Perspective image analysis method for bionic repair of bone tumor

Abstract

The invention discloses a perspective image analysis method for bionic repair of a bone tumor. The perspective image analysis method comprises the following steps: acquiring perspective image data of a bone tumor area; performing cross-modal feature alignment on the data, constructing a three-dimensional bone tumor image model in a three-dimensional voxel space coordinate system, and eliminating image artifacts; performing segmentation processing on the bone tumor through a deep learning segmentation network, and identifying a bone tumor infiltration boundary to obtain three-dimensional morphological parameters and bone tumor feature parameters of a bone tumor region; reconstructing a bionic scaffold structure of the bone defect area through biomechanical modeling, and implanting the bionic scaffold into the bone tumor area; and based on the postoperative image data, acquiring a bone-scaffold interface fusion rate, stress distribution uniformity and a new bone ingrowth depth index. The method has the advantages that through combination of voxel-level accurate segmentation and the convolutional neural network, the bone tumor region and form are accurately identified and divided, and reliable support is provided for subsequent tumor infiltration assessment and bionic stent design.

Description

Technical Field

[0001] The present invention relates to dynamic adjustment technology, and particularly to a method for analyzing fluoroscopic images for bionic repair of bone tumors. Background Art

[0002] The method for analyzing fluoroscopic images for bone tumor repair is a technology widely used in the field of medical imaging in recent years. It aims to assist doctors in accurately evaluating the detection, diagnosis, treatment, and repair processes of bone tumors by analyzing fluoroscopic images such as X-rays or CT. With the continuous progress of medical imaging technology, traditional manual diagnosis methods are gradually facing challenges in terms of efficiency and accuracy, especially in the identification and analysis of complex lesion areas. Fluoroscopic images can provide detailed internal structure and growth morphology information of bone tumors, so they have important value in the early detection and monitoring of bone tumors.

[0003] Most of the current methods for analyzing fluoroscopic images for bone tumor repair usually rely on relatively traditional image processing techniques, such as threshold-based segmentation, simple region growing algorithms, or manual annotation. These methods often have certain limitations when dealing with complex tumor morphologies and tiny structures. Due to the lack of voxel-level fine segmentation, they cannot accurately distinguish the tiny differences between tumors and surrounding healthy tissues, which may lead to inaccurate segmentation results, even misjudging the tumor boundary or missing tumor regions. In addition, traditional methods usually cannot effectively process multi-modal imaging data (such as CT, MRI, C-arm images, etc.), and have obvious shortcomings in cross-modal feature alignment and artifact elimination. These limitations make it difficult to meet the high-precision requirements for tumor morphology assessment, infiltration boundary identification, and subsequent bionic stent design, affecting the repair effect and prediction of patient prognosis. Summary of the Invention

[0004] In order to improve the existing method for analyzing fluoroscopic images for bone tumor repair, a method for analyzing fluoroscopic images for bionic repair of bone tumors is provided. This method improves the accuracy of bone tumor images through voxel-level precise segmentation, accurately identifies tumor regions and morphologies, combines convolutional neural networks with morphological operations, significantly improves the segmentation accuracy, and provides reliable support for subsequent tumor infiltration assessment and bionic stent design.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] A method for analyzing fluoroscopic images for bionic repair of bone tumors, comprising:

[0007] Obtaining fluoroscopic image data of the bone tumor region, including X-ray computed tomography (CT), magnetic resonance imaging (MRI), and intraoperative C-arm fluoroscopic images;

[0008] Perform cross-modal feature alignment based on the acquired multi-source image data, construct a three-dimensional bone tumor image model in the three-dimensional voxel space coordinate system, and eliminate image artifacts;

[0009] Based on the acquired three-dimensional bone tumor image data, segment the bone tumor through a deep learning segmentation network, identify the infiltration boundary of the bone tumor, and obtain the three-dimensional morphological parameters and bone tumor feature parameters of the bone tumor region;

[0010] Based on the acquired bone tumor region data, reconstruct the bionic scaffold structure of the bone defect area through biomechanical modeling and implant the bionic scaffold into the bone tumor region;

[0011] Based on the postoperative image data, obtain the bone-scaffold interface fusion rate, stress distribution uniformity, and new bone ingrowth depth indicators.

