Bone joint lesion area automatic labeling method and system based on image processing

Abstract

The invention discloses a bone joint lesion area automatic labeling method and system based on image processing, and belongs to the field of bone joint lesion area labeling. According to the invention, a bone joint reflection signal is obtained by combining the polarization-state-adjustable laser light source with polarization-sensitive optical coherence tomography, and a three-dimensional birefringence distribution diagram is generated through matrix analysis. And after non-local mean filtering is carried out on the distribution map, a lesion edge contour is extracted based on a dynamic gradient amplitude, and a binary mask is generated. Mask pixels are converted into network nodes containing space coordinates, gradient directions and birefringence, and a topological network is constructed based on the Euclidean distance of adjacent pixels. The weight is calculated through the node intersection density and the birefringence, and a graph attention network is utilized to aggregate multi-layer features to extract lesion topological features. The features are mapped to a three-dimensional space to construct a lesion probability distribution field, an interactive labeling result is generated through contour surface extraction and transparency mixed rendering, and the accuracy of labeling the complex lesion area of the bone joint can be improved.

Description

Technical Field

[0001] This application belongs to the field of bone and joint lesion area annotation, and particularly relates to an automatic annotation method and system for bone and joint lesion areas based on image processing. Background Art

[0002] Accurate annotation of bone and joint lesions is an important basis for the diagnosis and treatment planning of bone and joint diseases, and has key application value especially in the fields of personalized medicine and surgical navigation. Traditional annotation methods mostly rely on manual interpretation of imaging features or semi-automatic segmentation, but their efficiency and consistency are difficult to meet the refined clinical requirements, and there is an urgent need to develop high-precision and automated annotation technologies.

[0003] In the prior art, bone and joint lesion annotation is mainly based on threshold segmentation, region growing algorithms, and machine learning models. For example, imaging techniques such as X-ray and MRI often extract lesion areas through gray-scale threshold segmentation or use morphological operations to enhance feature contrast. Deep learning methods have also been used for bone and joint image segmentation.

[0004] However, the methods in the prior art have significant defects in the annotation of complex lesions such as multi-branch fissures and three-dimensional osteophytes. Threshold segmentation is easily affected by noise and is difficult to distinguish the gradual boundary between lesions and normal tissues. Edge detection has insufficient sensitivity to the texture features of micro-fissures, resulting in missed detections or breaks. Deep learning methods rely on large-scale annotated data and have limited modeling capabilities for complex topological structures. Therefore, there is a technical problem of poor annotation accuracy in the bone and joint lesion area annotation methods of the prior art when annotating complex lesion areas. Summary of the Invention

[0005] This application actually provides an automatic annotation method and system for bone and joint lesion areas based on image processing, which can improve the accuracy of annotating complex bone and joint lesion areas.

[0006] In a first aspect, this application provides an automatic annotation method for bone and joint lesion areas based on image processing, which is applicable to bone and joint lesion areas with osteophytes or cartilage fissures. The method includes: Emitting an optical signal to the bone and joint area through a polarization-state tunable laser light source, receiving the reflected signal through polarization-sensitive optical coherence tomography, and calculating the birefringence distribution using matrix analysis to obtain a three-dimensional birefringence distribution map including collagen fiber orientation information; Performing non-local means filtering on the birefringence distribution map, and extracting the edge contour of the target lesion area based on the gradient amplitude range to generate binary mask data that retains the texture features of micro-fissures. The gradient amplitude range is dynamically adjusted based on the local structural characteristics of the birefringence distribution map; Convert the pixel points in the binarized mask data into nodes, and construct a topological network representing the fissure branches and osteophyte morphology through the connection relationships between the nodes. The node attributes of the nodes include spatial coordinates, gradient directions, and birefringence. The connection relationships between the nodes are determined by the Euclidean distances of adjacent pixel points in the binarized mask data; Based on the connection relationships between the nodes, determine the node weights of the nodes according to the intersection density and birefringence of the nodes, and use a graph attention network for multi-layer feature aggregation to extract the lesion topological features of the target lesion area; Map the lesion topological features to the spatial coordinate system of the three-dimensional birefringence distribution map to generate a lesion probability distribution field, and based on the lesion probability distribution field, extract and blend-render through isosurfaces to generate an interactive annotation result that supports three-dimensional space rotation and multi-plane slice display.

[0007] In an implementable embodiment, based on the connection relationships between the nodes, determine the node weights of the nodes according to the intersection density and birefringence of the nodes, and use a graph attention network for multi-layer feature aggregation to extract the lesion topological features of the target lesion area, including: Based on the connection relationships between the nodes, determine the node weights of the nodes according to the intersection density, birefringence, and gradient direction consistency of the nodes, and use a two-channel graph attention network to aggregate the node attributes and geometric relationship features respectively. After fusion, extract the lesion topological features of the target lesion area. The intersection density is the ratio of the number of adjacent edges of the node to the preset maximum number of connection edges.

[0008] In an implementable embodiment, based on the connection relationships between the nodes, determine the node weights of the nodes according to the intersection density, birefringence, and gradient direction consistency of the nodes, and use a two-channel graph attention network to aggregate the node attributes and geometric relationship features respectively. After fusion, extract the lesion topological features of the target lesion area, including: Calculate the gradient direction consistency parameter of the node according to the included angle between the gradient direction of the node and the gradient directions of adjacent nodes; Based on the intersection density, birefringence, and gradient direction consistency parameter of the node, generate the initial node weight of the node by weighted summation; Based on the spatial coordinates, birefringence, and gradient direction of the initial node weight node, perform weighted aggregation to generate the node attribute features representing the biochemical characteristics of the lesion; According to the ratio of the Euclidean distance of the connection relationship between the nodes to the initial node weight, construct the geometric relationship strength of the node to generate the geometric relationship features representing the morphological distribution of the lesion; Input the node attribute features and geometric relationship features into the two-channel graph attention network, and perform weighted fusion on the node attribute features and geometric relationship features according to the initial node weight to generate the lesion topological features including the fissure branch topological structure and osteophyte morphology.

[0009] In an implementable embodiment, an optical signal is emitted to the bone and joint region by a polarization-state tunable laser light source, the reflected signal is received by polarization-sensitive optical coherence tomography, and the birefringence distribution is calculated by matrix analysis to obtain a three-dimensional birefringence distribution map including collagen fiber orientation information, including: By adjusting the wavelength and polarization direction parameters of the polarization-state tunable laser light source, multi-angle polarized light signals are emitted to the superficial and deep layers of cartilage in the bone and joint region, and the multi-angle polarized light signals cover the arrangement directions of collagen fibers in the superficial and deep layers of cartilage; The interference signals reflected by the multi-angle polarized light signals at different depth layers in the bone and joint region are received, and the local birefringence value of the cartilage is analyzed according to the phase delay amount and polarization-state change of the interference signals. The phase delay amount and polarization-state change are generated by the difference in the arrangement directions of collagen fibers in the cartilage; The local birefringence values are mapped to the three-dimensional space coordinate system of the bone and joint region, and a birefringence gradient distribution map associated with the collagen fiber orientation is generated to obtain a three-dimensional birefringence distribution map. The gradient direction in the birefringence gradient distribution map is orthogonal to the collagen fiber arrangement direction.

[0010] In an implementable embodiment, non-local means filtering is performed on the birefringence distribution map, and the edge contour of the target lesion area is extracted based on the gradient amplitude range to generate binary mask data retaining microcrack texture features, including: According to the consistency of the collagen fiber arrangement directions in adjacent regions in the birefringence distribution map, the local structural uniformity parameter of the birefringence distribution map is calculated. The local structural uniformity parameter is used to characterize the fiber continuity in the bone and joint region; Based on the local structural uniformity parameter, the gradient amplitude range is adjusted. The gradient amplitude range is expanded in the region where the local structural uniformity parameter is lower than the set threshold, and the gradient amplitude range is reduced in the region where the local structural uniformity parameter is higher than the set threshold; Traversing the birefringence distribution map according to the adjusted gradient amplitude range, the pixel points within the gradient amplitude range and matching the gradient direction are marked as edge contours, and binary mask data including the topological connection relationship of microcracks is generated.

