Positioning assisting method and device for femoral neck fracture reduction and electronic equipment

Abstract

The invention provides a positioning auxiliary method and device for femoral neck fracture reduction and electronic equipment, and relates to the technical field of medical treatment, and the method comprises the steps: obtaining three-dimensional image data of a femoral neck fracture part of a patient, and carrying out the image preprocessing of the three-dimensional image data, and obtaining first point cloud data after noise removal; performing feature extraction processing on the first point cloud data through a preset point cloud data segmentation network, determining a local feature corresponding to each pixel point, and determining a target implantation path of the internal fixture according to the local features, the fracture type and the fracture position; second point cloud data of the surgical area in the surgical process are obtained in real time, point cloud registration processing is conducted on the second point cloud data, the second point cloud data coincide with the first point cloud data, the internal fixture is implanted along the target implantation path, and femoral neck fracture reduction is completed. The femoral neck fracture reduction difficulty can be remarkably reduced, the reduction accuracy is improved, and therefore injuries caused in the reduction process are reduced.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a positioning auxiliary method, device and electronic equipment for femoral neck fracture reduction. Background Art

[0002] Femoral neck fracture is a common disease among hip fractures. Because the angle between the fracture line and the horizontal line is greater than 70 degrees, it has a large shear force and instability, which makes the treatment of this type of fracture difficult and has a high failure rate of reduction and internal fixation.

[0003] At present, relevant technologies have proposed that appropriate internal fixators can be selected according to the type and location of the fracture, and the internal fixators can be implanted into the fracture ends to provide sufficient fixation strength. The installation position of the internal fixators and the fracture alignment can be checked through X-ray fluoroscopy or intraoperative C-arm to ensure that the fracture ends are aligned and firmly fixed.

[0004] However, since this approach requires C-arm fluoroscopy, percutaneous puncture, manual positioning, or small incision exposure and positioning at the greater trochanter of the femur during the reduction process, it requires a high level of experience and is difficult to reduce, and repeated puncture fluoroscopy is required during the operation to adjust the entry point and direction of the screw, which can lead to increased radiation damage, local bone destruction, weak screw holding force, and even iatrogenic fractures. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a positioning auxiliary method, device and electronic equipment for femoral neck fracture reduction, which can significantly reduce the difficulty of femoral neck fracture reduction and improve the reduction accuracy, thereby reducing the damage caused during the reduction process.

[0006] In the first aspect, an embodiment of the present invention provides a positioning assistance method for femoral neck fracture reduction, the method comprising: acquiring three-dimensional image data of a patient's femoral neck fracture site, and performing image preprocessing on the three-dimensional image data to obtain first point cloud data after noise is removed; performing feature extraction processing on the first point cloud data through a preset point cloud data segmentation network to determine the local features corresponding to each pixel point, and determining the target implantation path of the internal fixation device based on the local features, fracture type and fracture location; acquiring second point cloud data of the surgical area during the operation in real time, and performing point cloud registration processing on the second point cloud data to make the second point cloud data coincide with the first point cloud data, so that the internal fixation device is implanted along the target implantation path to complete the femoral neck fracture reduction.

[0007] In one embodiment, the step of performing image preprocessing on three-dimensional image data to obtain first point cloud data after noise removal includes: performing image segmentation processing on the three-dimensional image data using a preset threshold segmentation algorithm to convert the three-dimensional image data into a three-dimensional mesh model, wherein the three-dimensional mesh model is composed of multiple triangular facets; performing denoising processing on point cloud data composed of triangle vertices of each triangular facet in the three-dimensional mesh model to determine the first point cloud data after noise removal.

[0008] In one embodiment, a preset point cloud data segmentation network includes: a sampling layer, a grouping layer and an extraction layer. The preset point cloud data segmentation network is used to perform feature extraction processing on the first point cloud data, and the step of determining the local features corresponding to each pixel point includes: downsampling the first point cloud data through the sampling layer, and dividing the downsampled point cloud data into a set of local areas through the grouping layer, wherein the downsampling includes: farthest point sampling and random sampling, and the set of local areas includes: multiple local areas, and the point clouds in each local area share similar local features; feature extraction processing is performed on each local area through the extraction layer to determine a target local feature set, wherein each target local feature includes: geometric distribution of local points, local surface shape, local density and local normal vector.

