System for automatically planning screw channel range

Abstract

The invention relates to the technical field of orthopedic surgery, in particular to a system for automatically planning a screw channel range, which comprises a medical image acquisition module used for receiving and preprocessing CT, MRI and other medical images from medical equipment; the three-dimensional reconstruction module is used for performing three-dimensional reconstruction on the preprocessed two-dimensional medical image to generate a three-dimensional skeleton model; the multi-constraint path planning module is used for planning a candidate path set meeting anatomical and mechanical constraints in the three-dimensional skeleton model; the parameter calculation module is used for calculating path information of the candidate path set; and the visualization module draws a screw channel based on the path information and displays related parameters. The method has the advantages that the candidate path set meeting various anatomy and mechanical constraints can be planned in the three-dimensional skeleton model, the path length and the path included angle are calculated according to the candidate path set, the screw channel is drawn based on the path information, and the accuracy and safety of surgical planning are improved.

Description

Technical Field

[0001] The invention relates to the technical field of orthopedic surgery, and in particular to a system for automatically planning a screw channel range. Background Art

[0002] In the past, external fixation or conservative treatment was often used to treat fractures, bone deformities or joint instability, but there were defects such as weak fixation, long healing period, and easy displacement or complications. With the advancement of medical technology, minimally invasive surgery has gradually become popular. Bone screws have become an important alternative to traditional open surgery due to their low trauma and precise fixation.

[0003] In the prior art, the screw channel is usually calculated and then inserted into the bone according to the screw channel. The screw channel uses a traditional path planning algorithm, which ignores bone density distribution and mechanical constraints, leading to postoperative loosening. In addition, there is a lack of quantitative analysis of the path angle, which cannot guide intraoperative angle fine-tuning. Summary of the invention

[0004] In response to the shortcomings of the existing technology, the present application proposes an automatic screw channel range planning system, which can plan a set of candidate paths that meet multiple anatomical and mechanical constraints in a three-dimensional bone model, calculate the path length and path angle based on the candidate path set, and draw the screw channel based on the path information, thereby improving the accuracy and safety of surgical planning.

[0005] The following is a technical solution of the present invention, a system for automatically planning the range of screw channels, comprising:

[0006] A medical image acquisition module is used to receive and pre-process medical images such as CT and MRI from medical equipment and connect to a three-dimensional reconstruction module;

[0007] A three-dimensional reconstruction module is used to reconstruct the preprocessed two-dimensional medical image into three dimensions, generate a three-dimensional skeleton model, and connect to the multi-constraint path planning module;

[0008] A multi-constraint path planning module is used to plan a set of candidate paths that meet anatomical and mechanical constraints in a three-dimensional bone model, and is connected to a parameter calculation module;

[0009] The parameter calculation module is used to calculate the path information of the candidate path set and connect to the visualization module;

[0010] Visualization module, which draws the screw channel based on the path information and displays the relevant parameters.

[0011] As a preferred solution of the present invention, medical image preprocessing includes noise suppression and tissue segmentation. Noise suppression uses adaptive anisotropic diffusion filtering, and tissue segmentation uses a deep residual U-Net network.

[0012] As a preferred solution of the present invention, the expression of the conduction coefficient of the adaptive anisotropic diffusion filter is as follows:

[0013]

[0014] In the above formula, c is the conduction coefficient, is the gradient amplitude of the image, σ=0.05.

[0015] As a preferred solution of the present invention, the loss function of the deep residual U-Net network is expressed as follows:

[0016] L lose =0.7L dice +0.3L ce

[0017] In the above formula, L lose is the loss function, L dice is the Dice coefficient loss function, L ce is the cross entropy loss function.

[0018] As a preferred solution of the present invention, the 3D reconstruction module uses an improved Marching Cubes algorithm for 3D reconstruction, and the energy function expression is as follows:

[0019] E total =0.6×E data +0.3×E smooth +0.1×E edge

[0020] In the above formula, E total is the energy function, E data is the data fidelity term, E smooth is the smoothing term, E edge This is an edge protection item.

