optimization objective, wherein the first optimization objective is to maximize the stability of the initial reconstruction anchoring base points, and the second optimization objective is to ensure that the spatial distribution uniformity of the initial reconstruction anchoring base points meets a preset uniformity threshold; and using the multi-objective optimization function to perform a greedy algorithm to optimize the initial reconstruction anchoring base points in the anchoring base point set to obtain at least one reconstruction anchoring base point.

[0052] The specific configuration of the anchoring base point set identification module 30 will be described in detail below. The anchoring base point set identification module 30 further includes: optimizing the initial reconstruction anchoring base points in the anchoring base point set using the multi-objective optimization function; wherein the candidate base point set has a preset priority sequence, and the priority sequence is arranged in descending order as follows: geometric feature points on the implant, skeletal anatomical feature points, feature points within the stable region of the soft tissue surface, and preset feature points on the user's anatomical structure mesh.

[0053] The specific configuration of the anchorage set identification module 30 will be described in detail below. The anchorage set identification module 30 further includes: detecting the reconstruction marker status data of the anchorage set, including the number of markers used for reconstruction and the marker frequency; and replacing the anchorage set with a failed one based on the reconstruction marker status data, wherein if the anchorage set is not marked for a preset number of consecutive times or the marker frequency is less than a preset threshold, the current anchorage set is determined to be failed.

[0054] The specific configuration of the 3D reconstruction module 40 will be described in detail below. The 3D reconstruction module 40 further includes: if the at least one reconstruction anchor point returns multiple points, inputting the at least one reconstruction anchor point into a rigid transformation solving algorithm to perform prior sparse 3D reconstruction on the at least one follow-up 3D image data to obtain a sparse 3D image reconstruction result; and performing posterior registration 3D reconstruction on the sparse 3D image reconstruction result to obtain a registered 3D image reconstruction result.

[0055] The specific configuration of the 3D reconstruction module 40 will be described in detail below. The 3D reconstruction module 40 further includes: obtaining a reference 3D image reconstruction result based on the reference 3D image data of the surgical area pixel mask; fusing the 3D image reconstruction result with the reference 3D image reconstruction result to output a fused 3D image reconstruction result; performing deformation analysis on the fused 3D image reconstruction result to output deformation characteristics, including average displacement, peak displacement, angle change, local volume change, and deformation rate; calculating a risk index for the at least one follow-up 3D image data based on the deformation characteristics, and sending a risk warning message to the user based on the risk index.

[0056] The image three-dimensional reconstruction system for postoperative monitoring in orthopedics provided in this embodiment of the invention can execute the image three-dimensional reconstruction method for postoperative monitoring in orthopedics provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the method.

[0057] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0058] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A three-dimensional image reconstruction method for postoperative monitoring in orthopedic surgery, characterized in that, The method includes: Locate the pixel mask of the user's surgical area and acquire the reference three-dimensional image data of the pixel mask of the surgical area; Candidate base point analysis is performed on the reference three-dimensional image data to obtain a set of candidate base points. The candidate base points include geometric feature points on the implant, skeletal anatomical feature points, feature points in the stable area of ​​the soft tissue surface, and preset feature points on the user's anatomical structure mesh. The candidate base point set is subjected to stability scoring. Based on the stability scoring results, the set of anchor base points with a value greater than a preset stability scoring threshold is identified from the candidate base point set, and at least one reconstructed anchor base point is selected. Acquire at least one follow-up three-dimensional image data, and perform three-dimensional reconstruction on the at least one follow-up three-dimensional image data based on the at least one reconstruction anchor point to obtain the three-dimensional image reconstruction result of the pixel mask of the surgical area; The method for performing stability scoring on the candidate base point set includes: Obtain stability training samples, which include multiple stability factors under different orthopedic postoperative samples. The multiple stability factors include the time variance of the base point at different time points, the average confidence level of the data detection, the structural rigidity score of the base point location, and the spatial distance between the base point and the lesion. Logistic regression training is performed based on the stability training samples to obtain multiple learning weights corresponding to the multiple stability factors. The multiple learning weights and the multiple stability factors are used to fit a stability scoring model, and the stability scoring model is used to score the stability of the candidate base point set.