[0009] In one embodiment, the step of determining a target implantation path of an internal fixation device based on local features, fracture type, and fracture location includes: performing feature fusion processing on local features of each target item through a preset multi-layer perceptron and a preset symmetry function to determine global features; performing classification and identification processing on global features through a preset point cloud data segmentation network to determine the fracture type and fracture location, and planning a target implantation path of an internal fixation device based on the global features, fracture type, and fracture location.

[0010] In one embodiment, the step of performing feature fusion processing on local features of each target item by using a preset multilayer perceptron and a preset symmetric function to determine the global features includes: performing nonlinear transformation processing on features corresponding to points in the local features of each target item by using a preset multilayer perceptron to extract point-by-point features; aggregating the point-by-point features by using a preset symmetric function to aggregate them into a feature vector of a fixed dimension to determine the global features, wherein the preset symmetric function includes: maximum pooling and average pooling.

[0011] In one embodiment, the second point cloud data is subjected to point cloud registration processing so that the second point cloud data overlaps with the first point cloud data so that the internal fixation device is implanted along the target implantation path to complete the step of femoral neck fracture reduction, including: dividing the first point cloud data into a plurality of voxels of the same size, and obtaining a normal distribution model corresponding to each voxel; performing rough registration on the second point cloud data to determine an initial transformation matrix; calculating the log-likelihood value between each point in the second point cloud data and the normal distribution of the same voxel in the first point cloud data through a preset iterative optimization algorithm, and determining the target transformation matrix corresponding to the maximum log-likelihood value by performing parameter update processing on the initial transformation matrix; and registering the second point cloud data using the target transformation matrix to track the bone position and the target implantation path in real time, so that the internal fixation device is implanted along the target implantation path to complete the femoral neck fracture reduction.

[0012] In one embodiment, the step of obtaining the normal distribution model corresponding to each voxel includes: calculating the normal distribution parameters corresponding to the points in each voxel respectively, wherein the normal distribution parameters include: mean and covariance matrix, the mean is the coordinate mean of all points in the voxel, and the covariance matrix is ​​the coordinate covariance of all points in the voxel; according to the mean and covariance matrix corresponding to each point in any voxel, the normal distribution model of the voxel is determined.

[0013] In a second aspect, an embodiment of the present invention further provides a positioning auxiliary device for femoral neck fracture reduction, the device comprising: a data preprocessing module, which acquires three-dimensional image data of a patient's femoral neck fracture site, and performs image preprocessing on the three-dimensional image data to obtain first point cloud data after noise removal; a path planning module, which performs feature extraction processing on the first point cloud data through a preset point cloud data segmentation network, determines the local features corresponding to each pixel point, and determines the target implantation path of the internal fixation device based on the local features, fracture type and fracture location; a registration and positioning module, which acquires second point cloud data of the surgical area during the operation in real time, and performs point cloud registration processing on the second point cloud data to make the second point cloud data coincide with the first point cloud data, so that the internal fixation device can be implanted along the target implantation path to complete the femoral neck fracture reduction.

[0014] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising a processor and a memory, wherein the memory stores computer-executable instructions that can be executed by the processor, and the processor executes the computer-executable instructions to implement any one of the methods provided in the first aspect.

[0015] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement any one of the methods provided in the first aspect.

[0016] The embodiments of the present invention bring the following beneficial effects:

[0017] The embodiments of the present invention provide a positioning auxiliary method, device and electronic device for femoral neck fracture reduction. The method first obtains three-dimensional image data of the femoral neck fracture site of the patient, performs image preprocessing on the three-dimensional image data to obtain first point cloud data after noise is removed, then performs feature extraction processing on the first point cloud data through a preset point cloud data segmentation network to determine the local features corresponding to each pixel point, and determines the target implantation path of the internal fixation device according to the local features, fracture type and fracture location, and finally obtains second point cloud data of the surgical area during the operation in real time, and performs point cloud registration processing on the second point cloud data to make the second point cloud data coincide with the first point cloud data, so that the internal fixation device is implanted along the target implantation path to complete the femoral neck fracture reduction. The embodiments of the present invention can significantly reduce the difficulty of femoral neck fracture reduction and improve the reduction accuracy, thereby reducing the damage caused during the reduction process.