[0021] As a preferred solution of the present invention, the multi-constraint path planning module adopts an improved bidirectional A* algorithm, and the cost function expression is as follows:

[0022] f(n)=g n +0.6h 欧式 (n)+0.3h 骨密度 (n)+0.1h 安全 (n)

[0023]

[0024] h 安全 (n) = min(d 骨皮质_n ,d 神经_n ,d 血管_n )

[0025] In the above formula, f(n) is the cost function, g n is the actual cost from the starting point to the target node n, h 欧式 (n) is the normalized Euclidean distance, h 骨密度 (n) is the bone density score, h 安全 (n) is the minimum safe distance of the screw channel at the target node n, n is the target node, g is the starting point, d max is the distance between the farthest two points, ρ n is the bone density value of the voxel where the target node n is located, ρ max is the maximum bone density value, d 骨皮质_n is the distance from the target node n to the nearest bone cortical surface, d 神经_n is the distance from the target node n to the nearest neural bundle, d 血管_n is the distance to the nearest vascular structure.

[0026] As a preferred solution of the present invention, the multi-constraint path planning module combines Monte Carlo sampling to generate several candidate paths, and accepts some paths to generate a candidate path set. The acceptance probability expression is as follows:

[0027]

[0028] S=α·C+β·D+γ·R

[0029]

[0030] In the above formula, P accept is the acceptance probability, S is the safety score of the path, C is the collision risk coefficient, D is the bone density adaptation coefficient, R is the path stability coefficient, α, β and γ are weight coefficients, and d i is the distance from the ith voxel point on the path to the nearest blood vessel or nerve, d thr is the safety distance threshold, k is the attenuation coefficient, μ path is the mean bone density of the area where the planning path passes, σ bone is the standard deviation of bone density in the target bone area, λ 1 is the curvature penalty factor, k max is the maximum curvature of the path.

[0031] As a preferred solution of the present invention, the parameter calculation module calculates the path information of the candidate path set, and the path information includes the path length and the path angle.

[0032] As a preferred solution of the present invention, the expression of path length is as follows:

[0033]

[0034] In the above formula, L is the path length, x end ,yend and z end is the three-dimensional coordinate of the end point of the path, x start ,y start and z start The three-dimensional coordinates of the starting point of the path.

[0035] As a preferred solution of the present invention, the expression of the path angle is as follows:

[0036]

[0037] In the above formula, θ is the path angle, is the direction vector of the path, is the reference vector of the shortest path.

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

[0039] 1. The present invention can receive and preprocess medical images such as CT and MRI from medical equipment, perform three-dimensional reconstruction on the preprocessed two-dimensional medical images, and generate a three-dimensional bone model with anatomical structure information, thereby improving applicability;

[0040] 2. In the present invention, a candidate path set that meets multiple anatomical and mechanical constraints can be planned in a three-dimensional bone model, and path information can be calculated based on the candidate path set, and a screw channel can be drawn based on the path information, thereby improving the accuracy and safety of surgical planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the system of the present invention;

[0042] Figure 2 It is a step diagram of the method of the present invention;

[0043] Figure 3 is a flow chart of the method of the present invention;

[0044] Figure 4 It is a schematic diagram of the path of the present invention;

[0045] In the figure: 1. Medical image acquisition module; 2. Three-dimensional reconstruction module; 3. Multi-constraint path planning module; 4. Parameter calculation module; 5. Visualization module. DETAILED DESCRIPTION

[0046] In order to make the technical problems solved by the present invention, the technical solutions adopted and the technical effects achieved clearer, the technical solutions of the embodiments of the present invention will be further described in detail below in conjunction with the accompanying drawings. 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0047] Embodiment 1:

[0048] like Figure 1 and Figure 4 As shown, a system for automatically planning the range of screw channels includes:

[0049] Medical image acquisition module 1, used for receiving and preprocessing medical images such as CT and MRI from medical equipment, and connected to 3D reconstruction module 2;

[0050] The three-dimensional reconstruction module 2 is used to perform three-dimensional reconstruction on the pre-processed two-dimensional medical image, generate a three-dimensional skeleton model with anatomical structure information, and connect to the multi-constraint path planning module 3;

[0051] A multi-constraint path planning module 3, used for planning a set of candidate paths satisfying multiple anatomical and mechanical constraints in a three-dimensional bone model, connected to a parameter calculation module 4;

[0052] The parameter calculation module 4 is used to calculate the path information of the candidate path set, such as the shortest path length, the maximum safe path length, the path angle, etc., and is connected to the visualization module 5;

[0053] Visualization module 5 draws the screw channel based on the path information, and displays the planned screw channel and related parameters to the doctor in the form of intuitive three-dimensional graphics and data tables, providing the doctor with an intuitive surgical planning reference.