[0011] In an implementable embodiment, the pixel points in the binary mask data are converted into nodes, and a topological network representing the crack branches and osteophyte morphology is constructed through the connection relationships between the nodes, including: Traversing the pixel points in the binary mask data, the pixel points satisfying the preset gradient direction continuity are marked as initial nodes. The preset gradient direction continuity is determined by the included angle between the gradient direction of the pixel point and the gradient direction of the adjacent pixel point and the preset angle threshold; Calculate the spatial distribution density and birefringence difference degree of the initial nodes based on the spatial coordinates, gradient directions, and birefringence indices of the initial nodes, and select the initial nodes with birefringence difference degrees greater than a preset difference threshold as target nodes; Taking the target nodes as the center, generate connection edges according to the Euclidean distances between adjacent pixel points in the binary mask data to construct a topological network representing the morphology of fissure branches and osteophytes. The maximum connection distance of the connection edges is determined according to the birefringence difference degree.

[0012] In an implementable embodiment, map the lesion topological features to the spatial coordinate system of the three-dimensional birefringence distribution map to generate a lesion probability distribution field, and based on the lesion probability distribution field, perform isosurface extraction and transparency blending rendering to generate an interactive annotation result supporting three-dimensional space rotation and multi-plane slicing display, including: According to the target node weights in the lesion topological features and the spatial coordinates of the three-dimensional birefringence distribution map, establish a probability mapping relationship between the target node weights and the spatial coordinates to generate a three-dimensional probability distribution field including lesion probability gradients; Based on the lesion probability gradient direction of the three-dimensional probability distribution field and the collagen fiber orientation information of the birefringence distribution map, determine the isosurface boundary of the lesion probability distribution field. The extraction range of the isosurface boundary is determined by the product of the lesion probability gradient and the birefringence difference degree; Align the isosurface boundary with the spatial coordinates of the three-dimensional birefringence distribution map, assign transparency parameters according to the birefringence difference degree, and perform transparency blending rendering on the area covered by the isosurface boundary to generate an interactive annotation result supporting multi-plane slicing display.

[0013] In a second aspect, the present application provides an automatic annotation system for osteoarticular lesion regions based on image processing, which is applicable to osteoarticular lesion regions with osteophytes or cartilage fissures. The system includes: An emission module for emitting an optical signal to the osteoarticular region through a polarization-state tunable laser light source, receiving the reflected signal through polarization-sensitive optical coherence tomography, and calculating the birefringence distribution using matrix analysis to obtain a three-dimensional birefringence distribution map including collagen fiber orientation information; An extraction module for performing non-local mean filtering on the birefringence distribution map and extracting the edge contour of the target lesion region based on the gradient amplitude range to generate binary mask data retaining the micro-fissure texture features. The gradient amplitude range is dynamically adjusted based on the local structural characteristics of the birefringence distribution map; A construction module for converting the pixel points in the binary mask data into nodes and constructing a topological network representing the morphology of fissure branches and osteophytes through the connection relationship between the nodes. The node attributes of the nodes include spatial coordinates, gradient directions, and birefringence, and the connection relationship between the nodes is determined by the Euclidean distances between adjacent pixel points in the binary mask data; The extraction module is further configured to determine the node weight of a node based on the connection relationship between nodes, the convergence density and the birefringence of the node, and perform multi-layer feature aggregation using a graph attention network to extract the lesion topological features of the target lesion area; The generation module is configured to map the lesion topological features to the spatial coordinate system of the three-dimensional birefringence distribution map to generate a lesion probability distribution field, and generate an interactive annotation result that supports three-dimensional space rotation and multi-plane slice display based on the lesion probability distribution field through isosurface extraction and transparency blending rendering.

[0014] In a third aspect, the present application provides an electronic device, which includes: a processor, and a memory storing computer program instructions; the processor reads and executes the computer program instructions to implement the automatic annotation method for bone and joint lesion areas based on image processing in any one of the embodiments of the first aspect.

[0015] In a fourth aspect, the present application provides a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the automatic annotation method for bone and joint lesion areas based on image processing in any one of the embodiments of the first aspect is implemented.

[0016] The automatic annotation method and system for bone and joint lesion areas based on image processing implemented in the present application can accurately capture the subtle changes in the orientation of collagen fibers and effectively distinguish the gradual boundary between the lesion tissue and the normal tissue by obtaining a three-dimensional birefringence distribution map through polarization-sensitive optical coherence tomography. By combining the dynamic gradient amplitude adjustment strategy with non-local mean filtering, while retaining the continuous texture features of microcracks, it suppresses the artifact interference and solves the problem of insufficient sensitivity of edge detection to fine structures. Through constructing a topological network and integrating the dual-channel graph attention mechanism of node attributes and geometric relationships, it realizes the adaptive modeling of complex structures such as multi-branched cracks and three-dimensional osteophytes, breaking through the limitations of traditional deep learning models on data scale and topological modeling capabilities. The three-dimensional visualization technology based on the lesion probability distribution field spatially correlates the pathological features with the anatomical structure and provides an interactive annotation result for clinical use. The present application reduces the need for manual intervention while improving the accuracy of complex lesion area annotation.

[0017] Furthermore, based on the joint calculation of the gradient direction consistency parameter and the intersection density, the continuous trend of microcracks and their branch intersection characteristics can be effectively identified. Secondly, by introducing a dual-channel graph attention network to aggregate node attribute features and geometric relationship features respectively, the collaborative modeling of the biochemical characteristics and morphological distribution of the lesion area is realized, breaking through the one-sidedness of single feature analysis in the characterization of complex structures. Through the adaptive fusion mechanism of geometric relationship strength and node weight, the discriminative ability of the network for the three-dimensional morphology of osteophytes and the spatial distribution of crack branches is enhanced. By means of refined feature extraction and multi-level topological correlation analysis, this method significantly improves the parsing ability for gradually changing boundaries and complex lesion structures while reducing the dependence on large-scale labeled data, and improves the annotation accuracy of the osteoarticular lesion area with multi-branch cracks and osteophyte morphology. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the accompanying drawings required for use in the embodiments of the present application will be briefly introduced below. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

[0019] Figure 1 is a schematic flowchart of a method for automatically annotating an osteoarticular lesion area based on image processing provided by an embodiment of the present application; Figure 2 is a schematic flowchart of a method for automatically annotating an osteoarticular lesion area based on image processing provided by an embodiment of the present application; Figure 3 is a schematic flowchart of a method for automatically annotating an osteoarticular lesion area based on image processing provided by an embodiment of the present application; Figure 4 is a schematic structural diagram of an automatic annotation system for an osteoarticular lesion area based on image processing provided by an embodiment of the present application; Figure 5 is a schematic hardware structure diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] The features and exemplary embodiments of all aspects of the present application will be described in detail below. In order to make the objectives, technical solutions and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only intended to explain the present application, rather than to limit the present application. For those skilled in the art, the present application can be implemented without some of these specific details. The following description of the embodiments is only intended to provide a better understanding of the present application by showing examples of the present application.

[0021] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the said element.

[0022] In the prior art, the fine and irregular branch structure of multi-branch fissures easily leads to the failure of traditional algorithms to capture local features, and the heterogeneous calcification morphology of three-dimensional osteophytes and the low contrast with surrounding tissues result in blurred boundaries in threshold-based segmentation. Moreover, deep learning methods are limited by the three-dimensional space topology modeling ability and the generalization of complex lesion features under small samples, and it is difficult to accurately analyze the multi-scale geometric characteristics and spatial continuity of such lesions. Therefore, there is a technical problem of poor annotation accuracy in the method for annotating osteoarticular lesion regions in the prior art when annotating complex lesion regions.