[0018] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0021] Figure 1 A schematic flow chart of a positioning auxiliary method for femoral neck fracture reduction provided by an embodiment of the present invention;

[0022] Figure 2 A schematic flow chart of another positioning auxiliary method for femoral neck fracture reduction provided by an embodiment of the present invention;

[0023] Figure 3 A schematic structural diagram of a positioning auxiliary device for femoral neck fracture reduction provided by an embodiment of the present invention;

[0024] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described in combination with the embodiments below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0026] At present, femoral neck fracture is a common disease among hip fractures. Since the angle between the fracture line and the horizontal line is greater than 70 degrees, it has a large shear force and instability, which makes the treatment of this type of fracture difficult. The failure rate of reduction and internal fixation is high. Patients undergoing internal fixation generally take the supine or lateral position, which is adjusted according to the needs of the operation, and the incision position is selected according to the fracture site and the surgical plan. Common incision positions are the lateral side of the femur or the anterior and lateral side of the hip joint.

[0027] Related technologies suggest that appropriate internal fixators (such as hollow screws, dynamic hip screws, intramedullary nails, etc.) can be selected according to the type and location of the fracture, and the internal fixators can be implanted into the fracture ends. For Pauwels type III fractures, multiple hollow screws or dynamic hip screws are usually used to provide sufficient fixation strength. The installation position of the internal fixators and the alignment of the fracture are checked by X-ray fluoroscopy or intraoperative C-arm to ensure that the fracture ends are aligned and firmly fixed.

[0028] However, since the femoral neck is located deep in the hip joint, the surgical operating space is limited. There are important blood vessels (such as the internal and external femoral circumflex arteries) and nerves near the femoral neck. The femoral neck has a rich blood supply. During surgery, it is necessary to avoid damaging the blood vessels to prevent fracture healing. In addition, special care must be taken during surgery to avoid injury. In addition, in Pauwels type III fractures, the angle between the fracture line and the horizontal line is large, making reduction difficult.

[0029] At the same time, this solution requires the use of C-arm fluoroscopy, percutaneous puncture, manual positioning or small incision exposure and positioning in the greater trochanter of the femur during the reduction process. This requires a high level of surgical experience for the doctor and makes reduction difficult. In addition, repeated puncture and fluoroscopy are required during the operation to adjust the entry point and direction of the screw, which can lead to increased radiation damage, local bone destruction, weak screw holding force and even iatrogenic fractures. Based on this, the positioning auxiliary method for femoral neck fracture reduction provided by the implementation of the present invention can significantly reduce the difficulty of femoral neck fracture reduction and improve the reduction accuracy, thereby reducing the damage caused during the reduction process.

[0030] See also Figure 1 The flowchart of a positioning auxiliary method for femoral neck fracture reduction is shown in FIG. 1 , and the method mainly includes the following steps S102 to S106:

[0031] Step S102, obtaining three-dimensional image data of the femoral neck fracture site of the patient, and performing image preprocessing on the three-dimensional image data to obtain the first point cloud data after noise is removed. In one embodiment, CT or MRI scanning can be used to obtain the three-dimensional image data of the fracture site of the patient, and the image data is converted into a three-dimensional mesh model through algorithms such as threshold segmentation. The triangular vertices of the mesh model constitute the point cloud data, and denoising, downsampling and alignment processing are performed to improve the consistency of data quality, wherein the three-dimensional mesh model is composed of multiple triangular facets.

[0032] Step S104, performing feature extraction processing on the first point cloud data through a preset point cloud data segmentation network, determining the local features corresponding to each pixel point, and determining the target implantation path of the internal fixation device according to the local features, fracture type and fracture location, wherein the preset point cloud data segmentation network is PointNet++, PointNet++ is a deep neural network for classifying and segmenting irregularly shaped point cloud data, and the preset point cloud data segmentation network includes: a sampling layer, a grouping layer and an extraction layer. In one embodiment, the sampling layer uses a certain sampling method (such as farthest point sampling, random sampling, etc.) to downsample the input point cloud data, and the grouping layer divides the downsampled point cloud data into several local areas, and the points in each area share similar local features. The extraction layer uses PointNet as a local feature extractor to extract features from each area.

[0033] Step S106, real-time acquisition of second point cloud data of the surgical area during the operation, and point cloud registration processing of the second point cloud data, so that the second point cloud data coincides with the first point cloud data, so that the internal fixation device is implanted along the target implantation path, and the femoral neck fracture reduction is completed. In one embodiment, since the registered point cloud data contains the position and angle of the surgical needle insertion planned before the operation, the doctor can use PointNet++ to process and register the point cloud data acquired in real time, and help track the bone position and the planned path position in real time. Therefore, PointNet++ is applied to femoral neck fracture internal fixation, which can improve the accuracy and success rate of the operation, reduce the amount of surgical radiation, provide more bone mass and blood supply protection, and optimize postoperative evaluation and rehabilitation management.

