Bone tumor boundary automatic extraction method comprising envelope and reaction zone and related device

Through CT image processing technology, the envelope and reaction area boundaries of bone tumors are extracted, which solves the problem of difficulty in border identification in traditional surgery, achieves accurate resection of bone tumors and minimal damage surgical boundaries, and improves the safety and effectiveness of the surgery.

CN119090910APending Publication Date: 2024-12-06BEIHANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411154441.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-22
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

It is difficult to accurately identify the boundaries of bone tumors during traditional bone tumor resection surgery, especially the invisible reaction areas of anatomy, resulting in insufficient resection or excessive resection, affecting the surgical effect.

Method used

Through CT image processing, the bone tumor envelope boundary and healthy bone surface model are generated using the trained segmentation model, and the envelope boundary is expanded outward to cross the surface, remove part of the model outside the healthy bone surface model, and extract the included model. Bone tumor boundary of envelope and reaction zone.

Benefits of technology

Accurate R0 resection of bone tumors is achieved, which reduces damage to surrounding tissues, ensures minimal damage to surgical boundaries, and improves the safety and effectiveness of the surgery.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090910A_ABST
    Figure CN119090910A_ABST
Patent Text Reader

Abstract

The invention discloses a method for automatically extracting a bone tumor boundary containing an envelope and a reaction area and a related device, relates to the technical field of medical image data processing, and aims at taking a CT image obtained by scanning a bone where a bone tumor is located as input. Respectively utilizing the trained first segmentation model and the trained second segmentation model to generate a bone tumor envelope boundary model and a healthy bone surface model, and outwards expanding the bone tumor envelope boundary model to obtain a bone tumor boundary preliminary model containing an envelope and a reaction area; and carrying out curved surface intersection on the bone tumor boundary preliminary model and the healthy bone surface model, and removing a part of the model positioned outside the healthy bone surface model in the bone tumor boundary preliminary model to obtain a bone tumor boundary model containing an envelope and a reaction region, thereby being beneficial to realizing accurate R0 excision of the bone tumor.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of medical image data processing, and in particular to a method for automatically extracting a bone tumor boundary including a capsule and a reaction zone and a related device. Background Art

[0002] Bone tumors refer to abnormal tissue masses or lumps formed by the uncontrolled division of cells within the bone. They can affect any bone in the body and can develop anywhere from the bone surface to the bone marrow. Bone tumors may weaken or destroy healthy bone tissue, making it more susceptible to fractures. Surgical removal of bone tumors is the main treatment. The resection margin of bone tumors directly affects the surgical quality of bone tumors. Inadequate resection may lead to tumor recurrence, secondary metastasis, amputation, or even death. Excessive resection can easily lead to functional impairment, prosthesis loosening, and even secondary surgery. Therefore, accurate and safe bone tumor resection is the focus of clinical treatment of bone tumors.

[0003] According to the Union for International Cancer Control (UICC), R0 (completely tumor-free margin) resection is used to indicate complete resection of a tumor. In order to achieve R0 resection with minimal tissue damage, accurate R0 resection has attracted increasing attention. However, traditional R0 resection is difficult to achieve accurate R0 resection due to the difficulty in identifying the resection boundary. Therefore, determining the minimally damaging surgical margin for complete tumor resection is the key to achieving accurate R0 resection.

[0004] At present, the automatic planning method of surgical path for bone tumor resection is mainly based on the automatic recognition of bone tumor features based on direct anatomical features. These anatomical features are mainly the primary lesion and capsule area of ​​bone tumor. In other words, the current bone tumor boundary is mainly based on the bone tumor capsule boundary model. However, in actual clinical bone tumor surgery, in addition to the anatomically visible primary lesion and capsule area of ​​bone tumor, the reaction area should also be considered when planning the surgical path to achieve accurate R0 resection. For the reaction area that is not visible in the anatomy, some researchers use near-infrared fluorescent probes to detect it, but its detection accuracy in clinical applications of bone tumors is still unstable.

[0005] Based on this, there is an urgent need for an automatic bone tumor boundary extraction technology that includes the capsule and reaction area. Summary of the invention

[0006] The purpose of this application is to provide a method and related device for automatically extracting the bone tumor boundary including the capsule and the reaction zone, which can automatically extract the bone tumor boundary including the capsule and the reaction zone, obtain the minimally damaging surgical boundary for complete tumor resection, and facilitate the accurate R0 resection of the bone tumor.

[0007] To achieve the above objectives, this application provides the following solutions:

[0008] In a first aspect, the present application provides a method for automatically extracting a bone tumor boundary including a capsule and a reaction zone, the method for automatically extracting a bone tumor boundary including a capsule and a reaction zone comprising:

[0009] Obtaining a CT image obtained by scanning the bone where the bone tumor is located;

[0010] Taking the CT image as input, generating a bone tumor capsule boundary model using the trained first segmentation model; the bone tumor capsule boundary model includes the primary lesion and capsule area of ​​the bone tumor;

[0011] Taking the CT image as input, generating a healthy bone surface model using the trained second segmentation model; the healthy bone surface model includes the surface of the bone where the bone tumor is located;

[0012] Expanding the bone tumor capsule boundary model outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area;

[0013] The preliminary bone tumor boundary model and the healthy bone surface model are intersected by surfaces, and a portion of the preliminary bone tumor boundary model located outside the healthy bone surface model is removed to obtain a bone tumor boundary model including a capsule and a reaction area.