[0023] To solve the problems of the prior art, the embodiments of the present application provide an automatic annotation method, system, device and computer storage medium for osteoarticular lesion regions based on image processing. First, the automatic annotation method for osteoarticular lesion regions based on image processing provided by the embodiments of the present application will be introduced below.

[0024] Figure 1 The flowchart of the automatic annotation method for osteoarticular lesion regions based on image processing provided by an embodiment of the present application is shown. As Figure 1 shown, it includes steps S110 to S150. This method is applicable to osteoarticular lesion regions with osteophytes or cartilage fissures, and this method includes: S110: Emitting an optical signal to the osteoarticular region through a polarization-state tunable laser light source, receiving the reflected signal through polarization-sensitive optical coherence tomography, and calculating the birefringence distribution by matrix analysis to obtain a three-dimensional birefringence distribution map including collagen fiber orientation information.

[0025] The osteoarthritic lesion area of osteophytes or cartilage fissures refers to the abnormal structures formed in articular cartilage or bone due to degeneration and injury. Osteophytes are manifested as abnormal hyperplasia at the bone margins, while cartilage fissures are microscopic cracks in cartilage tissue. A polarization-state tunable laser light source refers to an optical device that can adjust the polarization direction of the output light wave, and realizes multi-angle detection by changing the relative angle between the polarization state of the incident light and the orientation of cartilage fibers. Polarization-Sensitive Optical Coherence Tomography (PS-OCT) is a technique that combines optical coherence tomography and polarization-state analysis, and analyzes the birefringence characteristics of a sample by detecting the change in the polarization state of the reflected light. The birefringence distribution describes the difference in birefringence values at different positions in the tissue, reflecting the degree of order of collagen fiber arrangement. The collagen fiber orientation information refers to the fiber direction data derived from the birefringence gradient direction. The three-dimensional birefringence distribution map is a three-dimensional image formed by mapping birefringence values to a three-dimensional space coordinate system, and is used to characterize the three-dimensional abnormal area of fiber arrangement.

[0026] First, a detection beam with a preset wavelength and polarization direction is output by controlling the laser light source to perform multi-angle scanning on the osteoarthritic area. The reflected light signal is transmitted to the interferometer through a polarization-maintaining fiber, and after interfering with the reference light, a time-domain signal carrying polarization information is generated. Then, a polarization analysis algorithm based on the Jones matrix is used to calculate the birefringence value of each sampling point according to the phase difference of the interference signal and the change in polarization state. The probe is controlled to move in the transverse and axial directions by a three-dimensional scanning mechanism, and the birefringence values measured point by point are mapped to the space coordinate system. Finally, using the orthogonality between the birefringence gradient direction and the collagen fiber arrangement direction, a three-dimensional distribution map showing the fiber orientation is generated. Among them, the phase difference calculation is realized by comparing the displacement amount of the interference fringes of incident light and reflected light with different polarization states, and the birefringence gradient direction is determined by the maximum direction of the birefringence change of adjacent voxels.

[0027] S120: Perform non-local mean filtering on the birefringence distribution map, and extract the edge contour of the target lesion area based on the gradient amplitude range to generate binarized mask data that retains the texture features of microcracks. The gradient amplitude range is dynamically adjusted based on the local structural characteristics of the birefringence distribution map.

[0028] Non-local means filtering is a method for noise reduction by calculating the weighted average of similar regions in an image. Its core lies in utilizing the global structural similarity of the image rather than local neighborhood information. The gradient magnitude range refers to the effective interval of the gradient intensity values of pixels in the image, which is used to distinguish edges from background noise. The microcrack texture feature characterizes the fiber fracture regions with widths less than millimeter level in bone and joint lesions, manifested as local gradient direction mutations and numerical discontinuities in the birefringence distribution map. Binary mask data is a binary image that classifies image pixels into edge and non-edge regions, where edge pixels are marked as foreground. The local structural property refers to the spatial correlation of the arrangement direction and density of collagen fibers in adjacent regions in the birefringence distribution map, reflecting the continuity of tissue structure.

[0029] First, perform non-local means filtering on the birefringence distribution map. By calculating the weight coefficients between all similar blocks in the image, perform weighted average processing on the current pixel to suppress speckle noise while retaining the texture details of microcracks. Subsequently, dynamically adjust the gradient magnitude range based on the local structural properties of the filtered image: narrow the gradient threshold in regions with uniform collagen fiber arrangement to reduce noise interference, and expand the gradient threshold in regions with fiber fractures or disordered arrangements to capture the edges of microcracks. By traversing the entire image, extract the pixel points that meet the dynamic gradient threshold and have continuous directions as the edge contours. Finally, map the edge pixels to binary mask data, where the topological connection relationship of microcracks is maintained through the consistency of the gradient directions of adjacent edge pixels.

[0030] S130: Convert the pixel points in the binary mask data into nodes, and construct a topological network representing the morphology of crack branches and osteophytes through the connection relationships between nodes. The node attributes of nodes include spatial coordinates, gradient direction, and birefringence. The connection relationships between nodes are determined by the Euclidean distance between adjacent pixel points in the binary mask data.

[0031] The spatial coordinates are the position information of the node in the three-dimensional coordinate system. The gradient direction represents the orthogonal direction of the collagen fiber arrangement direction at this node in the birefringence distribution map, characterizing the anisotropic characteristics of local tissue. The Euclidean distance between adjacent pixel points refers to the geometric distance between adjacent edge pixels in the three-dimensional space in the binary mask data, which is used to quantify the spatial correlation strength between nodes.

[0032] First, traverse the edge pixel points in the binarized mask data, convert each valid pixel into a graph node, and the node attributes inherit the spatial coordinates, gradient direction, and birefringence value at the corresponding position in the original birefringence distribution map. Subsequently, construct the connection relationship based on the Euclidean distance between adjacent pixels, calculate the three-dimensional spatial distance between each node and its neighboring pixels, and establish a connection edge if the distance is less than the preset threshold. Further optimize the connection relationship. Strengthen the connection strength for adjacent nodes with continuous gradient directions and significant birefringence differences, while weaken the connection for nodes with sudden changes in gradient directions or similar birefringence values. By iteratively traversing all nodes, a topological network representing the morphology of fissure branches and osteophytes is formed. Among them, the fissure area shows a long-chain connection structure, while the osteophyte area presents a dense network connection pattern. The threshold setting of the connection edge needs to meet the spatial scale smaller than the typical fissure width, which can be dynamically adjusted according to the spatial resolution of the birefringence distribution map to ensure that the topological network can accurately reflect the lesion morphology.

[0033] S140: Based on the connection relationship between nodes, determine the node weights of the nodes according to the intersection density and birefringence of the nodes, and use a graph attention network for multi-layer feature aggregation to extract the lesion topological features of the target lesion area.

[0034] The intersection density is the ratio of the number of connection edges of a node in the topological network to the preset maximum possible number of connection edges, reflecting the key degree of the node in the lesion structure. The graph attention network is a deep learning model based on the graph structure, which focuses on the key node features through an adaptive weight allocation mechanism. The lesion topological features refer to the high-dimensional feature vectors generated by fusing node attributes and geometric relationships through a dual-channel graph attention network, representing the topological structure of fissure branches and the three-dimensional morphology of osteophytes. It is obtained by weighted aggregation of the features of key nodes through the self-attention mechanism and can include the geometric connection pattern of fissure branches and the density distribution information of osteophytes.

[0035] First, calculate the intersection density according to the number of adjacent edges of the node, and generate the initial weight in combination with the birefringence value of the node. Assign higher weights to nodes with high intersection density and abnormal birefringence to reflect their key role in the lesion structure. Subsequently, use a multi-layer graph attention network for feature aggregation. In each layer, calculate the feature correlation between adjacent nodes through the attention mechanism and dynamically adjust the feature propagation weight. Through layer-by-layer iteration, the node features gradually fuse the local neighborhood information and the global topological structure. Finally, extract the high-order topological features including the extension trend of fissure branches and the growth pattern of osteophytes. Among them, the birefringence difference is used to enhance the characterization of the biochemical properties of the lesion area, while the intersection density strengthens the feature expression of the key nodes of the topological connection.