[0034] The positioning auxiliary method for femoral neck fracture reduction provided by the embodiment of the present invention can significantly reduce the difficulty of femoral neck fracture reduction and improve the reduction accuracy, thereby reducing the damage caused during the reduction process.

[0035] The embodiment of the present invention also provides an implementation method of performing image segmentation processing on three-dimensional image data using a preset threshold segmentation algorithm, converting the three-dimensional image data into a three-dimensional mesh model, and performing denoising processing on point cloud data composed of triangle vertices of various triangular facets in the three-dimensional mesh model to determine the first point cloud data after the noise is removed. Specifically, the noise points in the first point cloud data set usually refer to points that are far away from surrounding points or have significantly different statistical characteristics from other points. The noise is regarded as an outlier and a noise point. The noise point cloud can be expressed as:

[0036]

[0037] Where P' is a noise point cloud, where each point is p' i ; P is the real (clean) point cloud, where each point is p i ; Ο is a set of outliers, where each point is o j ;n i is added to each point p i The interference on p i For vectors of the same dimension, the denoising process is actually to learn the noise offset vector of each point from the noise point cloud and move each point to the appropriate position. The process can be expressed as:

[0038]

[0039] in, is the output point cloud, d i is the learned offset vector for each point, The set of outliers predicted by the model, The noisy point cloud after removing outliers is also recorded as For each point p′ i ∈P', which only comes from the local neighborhood P of the point i ′, this neighborhood is based on p′ i A sphere with centered at r and radius r. By adjusting the radius, we can adjust whether the denoising result depends more on local information or global information.

[0040] In addition, the denoising task can be divided into two subtasks: one is to eliminate outliers, and the other is to learn the offset vector. In the process of eliminating outliers, each point is first calculated as an outlier p′ i The probability of the whole process can be expressed by function g:

[0041]

[0042] in, Representative point p i The set of points in the r neighborhood of ; Represents point p′ i The probability of belonging to an outlier point. After removing the outliers, the offset is estimated for each remaining point to correct their position and estimate each point p i The offset process is called function f:

[0043]

[0044] in, Indicates that at point p i The point cloud after removing outliers in the r neighborhood; d i is for point p i The estimated offset is obtained by comparing it with the offset d i The final result for each point p can be obtained by adding i The estimated result is p i The outliers are no longer included, p i It can be expressed as:

[0045]

[0046] in is the point after removing outliers; is the final point p i The prediction results of all The set of points is the point cloud after denoising, that is, the first point cloud data.

[0047] The embodiment of the present invention also provides an implementation method for determining a target implantation path, and specifically, see the following steps (A) to (D):

[0048] (A) Downsampling the first point cloud data through the sampling layer, and dividing the downsampled point cloud data into a set of local regions through the grouping layer, wherein the downsampling includes: farthest point sampling and random sampling, and the local region set includes: multiple local regions, and the point clouds in each local region share similar local features, that is, the local features in the grouping layer refer to the common features of an internal region, and the local features of the points in the internal region are the same and shared, and the extraction layer calculates the specific parameters of these features and distinguishes them according to different types.

[0049] (B) Performing feature extraction processing on each local area through the extraction layer to determine the target local feature set, wherein each target local feature includes: geometric distribution of local points, local surface shape, local density and local normal vector. For details, see (1) to (4) below:

[0050] (1) Geometric distribution of local points: Through the PointNet layer, the geometric distribution features of points in the local area can be extracted. These features can reflect the relative position relationship and spatial structure between points.

[0051] (2) Local surface shape: By extracting features from points in the local area, the local shape information of the point cloud surface can be captured, such as planes, curved surfaces, edges, etc.

[0052] (3) Local density: The density feature of the local point cloud can reflect the distribution of points in the local area, such as dense areas and sparse areas.

[0053] (4) Local normal vectors: If the normal vectors of the point cloud are calculated in the preprocessing stage, these normal vectors can be input into PointNet++ as additional features to further enhance the model's ability to capture local geometric features.

[0054] (C) By using a preset multilayer perceptron and a preset symmetric function, feature fusion processing is performed on the local features of each target to determine the global features. Specifically, by using a preset multilayer perceptron, nonlinear transformation processing is performed on the features corresponding to the points in the local features of each target to extract point-by-point features, and the point-by-point features are aggregated by using a preset symmetric function to aggregate them into a feature vector of a fixed dimension to determine the global features. The preset symmetric function includes: maximum pooling and average pooling. A multilayer perceptron (MLP) is a feedforward artificial neural network model, which is used to map multiple input data sets to a single output data set.