[0012] Preferably, the acquisition of the perspective image data of the bone tumor region includes X-ray computed tomography (CT), magnetic resonance imaging (MRI), and intraoperative C-arm fluoroscopy images, specifically including:

[0013] Through high-resolution spiral CT, obtain the fine structure and characteristics of bone tissue and tumors, and generate axial tomographic images in DICOM format;

[0014] Through the T1-weighted, T2-weighted, and enhanced sequences of MRI scans, highlight the contrast between tumors and soft tissues;

[0015] Intraoperatively collect two-dimensional X-ray images and transmit data through the DICOM interface of the C-arm.

[0016] Preferably, the cross-modal feature alignment based on the acquired multi-source image data, construction of a three-dimensional bone tumor image model in the three-dimensional voxel space coordinate system, and elimination of image artifacts specifically include:

[0017] Perform global alignment of multi-source image data based on the landmark points in CT image data and the soft tissue contours of MRI;

[0018] Perform local alignment of multi-source image data by compensating for the soft tissue deformation differences between CT and MRI;

[0019] Taking the DICOM coordinate system of preoperative CT as the reference, map the MRI and C-arm data to the CT coordinate system through a registration matrix to construct a three-dimensional voxel space coordinate system;

[0020] Perform voxel-level weighted fusion based on the proportion of the original structure and tumor soft tissue in the bone tumor;

[0021] Based on the fused image data after fusion, match it with the intraoperative C-arm data, and eliminate image artifacts through deep learning;

[0022] Based on the fused image data after artifact elimination processing, a three-dimensional bone tumor image model is constructed in the three-dimensional voxel space coordinate system.

[0023] Preferably, based on the obtained three-dimensional bone tumor image data, the bone tumor is segmented by a deep learning segmentation network, and the bone tumor infiltration boundary is identified. The specific steps for obtaining the three-dimensional morphological parameters and bone tumor feature parameters of the bone tumor region include:

[0024] Based on the obtained three-dimensional bone tumor image data, the bone tumor region is preliminarily segmented by an image segmentation algorithm to obtain the image part containing the bone tumor.

[0025] Based on morphological operations, the image part containing the bone tumor is further refined and segmented, and the segmentation accuracy is improved by a convolutional neural network to obtain voxel-level bone tumor image data.

[0026] Based on the bone tumor characteristics, it is judged whether each voxel is a bone tumor voxel and classified.

[0027] Based on the voxel judgment results of all the image parts containing the bone tumor, the parts judged as non-bone tumor voxels are removed to obtain image data containing only the bone tumor.

[0028] Based on the obtained image data containing only the bone tumor, the bone tumor infiltration boundary is identified, and by analyzing the smoothness of the boundary, it is evaluated whether the tumor invades adjacent regions.

[0029] Based on the identified bone tumor infiltration boundary, the infiltration boundary is marked in the three-dimensional bone tumor image model.

[0030] Based on the marked bone tumor image data, the bone tumor morphological parameters and bone tumor feature parameters are extracted.

[0031] Preferably, the specific steps for judging whether each voxel is a bone tumor voxel and classifying based on the bone tumor characteristics include:

[0032] Calculate the spatial consistency of each voxel and its neighboring voxels. If the spatial consistency between voxels changes greatly, the voxel is located at the boundary of the bone tumor. If the spatial consistency changes little, the voxel is located inside the bone tumor.

[0033] Based on the voxels located at the boundary of the bone tumor, the neighboring regions are analyzed, and it is judged whether the voxel belongs to the tumor by comparing with the attribute characteristics of the surrounding voxels. The characteristics include: gray level difference, texture contrast, and uniformity.

[0034] Preferably, based on the acquired data of the bone tumor region, reconstructing the bionic scaffold structure of the bone defect area through biomechanical modeling and implanting the bionic scaffold into the bone tumor region specifically includes:

[0035] Based on the acquired morphological parameters and characteristic parameters of the bone tumor, obtain the mechanical properties of the bone defect area through biomechanical modeling;

[0036] Based on the mechanical properties and mechanical requirements of the bone tumor region, select bionic materials and design the bionic scaffold structure;

[0037] Perform mechanical simulation through finite element analysis, simulate the load situation of the scaffold in the human body, and optimize the stress, strain distribution and the stability of the scaffold.