[0036] S150: Map the topological features of the lesion to the spatial coordinate system of the three-dimensional birefringence distribution map to generate a lesion probability distribution field. Based on the lesion probability distribution field, extract the isosurface and perform transparency blending rendering to generate an interactive annotation result that supports three-dimensional spatial rotation and multi-plane slicing display.

[0037] The lesion probability distribution field is a probability density field formed by mapping the topological features of the lesion to three-dimensional space, representing the possibility of each spatial position belonging to the lesion area. Isosurface extraction is to extract the surface boundary with the same probability value from the three-dimensional field by setting a probability threshold, which is used to define the spatial range of the lesion area. Transparency blending rendering is a technology that superimposes and displays the three-dimensional annotation result and the original image data, and realizes multi-level visualization by adjusting the transparency of different regions. The interactive annotation result is a visualization model that supports three-dimensional rotation and slicing operations, allowing users to observe the lesion morphology from multiple angles.

[0038] First, associate the node weights in the topological features of the lesion with the three-dimensional spatial coordinates, and generate a continuous probability distribution field through an interpolation algorithm. Based on the correlation between the probability gradient direction and the collagen fiber orientation, use the isosurface extraction algorithm to determine the lesion boundary. Among them, the probability gradient region corresponds to the tip of the crack or the edge of the osteophyte. Subsequently, spatially align the extracted isosurface with the original birefringence distribution map, and assign transparency parameters according to the birefringence difference. A gradual transparency is adopted in the junction area between the lesion and the normal tissue to highlight the boundary, and the transparency is reduced in the core area of the lesion to enhance the display. Finally, through the graphics rendering engine, the three-dimensional model can be interactively operated to support multi-plane slicing to view the internal structure of the lesion.

[0039] In this embodiment, a three-dimensional birefringence distribution map is obtained through polarization-sensitive optical coherence tomography, which can accurately capture the subtle changes in the orientation of collagen fibers and effectively distinguish the gradual boundary between the lesion tissue and the normal tissue. By combining the dynamic gradient amplitude adjustment strategy and non-local mean filtering, while retaining the continuous texture features of microcracks, it suppresses the artifact interference and solves the problem of insufficient sensitivity of edge detection to fine structures. By constructing a topological network and fusing the dual-channel graph attention mechanism of node attributes and geometric relationships, the adaptive modeling of complex structures such as multi-branched cracks and three-dimensional osteophytes is realized, breaking through the limitations of traditional deep learning models on data scale and topological modeling ability. The three-dimensional visualization technology based on the lesion probability distribution field spatially correlates the pathological features with the anatomical structure, providing an interactive annotation result for clinical practice. This application reduces the need for manual intervention while improving the accuracy of annotating complex lesion areas.

[0040] In the prior art, due to the inability to effectively distinguish the gradual boundary between the lesion area and normal tissues and the insufficient ability to model complex topological structures such as multi-branched fissures, missed detections or fractures may occur. To address the above deficiencies, a gradient direction consistency parameter is introduced to quantify the fiber fracture mode, and a dual-channel graph attention network is used to process the biochemical characteristics and spatial distribution characteristics in parallel, strengthening the accurate modeling of complex lesion topological structures.

[0041] In an implementable embodiment, step S140: Based on the connection relationship between nodes, determine the node weight of each node according to the intersection density and birefringence of the nodes, and use a graph attention network for multi-layer feature aggregation to extract the lesion topological features of the target lesion area, including: Based on the connection relationship between nodes, determine the node weight of each node according to the intersection density, birefringence and gradient direction consistency of the nodes, and use a dual-channel graph attention network to aggregate the node attribute and geometric relationship features respectively. After fusion, extract the lesion topological features of the target lesion area. The intersection density is the ratio of the number of adjacent edges of the node to the preset maximum number of connected edges.

[0042] The gradient direction consistency refers to the cosine value of the angle between the gradient directions of a node and its adjacent nodes, which is used to quantify the local direction continuity of the fiber fracture area. The dual-channel graph attention network is a deep learning model that processes node attributes and geometric relationships in parallel. The attribute channel focuses on biochemical characteristics, and the geometric channel learns the spatial distribution law, and dynamically adjusts the feature contribution degree through the attention mechanism. Node attributes include inherent features such as spatial coordinates, gradient directions and birefringence, reflecting the physical properties of local tissues. The geometric relationship features are constructed by the distance and weight of the connecting edges between nodes, characterizing the spatial distribution law of the lesion morphology.

[0043] First, calculate the angle between the gradient directions of each node and its adjacent nodes, and generate a direction consistency parameter as an evaluation index for the topological connection quality. Weightedly fuse the intersection density, birefringence abnormality and direction consistency parameter to generate an initial node weight. The node attribute channel extracts biochemical features reflecting the degree of collagen fiber fracture by weightedly aggregating spatial coordinates, gradient directions and birefringence. The geometric relationship channel constructs the relationship strength based on the inverse ratio of the Euclidean distance of the connecting edges to the initial weight, and captures the extension mode of the fissure branches. The dual-channel graph attention network performs attention weighting on the two types of features respectively, and generates topological features including fissure bifurcation patterns and osteophyte density distributions through feature splicing and fully connected layers.

[0044] Figure 2 The flowchart of the method for extracting the lesion topological features of the target lesion area provided by an embodiment of the present application is shown. As Figure 2 shown, it includes steps S210 to S250.

[0045] In an implementable embodiment, based on the connection relationship between nodes, the node weight of a node is determined according to the intersection density, birefringence, and gradient direction consistency of the node, and a dual-channel graph attention network is used to aggregate the node attributes and geometric relationship features respectively. After fusion, the lesion topological features of the target lesion area are extracted, including: S210: Calculate the gradient direction consistency parameter of a node according to the included angle between the gradient direction of the node and the gradient directions of adjacent nodes.

[0046] The gradient direction consistency parameter is an index that quantifies the continuity of the collagen fiber arrangement direction between a node and its neighborhood nodes. In the osteoarticular lesion area, the fracture or disordered arrangement of collagen fibers will cause local gradient direction mutations. This parameter reflects the local direction consistency of the fiber fracture area by calculating the cosine value of the included angle between the gradient directions of the node and adjacent nodes.

[0047] First, traverse each node in the topological network to obtain its gradient direction vector and the gradient direction vectors of all adjacent nodes. Calculate the cosine value of the included angle between the gradient direction vector of the current node and the gradient direction vectors of each adjacent node. The calculation formula is shown in formula (1).

[0048] (1) Among them, is the cosine value of the included angle between the gradient directions of the node and adjacent nodes, and its value range is [-1, 1]. The closer the value is to 1, the higher the direction consistency. is the gradient direction vector of the current node, which characterizes the arrangement direction of collagen fibers in three-dimensional space. is the gradient direction vector of the adjacent node, and the local direction continuity can be quantified by comparing the direction vectors of adjacent nodes.

[0049] After that, take the average of the cosine values of all adjacent nodes to generate the gradient direction consistency parameter of this node. The value range is [-1, 1]. The closer the value is to 1, the higher the direction consistency.

[0050] S220: Based on the intersection density, birefringence, and gradient direction consistency parameter of the node, generate the initial node weight of the node by weighted summation.

[0051] The birefringence difference degree refers to the absolute deviation between the birefringence value of a node and the neighborhood mean value, which reflects the degree of local biochemical property abnormality.

[0052] First, count the actual number of connected edges of the node and divide it by the preset maximum possible number of connected edges to calculate the node intersection density D; then compare the birefringence value of the node with the normal tissue reference value to calculate the birefringence abnormality B; finally, perform linear weighted summation on the intersection density, birefringence abnormality, and gradient direction consistency parameter to obtain the initial node weight . Specifically, the process of calculating the initial node weight is shown in Formula (2).