[0055] In one embodiment, a multilayer perceptron is used to extract and process local features in a point cloud layer by layer. In each local region, the features of the points are nonlinearly transformed by MLP, and a symmetric function is used to aggregate the point features in the local region to ensure that the aggregation process is independent of the order of the points. Commonly used symmetric functions include maximum pooling and average pooling, that is, finding the maximum value and the average value. The fusion process of local features generally includes the following steps (a) to (c):

[0056] (a) Local region division: The point cloud is sampled and grouped to obtain multiple local regions.

[0057] (b) Point-wise feature extraction: MLP is applied to the points in each local region to extract point-wise features.

[0058] (c) Feature aggregation: Apply a symmetric function (such as maximum pooling) to the point features in each local area, and aggregate the point-by-point features into a feature vector of fixed dimension. Finally, the classification / segmentation layer outputs the global features or the local features of each point. It is used to segment and classify point cloud data.

[0059] (D) Through the preset point cloud data segmentation network, the global features are classified and identified to determine the fracture type and fracture location. Based on the global features, fracture type and fracture location, the target implantation path of the internal fixation device is planned. Specifically, based on the PointNet++ model, a certain amount of labeled data training can be used to enable it to accurately identify and classify different types of femoral neck fractures. According to the features and classification results output by the model, the fracture ends are accurately aligned. According to the fracture type and location, combined with the three-dimensional analysis results of the PointNet model, the optimal implantation position and angle of the internal fixation device (such as hollow nails and dynamic hip screws) are planned.

[0060] In one embodiment, a laser scanner or a camera with depth information can be used during surgery to collect point cloud data of the patient's surgical area to collect second point cloud data. The first point cloud data of the lesion area extracted from CT data before surgery contains position and direction information of the surgical plan. During the operation, the first point cloud data can be divided into multiple voxels of the same size, and the normal distribution model corresponding to each voxel can be obtained. The second point cloud data is roughly aligned to determine the initial transformation matrix. Then, a preset iterative optimization algorithm is used to calculate the log-likelihood value between each point in the second point cloud data and the normal distribution of the same voxel in the first point cloud data. The initial transformation matrix is ​​updated with parameters to determine the target transformation matrix corresponding to the maximum log-likelihood value. In actual applications, due to software and hardware system errors, the real position cannot be aligned. Therefore, the above scheme is used to perform a rough match first to match to a close position, and then continuous fine-tuning is performed to reach the final error range.

[0061] Furthermore, the target transformation matrix is ​​used to register the second point cloud data, that is, the second point cloud data and the first point cloud data are made to coincide with each other through NDT point cloud registration, so that accurate surgical operations can be performed according to the planned path, so as to track the bone position and the target implantation path in real time, so that the internal fixation device can be implanted along the target implantation path to complete the reduction of the femoral neck fracture, thereby combining the intraoperative point cloud matching and the real-time processing capability of the PointNet++ model to assist surgeons in accurate fracture reduction and internal fixation device implantation. Among them, nondestructive testing (NDT), also known as nondestructive flaw detection, refers to the use of changes in physical quantities such as heat, sound, light, electricity, and magnetism caused by abnormal internal structure or defects of the material without damaging the object being tested, to detect internal and surface defects of various engineering materials, parts, structural parts, etc. In practical applications, due to software and hardware system errors, the real position cannot be registered, so the above scheme is used to roughly match the close position first, and then the final error range is reached through continuous fine-tuning.

[0062] In addition, when determining the normal distribution equation of the voxel, the normal distribution parameters corresponding to the points in each voxel can be calculated respectively, and the normal distribution model of the voxel can be determined according to the mean and covariance matrix corresponding to each point in any voxel. The normal distribution parameters include: mean and covariance matrix, the mean is the coordinate mean of all points in the voxel, and the covariance matrix is ​​the coordinate covariance of all points in the voxel. For details, see Figure 2 The flowchart of another positioning auxiliary method for femoral neck fracture reduction is shown, wherein the NDT point cloud registration is used to make the point cloud data of the lesion area extracted by CT data coincide with the point cloud data collected during the operation, so as to perform accurate surgical operations according to the planned path, including the following steps S202 to S214:

[0063] Step S202, point cloud data preprocessing, includes: preprocessing the input point cloud data, including noise reduction, downsampling and other operations, to reduce the amount of calculation and improve the stability of the registration.