[0014] Optionally, the bone tumor capsule boundary model is expanded outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area, specifically including:

[0015] Calculating the normal vector of each vertex in the bone tumor capsule boundary model;

[0016] For each of the vertices, the vertex is extended outward by a preset offset along the normal vector direction of the vertex to obtain a new vertex corresponding to the vertex;

[0017] Based on the connection relationship between the vertices in the bone tumor capsule boundary model, all the new vertices are connected to obtain a preliminary bone tumor boundary model including the capsule and the reaction area.

[0018] Optionally, calculating the normal vector of each vertex in the bone tumor capsule boundary model specifically includes:

[0019] Calculating the normal vector of each plane in the bone tumor capsule boundary model;

[0020] For each vertex in the bone tumor capsule boundary model, the sum of the normal vectors of all planes containing the vertex is calculated to obtain an accumulated normal vector, and the ratio of the accumulated normal vector to the modulus of the accumulated normal vector is calculated to obtain the normal vector of the vertex.

[0021] Optionally, before the bone tumor capsule boundary model is expanded outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area, the method for automatically extracting the bone tumor boundary including the capsule and the reaction area further includes:

[0022] The bone tumor capsule boundary model and the healthy bone surface model are mesh simplified respectively by using a quadratic edge folding extraction algorithm to obtain a simplified bone tumor capsule boundary model and a simplified healthy bone surface model, and the simplified bone tumor capsule boundary model is used as a new bone tumor capsule boundary model, and the simplified healthy bone surface model is used as a new healthy bone surface model.

[0023] Optionally, performing surface intersection on the preliminary bone tumor boundary model and the healthy bone surface model, removing a portion of the preliminary bone tumor boundary model located outside the healthy bone surface model, and obtaining a bone tumor boundary model including a capsule and a reaction area, specifically includes:

[0024] Calculating the minimum distance from each vertex in the preliminary model of the bone tumor boundary to the healthy bone surface model;

[0025] For each of the vertices, determining whether the minimum distance is less than a distance threshold; if not, recording the vertex as the first vertex;

[0026] connecting all the first vertices to form a first model;

[0027] The first model is removed from the preliminary model of the bone tumor boundary to obtain a second model, and the second model is intersected with the healthy bone surface model to remove a portion of the second model located outside the healthy bone surface model to obtain a third model;

[0028] The first model and the third model are spliced ​​to obtain a bone tumor boundary model including a capsule and a reaction area.

[0029] Optionally, before using the CT image as input and using the trained first segmentation model to generate a bone tumor capsule boundary model, the method for automatically extracting a bone tumor boundary including a capsule and a reaction area further includes:

[0030] Acquire a data set; the data set includes a sample CT image and a first label and a second label corresponding to each of the sample CT images, the first label is a sample bone tumor capsule boundary model, and the second label is a sample healthy bone surface model; the first label and the second label are both obtained by combining the sample CT image and the sample MRI image and annotating the sample CT image; the sample CT image and the sample MRI image are both obtained by scanning the bone where the bone tumor is located;

[0031] constructing an initial first segmentation model and an initial second segmentation model;

[0032] Taking the sample CT image as input and the first label as a label, training the initial first segmentation model to obtain a trained first segmentation model;

[0033] The sample CT image is used as input and the second label is used as a label to train the initial second segmentation model to obtain a trained second segmentation model.

[0034] In a second aspect, the present application provides an automatic bone tumor boundary extraction device including a capsule and a reaction zone, the automatic bone tumor boundary extraction device including a capsule and a reaction zone comprising:

[0035] An image acquisition module, used to acquire a CT image obtained by scanning the bone where the bone tumor is located;

[0036] A first segmentation module is used to use the CT image as input and generate a bone tumor capsule boundary model using a trained first segmentation model; the bone tumor capsule boundary model includes a primary lesion and a capsule area of ​​the bone tumor;

[0037] A second segmentation module is used to use the CT image as input and generate a healthy bone surface model using the trained second segmentation model; the healthy bone surface model includes the surface of the bone where the bone tumor is located;

[0038] An expansion module, used to expand the bone tumor capsule boundary model outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area;

[0039] The surface intersection module is used to perform surface intersection on the preliminary bone tumor boundary model and the healthy bone surface model, remove the partial model of the preliminary bone tumor boundary model located outside the healthy bone surface model, and obtain the bone tumor boundary model including the capsule and the reaction area.

[0040] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described methods for automatically extracting bone tumor boundaries including capsules and reaction zones.

[0041] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for automatically extracting the boundary of a bone tumor including a capsule and a reaction zone.

[0042] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-mentioned methods for automatically extracting the bone tumor boundary including the capsule and the reaction zone.

[0043] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0044] The present application provides a method and a related device for automatically extracting a bone tumor boundary including a capsule and a reaction zone. The method uses a CT image obtained by scanning a bone where a bone tumor is located as an input, and uses a trained first segmentation model and a trained second segmentation model to generate a bone tumor capsule boundary model and a healthy bone surface model respectively. The bone tumor capsule boundary model is expanded outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction zone. The preliminary bone tumor boundary model and the healthy bone surface model are intersected by surfaces, and the part of the preliminary bone tumor boundary model that is outside the healthy bone surface model is removed to obtain a bone tumor boundary model including the capsule and the reaction zone. In this way, the reaction zone is included and damage to surrounding normal tissues is avoided. The bone tumor boundary including the capsule and the reaction zone can be automatically extracted to obtain a minimally damaging surgical boundary for complete tumor resection, which is conducive to achieving accurate R0 resection of the bone tumor. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 This is a diagram of the application environment of a method for automatically extracting bone tumor boundaries including a capsule and a reaction zone provided in Example 1 of the present application.