[0038] Preferably, based on the postoperative imaging data, obtaining the bone-scaffold interface fusion rate, stress distribution uniformity and new bone ingrowth depth index specifically includes:

[0039] Through the acquired bone-scaffold interface fusion imaging data, extract the bone-scaffold interface and identify the area where the bone contacts the scaffold;

[0040] Based on the ratio of the bone-scaffold contact area to the total bone defect area, calculate and obtain the bone-scaffold interface fusion rate;

[0041] Perform simulation on the postoperative bone mechanics through finite element analysis, calculate and obtain the stress distribution uniformity index, including stress standard deviation and stress concentration index.

[0042] Compared with the prior art, the advantages of the present invention are:

[0043] Precise voxel-level segmentation processing can achieve high-resolution recognition and classification of the bone tumor region. By combining multi-source imaging data (such as CT, MRI and C-arm fluoroscopy images), this method first constructs an accurate bone tumor imaging model in the three-dimensional voxel space, effectively eliminating imaging artifacts. On this basis, morphological operations and convolutional neural network (CNN) are used to further optimize the segmentation accuracy of the tumor region, realizing the precise classification of bone tumor voxels. This voxel-level segmentation can accurately distinguish bone tumors from surrounding healthy tissues, eliminate irrelevant background information, thereby improving the accuracy of subsequent analysis and treatment planning. In addition, through careful identification of the infiltration boundary, it is possible to evaluate whether the tumor has invaded adjacent bone regions, providing strong support for formulating personalized surgical plans and implanting scaffold designs. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 It is a schematic diagram of the method proposed by the present invention;

[0045] Figure 2Schematic diagram of obtaining bone tumor imaging data proposed by the present invention;

[0046] Figure 3 Schematic diagram of constructing a three-dimensional bone tumor imaging model proposed by the present invention;

[0047] Figure 4 Schematic diagram of bone tumor image processing proposed by the present invention;

[0048] Figure 5 Schematic diagram of bone tumor voxel judgment proposed by the present invention;

[0049] Figure 6 Schematic diagram of reconstructing the bone defect area proposed by the present invention;

[0050] Figure 7 Schematic diagram of postoperative image analysis proposed by the present invention;

[0051] Figure 8 Architecture diagram of the electronic device in this solution;

[0052] Figure 9 Schematic diagram of the structure of the computer-readable storage medium in this solution. Specific implementation manners

[0053] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variations.

[0054] Example 1:

[0055] Referring to Figure 1 As shown, a perspective image analysis method for bone tumor bionic repair includes:

[0056] Step 1: Obtain perspective image data of the bone tumor area, including X-ray computed tomography (CT), magnetic resonance imaging (MRI), and intraoperative C-arm fluoroscopy images;

[0057] Step 2: Perform cross-modal feature alignment based on the obtained multi-source image data, construct a three-dimensional bone tumor imaging model in a three-dimensional voxel space coordinate system, and eliminate image artifacts;

[0058] Step 3: Based on the obtained three-dimensional bone tumor image data, segment the bone tumor through a deep learning segmentation network, identify the bone tumor infiltration boundary, and obtain the three-dimensional morphological parameters and bone tumor feature parameters of the bone tumor area;

[0059] Step 4: Based on the obtained bone tumor area data, reconstruct the bionic scaffold structure of the bone defect area through biomechanical modeling, and implant the bionic scaffold into the bone tumor area;

[0060] Step Five: Based on the postoperative imaging data, obtain the indexes of bone-stent interface fusion rate, stress distribution uniformity, and new bone ingrowth depth.

[0061] Refer to Figure 2 As shown, obtain the fluoroscopic imaging data of the bone tumor area, including X-ray computed tomography (CT), magnetic resonance imaging (MRI), and intraoperative C-arm fluoroscopy images. Specifically include:

[0062] Through high-resolution spiral CT, obtain the fine structure and characteristics of bone tissue and tumors, and generate axial tomographic images in DICOM format;

[0063] Through the T1-weighted, T2-weighted, and enhanced sequences of MRI scans, highlight the contrast between tumors and soft tissues;

[0064] Intraoperatively collect two-dimensional X-ray images in real time, and transmit data through the DICOM interface of the C-arm.