2. The three-dimensional image reconstruction method for postoperative monitoring in orthopedics as described in claim 1, characterized in that, Based on the stability score results, identify a set of anchor points from the candidate anchor point set that are greater than a preset stability score threshold, and select at least one reconstructed anchor point. The method includes: Randomly select initialization and reconstruction anchor points from the set of anchor points; Define a multi-objective optimization function, which includes a first optimization objective and a second optimization objective. The first optimization objective is to maximize the stability of the initial reconstructed anchor points, and the second optimization objective is to ensure that the spatial distribution uniformity of the initial reconstructed anchor points meets a preset uniformity threshold. The initialized reconstructed anchoring base points are optimized using a greedy algorithm within the set of anchoring base points through the multi-objective optimization function, resulting in at least one reconstructed anchoring base point.

3. The three-dimensional image reconstruction method for postoperative monitoring in orthopedics as described in claim 2, characterized in that, The initialization and reconstruction anchor points are optimized in the set of anchor points using the multi-objective optimization function. The candidate base point set is pre-set with a priority sequence, which is arranged in descending order as follows: geometric feature points on the implant, skeletal anatomical feature points, feature points in the stable area of ​​the soft tissue surface, and pre-set feature points on the user's anatomical structure grid.

4. The three-dimensional image reconstruction method for postoperative monitoring in orthopedics as described in claim 1, characterized in that, The method further includes identifying a set of anchor points with values ​​greater than a preset stability score threshold from the candidate anchor point set based on the stability score results. The reconstructed marker status data of the anchoring base point set is detected, including the number of markers used for reconstruction and the marker frequency; The set of anchor points is replaced with invalid ones based on the reconstructed marker status data. If a preset number of consecutive times no marker is used or the marker frequency is less than a preset threshold, the current anchor point is determined to be invalid.

5. The three-dimensional image reconstruction method for postoperative monitoring in orthopedics as described in claim 1, characterized in that, The method for performing three-dimensional reconstruction of the at least one follow-up three-dimensional image data based on the at least one reconstruction anchor point includes: If the at least one reconstruction anchor point returns multiple points, the at least one reconstruction anchor point is input into the rigid transformation solution algorithm to perform prior sparse 3D reconstruction on the at least one follow-up 3D image data to obtain the sparse 3D image reconstruction result. The sparse 3D image reconstruction results are subjected to posterior registration 3D reconstruction to obtain the registered 3D image reconstruction results.

6. The three-dimensional image reconstruction method for postoperative monitoring in orthopedics as described in claim 1, characterized in that, After obtaining the three-dimensional image reconstruction result of the pixel mask of the surgical area, the method further includes: Based on the reference three-dimensional image data of the pixel mask of the surgical area, obtain the reference three-dimensional image reconstruction result; The three-dimensional image reconstruction result is fused with the baseline three-dimensional image reconstruction result to output the fused three-dimensional image reconstruction result; The fused 3D image reconstruction results are subjected to deformation analysis, and deformation features are output, including average displacement, peak displacement, angle change, local volume change and deformation rate. Based on the deformable features, a risk index is calculated for the at least one follow-up 3D image data, and a risk alert is sent to the user based on the risk index.

7. A three-dimensional image reconstruction system for postoperative monitoring in orthopedics, characterized in that, The system is used to implement the three-dimensional image reconstruction method for postoperative monitoring in orthopedics as described in any one of claims 1 to 6, the system comprising: The three-dimensional image data acquisition module is used to locate the pixel mask of the user's surgical area and acquire the reference three-dimensional image data of the pixel mask of the surgical area. The candidate base point analysis module is used to perform candidate base point analysis on the benchmark three-dimensional image data to obtain a set of candidate base points. The candidate base points include geometric feature points on the implant, skeletal anatomical feature points, feature points in the stable area of ​​the soft tissue surface, and preset feature points on the user's anatomical structure mesh. An anchoring base point set identification module is used to perform stability scoring on the candidate base point set, identify anchoring base points with a value greater than a preset stability score threshold from the candidate base point set based on the stability score result, and select at least one reconstructed anchoring base point. The three-dimensional reconstruction module is used to acquire at least one follow-up three-dimensional image data, and to perform three-dimensional reconstruction on the at least one follow-up three-dimensional image data based on the at least one reconstruction anchor point to obtain the three-dimensional image reconstruction result of the pixel mask of the surgical area.

Images