[0054] In this embodiment, the medical image acquisition module 1 receives CT and MRI image data that conforms to the DICOM 3.0 protocol. The medical image acquisition module 1 preprocesses the image, and the image preprocessing includes noise suppression and tissue segmentation. Noise suppression uses adaptive anisotropic diffusion filtering, and the expression for calculating the conduction coefficient is as follows:

[0055]

[0056] In the above formula, c is the conduction coefficient, is the gradient amplitude of the image, σ=0.05.

[0057] Tissue segmentation uses a deep residual U-Net network, and the loss function is expressed as follows:

[0058] L lose =0.7L dice +0.3L ce

[0059] In the above formula, L lose is the loss function, L dice is the Dice coefficient loss function, L ce is the cross entropy loss function.

[0060] The processed image data is transmitted to the three-dimensional reconstruction module 2.

[0061] In this embodiment, after receiving the pre-processed image data, the 3D reconstruction module 2 uses an improved Marching Cubes algorithm to perform 3D reconstruction. The energy function expression is as follows:

[0062] E total =0.6×E data +0.3×E smooth +0.1×E edge

[0063] In the above formula, E total is the energy function, E data It is the data fidelity item, ensuring that the model fits the original data. smooth is a smoothing term to suppress staircase artifacts, E edge It is an edge protection item that strengthens the boundaries of anatomical structures.

[0064] Construct an octree spatial index with a voxel resolution of 0.5mm 3 , and pass the reconstructed three-dimensional skeleton model to the multi-constraint path planning module 3.

[0065] In this embodiment, the multi-constraint path planning module 3 uses the received three-dimensional skeleton model to perform path planning. The improved bidirectional A* algorithm is adopted, and the cost function expression is as follows:

[0066] f(n)=g n +0.6h 欧式 (n)+0.3h 骨密度 (n)+0.1h 安全 (n)

[0067]

[0068] h 安全 (n) = min(d 骨皮质_n ,d 神经_n ,d 血管_n )

[0069] In the above formula, f(n) is the cost function, g n is the actual cost from the starting point to the target node n, h 欧式 (n) is the normalized Euclidean distance, h 骨密度 (n) is the bone density score, h 安全 (n) is the minimum safe distance of the screw channel at the target node n, n is the target node, g is the starting point, d max is the distance between the farthest two points, ρ n is the bone density value of the voxel where the target node n is located, ρmax is the maximum bone density value, d 骨皮质_n is the distance from the target node n to the nearest bone cortical surface, d 神经_n is the distance from the target node n to the nearest neural bundle, d 血管_n is the distance to the nearest vascular structure.

[0070] d 骨皮质_n Through image segmentation and distance field calculation, d 神经_n Based on preoperative MRI / CT multimodality registration, d 血管_n Extracted by enhanced CT angiography. 安全 (n) represents the minimum safe distance of the screw channel at the target node n, which must satisfy h 安全 (n)≥2mm, if h 安全 (n) If the value is below the threshold, the planning system automatically adjusts the path or marks high-risk areas for the doctor to review.

[0071] Combined with Monte Carlo sampling, 500-800 candidate paths are generated, and some paths are accepted to generate a candidate path set. The acceptance probability expression is as follows:

[0072]

[0073] S=α·C+β·D+γ·R

[0074]

[0075] In the above formula, P accept is the acceptance probability, S is the safety score of the path, C is the collision risk coefficient, D is the bone density adaptation coefficient, R is the path stability coefficient, α, β and γ are weight coefficients, and d i is the distance from the ith voxel point on the path to the nearest blood vessel or nerve, d thr is the safety distance threshold, k is the attenuation coefficient, μ path is the mean bone density of the area where the planning path passes, σ bone is the standard deviation of bone density in the target bone area, λ 1 is the curvature penalty factor, k max is the maximum curvature of the path, calculated by differential geometry.

[0076] The planned candidate path set P = {P 1 ,P 2 ,...,P n} is passed to parameter calculation module 4.

[0077] In this embodiment, the parameter calculation module 4 calculates the path information according to the received candidate path set.

[0078] Calculate the path length. The expression of path length is as follows:

[0079]

[0080] In the above formula, L is the path length, x end ,y end and z end is the three-dimensional coordinate of the end point of the path, x start ,y start and z start The three-dimensional coordinates of the starting point of the path.