[0065] Specifically, based on the density differences in CT images, use Hounsfield units (HU) to distinguish different tissues: bone tissue usually presents high HU values (e.g., above 1000 HU); tumors (such as malignant tumors) may have low HU values (usually between 0 and 20 HU, depending on the tumor type and its internal components);

[0066] MRI highlights the contrast between tumors and soft tissues through different weighted imaging methods, including T1-weighted imaging, T2-weighted imaging, and enhanced sequences;

[0067] Among them, T1-weighted imaging can display the details of anatomical structures, with good contrast in soft tissues. T1-weighted images are suitable for observing the size, edges, and relationships with surrounding tissues of tumors; T2-weighted imaging has good contrast for tissues with higher water content (such as tumors), and tumors usually appear as high signals (bright white) in T2 images; the enhanced sequence enhances the signals in the tumor area by injecting contrast agents (such as Gd-DTPA), especially suitable for observing tumor vascularization;

[0068] Intraoperative real-time imaging is used for tumor localization and navigation during the operation. The C-arm device displays the patient's internal structures in real time through two-dimensional X-ray images.

[0069] Refer to Figure 3 As shown, perform cross-modal feature alignment based on the obtained multi-source imaging data, construct a three-dimensional bone tumor imaging model in the three-dimensional voxel space coordinate system, and eliminate imaging artifacts. Specifically include:

[0070] Perform global alignment of multi-source imaging data based on the landmark points in CT imaging data and the soft tissue contours of MRI;

[0071] Compensate for the soft tissue deformation differences between CT and MRI to perform local alignment of multi-modal image data;

[0072] Taking the DICOM coordinate system of preoperative CT as the reference, map the MRI and C-arm data to the CT coordinate system through a registration matrix to construct a three-dimensional voxel space coordinate system;

[0073] Perform voxel-level weighted fusion based on the proportions of the original structure and tumor soft tissue in bone tumors;

[0074] Based on the fused image data after fusion, match it with the intraoperative C-arm data, and eliminate image artifacts through deep learning;

[0075] Based on the fused image data after artifact elimination processing, construct a three-dimensional bone tumor image model in the three-dimensional voxel space coordinate system.

[0076] Specifically, the selection of landmark points in CT images is mainly the edges of bone structures or tumor boundaries. During global alignment, through rigid registration methods, rotation and translation are used for global alignment to maintain the positional relationship of landmark points between the two images. Affine registration can also be used, adopting scaling and shear transformations to increase the flexibility of registration;

[0077] During local alignment, a non-rigid registration method is used to compensate for the soft tissue deformation differences between CT and MRI, calculate the deformation field between images, and locally adjust the alignment of MRI and CT images. The calculation of the deformation field is usually based on image similarity metrics (such as mutual information, mean square error), and the formula is: Y = X + D(X), where Y is the deformed point, X is the original point, and D(X) is the deformation field;

[0078] Through the aforementioned rigid and non-rigid registration steps, obtain the registration matrix that maps the MRI and C-arm data to the CT coordinate system, and transform the MRI and C-arm images from their original coordinate systems to the coordinate system of preoperative CT;

[0079] According to the proportions of bone tissue and tumor soft tissue in CT images, define the weighted coefficients for different tissue types,

[0080] For bone tissue, the high-density areas of CT have higher weights; for soft tissue, the high-signal areas of MRI are given higher weights. The weighted fusion formula is:

[0081] V fused =w bone ·V CT +w tumor ·V MRI

[0082] where, w bone 、w tumoris the weighting coefficient for bone tissue and tumor soft tissue, V CT , V MRI are the voxel values in CT and MRI images;

[0083] Based on the processed fused image data, a three-dimensional model is constructed through a three-dimensional reconstruction algorithm, and further analysis is performed on the reconstructed three-dimensional bone tumor image model.