[0064] Step S204, dividing voxels (Voxel), includes: dividing the point cloud into voxels of equal size, and the points in each voxel will be used to calculate the normal distribution parameters (mean and covariance matrix) of the voxel.

[0065] Step S206, calculating the normal distribution model, including: calculating the mean and covariance matrix of the points in each voxel, thereby obtaining the normal distribution model of the points in the voxel, wherein the mean is the coordinate mean of all points in the voxel, and the covariance matrix is ​​the coordinate covariance of all points in the voxel.

[0066] Step S208, initial transformation estimation, includes: obtaining an initial transformation matrix according to the initial estimation or by rough registration, and roughly aligning the source point cloud to the target point cloud.

[0067] Step S210, iterative optimization, includes: using an iterative optimization algorithm (such as gradient descent) to maximize the similarity measure between the point cloud and the normal distribution, and in each iteration, updating the transformation parameters (translation and rotation) of the point cloud by calculating the log-likelihood value of the point in the target point cloud in the normal distribution of the reference point cloud, wherein calculating the similarity is calculating the log-likelihood value of the normal distribution of each point in the source point cloud and the corresponding voxel in the target point cloud, and updating the transformation parameters is updating the transformation matrix through the optimization algorithm to maximize the similarity.

[0068] Step S212, judging convergence, includes: checking whether the optimization process has converged, usually by setting an upper limit on the number of iterations or a similarity increment threshold to judge whether convergence has occurred.

[0069] Step S214, outputting the results, including: outputting the final transformation matrix and the registered point cloud.

[0070] For the positioning auxiliary method for femoral neck fracture reduction provided in the above embodiment, the embodiment of the present invention provides a positioning auxiliary device for femoral neck fracture reduction, see Figure 3 The schematic diagram of the structure of a positioning auxiliary device for femoral neck fracture reduction is shown, and the device includes the following parts:

[0071] The data preprocessing module 302 acquires the three-dimensional image data of the femoral neck fracture site of the patient, and performs image preprocessing on the three-dimensional image data to obtain the first point cloud data after noise removal;

[0072] The path planning module 304 performs feature extraction processing on the first point cloud data through a preset point cloud data segmentation network, determines the local features corresponding to each pixel point, and determines the target implantation path of the internal fixation device according to the local features, fracture type and fracture location;

[0073] The registration and positioning module 306 acquires the second point cloud data of the surgical area in real time during the operation, and performs point cloud registration processing on the second point cloud data to make the second point cloud data coincide with the first point cloud data, so that the internal fixation device is implanted along the target implantation path to complete the reduction of the femoral neck fracture.

[0074] The positioning auxiliary device for femoral neck fracture reduction provided in the embodiment of the present application can significantly reduce the difficulty of femoral neck fracture reduction and improve the reduction accuracy, thereby reducing the damage caused during the reduction process.

[0075] In one embodiment, when performing image preprocessing on three-dimensional image data to obtain the first point cloud data after noise removal, the above-mentioned data preprocessing module 302 is also used to: perform image segmentation processing on the three-dimensional image data using a preset threshold segmentation algorithm, and convert the three-dimensional image data into a three-dimensional mesh model, wherein the three-dimensional mesh model is composed of multiple triangular facets; perform denoising processing on the point cloud data composed of triangle vertices of each triangular facet in the three-dimensional mesh model, and determine the first point cloud data after noise removal.

[0076] In one embodiment, the preset point cloud data segmentation network includes: a sampling layer, a grouping layer and an extraction layer. When performing feature extraction processing on the first point cloud data through the preset point cloud data segmentation network to determine the local features corresponding to each pixel point, the above-mentioned path planning module 304 is also used to: downsample the first point cloud data through the sampling layer, and divide the downsampled point cloud data into a set of local areas through the grouping layer, wherein the downsampling includes: farthest point sampling and random sampling, and the local area set includes: multiple local areas, and the point clouds in each local area share similar local features; perform feature extraction processing on each local area respectively through the extraction layer to determine a target local feature set, wherein each target local feature includes: the geometric distribution of local points, the local surface shape, the local density and the local normal vector.

[0077] In one embodiment, when determining the target implantation path of the internal fixation device according to the local features, fracture type and fracture location, the path planning module 304 is further used to: perform feature fusion processing on the local features of each target through a preset multi-layer perceptron and a preset symmetry function to determine the global features; perform classification and identification processing on the global features through a preset point cloud data segmentation network to determine the fracture type and fracture location, and plan the target implantation path of the internal fixation device according to the global features, fracture type and fracture location.