[0047] Figure 2 A schematic flow chart of a method for automatically extracting the boundary of a bone tumor including a capsule and a reaction zone provided in Example 1 of the present application.

[0048] Figure 3 A detailed flowchart of a method for automatically extracting bone tumor boundaries including capsules and reaction zones provided in Example 1 of the present application.

[0049] Figure 4 This is a schematic diagram of bone tumor characteristics and boundaries provided in Example 1 of the present application.

[0050] Figure 5 Schematic diagram of the automatic extraction process of the femoral bone tumor boundary model provided in Example 1 of the present application.

[0051] Figure 6 Schematic diagram of the preliminary model of the bone tumor boundary of the femur including the capsule and reaction area provided in Example 1 of the present application.

[0052] Figure 7 Schematic diagram of the bone tumor boundary model of the femur including the capsule and reaction area provided in Example 1 of the present application.

[0053] Figure 8 A schematic diagram of the functional modules of an automatic bone tumor boundary extraction device including a capsule and a reaction zone provided in Example 2 of the present application.

[0054] Fig. 9 A schematic diagram of the structure of a computer device provided in Example 3 of the present application. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0056] Example 1

[0057] The method for automatically extracting the bone tumor boundary including the capsule and the reaction area provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal communicates with the server through the network. The data storage system can store the data that the server needs to process. The data storage system can be set up separately, integrated on the server, or placed on the cloud or other servers. The terminal can send the CT image to be processed to the server. After the server receives the CT image to be processed, for the CT image to be processed, the server uses the CT image as input and generates a bone tumor capsule boundary model using the trained first segmentation model; uses the CT image as input and generates a healthy bone surface model using the trained second segmentation model; expands the bone tumor capsule boundary model outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area; performs surface intersection on the preliminary bone tumor boundary model and the healthy bone surface model, removes the part of the preliminary bone tumor boundary model located outside the healthy bone surface model, and obtains a bone tumor boundary model including the capsule and the reaction area. The server can feed back the obtained bone tumor boundary model to the terminal.

[0058] In addition, in some embodiments, the method for automatically extracting the bone tumor boundary including the capsule and the reaction area can also be implemented independently by a server or a terminal. For example, the terminal can directly process the CT image to be processed, or the server can obtain the CT image to be processed from the data storage system and process the CT image to be processed.

[0059] The terminal may be, but is not limited to, various desktop computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. The server may be implemented as an independent server or a server cluster consisting of multiple servers, or may be a cloud server.

[0060] like Figure 2 and Figure 3 As shown, a method for automatically extracting the boundary of a bone tumor including a capsule and a reaction area is provided. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or by a terminal and a server together. In the embodiment of the present application, the method is applied to Figure 1 Taking the server in the example as an example, the method for automatically extracting the bone tumor boundary including the capsule and the reaction area includes the following steps:

[0061] Step S1, obtaining a CT image obtained by scanning the bone where the bone tumor is located.

[0062] Step S2, using the CT image as input, and using the trained first segmentation model to generate a bone tumor capsule boundary model; the bone tumor capsule boundary model includes the primary lesion and capsule area of ​​the bone tumor.

[0063] Step S3, using the CT image as input and using the trained second segmentation model to generate a healthy bone surface model; the healthy bone surface model includes the surface of the bone where the bone tumor is located.

[0064] Step S4, expanding the bone tumor capsule boundary model outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area.

[0065] Step S5, performing surface intersection on the preliminary bone tumor boundary model and the healthy bone surface model, removing the partial model of the preliminary bone tumor boundary model located outside the healthy bone surface model, and obtaining a bone tumor boundary model including a capsule and a reaction area.

[0066] By implementing the above steps S1 to S5, this embodiment can automatically extract the bone tumor boundary including the capsule and the reaction area, which is conducive to achieving accurate R0 resection.

[0067] like Figure 4As shown, the inner edge of the reaction zone of the bone tumor is attached to the surface of the capsule, and the thickness is smooth and does not change much. Therefore, the idea of ​​this embodiment is to expand the bone tumor capsule boundary model along the normal direction of the bone tumor capsule boundary model by a certain offset distance, so as to include the capsule and the reaction zone, and form a new preliminary model of the bone tumor boundary. The preliminary model of the bone tumor boundary is further processed to obtain a bone tumor boundary model, which can provide a safe and effective bone tumor boundary model for the path planning of robot-assisted or computer-assisted precise R0 resection of bone tumors. Based on this idea, this embodiment proposes a method for automatically extracting the bone tumor boundary including the capsule and the reaction zone. By expanding outward along the surface normal of the bone tumor capsule boundary model for a certain distance, the reaction zone is included in the bone tumor recognition range to obtain a bone tumor boundary model that includes both the primary lesion and the capsule area and the reaction area. The method includes three steps: Step 1, using a deep learning network to automatically generate a bone tumor capsule boundary model and a healthy bone surface model for surface intersection, and then simplify the mesh; Step 2, expanding the normal of each point on the bone tumor capsule boundary model outward for a certain distance to obtain a preliminary bone tumor boundary model including the capsule and the reaction area; Step 3, along the direction where the preliminary bone tumor boundary model may interfere with the surrounding bone tissue, the preliminary bone tumor boundary model and the healthy bone surface model of the bone where the bone tumor is located are intersected to obtain the final bone tumor boundary model. This method can provide a tumor safety boundary for path planning of precise bone tumor resection surgery.