[0053] (2) Where is the initial node weight, is the preset weight coefficient, C is the gradient direction consistency parameter; D is the convergence density, , E is the actual number of connected edges of the node, is the preset maximum possible number of connected edges; B is the birefringence anomaly, , b is the birefringence value of the node, is the birefringence reference value of normal tissue.

[0054] S230: Based on the spatial coordinates, birefringence, and gradient direction of the nodes with initial node weights, perform weighted aggregation to generate node attribute features representing the biochemical characteristics of the lesion.

[0055] The node attribute feature is a multi-dimensional feature vector that fuses the spatial position, biochemical characteristics, and direction information of the node. The spatial coordinates reflect the three-dimensional distribution of the lesion area, the birefringence is related to the orderliness of collagen fiber arrangement, and the gradient direction characterizes the fiber fracture pattern.

[0056] First, standardize the spatial coordinates of the nodes to eliminate the dimensional difference. Then, construct a node attribute matrix, where each row contains the standardized coordinates, birefringence value, and gradient direction vector. Finally, perform weighted aggregation on the attribute matrix based on the initial node weights to obtain the node attribute features. The calculation formula is shown in Formula (3).

[0057] (3) Where is the node attribute feature, a multi-dimensional feature vector that fuses spatial coordinates, birefringence, and gradient direction information; is the initial node weight of the i-th node, used to adjust the contribution of the node attribute during the aggregation process; is the node attribute vector, containing the standardized spatial coordinates, birefringence value, and gradient direction vector.

[0058] S240: According to the ratio of the Euclidean distance of the connection relationship between nodes to the initial node weight, construct the geometric relationship strength of the nodes to generate geometric relationship features representing the morphological distribution of the lesion.

[0059] The geometric relationship strength refers to quantifying the tightness of the spatial connection between nodes, and reflects the spatial constraint relationship of the lesion morphology through the ratio of the Euclidean distance to the initial weight.

[0060] First, calculate the three-dimensional geometric distance based on the normalized spatial coordinates of the nodes to obtain the Euclidean distance between the nodes. Then, construct a geometric relationship strength matrix based on the initial node weights and the Euclidean distance between the nodes. Finally, generate geometric relationship features through normalization processing of the geometric relationship strength matrix. Specifically, the process of calculating the geometric relationship strength is shown in Equation (4).

[0061] (4) where is the geometric relationship strength between the i-th node and the j-th node, is the initial node weight of the i-th node, is the Euclidean distance between the i-th node and the j-th node, , is the node spatial coordinate.

[0062] S250: Input the node attribute features and geometric relationship features into a dual-channel graph attention network, and perform weighted fusion on the node attribute features and geometric relationship features according to the initial node weights to generate lesion topological features including fissure branch topologies and osteophyte morphologies.

[0063] First, initialize the dual-channel graph attention network and set two independent attention heads for the two channels. Then, input the node attribute feature matrix into the attribute channel and the geometric relationship feature matrix into the geometric channel. Within each channel, calculate the feature correlation through the self-attention mechanism, as shown in Equation (5).

[0064] (5) where is the attention weight, and the feature correlation between nodes is calculated through the self-attention mechanism, and are the query vector and key vector respectively, mapped from the node attribute features and geometric relationship features; d is the feature dimension, used to scale the attention score to avoid gradient vanishing.

[0065] After that, perform weighted fusion on the outputs of the two channels based on the initial node weights to obtain the lesion topological features. The specific process is shown in Equation (6).

[0066] (6) where is the fusion weight coefficient, is the weighted fusion feature, is the node attribute feature, is the geometric relationship feature.

[0067] During the processing of the dual-channel graph attention network, the node attribute features and geometric relationship features are first input into independent channels for feature learning. The node attribute channel focuses on the biochemical characteristics and spatial distribution of the nodes themselves. Through the self-attention mechanism, the correlation between node attributes is calculated, and the feature propagation weights are dynamically adjusted to strengthen the influence of key biochemical indicators on the lesion area. The geometric relationship channel focuses on the spatial connection patterns between nodes. Using the self-attention mechanism, it captures the extension rules of fissure branches and the reticular distribution characteristics of osteophytes, and quantifies the spatial constraint relationship of the lesion morphology through the geometric relationship intensity matrix. Inside each channel, multiple attention heads are used in parallel to extract feature representations at different abstraction levels, forming multi-dimensional feature vectors.

[0068] Subsequently, adaptive weighted fusion is performed on the outputs of the two channels based on the initial node weights. The initial weights integrate the intersection density, birefringence abnormality, and gradient direction consistency of the nodes, and can reflect the importance of the nodes in the lesion structure. During the fusion process, higher weights are assigned to the attribute channel to highlight the differences in biochemical characteristics, and at the same time, the spatial connection patterns of the geometric channel are combined to generate composite features that integrate the topological structure of the fissure branches and the three-dimensional morphology of osteophytes. The finally output lesion topological features include both the fiber fracture pattern and bifurcation angle information of microfissures, and also integrate the density gradient and spatial growth direction of the osteophyte area, forming a multi-dimensional representation of the complex lesion area.

[0069] Exemplarily, taking the knee joint cartilage fissure as an example, the process of extracting the lesion topological features of the target lesion area is as follows. First, traverse the nodes of the topological network to calculate the gradient direction consistency parameter. For each node and its adjacent nodes, the cosine value of the included angle of the gradient direction vectors is calculated through formula (1), and the mean value is taken to generate the direction continuity index. Assuming that the cosine values of the included angles of the gradient directions with three adjacent nodes are (0.45, 0.38, 0.51), the gradient direction consistency parameter 0.45 is obtained after averaging. Subsequently, based on the intersection density, that is, the ratio of the actual number of connected edges to the preset maximum number of edges, the birefringence abnormality, that is, the absolute deviation of the node value from the normal tissue reference value, and the gradient direction consistency parameter, the initial node weights are linearly weighted according to formula (2). Set the intersection density weight α = 0.4, the birefringence abnormality weight β = 0.5, and the direction consistency weight γ = 0.1. Combining the intersection density 0.8 and the birefringence abnormality 0.3, the initial weight 0.4×0.8 + 0.5×0.3 + 0.1×(1 - 0.45) = 0.61 is generated.

[0070] Next, construct a node attribute feature matrix, which includes normalized spatial coordinates, birefringence values, and normalized gradient direction vectors, and aggregate the features by weighting with the initial weights in formula (3). Based on these weights, aggregate the three-dimensional coordinates of the node (12.5, 7.2, 3.8), the birefringence of 1.0, and the gradient direction vector (0.7, -0.1, 0.7) to generate an attribute feature vector [0.32, 0.18, 0.24, 0.61]. At the same time, calculate the geometric relationship strength matrix between nodes according to formula (4), and construct a spatial constraint relationship using the inverse ratio of the Euclidean distance and the initial weights. Calculate the Euclidean distances (2.1μm, 1.8μm, 2.3μm) between this node and its adjacent nodes, and construct a geometric relationship strength matrix [0.61 / 2.1, 0.61 / 1.8, 0.61 / 2.3]. Finally, input the attribute features and geometric relationship features into a two-channel graph attention network. Capture birefringence anomalies in the attribute channel, strengthen short-distance connections in the geometric channel, and fuse the features to identify this bifurcation point as a key node for crack extension. After extracting the feature correlations respectively through the self-attention mechanism in formula (5), perform weighted splicing according to the fusion weight of λ = 0.6 in formula (6) to generate high-order topological features that integrate biochemical characteristics and morphological distributions.