[0084] Example two:

[0085] In this embodiment, referring to Figure 4 as shown, based on the acquired three-dimensional bone tumor image data, the bone tumor is segmented through a deep learning segmentation network, and the bone tumor infiltration boundary is identified. The specific three-dimensional morphological parameters and bone tumor characteristic parameters of the bone tumor region include:

[0086] Based on the acquired three-dimensional bone tumor image data, the bone tumor region is initially segmented through an image segmentation algorithm to obtain the image part containing the bone tumor;

[0087] Based on morphological operations, the image part containing the bone tumor is further refined and segmented, and the segmentation accuracy is improved through a convolutional neural network to obtain voxel-level bone tumor image data;

[0088] Based on the bone tumor characteristics, each voxel is judged whether it is a bone tumor voxel and classified;

[0089] Based on the voxel judgment results of all image parts containing the bone tumor, the parts judged as non-bone tumor voxels are removed to obtain image data containing only the bone tumor;

[0090] Based on the acquired image data containing only the bone tumor, the bone tumor infiltration boundary is identified, and by analyzing the smoothness of the boundary, it is evaluated whether the tumor has invaded adjacent regions;

[0091] Based on the identified bone tumor infiltration boundary, the infiltration boundary is marked in the three-dimensional bone tumor image model;

[0092] Based on the marked bone tumor image data, the bone tumor morphological parameters and bone tumor characteristic parameters are extracted.

[0093] Specifically, the initially segmented image data containing the bone tumor is obtained through an image segmentation algorithm, and then small artifacts and noises are removed through morphological operations such as dilation, erosion, opening operation, and closing operation, the boundary is smoothed, and the segmentation quality is improved. Through the deep learning method of the 3D U-Net model, the refined segmentation of the bone tumor special image data is trained and output as voxel-level bone tumor image segmentation data;

[0094] Based on spatial consistency, determine whether each voxel is a bone tumor voxel. Judge whether the tumor invades adjacent tissues through the infiltration boundary of the tumor. Extract the boundary of the tumor through the edge detection algorithm of the Sobel operator. The formula is as follows:

[0095]

[0096] Among them, G x and G y are the gradients of the image in the horizontal and vertical directions;

[0097] Evaluate whether there is a phenomenon of the tumor invading adjacent regions through the boundary smoothness index. The smoothness can be represented by the boundary curvature. The formula is as follows:

[0098]

[0099] Among them, y(s) is the position function on the boundary, and s is the arc length parameter;

[0100] Based on the labeled bone tumor image data, the extraction of morphological parameters includes extracting morphological features such as the volume, surface area, aspect ratio, roundness, and surface complexity of the tumor. The extraction of characteristic parameters includes extracting the texture features, density distribution, heterogeneity, etc. of the tumor, and further analyzing the invasiveness and malignancy of the tumor.

[0101] Refer to Figure 5 As shown, judging whether each voxel is a bone tumor voxel and classifying based on bone tumor characteristics specifically include:

[0102] Calculate the spatial consistency of each voxel and its neighboring voxels. If the spatial consistency between voxels changes greatly, then the voxel is located at the boundary of the bone tumor. If the spatial consistency changes little, then the voxel is located inside the bone tumor;

[0103] Based on the voxels located at the boundary of the bone tumor, analyze the aligned neighborhood, and judge whether the voxel belongs to the tumor by comparing with the attribute characteristics of the surrounding voxels. The characteristics include: gray level difference, texture contrast, and uniformity.

[0104] Specifically, when judging the position of a voxel in a bone tumor, for each voxel, calculate the gray level difference between it and its neighboring voxels. The gray level value can be represented by the pixel value of the image (such as the Hounsfield unit of CT or the signal intensity of MRI). The gray level difference ΔG is defined as:

[0105]

[0106] Among them, I(v) is the gray level value of voxel v, N(v) is the neighborhood of voxel v, |N(v)| is the number of voxels in the neighborhood, and u is the neighboring voxel;

[0107] Spatial consistency measures the gray-scale consistency between a voxel and its neighboring voxels, which can be represented by the mean or variance of the gray-scale differences. The formula is as follows:

[0108]

[0109] where max(ΔG) is the maximum gray-scale difference between all voxels and their neighbors;

[0110] Higher spatial consistency (close to 1) means that the voxel is located in the internal region of the tumor, while lower spatial consistency (close to 0) means that the voxel is located at the boundary of the tumor;

[0111] Texture contrast is obtained through the contrast index in the gray-level co-occurrence matrix. The formula is as follows:

[0112]

[0113] where p(i,j) is an element in the gray-level co-occurrence matrix, and i and j are the gray levels respectively, which measure the intensity of gray-scale changes;

[0114] Homogeneity is the uniformity of the gray-scale distribution, indicating the degree of concentration of pixel gray-scale values in the image region. It is calculated through the homogeneity in the gray-level co-occurrence matrix. The formula is as follows:

[0115]

[0116] Combining features such as gray-scale differences, texture contrast, and homogeneity, machine learning classification methods (such as support vector machines, decision trees, etc.) are used to determine whether the voxel v belongs to the tumor based on these features.