[0068] In S1, a CT (Computed Tomography) image is obtained by scanning the bone where the bone tumor is located. The bone where the bone tumor is located can be any bone of the human body containing the bone tumor, such as the sternum, coccyx, phalanges, femur, etc. The bone where the bone tumor is located is scanned to obtain a CT image, which is a three-dimensional image including multiple slices.

[0069] In S2, the CT image is used as input, and the trained first segmentation model is used to generate a bone tumor capsule boundary model, wherein the bone tumor capsule boundary model includes the primary lesion and capsule area of ​​the bone tumor, wherein the bone tumor capsule boundary model is a three-dimensional model.

[0070] In S3, the CT image is used as input, and the trained second segmentation model is used to generate a healthy bone surface model, where the healthy bone surface model includes the surface of the bone where the bone tumor is located. The healthy bone surface model is a three-dimensional model.

[0071] Before using the CT image as input and using the trained first segmentation model to generate a bone tumor capsule boundary model, the method for automatically extracting the bone tumor boundary including the capsule and the reaction area in this embodiment further includes:

[0072] (1) Obtain a data set, the data set including sample CT images and a first label and a second label corresponding to each sample CT image, the first label being a sample bone tumor capsule boundary model, and the second label being a sample healthy bone surface model. The first label and the second label are both obtained by combining the sample CT image and the sample MRI image and annotating the sample CT image, and the sample CT image and the sample MRI image are both obtained by scanning the bone where the bone tumor is located.

[0073] The bone where the bone tumor is located is scanned to obtain a sample CT image and a sample MRI (Magnetic Resonance Imaging) image. The bone where the bone tumor is located at this time is the same as the bone where the bone tumor is located in S1. The sample CT image and the sample MRI image are merged to obtain a merged image. Based on the merged image, the sample CT image is manually annotated to obtain a first label and a second label corresponding to the sample CT image.

[0074] (2) Construct an initial first segmentation model and an initial second segmentation model.

[0075] The initial first segmentation model and the initial second segmentation model may select any existing model with a segmentation function, such as a deep learning neural network model, and specifically, the nnU-Net model may be used.

[0076] (3) Using the sample CT image as input and the first label as a label, the initial first segmentation model is trained to obtain a trained first segmentation model.

[0077] (4) Using the sample CT image as input and the second label as the label, the initial second segmentation model is trained to obtain a trained second segmentation model.

[0078] After obtaining the trained first segmentation model and the trained second segmentation model, the present embodiment can use the deep learning neural network model to automatically generate a bone tumor capsule boundary model including the primary lesion and the capsule area and a healthy bone surface model of the bone where the bone tumor is located for surface intersection. Specifically, the bone tumor capsule boundary model and the healthy bone surface model are automatically extracted based on image segmentation networks such as nnU-Net. Both the bone tumor capsule boundary model and the healthy bone surface model are three-dimensional models, and the surfaces of both can adopt triangular meshes. At this time, the present embodiment can further use the triangular mesh simplification algorithm to compress the number of triangular meshes of the two models to improve subsequent calculation and display speeds.

[0079] Before the bone tumor capsule boundary model is expanded outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area, the automatic bone tumor boundary extraction method including the capsule and the reaction area of ​​this embodiment also includes: using a quadric edge collapse decimal algorithm (quadric edge collapse decimalization algorithm, QECD) to respectively simplify the bone tumor capsule boundary model and the healthy bone surface model to obtain a simplified bone tumor capsule boundary model and a simplified healthy bone surface model, and using the simplified bone tumor capsule boundary model as a new bone tumor capsule boundary model and the simplified healthy bone surface model as a new healthy bone surface model to execute subsequent S4 and S5.

[0080] Specifically, this embodiment uses a quadratic edge folding extraction algorithm to complete mesh simplification, which reduces the number of vertices and faces by merging the edges in the mesh, while trying to maintain the geometric characteristics of the original mesh. This algorithm is widely used in the fields of computer graphics and computer vision, especially in 3D model processing and optimization. The core is to calculate the cost function of each edge. This cost function takes into account the geometric properties of the edge, such as the length, angle and distribution of adjacent vertices of the edge, so as to determine which edges should be merged, thereby achieving the purpose of simplifying the model. The quadratic edge folding extraction algorithm optimizes the vertex selection strategy based on the mesh simplification method based on the quadratic error metric, takes the shrinkage cost of the vertex pair as the sorting criterion, and provides a more flexible simplified path selection algorithm; and updates the error matrix during the folding process, so that the simplification process can more accurately maintain the geometric shape of the original model and reduce the simplified geometric error. In this embodiment, the maximum error, average error and mean square error can be used to evaluate the simplified error of the model geometry.

[0081] Compared with the existing quadratic edge folding extraction algorithm, this embodiment can jointly use multiple local geometric features and corresponding weights in the calculation of the error matrix. That is, the error metrics of the existing quadratic edge folding extraction algorithm are mostly calculated using fixed geometric features (distance metrics or triangle importance). This embodiment further introduces multiple local geometric features of vertices and surfaces on the basis of distance metrics, and obtains the error matrix by jointly weighting them based on weights, wherein the types and weights of the local geometric features are determined according to user needs, and an adaptive error matrix is ​​further used to calculate the shrinkage cost, thereby better retaining the key geometric features while ensuring the simplification effect, so as to provide an improved quadratic edge folding extraction algorithm.