[0071] In an implementable embodiment, step S110: Emits an optical signal to the bone and joint area through a polarization-state tunable laser light source, receives the reflected signal through polarization-sensitive optical coherence tomography, and calculates the birefringence distribution using matrix analysis to obtain a three-dimensional birefringence distribution map including collagen fiber orientation information, including: By adjusting the wavelength and polarization direction parameters of the polarization-state tunable laser light source, emit multi-angle polarized optical signals to the surface layer and deep layer of the cartilage in the bone and joint area. The multi-angle polarized optical signals cover the arrangement directions of the collagen fibers in the surface layer and deep layer of the cartilage.

[0072] The multi-angle polarized optical signal refers to a beam sequence with different polarization states formed by adjusting the polarization direction of the light source output. In the bone and joint detection scenario, the collagen fibers in the surface layer of the cartilage are arranged horizontally, and the fibers in the deep layer gradually transition to vertical arrangement. The multi-angle polarized light can cover the fiber orientations of different depth layers. The polarization direction parameter refers to the electric field vibration direction of the beam, which is continuously adjusted by rotating a wave plate or a liquid crystal modulator to make the polarization direction of the incident light form a specific angle with the fiber orientation, thereby enhancing the difference in the reflected signals of fibers with different orientations.

[0073] First, configure a laser light source with adjustable polarization state. Select the near-infrared band of 840 nm to balance tissue penetration depth and resolution. Generate incident light with three sets of orthogonal polarization directions of 0°, 45°, and 90° through a polarization controller. Each set of polarized light covers the surface to deep regions of cartilage in a fan-shaped scanning mode. During the scanning process, dynamically adjust the polarization direction according to the preset depth stratification strategy: horizontal polarized light is used for surface scanning to match the surface fiber arrangement direction, and vertical polarized light is switched to for deep scanning to enhance the deep fiber reflection signal. The light source output power is dynamically adjusted according to tissue transmittance to ensure balanced signal intensity at each depth layer.

[0074] Receive the interference signals reflected from different depth layers in the osteoarticular region for multi-angle polarized light signals. Analyze the local birefringence value of cartilage based on the phase delay amount and polarization state change of the interference signals. The phase delay amount and polarization state change are generated by the difference in the arrangement direction of collagen fibers in cartilage.

[0075] The phase delay amount refers to the phase difference between two orthogonal polarization components caused by the birefringence effect when polarized light propagates in tissue. This difference is directly related to the local collagen fiber arrangement density and direction. The polarization state change refers to the change amount of the polarization direction of the reflected light relative to the incident light, and the polarization modulation effect caused by birefringence can be quantitatively described by the Jones matrix.

[0076] The interference signal is generated by the superposition of the reflected light from the reference arm and the sample arm. After being split by a spectrometer, the time-domain interference fringes are obtained. Adopt a polarization analysis algorithm based on the Jones matrix to decompose the interference signals corresponding to the incident light of each polarization direction: First, extract the interference intensities of the orthogonal polarization channels and calculate the phase difference at each sampling point; then solve the local birefringence value through matrix inverse operation, and the calculation process is shown in formula (7).

[0077] Δn = λΔφ / (4πd) (7) Among them, Δn is the local birefringence value, Δφ is the phase delay amount, d is the optical path difference, and λ is the light source wavelength.

[0078] For multi-angle polarization data, adopt the maximum likelihood estimation method to fuse the measurement results of different polarization directions, eliminate the measurement deviation caused by anisotropy, and finally generate a depth-resolved birefringence distribution curve.

[0079] Map the local birefringence value to the three-dimensional space coordinate system of the osteoarticular region to generate a birefringence gradient distribution map associated with the collagen fiber orientation to obtain a three-dimensional birefringence distribution map. The gradient direction in the birefringence gradient distribution map is orthogonal to the collagen fiber arrangement direction.

[0080] The birefringence gradient distribution map is a vector field generated by calculating the change rate of birefringence between adjacent voxels in three-dimensional space, and its gradient direction is orthogonal to the orientation of collagen fibers. This map can visually display areas with disordered fiber arrangements. For example, at the crack, the gradient direction changes abruptly due to fiber breakage.

[0081] Map the calculated birefringence values to a three-dimensional Cartesian coordinate system, and perform a three-dimensional gradient operation on each voxel: calculate the difference values of birefringence in the X / Y / Z directions respectively, and synthesize the gradient vectors. Use a Gaussian kernel function to smooth the gradient field to suppress noise interference. Generate a pseudo-color coded image according to the magnitude and direction of the gradient vectors. Among them, the magnitude represents the orderliness of fiber arrangement, and a high magnitude corresponds to a highly ordered area. The direction is represented in the HSV color space, where the hue corresponds to the azimuth angle and the saturation corresponds to the elevation angle. In the finally output three-dimensional distribution map, the normal cartilage area shows continuous and gradual colors, while the diseased area shows discrete color blocks or sharp color bands.

[0082] Exemplarily, taking the diseased area of the knee joint as an example, first, use an 840nm superluminescent diode light source, and a polarization-maintaining fiber is coupled to a handheld PS-OCT probe, equipped with a high-speed galvanometer to achieve three-dimensional scanning, with a lateral resolution of 15μm and an axial resolution of 10μm. Then, data acquisition is carried out. The sample is scanned in layers, with each layer having a thickness of 50μm, and a total of 200 layers are acquired. Each layer is irradiated in turn with polarization directions of 0°, 45°, and 90°, and interference signals of 1024×1024 pixels are acquired in each direction. Then, signal analysis is carried out. The Jones matrix elements are calculated for each pixel point, and the birefringence value is extracted through singular value decomposition. After fusing the polarization data of three layers, the birefringence of the cartilage surface layer is 0.002±0.0003, and that of the deep layer is 0.0015±0.0002. Abnormal values below 0.0008 appear in the crack area. And gradient calculation is carried out. The Sobel operator is used to calculate the three-dimensional gradient, and a peak value of the gradient magnitude (>0.0015 / μm) is detected at the crack edge, and the gradient direction has a 90° difference from the surrounding normal area. Finally, the generated three-dimensional birefringence gradient distribution map is displayed. The normal cartilage shows a blue gradient, that is, the gradient directions are consistent, while the crack area suddenly changes to a red patch, that is, the gradient directions are orthogonal, and the osteophyte area shows an annular gradient distribution. This embodiment realizes the high-precision three-dimensional characterization of the orientation of collagen fibers and provides reliable joint feature data for subsequent lesion annotation.

[0083] In an implementable embodiment, step S120: Perform non-local mean filtering on the birefringence distribution map, and extract the edge contour of the target lesion area based on the gradient magnitude range to generate binary mask data that retains the microcrack texture features, including: Calculate the local structural uniformity parameter of the birefringence distribution map based on the consistency of the collagen fiber arrangement directions in adjacent regions of the birefringence distribution map. The local structural uniformity parameter is used to characterize the fiber continuity in the bone and joint region.

[0084] The local structural uniformity parameter is an index that quantifies the consistency of the collagen fiber arrangement directions in adjacent regions of the birefringence distribution map. In the bone and joint detection scenario, the collagen fibers in normal cartilage are arranged with directional continuity, while the diseased regions (such as microcracks) will cause local fiber orientation disorders. This parameter reflects the degree of tissue structure continuity by statistically analyzing the similarity of the gradient directions of adjacent pixels. A higher value indicates more orderly fiber arrangement and a lower risk of disease.

[0085] First, define a local neighborhood window (such as 5×5 pixels) in the birefringence distribution map, and calculate the gradient direction vector for each pixel within the window. Then, use the cosine similarity to measure the consistency of the gradient directions between the central pixel and the neighborhood pixels, and calculate the mean cosine of the angles between all neighborhood pixels and the central pixel as the uniformity parameter of this window. For three-dimensional data, the same operation needs to be performed along the X / Y / Z three axes, and finally a three-dimensional uniformity parameter matrix is generated. The parameter value range is from -1 to 1, with values close to 1 indicating high uniformity and values close to -1 indicating severe disorder.

[0086] Adjust the gradient amplitude range based on the local structural uniformity parameter. Expand the gradient amplitude range in the regions where the local structural uniformity parameter is lower than the set threshold, and shrink the gradient amplitude range in the regions where the local structural uniformity parameter is higher than the set threshold.