[0081] Calculate the path angle, which is the angle between the path and the shortest path. The expression of the path angle is as follows:

[0082]

[0083] In the above formula, θ is the path angle, is the direction vector of the path, is the reference vector of the shortest path.

[0084] The length and angle of the path affect the difficulty and safety of the surgical operation. The parameters and various factors considered in the path planning process are also interrelated. The calculated parameters are transmitted to the visualization module 5 for display.

[0085] In this embodiment, after receiving the parameters from the parameter calculation module 4, the visualization module 5 draws the screw channel based on the path information, and displays the planned screw channel and related parameters to the doctor in the form of intuitive three-dimensional graphics and data tables. The doctor can plan and make decisions for the operation based on the displayed information, thereby improving the accuracy and safety of the operation.

[0086] Embodiment 2:

[0087] like Figure 2 and Figure 3 As shown, a method for automatically planning the range of a screw channel comprises the following steps:

[0088] S1, acquiring medical images and performing image preprocessing;

[0089] S2, reconstructing a three-dimensional bone model;

[0090] S3, planning multi-constrained paths and outputting a set of candidate paths;

[0091] S4, calculating path information based on the candidate path set;

[0092] S5. Draw a screw channel based on the path information, and determine the surgical plan based on the screw channel.

[0093] In step S1, a medical image is acquired and image preprocessing is performed. Specifically, the medical image acquisition module 1 preprocesses the image, and the image preprocessing includes noise suppression and tissue segmentation. Noise suppression uses adaptive anisotropic diffusion filtering to calculate its conduction coefficient; tissue segmentation uses a deep residual U-Net network to calculate and set its loss function. The processed image data is passed to the three-dimensional reconstruction module 2.

[0094] In step S2, a three-dimensional bone model is reconstructed. Specifically, the preprocessed two-dimensional medical image is reconstructed using an improved Marching Cubes algorithm, and its energy function is calculated. An octree spatial index is constructed with a voxel resolution of 0.5 mm. 3 , generate a three-dimensional bone model with anatomical structure information.

[0095] In step S3, a multi-constrained path is planned and a candidate path set is output. Specifically, an improved bidirectional A* algorithm is used in the three-dimensional skeleton model to calculate its cost function. Combined with Monte Carlo sampling, 500-800 candidate paths are generated, and some candidate paths are accepted by probability to form a candidate path set, and the candidate path set is passed to the parameter calculation module 4.

[0096] In step S4, the path information is calculated based on the candidate path set. Specifically, the parameter calculation module 4 calculates the path information based on the received path information, and the path information includes the path length and the path angle. The path length and angle affect the difficulty and safety of the surgical operation. The parameters and various factors considered in the path planning process are also interrelated. The calculated parameters are passed to the visualization module 5 for display.

[0097] In step S5, a screw channel is drawn based on the path information, and the surgical plan is determined based on the screw channel. Specifically, after the visualization module 5 receives the parameters transmitted by the parameter calculation module 4, it draws the screw channel based on the path information, and displays the planned screw channel and related parameters to the doctor in the form of intuitive three-dimensional graphics and data tables. The doctor can make surgical plans and decisions based on the displayed information, thereby improving the accuracy and safety of the surgery.

[0098] The present invention can receive and preprocess medical images such as CT and MRI from medical equipment, perform three-dimensional reconstruction of the preprocessed two-dimensional medical images, and generate a three-dimensional bone model with anatomical structure information, thereby improving applicability; it can plan a set of candidate paths that meet multiple anatomical and mechanical constraints in the three-dimensional bone model, calculate path information based on the candidate path set, and draw a screw channel based on the path information, thereby improving the accuracy and safety of surgical planning.

[0099] Although preferred embodiments of the present invention have been described, additional changes and modifications may be made to these embodiments by those skilled in the art once the basic inventive concepts are known. Obviously, various changes and modifications may be made to the present invention by those skilled in the art without departing from the spirit and scope of the present invention. Thus, the present invention is intended to include such changes and modifications if they fall within the scope of the technical equivalents of the present invention.