[0082] In S4, the bone tumor capsule boundary model is extended outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction zone. The direction from the primary lesion to the capsule area is defined as the outside, and the normals of each vertex in the bone tumor capsule boundary model are automatically calculated, and each vertex is extended outward for a certain distance along the normal direction to extend the bone tumor capsule boundary model outward, so as to include the reaction zone within the bone tumor boundary range, that is, to include the reaction zone within the resection boundary range, and obtain a preliminary bone tumor boundary model including the capsule and the reaction zone, so as to extend the bone tumor capsule boundary model to the minimum damage boundary that meets the complete resection of the bone tumor, where the minimum damage boundary includes both the primary lesion and capsule area of ​​the bone tumor and the reaction area of ​​the bone tumor.

[0083] The preliminary model of the bone tumor boundary is expanded outward on the basis of the bone tumor capsule boundary model automatically segmented from the CT image. In this embodiment, the normal direction is characterized by a normal vector. The normal vectors of adjacent triangular planes in the bone tumor capsule boundary model are averaged and normalized to obtain the normal vectors of each vertex in the bone tumor capsule boundary model, and each vertex in the bone tumor capsule boundary model is increased outward by a set offset along the direction of the normal vector to obtain the preliminary model of the bone tumor boundary. At this time, the preliminary model of the bone tumor boundary further includes a reaction area, achieving full coverage of the primary lesion, capsule area and reaction area of ​​the bone tumor, and realizing the expansion of the boundary of complete tumor resection.

[0084] In S4, the bone tumor capsule boundary model is expanded outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area, which specifically includes:

[0085] (1) Calculate the normal vector of each vertex in the bone tumor capsule boundary model.

[0086] Calculating the normal vector of each vertex in the bone tumor capsule boundary model specifically includes: calculating the normal vector of each plane in the bone tumor capsule boundary model; for each vertex in the bone tumor capsule boundary model, calculating the sum of the normal vectors of all planes containing the vertex to obtain the accumulated normal vector, and calculating the ratio of the accumulated normal vector and the modulus of the accumulated normal vector to obtain the normal vector of the vertex.

[0087] (2) For each vertex, expand the vertex outward by a preset offset along the direction of the vertex's normal vector to obtain a new vertex corresponding to the vertex.

[0088] The preset offset may be 5 mm.

[0089] (3) Based on the connection relationship between each vertex in the bone tumor capsule boundary model, all new vertices are connected to obtain a preliminary model of the bone tumor boundary including the capsule and the reaction area.

[0090] Since there is a corresponding relationship between the vertices and the new vertices, all the new vertices can be connected according to the connection relationship of each vertex to obtain a preliminary model of the bone tumor boundary.

[0091] In S5, the preliminary model of the bone tumor boundary and the healthy bone surface model are intersected, and the part of the preliminary model of the bone tumor boundary that is outside the healthy bone surface model is removed to obtain a bone tumor boundary model that includes the capsule and the reaction area. In this embodiment, the preliminary model of the bone tumor boundary and the healthy bone surface model of the bone where the bone tumor is located are intersected along the direction in which the preliminary model of the bone tumor boundary and the surrounding bone tissue may interfere, and the final bone tumor boundary model that includes the capsule and the reaction area is obtained, which solves the problem that the preliminary model of the bone tumor boundary may damage other surrounding bone tissues and ensures the safety of the precise surgical planning for bone tumor resection.

[0092] During the surface intersection process, the distance between the surface of the preliminary model of the bone tumor boundary and the bone surface that may interfere is calculated, the part of the model that exceeds the set distance threshold is filtered, the intersection of the remaining model and the healthy bone surface model is calculated, and the filtered part of the model is merged with the intersection to obtain the final bone tumor boundary model, ensuring that the final bone tumor boundary model does not interfere with other surrounding bone tissues, minimizing tissue damage during surgical resection, and ensuring the effectiveness of the resection range.

[0093] In S5, the surface intersection of the preliminary bone tumor boundary model and the healthy bone surface model is performed, and the part of the preliminary bone tumor boundary model located outside the healthy bone surface model is removed to obtain the bone tumor boundary model including the capsule and the reaction area, which specifically includes:

[0094] (1) Calculate the minimum distance from each vertex in the preliminary model of the bone tumor boundary to the healthy bone surface model.

[0095] Since the preliminary model of the bone tumor boundary and the healthy bone surface model are both known, and the coordinates of each point therein are also known, the present embodiment can directly calculate the minimum distance from each vertex in the preliminary model of the bone tumor boundary to the healthy bone surface model. In order to further improve the calculation efficiency, the present embodiment can use a multidimensional binary search tree to generate a first spatial structure of the preliminary model of the bone tumor boundary, the first spatial structure includes the number and position of each vertex of the preliminary model of the bone tumor boundary, the number of each plane, the association relationship between each vertex and each plane, and the connection relationship between each vertex, and the association relationship refers to which plane the vertex belongs to; use a multidimensional binary search tree to generate a second spatial structure of the healthy bone surface model, the second spatial structure includes the number and position of each vertex of the healthy bone surface model, the number of each plane, the association relationship between each vertex and each plane, and the connection relationship between each vertex; based on the first spatial structure and the second spatial structure, calculate the minimum distance from each vertex in the preliminary model of the bone tumor boundary to the healthy bone surface model.

[0096] (2) For each vertex, determine whether the minimum distance is less than the distance threshold; if not, record the vertex as the first vertex.