[0117] Refer to Figure 6 As shown, based on the obtained bone tumor region data, reconstructing the bionic scaffold structure of the bone defect area through biomechanical modeling and implanting the bionic scaffold into the bone tumor region specifically includes:

[0118] Based on the obtained bone tumor morphological parameters and bone tumor feature parameters, through biomechanical modeling, obtain the mechanical properties of the bone defect area;

[0119] Based on the mechanical properties and mechanical requirements of the bone tumor region, select bionic materials and design the bionic scaffold structure;

[0120] Through finite element analysis for mechanical simulation, simulate the load situation of the scaffold in the human body, and optimize the stress, strain distribution, and the stability of the scaffold.

[0121] Specifically, the mechanical properties of bone tissue are usually represented by the elastic modulus and Poisson's ratio. Assuming that the elastic modulus of the tumor region is obtained through derivation by a biomechanical model, a commonly used model is to estimate based on tissue density. The formula is as follows:

[0122]

[0123] Among them, E tumor is the elastic modulus, ρ tumor is the density of the tumor region, and α and β are empirical parameters;

[0124] Through finite element analysis, the geometry of the bone tumor is transformed into a finite element mesh, and the material properties are defined. Necessary boundary conditions and external loads are applied to the model to simulate the actual use environment. The finite element method is used to solve the displacement, stress, and strain distributions of the structure under the loading conditions. The stress of each voxel can be calculated through the stress-strain relationship, and the formula is:

[0125] σ = E material ·ε

[0126] Among them, ε is the strain, which can be derived from the displacement field and geometric relationship: Δu is the voxel displacement, and Δx is the length of the voxel;

[0127] Based on the model, evaluate its stress, strain, displacement and other results, and analyze the strength and stability of the structure.

[0128] Example 3:

[0129] Refer to Figure 7 As shown, on the basis of Example 2, this example further proposes to obtain the bone-stent interface fusion rate, stress distribution uniformity, and new bone ingrowth depth index based on postoperative imaging data, specifically including:

[0130] Through the obtained bone-stent interface fusion imaging data, extract the bone-stent interface and identify the area where the bone contacts the stent;

[0131] Based on the ratio of the bone-stent contact area to the total bone defect area, calculate the bone-stent interface fusion rate;

[0132] Through finite element analysis, simulate the postoperative bone mechanics, and calculate the uniformity index of the stress distribution, including the stress standard deviation and the stress concentration index.

[0133] Specifically, based on the segmentation result, extract the area where the bone contacts the stent. In the three-dimensional image, by constructing the bone-stent contact surface, extract the part where the bone tissue intersects the stent surface. Based on the extracted bone-stent contact area and the total bone defect area, calculate the bone-stent interface fusion rate;

[0134] The stress standard deviation in the uniformity index of the stress distribution describes the uniformity of the stress distribution, and the calculation formula is:

[0135]

[0136] where σ i is the stress value of the i-th voxel, is the average value of the stress, and N is the total number of voxels;

[0137] The stress concentration index in the uniformity index of stress distribution measures the stress concentration phenomenon, which is usually defined as the ratio of the maximum stress to the average stress, and the formula is:

[0138]

[0139] where σ max is the maximum stress value;

[0140] The uniformity of the stress distribution is evaluated by the stress standard deviation and the stress concentration index. Ideally, a smaller stress standard deviation and a lower stress concentration index indicate a uniform load distribution.