[0078] In one embodiment, when performing feature fusion processing on local features of each target through a preset multi-layer perceptron and a preset symmetric function to determine the global features, the above-mentioned path planning module 304 is also used to: perform nonlinear transformation processing on features corresponding to points in the local features of each target through a preset multi-layer perceptron to extract point-by-point features; aggregate the point-by-point features through a preset symmetric function to aggregate them into a feature vector of a fixed dimension to determine the global features, wherein the preset symmetric function includes: maximum pooling and average pooling.

[0079] In one embodiment, when performing point cloud registration processing on the second point cloud data so that the second point cloud data overlaps with the first point cloud data so that the internal fixation device is implanted along the target implantation path to complete the step of femoral neck fracture reduction, the above-mentioned registration and positioning module 306 is also used to: divide the first point cloud data into multiple voxels of the same size, and obtain the normal distribution model corresponding to each voxel; perform rough registration on the second point cloud data to determine the initial transformation matrix; calculate the logarithmic likelihood value between each point in the second point cloud data and the normal distribution of the same voxel in the first point cloud data through a preset iterative optimization algorithm, and determine the target transformation matrix corresponding to the maximum logarithmic likelihood value by performing parameter update processing on the initial transformation matrix; register the second point cloud data using the target transformation matrix to track the bone position and the target implantation path in real time, so that the internal fixation device is implanted along the target implantation path to complete the femoral neck fracture reduction.

[0080] In one embodiment, when performing the step of obtaining the normal distribution model corresponding to each voxel, the above-mentioned registration and positioning module 306 is also used to: calculate the normal distribution parameters corresponding to the points in each voxel respectively, wherein the normal distribution parameters include: mean and covariance matrix, the mean is the coordinate mean of all points in the voxel, and the covariance matrix is ​​the coordinate covariance of all points in the voxel; determine the normal distribution model of any voxel based on the mean and covariance matrix corresponding to each point in any voxel.

[0081] The device provided in the embodiment of the present invention has the same implementation principle and technical effects as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0082] An embodiment of the present invention provides an electronic device. Specifically, the electronic device includes a processor and a storage device. The storage device stores a computer program, and when the computer program is executed by the processor, it executes the method described in any one of the above-mentioned implementation methods.

[0083] Figure 4 A structural diagram of an electronic device provided in an embodiment of the present invention, the electronic device 100 includes: a processor 40, a memory 41, a bus 42 and a communication interface 43, wherein the processor 40, the communication interface 43 and the memory 41 are connected via the bus 42; the processor 40 is used to execute an executable module stored in the memory 41, such as a computer program.

[0084] The memory 41 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 43 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0085] The bus 42 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0086] Among them, the memory 41 is used to store programs, and the processor 40 executes the program after receiving the execution instruction. The method executed by the device for flow process definition disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 40 or implemented by the processor 40.

[0087] The processor 40 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 40. The above processor 40 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor to execute, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 41, and the processor 40 reads the information in the memory 41 and completes the steps of the above method in combination with its hardware.

[0088] The computer program product of the readable storage medium provided in the embodiment of the present invention includes a computer-readable storage medium storing program code, and the instructions included in the program code can be used to execute the methods described in the previous method embodiments. The specific implementation can be referred to the previous method embodiments, which will not be repeated here.

[0089] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, electronic device, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0090] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A positioning auxiliary method for femoral neck fracture reduction, characterized in that: The method comprises: Acquire three-dimensional image data of the femoral neck fracture site of the patient, and perform image preprocessing on the three-dimensional image data to obtain first point cloud data after noise removal; Performing feature extraction processing on the first point cloud data through a preset point cloud data segmentation network to determine the local features corresponding to each pixel point, and determining a target implantation path of the internal fixation device according to the local features, the fracture type and the fracture location; The second point cloud data of the surgical area during the operation is acquired in real time, and the second point cloud data is processed by point cloud registration so that the second point cloud data coincides with the first point cloud data, so that the internal fixation device is implanted along the target implantation path to complete the reduction of the femoral neck fracture.

2. The positioning auxiliary method for femoral neck fracture reduction according to claim 1, characterized in that: The step of performing image preprocessing on the three-dimensional image data to obtain first point cloud data after noise removal includes: Performing image segmentation processing on the three-dimensional image data using a preset threshold segmentation algorithm, and converting the three-dimensional image data into a three-dimensional mesh model, wherein the three-dimensional mesh model is composed of a plurality of triangular facets; De-noising is performed on the point cloud data composed of triangle vertices of each of the triangular facets in the three-dimensional mesh model to determine the first point cloud data after the noise is removed.