[0097] (3) Connect all first vertices to form a first model.

[0098] (4) The first model is removed from the preliminary model of the bone tumor boundary to obtain a second model, and the second model is intersected with the healthy bone surface model to remove the part of the second model that is outside the healthy bone surface model to obtain a third model.

[0099] This embodiment can utilize the intersection function in Python to complete the surface intersection, and the third model is the intersection of the second model and the healthy bone surface model. After removing the part model outside the healthy bone surface model in the second model, the second model becomes the third model.

[0100] (5) The first model and the third model are spliced ​​to obtain a bone tumor boundary model including the capsule and the reaction area.

[0101] This embodiment proposes a method for automatically extracting a bone tumor boundary including a capsule and a reaction zone. First, the bone tumor capsule boundary model is extended outward by a certain distance along the normal direction of the capsule surface, so that the reaction zone is included in the bone tumor boundary range, and a preliminary bone tumor boundary model including both the primary lesion and the capsule area and the reaction area is formed; then, in order to avoid possible interference between the preliminary bone tumor boundary model and other surrounding bone tissues, it is necessary to perform a surface intersection operation on the preliminary bone tumor boundary model and the healthy bone surface model of the bone where the bone tumor is located along the direction where the interference may occur, so as to form a final bone tumor boundary model. The bone tumor boundary model includes the primary lesion, capsule area and reaction area of ​​the bone tumor in the boundary recognition range, thereby achieving complete resection of the bone tumor. The bone tumor is subsequently mainly used for robot-assisted or computer-assisted surgical path planning for precise R0 resection of bone tumors, so as to ensure the effectiveness and safety of surgical path planning for precise R0 resection of bone tumors.

[0102] This embodiment is suitable for automatic extraction of accurate and safe boundaries of bone tumors, and has been tested on clinical CT scan images. In the following, this example is described by taking the automatic extraction of the bone tumor boundary model of the femur as an example.

[0103] Considering that the direction in which the preliminary model of the bone tumor boundary may interfere with other surrounding bone tissues is the distal axis, and the object of the surface intersection is the femoral condyle surface at the distal end of the femur, this example no longer establishes a healthy bone surface model of the entire femur, but instead establishes a healthy bone surface model of the femoral condyle surface at the distal end of the femur. In actual applications, if the user can confirm the direction in which interference with other surrounding bone tissues may occur, a healthy surface model of part of the bone where the bone tumor is located can be established; otherwise, a healthy surface model of the entire bone where the bone tumor is located can be established.

[0104] The automatic extraction method of bone tumor boundary including capsule and reaction area of ​​femur includes the following six steps:

[0105] (1) Design an automatic extraction scheme for bone tumor boundaries including capsule and reaction area

[0106] like Figure 5 As shown in the figure, with the goal of obtaining the minimum damage boundary that meets the complete resection of the bone tumor of the femur, first, the bone tumor capsule range in the input original CT image data is identified, the bone tumor capsule boundary model is automatically segmented and generated, and the mesh simplification is completed. The distal femur is subjected to the same segmentation and mesh simplification operations as the bone tumor capsule boundary model to obtain the femoral condyle surface model; then, the bone tumor capsule boundary model is extended outward by a certain distance to include the reaction area, and the minimum reference value of the distance here can be 5mm; finally, the expanded preliminary bone tumor boundary model is intersected with the femoral condyle surface model to obtain the final bone tumor boundary model.

[0107] (2) Automatically generate the femoral bone tumor capsule boundary model

[0108] This example uses a neural network model to implement image segmentation and generate a bone tumor capsule boundary model of the femur. First, the sample CT image and sample MRI image of the femur are merged in the ITK-SNAP software (version 4.0.2); then, the clinician performs semi-automatic segmentation and image annotation to form a data set; finally, the annotated data set is used to train nnU-Net to obtain the trained first segmentation model, and the trained first segmentation model is used for the automatic segmentation of the bone tumor capsule boundary model. The automatically segmented bone tumor capsule boundary model is stored as a binary image in NIFTI format and as a three-dimensional geometric model in STL format.

[0109] (3) Mesh simplified femoral bone tumor capsule boundary model

[0110] For the bone tumor capsule boundary model in the STL format obtained above, the quadratic edge collapse extraction algorithm is used to simplify the mesh. First, the error matrix of each vertex is calculated, and the error tetrahedron of each vertex is further calculated. The square distance between the vertex and its adjacent surface is accumulated to quantify the impact of vertex movement on the model shape; then, the optimal shrinkage target is calculated for each vertex pair to minimize the sum of the error tetrahedrons of the two vertices, and all vertex pairs are inserted into the priority queue in ascending order of shrinkage cost. The vertex pair with the smallest shrinkage cost is selected from the priority queue for shrinkage until the set simplification target is reached. The simplified bone tumor capsule boundary model is obtained and stored in STL format.