[0141] Furthermore, the method according to the embodiment of the present application can also be implemented by means of Figure 8 the architecture of the electronic device shown. As Figure 8 shown, the electronic device 500 may include a bus 501, one or more CPUs 502, a read-only memory (ROM) 503, a random access memory (RAM) 504, a communication port 505 connected to the network, an input / output component 506, a hard disk 507, etc. The storage device in the electronic device 500, such as the ROM 503 or the hard disk 507, can store a method for analyzing fluoroscopic images for bone tumor bionic repair provided by the present application. The electronic device 500 may also include a terminal interface 508. Of course, Figure 8 the architecture shown is only exemplary, and when implementing different devices, one or more components shown in the Figure 8 electronic device may be omitted according to actual needs.

[0142] Figure 9 is a schematic diagram of the structure of a computer-readable storage medium provided by an embodiment of the present application. As Figure 9 shown, it is a computer-readable storage medium 600 according to an embodiment of the present application. Computer-readable instructions are stored on the computer-readable storage medium 600. When the computer-readable instructions are run by a processor, a method for analyzing fluoroscopic images for bone tumor bionic repair according to the embodiment of the present application described with reference to the above drawings can be executed. The storage medium 600 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory, etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0143] The advantages of the present invention are as follows: Through precise voxel-level segmentation processing, high-resolution recognition and classification of bone tumor regions can be achieved. By combining multi-source imaging data (such as CT, MRI, and C-arm fluoroscopy images), the method first constructs an accurate bone tumor imaging model in the three-dimensional voxel space, effectively eliminating image artifacts. On this basis, morphological operations and convolutional neural networks (CNNs) are used to further optimize the segmentation accuracy of the tumor region, realizing the accurate classification of bone tumor voxels. This voxel-level segmentation can accurately distinguish bone tumors from surrounding healthy tissues, eliminating irrelevant background information, thereby improving the accuracy of subsequent analysis and treatment planning. In addition, through careful identification of the infiltration boundary, it is possible to evaluate whether the tumor has invaded adjacent bone regions, providing strong support for formulating personalized surgical plans and implant stent designs.

[0144] It should be noted that: The above sequence of embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. And the above specific embodiments of this specification have been described. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0145] Each embodiment in this specification is described in a progressive manner. For the same or similar parts among the embodiments, reference can be made to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

[0146] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A perspective image analysis method for bionic repair of bone tumors, characterized in that Including: Obtaining fluoroscopic image data of the bone tumor region, including X-ray computed tomography (CT), magnetic resonance imaging (MRI), and intraoperative C-arm fluoroscopic images; Performing cross-modal feature alignment based on the obtained multi-source image data, constructing a three-dimensional bone tumor image model in a three-dimensional voxel space coordinate system, and eliminating image artifacts; Based on the obtained three-dimensional bone tumor image data, segmenting the bone tumor through a deep learning segmentation network, identifying the infiltration boundary of the bone tumor, and obtaining three-dimensional morphological parameters and bone tumor feature parameters of the bone tumor region; Based on the obtained bone tumor region data, reconstructing the bionic scaffold structure of the bone defect area through biomechanical modeling and implanting the bionic scaffold into the bone tumor region; Based on the postoperative image data, obtaining indicators such as the bone-scaffold interface fusion rate, stress distribution uniformity, and new bone ingrowth depth.

2. The perspective image analysis method for bionic repair of bone tumors according to claim 1, characterized in that, The obtaining of the fluoroscopic image data of the bone tumor region, including X-ray computed tomography (CT), magnetic resonance imaging (MRI), and intraoperative C-arm fluoroscopic images specifically includes: Through high-resolution spiral CT, obtaining the fine structure and characteristics of bone tissue and tumors, and generating axial tomographic images in DICOM format; Through the T1-weighted, T2-weighted, and enhanced sequences of MRI scans, highlighting the contrast between tumors and soft tissues; During the operation, two-dimensional X-ray images are collected in real time, and data is transmitted through the DICOM interface of the C-arm.