3. The positioning auxiliary method for femoral neck fracture reduction according to claim 1, characterized in that: The preset point cloud data segmentation network includes: a sampling layer, a grouping layer and an extraction layer. The step of performing feature extraction processing on the first point cloud data through the preset point cloud data segmentation network to determine the local features corresponding to each pixel point includes: Downsampling the first point cloud data through the sampling layer, and dividing the downsampled point cloud data into a set of local regions through the grouping layer, wherein the downsampling includes: farthest point sampling and random sampling, and the set of local regions includes: a plurality of local regions, and point clouds in each of the local regions share similar local features; The extraction layer performs feature extraction processing on each of the local areas to determine a target local feature set, wherein each of the target local features includes: geometric distribution of local points, local surface shape, local density and local normal vector.

4. The positioning auxiliary method for femoral neck fracture reduction according to claim 1, characterized in that: The step of determining a target implantation path of the internal fixation device according to the local characteristics, fracture type and fracture location comprises: By presetting a multi-layer perceptron and a pre-set symmetric function, the local features of each target are fused to determine the global features; The preset point cloud data segmentation network is used to classify and identify the global features, determine the fracture type and fracture location, and plan the target implantation path of the internal fixation device based on the global features, the fracture type and the fracture location.

5. The positioning auxiliary method for femoral neck fracture reduction according to claim 4, characterized in that: The step of performing feature fusion processing on the local features of each target by using a preset multi-layer perceptron and a preset symmetric function to determine the global features includes: Through the preset multi-layer perceptron, nonlinear transformation processing is performed on the features corresponding to the points in each of the target local features to extract point-by-point features; The point-by-point features are aggregated through the preset symmetric function to form a feature vector of a fixed dimension, and the global feature is determined, wherein the preset symmetric function includes: maximum pooling and average pooling.

6. The positioning auxiliary method for femoral neck fracture reduction according to claim 1, characterized in that: The step of performing point cloud registration processing on the second point cloud data so that the second point cloud data overlaps with the first point cloud data, so that the internal fixation device is implanted along the target implantation path to complete the femoral neck fracture reduction includes: Dividing the first point cloud data into a plurality of voxels of the same size, and obtaining a normal distribution model corresponding to each voxel; Performing rough registration on the second point cloud data to determine an initial transformation matrix; Calculating the log-likelihood value between each point in the second point cloud data and the normal distribution of the same voxel in the first point cloud data by using a preset iterative optimization algorithm, and determining the target transformation matrix corresponding to the maximum log-likelihood value by performing parameter updating processing on the initial transformation matrix; The target transformation matrix is ​​used to register the second point cloud data to track the bone position and the target implantation path in real time, so that the internal fixation device is implanted along the target implantation path to complete the reduction of the femoral neck fracture.

7. The positioning auxiliary method for femoral neck fracture reduction according to claim 6, characterized in that: The step of obtaining the normal distribution model corresponding to each voxel includes: Calculate respectively the normal distribution parameters corresponding to the points in each voxel, wherein the normal distribution parameters include: a mean and a covariance matrix, the mean is the coordinate mean of all points in the voxel, and the covariance matrix is ​​the coordinate covariance of all points in the voxel; The normal distribution model of any voxel is determined according to the mean value and the covariance matrix corresponding to each point in the voxel.

8. A positioning auxiliary device for femoral neck fracture reduction, characterized in that: The device comprises: A data preprocessing module, which obtains three-dimensional image data of the femoral neck fracture site of the patient, and performs image preprocessing on the three-dimensional image data to obtain first point cloud data after noise is removed; a path planning module, which performs feature extraction processing on the first point cloud data through a preset point cloud data segmentation network, determines the local features corresponding to each pixel point, and determines the target implantation path of the internal fixation device according to the local features, fracture type and fracture location; The registration and positioning module acquires the second point cloud data of the surgical area in real time during the operation, and performs point cloud registration processing on the second point cloud data so that the second point cloud data coincides with the first point cloud data, so that the internal fixation device is implanted along the target implantation path to complete the reduction of the femoral neck fracture.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer executable instructions that can be executed by the processor, and the processor executes the computer executable instructions to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to implement the method according to any one of claims 1 to 7.

Images