[0111] (4) Normally expand the bone tumor capsule boundary model to obtain a preliminary model of the bone tumor boundary

[0112] For the simplified bone tumor capsule boundary model obtained in STL format, firstly follow the formula The normal vectors of all adjacent triangular planes in the simplified bone tumor capsule boundary model are averaged and normalized to obtain the normal vector of each vertex, n vertax (v i ) is the i-th vertex v i The normal vector, n facet (j) is the number of nodes that contain the i-th vertex v i The normal vector of the jth plane of . Specifically, first, by traversing each plane in the simplified bone tumor capsule boundary model, the normal vector of each plane is added to each vertex it contains, thereby calculating the cumulative normal vector of the vertex; then, the accumulated normal vector of the vertex is normalized to obtain the normal vector of each vertex. Based on the calculated normal vectors of the surface vertices of the simplified bone tumor capsule boundary model, the new position of each vertex after extending 5mm along the direction of its normal vector is calculated, thereby obtaining a preliminary bone tumor boundary model that can further cover the reaction area, such as Figure 6 In this example, the 5 mm offset is a reference value set based on the clinician's experience, which can effectively cover the reaction area of ​​most bone tumors.

[0113] (5) Automatically generate and simplify the femoral condyle surface model

[0114] First, ITK-SNAP software (version 4.0.2) was used to merge the sample CT images and sample MRI images of the femur. Then, clinicians performed semi-automatic segmentation and image annotation to form a data set. The annotated data set was used to train nnU-Net to obtain a trained second segmentation model, which was then used for automatic segmentation of the femoral condyle surface model. Finally, the NIFTI format binary image and STL format 3D geometric model were output. The mesh was simplified using the secondary edge collapse extraction algorithm to obtain a simplified femoral condyle surface model.

[0115] (6) Surface intersection of the preliminary bone tumor boundary model and the simplified femoral condyle surface model to obtain the final femoral bone tumor boundary model

[0116] Read the STL files of the models in step (4) and step (5), use a multidimensional binary search tree to implement spatial indexing, calculate the minimum distance from each vertex in the preliminary model of the bone tumor boundary to the surface of the femoral condyle, and set a distance threshold to filter the preliminary model of the bone tumor boundary, remove the part of the model that exceeds the set distance threshold, further intersect the remaining model with the simplified femoral condyle surface model, merge the removed part of the model and the intersection, so that the surface of the preliminary model of the bone tumor boundary does not exceed the surface of the femoral condyle, thereby obtaining the final bone tumor boundary model that includes the capsule and the reaction area and meets the requirements of minimum tissue damage for safe resection, such as Figure 7 shown.

[0117] The embodiment of the present application also provides an application scenario, which applies the above-mentioned method for automatically extracting the bone tumor boundary including the capsule and the reaction zone. Specifically, the method for automatically extracting the bone tumor boundary including the capsule and the reaction zone provided in this embodiment can be applied in the bone tumor boundary display scenario. The bone tumor boundary display scenario includes an image acquisition link, an image processing link, and a result display link. The image acquisition link is used to acquire CT images, the data processing link is used to process the CT images to obtain a bone tumor boundary model, and the result display link is used to display the bone tumor boundary model to the user. The method for automatically extracting the bone tumor boundary including the capsule and the reaction zone provided in this embodiment belongs to the image processing link.

[0118] Example 2

[0119] Based on the same inventive concept, the embodiment of the present application also provides a device for automatically extracting the boundary of a bone tumor including a capsule and a reaction zone, which is used to implement the method for automatically extracting the boundary of a bone tumor including a capsule and a reaction zone. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in the embodiment of the device for automatically extracting the boundary of a bone tumor including a capsule and a reaction zone provided below can refer to the limitations of the method for automatically extracting the boundary of a bone tumor including a capsule and a reaction zone above, and will not be repeated here.

[0120] like Figure 8 As shown, a device for automatically extracting the boundary of a bone tumor including a capsule and a reaction zone is provided, and the device for automatically extracting the boundary of a bone tumor including a capsule and a reaction zone comprises:

[0121] The image acquisition module M1 is used to acquire a CT image obtained by scanning the bone where the bone tumor is located.

[0122] The first segmentation module M2 is used to use the CT image as input and generate a bone tumor capsule boundary model using the trained first segmentation model; the bone tumor capsule boundary model includes the primary lesion and capsule area of ​​the bone tumor.

[0123] The second segmentation module M3 is used to use the CT image as input and generate a healthy bone surface model using the trained second segmentation model; the healthy bone surface model includes the surface of the bone where the bone tumor is located.

[0124] The expansion module M4 is used to expand the bone tumor capsule boundary model outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area.

[0125] The surface intersection module M5 is used to perform surface intersection on the preliminary bone tumor boundary model and the healthy bone surface model, remove the partial model of the preliminary bone tumor boundary model located outside the healthy bone surface model, and obtain the bone tumor boundary model including the capsule and the reaction area.

[0126] Example 3

[0127] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Fig. 9 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store image data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for automatically extracting the boundary of a bone tumor including a capsule and a reaction zone is implemented.

[0128] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0129] In an exemplary embodiment, a computer device is also provided, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for automatically extracting the bone tumor boundary including the capsule and the reaction zone described in Example 1.

[0130] Example 4

[0131] The embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for automatically extracting the bone tumor boundary including the capsule and the reaction zone described in Example 1 is implemented.

[0132] Example 5

[0133] The embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method for automatically extracting the bone tumor boundary including the capsule and the reaction zone described in Example 1.

[0134] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0135] Any reference to memory, database or other media used in the embodiments provided in the present application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0136] The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, etc., but is not limited thereto.