3. A perspective image analysis method for bone tumor bionic repair according to claim 1, characterized in that, The performing of cross-modal feature alignment based on the obtained multi-source image data, constructing a three-dimensional bone tumor image model in a three-dimensional voxel space coordinate system, and eliminating image artifacts specifically includes: Performing global alignment of multi-source image data based on the landmark points in CT image data and the soft tissue contours of MRI; Performing local alignment of multi-source image data by compensating for the soft tissue deformation differences between CT and MRI; Taking the DICOM coordinate system of preoperative CT as the reference, mapping MRI and C-arm data to the CT coordinate system through a registration matrix to construct a three-dimensional voxel space coordinate system; Performing voxel-level weighted fusion based on the proportion of the old structure and tumor soft tissue in the bone tumor; Based on the fused image data after fusion, matching it with the intraoperative C-arm data, and eliminating image artifacts through deep learning; Based on the fused image data after artifact elimination processing, constructing a three-dimensional bone tumor image model in a three-dimensional voxel space coordinate system.

4. A perspective image analysis method for bionic repair of bone tumors according to claim 1, characterized in that The segmenting of the bone tumor through a deep learning segmentation network based on the obtained three-dimensional bone tumor image data, identifying the infiltration boundary of the bone tumor, and obtaining three-dimensional morphological parameters and bone tumor feature parameters of the bone tumor region specifically includes: Based on the obtained three-dimensional bone tumor image data, initially segmenting the bone tumor region through an image segmentation algorithm to obtain the image part containing the bone tumor; Performing further refined segmentation on the image part containing the bone tumor based on morphological operations, improving the segmentation accuracy through a convolutional neural network, and obtaining voxel-level bone tumor image data; Judging whether each voxel is a bone tumor voxel and classifying it based on bone tumor characteristics; Based on the voxel judgment results of all image parts containing bone tumors, the parts judged as non-bone tumor voxels are removed to obtain image data containing only bone tumors; Based on the obtained image data containing only bone tumors, the infiltration boundary of the bone tumor is identified, and by analyzing the smoothness of the boundary, it is evaluated whether the tumor shows the phenomenon of invading adjacent areas; Based on the identified infiltration boundary of the bone tumor, the infiltration boundary is marked in the three-dimensional bone tumor image model; Based on the marked bone tumor image data, the morphological parameters and characteristic parameters of the bone tumor are extracted.

5. The perspective image analysis method for bionic repair of bone tumors according to claim 4, wherein, The judgment and classification of each voxel as a bone tumor voxel based on bone tumor characteristics specifically include: Calculate the spatial consistency of each voxel and its neighboring voxels. If the change in spatial consistency between voxels is large, the voxel is located at the boundary of the bone tumor; if the change in spatial consistency is small, the voxel is located inside the bone tumor; Based on the voxels located at the boundary of the bone tumor, analyze the aligned neighborhood, and judge whether the voxel belongs to the tumor by comparing with the attribute characteristics of the surrounding voxels. The characteristics include: gray difference, texture contrast, uniformity.

6. A perspective image analysis method for bone tumor bionic repair according to claim 1, characterized in that The reconstruction of the bionic scaffold structure of the bone defect area and the implantation of the bionic scaffold into the bone tumor area based on the obtained bone tumor region data specifically include: Based on the obtained morphological parameters and characteristic parameters of the bone tumor, obtain the mechanical properties of the bone defect area through biomechanical modeling; Based on the mechanical properties and mechanical requirements of the bone tumor area, select bionic materials and design the bionic scaffold structure; Perform mechanical simulation through finite element analysis to simulate the load situation of the scaffold in the human body and optimize the stress, strain distribution and the stability of the scaffold.

7. A perspective image analysis method for bionic repair of bone tumors according to claim 1, characterized in that The obtaining of the bone-scaffold interface fusion rate, stress distribution uniformity and new bone ingrowth depth index based on the postoperative image data specifically includes: Through the obtained bone-scaffold interface fusion image data, extract the bone-scaffold interface and identify the area where the bone contacts the scaffold; Based on the ratio of the bone-scaffold contact area to the total bone defect area, calculate and obtain the bone-scaffold interface fusion rate; Perform simulation on the postoperative bone mechanics through finite element analysis, and calculate and obtain the uniformity index of the stress distribution, including stress standard deviation and stress concentration index.

8. An electronic device, characterized in that, Including: At least one processor; And a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute a method for analyzing perspective images for bionic repair of bone tumors as described in any one of claims 1-7.

9. A computer-readable storage medium storing computer-readable instructions, characterized in that, When the computer-readable instructions are executed by the processor, a method for analyzing perspective images for bionic repair of bone tumors as described in any one of claims 1-7 is implemented.

Images