[0087] The gradient amplitude range refers to the effective interval of the gradient intensity values of the pixel points in the image, which is used to distinguish edges from background noise. Its dynamic adjustment strategy is realized based on the local structural uniformity parameter. Narrow the threshold in the regions with uniform fiber arrangement to reduce false detection of noise, and expand the threshold in the regions with disordered structures to capture real lesion edges.

[0088] Divide the regions according to the uniformity parameter matrix, set a uniformity threshold. For example, the threshold is 0.7. Regions above the threshold are determined to be normal tissues, and a fixed narrow threshold, such as the top 10% of the gradient amplitude, is used as the upper limit; regions below the threshold are determined to be potential lesions, and an adaptive wide threshold, such as the top 30% as the upper limit. Specifically, for each pixel, determine whether it belongs to the edge according to the threshold range of the region it belongs to. If the pixel gradient amplitude is within the threshold interval of this region, it is marked as a candidate edge point. This process needs to traverse the entire three-dimensional distribution map and adjust the threshold layer by layer.

[0089] Traverse the birefringence distribution map according to the adjusted gradient amplitude range, and mark the pixel points within the gradient amplitude range and matching the gradient direction as the edge contour, generating binary mask data including the topological connection relationship of microcracks.

[0090] The topological connection relationship of microcracks refers to the crack branch structure reconstructed through the spatial connectivity of edge pixels. The binary mask data is a three-dimensional binary image that marks edge pixels, where foreground pixels (value 1) represent the lesion edge and background pixels (value 0) represent normal tissue.

[0091] Perform direction consistency verification on the candidate edge points after dynamic threshold processing. Calculate the gradient direction angle between each edge point and its adjacent edge points. If the angle is less than the preset angle tolerance, such as the preset angle tolerance is 30°, it is determined as the continuous edge of the same crack branch. Connect the edge points that meet the direction consistency through the three-dimensional region growing algorithm, and remove isolated noise points. In the finally generated three-dimensional mask data, each connected region represents a crack or osteophyte boundary, and its topological structure is completely retained through the pixel adjacency relationship.

[0092] Exemplarily, taking the detection of microcracks in knee joint cartilage as an example, first preprocess the three-dimensional birefringence distribution map generated by polarization-sensitive optical coherence tomography: define a local neighborhood window of 5×5×3 (X / Y / Z), calculate the gradient direction vector of the pixels in each window, such as using the Sobel operator, and through cosine similarity, for example, when the gradient direction vector of the central pixel is [0.8, 0.6, 0] and the neighboring pixel is [0.7, 0.7, 0], the cosine similarity is 0.98, statistically analyze the direction consistency of adjacent pixels, and generate a three-dimensional uniformity parameter matrix; set the uniformity threshold to 0.7. When the parameter of a certain region is lower than the threshold, such as the parameter around the microcrack is 0.3 - 0.5, expand the upper limit of the gradient amplitude range of this region from the default 10% to 30%. After traversing and adjusting, screen the candidate edge points with gradient amplitude in [15, 85] (normalized value) and direction angle less than 30°, for example, the gradient direction of the crack edge points is concentrated in 45°±10°. Subsequently, connect adjacent points based on the three-dimensional region growing algorithm. For example, a crack branch needs to meet the direction consistency of 5 consecutive pixels. Finally, generate binary mask data containing 12 independent crack branches, and the error between its topological structure and the actual microscopic observation of the cartilage microcrack network is less than 5μm. This embodiment realizes the high-precision extraction of the lesion edge and provides reliable input for subsequent topological network construction and feature enhancement.

[0093] In an implementable embodiment, step S130: Convert the pixel points in the binary mask data into nodes, and construct a topological network representing the crack branch and osteophyte morphology through the connection relationship between the nodes, including: Traverse the pixel points in the binary mask data, and mark the pixel points that meet the preset gradient direction continuity as initial nodes. The preset gradient direction continuity is determined by the gradient direction angle between the pixel point and the gradient direction of the adjacent pixel point and the preset angle threshold.

[0094] Preset gradient direction continuity means screening pixel points with continuous fiber arrangement by setting a threshold for the included angle of gradient directions of adjacent pixel points. In the scenario of bone and joint lesions, the sudden change of gradient direction in the microfracture or osteophyte area will lead to an increase in the difference in adjacent pixel directions, while the fiber arrangement in the normal area shows direction consistency.

[0095] Traverse each edge pixel in the binary mask data, and calculate the included angle between its gradient direction vector and the adjacent pixel in the neighborhood. For two dimensions, it is an 8-neighborhood, and for three dimensions, it is a 26-neighborhood. If the included angles of all neighborhood pixels are less than the preset threshold, mark this pixel as an initial node. Specifically, the cosine value is calculated using the dot product of vectors, and the angle is obtained by taking the inverse cosine. For example, when the gradient direction vectors are 0.8, 0.6, 0 and 0.7, 0.7, 0, the cosine value is 0.98, corresponding to an included angle of about 12°, meeting the continuity requirement. The finally generated initial node set contains edge pixels with continuous direction and spatial coherence.

[0096] Based on the spatial coordinates, gradient direction, and birefringence of the initial nodes, calculate the spatial distribution density and birefringence difference degree of the initial nodes, and screen out the initial nodes with a birefringence difference degree greater than the preset difference threshold as target nodes.

[0097] The spatial distribution density is an index to measure the aggregation degree of initial nodes in three-dimensional space, calculated by kernel density estimation or neighborhood counting method. The birefringence difference degree refers to the absolute deviation of the birefringence value of the node from the neighborhood mean, reflecting the degree of local biochemical characteristic abnormality. When screening target nodes, nodes with high birefringence difference degree need to be retained to focus on the lesion area.

[0098] First, calculate the spatial distribution density. Taking each initial node as the center, count the number of nodes within a radius of 50 μm. The higher the density value, the stronger the aggregation. The statistical radius can be adjusted according to the resolution. Then, calculate the birefringence difference degree. For each node, calculate the absolute difference between its birefringence value and the mean value within a 10×10×5 voxel neighborhood around it. Finally, perform screening. Set the birefringence difference degree threshold (such as 0.001), and only retain the nodes with a difference degree higher than the threshold. For example, the difference degree of nodes in the normal area is 0.0002, the difference degree of nodes at the edge of the fissure can reach 0.0015, and the difference degree in the osteophyte area is about 0.002.

[0099] Taking the target nodes as the center, generate connection edges according to the Euclidean distance between adjacent pixel points in the binary mask data, and construct a topological network representing the morphology of fissure branches and osteophytes. The maximum connection distance of the connection edges is determined according to the birefringence difference degree.

[0100] The maximum connection distance of the connection edges is the farthest Euclidean distance allowed for connection between nodes that is dynamically adjusted according to the birefringence difference degree. In regions with a high birefringence difference degree, such as the tip of a crack, the connection distance needs to be reduced for accurate modeling, while in regions with a low difference degree, the distance can be appropriately relaxed to capture the overall morphology.

[0101] First, for each target node, calculate the maximum connection distance D_max according to its birefringence difference degree (B). The formula is D_max = D_base × (1 + k×B), where D_base is the base distance, for example, the base distance is 20μm, and k is the adjustment coefficient. Then, with the target node as the center, search for adjacent nodes within its D_max range. If the included angle between the gradient directions of the two is less than the threshold and the birefringence difference is significant (such as the difference degree > 0.0005), then establish a connection edge. Finally, optimize the network structure, use the Delaunay triangulation algorithm to supplement long-distance connections to ensure the continuity of crack branches; perform grid densification processing on the osteophyte region to increase the node connection density.