[0137] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0138] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for automatically extracting the boundary of a bone tumor including the capsule and the reaction area. It is characterized in that The method for automatically extracting the bone tumor boundary including the capsule and the reaction area comprises: Obtaining a CT image obtained by scanning the bone where the bone tumor is located; Taking the CT image as input, generating a bone tumor capsule boundary model using the trained first segmentation model; the bone tumor capsule boundary model includes the primary lesion and capsule area of ​​the bone tumor; Taking the CT image as input, generating a healthy bone surface model using the trained second segmentation model; the healthy bone surface model includes the surface of the bone where the bone tumor is located; Expanding the bone tumor capsule boundary model outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area; The preliminary bone tumor boundary model and the healthy bone surface model are intersected by surfaces, and a portion of the preliminary bone tumor boundary model located outside the healthy bone surface model is removed to obtain a bone tumor boundary model including a capsule and a reaction area.

2. The method for automatically extracting the bone tumor boundary including the capsule and the reaction area according to claim 1, It is characterized in that The bone tumor capsule boundary model is expanded outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area, specifically including: Calculating the normal vector of each vertex in the bone tumor capsule boundary model; For each of the vertices, the vertex is extended outward by a preset offset along the normal vector direction of the vertex to obtain a new vertex corresponding to the vertex; Based on the connection relationship between the vertices in the bone tumor capsule boundary model, all the new vertices are connected to obtain a preliminary bone tumor boundary model including the capsule and the reaction area.

3. The method for automatically extracting the bone tumor boundary including the capsule and the reaction area according to claim 2, It is characterized in that Calculating the normal vector of each vertex in the bone tumor capsule boundary model specifically includes: Calculating the normal vector of each plane in the bone tumor capsule boundary model; For each vertex in the bone tumor capsule boundary model, the sum of the normal vectors of all planes containing the vertex is calculated to obtain an accumulated normal vector, and the ratio of the accumulated normal vector to the modulus of the accumulated normal vector is calculated to obtain the normal vector of the vertex.

4. The method for automatically extracting the bone tumor boundary including the capsule and the reaction area according to claim 1, It is characterized in that Before the bone tumor capsule boundary model is expanded outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area, the method for automatically extracting the bone tumor boundary including the capsule and the reaction area further includes: The bone tumor capsule boundary model and the healthy bone surface model are mesh simplified respectively by using a quadratic edge folding extraction algorithm to obtain a simplified bone tumor capsule boundary model and a simplified healthy bone surface model, and the simplified bone tumor capsule boundary model is used as a new bone tumor capsule boundary model, and the simplified healthy bone surface model is used as a new healthy bone surface model.

5. The method for automatically extracting the bone tumor boundary including the capsule and the reaction area according to claim 1, It is characterized in that Performing surface intersection on the preliminary bone tumor boundary model and the healthy bone surface model, removing a portion of the preliminary bone tumor boundary model located outside the healthy bone surface model, and obtaining a bone tumor boundary model including a capsule and a reaction area, specifically comprising: Calculating the minimum distance from each vertex in the preliminary model of the bone tumor boundary to the healthy bone surface model; For each of the vertices, determining whether the minimum distance is less than a distance threshold; if not, recording the vertex as the first vertex; connecting all the first vertices to form a first model; The first model is removed from the preliminary model of the bone tumor boundary to obtain a second model, and the second model is intersected with the healthy bone surface model to remove a portion of the second model located outside the healthy bone surface model to obtain a third model; The first model and the third model are spliced ​​to obtain a bone tumor boundary model including a capsule and a reaction area.

6. The method for automatically extracting the bone tumor boundary including the capsule and the reaction area according to claim 1, It is characterized in that Before generating a bone tumor capsule boundary model using the trained first segmentation model with the CT image as input, the method for automatically extracting a bone tumor boundary including the capsule and the reaction area further includes: Acquire a data set; the data set includes a sample CT image and a first label and a second label corresponding to each of the sample CT images, the first label is a sample bone tumor capsule boundary model, and the second label is a sample healthy bone surface model; the first label and the second label are both obtained by combining the sample CT image and the sample MRI image and annotating the sample CT image; the sample CT image and the sample MRI image are both obtained by scanning the bone where the bone tumor is located; constructing an initial first segmentation model and an initial second segmentation model; Taking the sample CT image as input and the first label as a label, training the initial first segmentation model to obtain a trained first segmentation model; The sample CT image is used as input and the second label is used as a label to train the initial second segmentation model to obtain a trained second segmentation model.

7. An automatic bone tumor boundary extraction device including capsule and reaction area, It is characterized in that The automatic extraction device for bone tumor boundary including capsule and reaction area comprises: An image acquisition module, used to acquire a CT image obtained by scanning the bone where the bone tumor is located; A first segmentation module is used to use the CT image as input and generate a bone tumor capsule boundary model using a trained first segmentation model; the bone tumor capsule boundary model includes a primary lesion and a capsule area of ​​the bone tumor; A second segmentation module is used to use the CT image as input and generate a healthy bone surface model using the trained second segmentation model; the healthy bone surface model includes the surface of the bone where the bone tumor is located; An expansion module, used to expand the bone tumor capsule boundary model outward to obtain a preliminary bone tumor boundary model including the capsule and the reaction area; The surface intersection module is used to perform surface intersection on the preliminary bone tumor boundary model and the healthy bone surface model, remove the part of the preliminary bone tumor boundary model located outside the healthy bone surface model, and obtain the bone tumor boundary model including the capsule and the reaction area.

8. A computer device, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for automatically extracting the bone tumor boundary including the capsule and the reaction zone according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the method for automatically extracting the bone tumor boundary including the capsule and the reaction area according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, It is characterized in that When the computer program is executed by a processor, the method for automatically extracting the bone tumor boundary including the capsule and the reaction area according to any one of claims 1 to 6 is implemented.