Claims

1. A system for automatically planning the range of screw channels, characterized in that: include: A medical image acquisition module is used to receive and pre-process medical images such as CT and MRI from medical equipment and connect to a three-dimensional reconstruction module; A three-dimensional reconstruction module is used to reconstruct the preprocessed two-dimensional medical image into three dimensions, generate a three-dimensional skeleton model, and connect to the multi-constraint path planning module; A multi-constraint path planning module is used to plan a set of candidate paths that meet anatomical and mechanical constraints in a three-dimensional bone model, and is connected to a parameter calculation module; The parameter calculation module is used to calculate the path information of the candidate path set and connect to the visualization module; Visualization module, which draws the screw channel based on the path information and displays the relevant parameters.

2. The automatic screw channel range planning system according to claim 1, characterized in that: Medical image preprocessing includes noise suppression and tissue segmentation. Noise suppression uses adaptive anisotropic diffusion filtering, and tissue segmentation uses a deep residual U-Net network.

3. The automatic screw channel range planning system according to claim 2, characterized in that: The expression of the transmission coefficient of adaptive anisotropic diffusion filtering is as follows: In the above formula, c is the conduction coefficient, is the gradient amplitude of the image, σ=0.

05.

4. The automatic screw channel range planning system according to claim 2, characterized in that: The loss function of the deep residual U-Net network is expressed as follows: <h2 style=";text-align:left;direction:ltr">L<h2 style=";text-align:left;direction:ltr"> lose <h2 style=";text-align:left;direction:ltr"> <0.7L<h2 style=";text-align:left;direction:ltr"> dice <h2 style=";text-align:left;direction:ltr"> +0.3L<h2 style=";text-align:left;direction:ltr"> ce In the above formula, L lose is the loss function, L dice is the Dice coefficient loss function, L ce is the cross entropy loss function.

5. The automatic screw channel range planning system according to claim 1, characterized in that: The 3D reconstruction module uses the improved Marching Cubes algorithm for 3D reconstruction. The energy function expression is as follows: AND total =0.6×E data +0.3×E smooth +0.1×E edge In the above formula, E total is the energy function, E data is the data fidelity term, E smooth is the smoothing term, E edge This is an edge protection item.

6. The automatic screw channel range planning system according to claim 1, characterized in that: The multi-constraint path planning module uses an improved bidirectional A* algorithm, and the cost function expression is as follows: f(n)=g n +0.6h 欧式 (n)+0.3h 骨密度 (n)+0.1h 安全 (a) h 安全 (n)=min(d) 骨皮质_n ,d 神经_n ,d 血管_n ) In the above formula, f(n) is the cost function, g n is the actual cost from the starting point to the target node n, h 欧式 (n) is the normalized Euclidean distance, h 骨密度 (n) is the bone density score, h 安全 (n) is the minimum safe distance of the screw channel at the target node n, n is the target node, g is the starting point, d max is the distance between the farthest two points, ρ n is the bone density value of the voxel where the target node n is located, ρ max is the maximum bone density value, d 骨皮质_n is the distance from the target node n to the nearest bone cortical surface, d 神经_n is the distance from the target node n to the nearest neural bundle, d 血管_n is the distance to the nearest vascular structure.

7. The automatic screw channel range planning system according to claim 6, characterized in that: The multi-constrained path planning module combines Monte Carlo sampling to generate several candidate paths, and accepts some paths to generate a candidate path set. The acceptance probability expression is as follows: S=α·C+β·D+γ·R In the above formula, P accept is the acceptance probability, S is the safety score of the path, C is the collision risk coefficient, D is the bone density adaptation coefficient, R is the path stability coefficient, α, β and γ are weight coefficients, and d i is the distance from the ith voxel point on the path to the nearest blood vessel or nerve, d thr is the safety distance threshold, k is the attenuation coefficient, μ path is the mean bone density of the area where the planning path passes, σ bone is the standard deviation of bone density in the target bone area, λ1 is the curvature penalty factor, k max is the maximum curvature of the path.

8. The automatic screw channel range planning system according to claim 1, characterized in that: The parameter calculation module calculates the path information of the candidate path set, and the path information includes the path length and the path angle.

9. The automatic screw channel range planning system according to claim 8, characterized in that: The expression for path length is as follows: In the above formula, L is the path length, x end ,y end and z end is the three-dimensional coordinate of the end point of the path, x start ,y start and z start The three-dimensional coordinates of the starting point of the path.

10. The automatic screw channel range planning system according to claim 8, characterized in that: The expression of the path angle is as follows: In the above formula, θ is the path angle, is the direction vector of the path, is the reference vector of the shortest path.

Images