[0102] Exemplarily, taking the detection of microcracks in knee joint cartilage as an example, first perform initial node marking. Traverse the binary mask data and set the threshold of the included angle of the gradient direction to 30°. The gradient direction of a pixel at the edge of a certain crack is 0.7, 0.7, 00.7, 0.7, 0, and the direction of its neighboring pixels is 0.68, 0.72, 0.10.68, 0.72, 0.1, with an included angle of 25°, which meets the conditions and is marked as an initial node. Then, screen the target nodes. When calculating the spatial distribution density, the density of the neighborhood of a node at a crack bifurcation is 8 nodes / 100μm. When the neighborhood mean is 0.001 and the node value is 0.0023, the birefringence difference degree is 0.0013, which exceeds the threshold of 0.001 and is retained. Finally, perform topological network construction. The birefringence difference degree of this node is 0.0013. Set D_base = 20μm and k = 10, then D_max = 20×(1 + 10×0.0013) = 22.6μm. Find 3 adjacent nodes within the search range, establish connection edges, and form the topology of the crack branch. The final network contains 120 nodes and 350 edges, accurately restoring the three-dimensional morphology of the crack, with an error less than 8μm compared with the microscopic imaging results.

[0103] Figure 3 The flowchart of the method for generating an interactive annotation result provided by an embodiment of the present application is shown. As Figure 3 shown, it includes steps S310 to S330.

[0104] In an implementable embodiment, step S150: Map the lesion topological features to the spatial coordinate system of the three-dimensional birefringence distribution map to generate a lesion probability distribution field, and based on the lesion probability distribution field, extract and blend-render through isosurfaces to generate an interactive annotation result that supports three-dimensional space rotation and multi-plane slicing display, including: S310: According to the target node weights in the lesion topological features and the spatial coordinates of the three-dimensional birefringence distribution map, establish a probability mapping relationship between the target node weights and the spatial coordinates to generate a three-dimensional probability distribution field including the lesion probability gradient.

[0105] The target node weights are high-order feature weights obtained by optimizing and adjusting the initial weights in the dual-channel graph attention network, which integrate the context information of node attributes and geometric relationships and are dynamically allocated through the self-attention mechanism, and are used to identify the spatial distribution weights of fissure bifurcation points and osteophyte core areas. The formation of the lesion topological features directly depends on the target node weights. The weights determine the contribution degree of the nodes in feature aggregation. High-weight nodes dominate the feature expression of fissure bifurcation points and osteophyte core areas, thereby affecting the modeling accuracy of the topological network for lesion structures.

[0106] First, bind the node weights in the lesion topological features to their corresponding three-dimensional spatial coordinates (X / Y / Z) to form weighted discrete point cloud data. Then perform spatial interpolation. Using the inverse distance weighted interpolation algorithm, calculate the probability value for each voxel in the three-dimensional space. The calculation is as shown in formula (8).

[0107] (8) Where, is the node weight, is the Euclidean distance from the voxel to the i-th node, and k is the attenuation coefficient.

[0108] Then perform gradient calculation. Perform a three-dimensional gradient operation on the interpolated probability field to generate a probability gradient vector field. The gradient direction points to the region where the probability increases fastest, usually corresponding to the fissure extension direction or the osteophyte growth edge.

[0109] S320: Based on the lesion probability gradient direction of the three-dimensional probability distribution field and the collagen fiber orientation information of the birefringence distribution map, determine the isosurface boundary of the lesion probability distribution field. The extraction range of the isosurface boundary is determined by the product of the lesion probability gradient and the birefringence difference degree.

[0110] The isosurface boundary is an isosurface extracted from the three-dimensional probability distribution field and is used to define the spatial range of the lesion area. Its extraction range is dynamically adjusted by the product of the lesion probability gradient and the birefringence difference degree: The boundary range is expanded in high-gradient and high-birefringence difference regions, and the range is reduced in low-difference regions to accurately match the pathological features.

[0111] First, perform dynamic threshold setting and calculate the adjustment factor for each voxel. , where is the probability gradient magnitude and B is the birefringence difference degree. Then, use the Marching Cubes algorithm to traverse the probability field and extract the isosurface that meets the threshold condition , where T is the basic probability threshold. Finally, perform boundary optimization, smooth the extracted isosurface, such as Laplacian smoothing, remove isolated patches, and retain continuous surfaces.

[0112] S330: Align the isosurface boundary with the spatial coordinates of the three-dimensional birefringence distribution map, assign transparency parameters according to the birefringence difference degree, and perform transparency blending rendering on the area covered by the isosurface boundary to generate an interactive annotation result that supports multi-planar slice display.

[0113] The transparency parameter is a rendering attribute assigned according to the birefringence difference degree, which is used to control the overlay display effect of the annotation result and the original image. The high-difference area (lesion core) has a low transparency to highlight it, and the low-difference area (transition zone) has a high transparency to weaken the boundary. The interactive annotation result supports three-dimensional rotation, multi-planar slice (MPR), and zoom operations, facilitating multi-angle clinical observation.

[0114] First, perform transparency mapping and establish a linear mapping relationship from the birefringence difference degree B to the transparency : , where and are the ratios of the minimum and maximum difference degrees. Then, perform data overlay, align the isosurface grid with the original OCT volume data space, and use Alpha blending technology for rendering. Among them, the lesion core area ( ): (low transparency); the transition area ( ): ; the normal area ( ): (high transparency). Finally, through a graphics engine, such as VTK, Three.js, support the rotation, slicing, and transparency adjustment of the three-dimensional model to obtain an interactive annotation result that supports multi-planar slice display.

[0115] Exemplarily, the interactive annotation result includes the multi-planar fusion display of the three-dimensional rendered lesion area model and the original OCT image. The annotation result is presented as a semi-transparent colored surface mesh, for example, red represents the crack tip and blue represents the osteophyte edge, which is superimposed on the grayscale OCT volume data, supporting free rotation, zooming, and arbitrary plane slicing operations in three-dimensional space.

[0116] Exemplarily, in the scenario of knee joint cartilage microfracture detection, first, the node weights in the lesion topological features are spatially bound to the spatial coordinates of the three-dimensional birefringence distribution map, and the inverse distance weighted interpolation algorithm is used to calculate the probability value for each voxel. Among them, the weight of the fissure bifurcation node is 0.85, and the weight of the osteophyte core node is 0.92. A three-dimensional field containing probability gradients is generated by setting the attenuation coefficient k = 2. Subsequently, based on the orthogonal relationship between the probability gradient direction and the collagen fiber orientation, the dynamic adjustment factor R = |P| × B is calculated. When the magnitude of the probability gradient |P| = 0.035 / μm and the birefringence difference degree B = 0.0023 in the fissure tip region, the isosurface extraction threshold is automatically extended to T × R = 0.7 × 0.0000805. Finally, the Alpha blending technology is used to overlay the isosurface mesh with the original OCT volume data. The transparency α = 0.2 is assigned to the fissure core region with a birefringence difference degree B > 0.001, α = 0.5 for the transition region, and α = 0.8 for the normal region, generating an interactive annotation model that supports three-dimensional rotation and multi-plane slicing, and this model is used to display the interactive annotation results.

[0117] Based on the same concept, the embodiment of the present application provides an automatic annotation system for osteoarticular lesion regions based on image processing. The following combines Figure 4 to elaborate in detail on the automatic annotation system for osteoarticular lesion regions based on image processing provided by the embodiment of the present application.

[0118] Figure 4 is the structural block diagram of the automatic annotation system for osteoarticular lesion regions based on image processing shown in the embodiment of the present application.

[0119] As Figure 4 shown, this automatic annotation system for osteoarticular lesion regions based on image processing is applicable to osteoarticular lesion regions with osteophytes or cartilage fissures. This system includes: A transmitting module 410, configured to emit an optical signal to the osteoarticular region through a polarization-state adjustable laser light source, receive the reflected signal through polarization-sensitive optical coherence tomography, and use matrix analysis to calculate the birefringence distribution, obtaining a three-dimensional birefringence distribution map including collagen fiber orientation information; An extraction module 420, configured to perform non-local mean filtering on the birefringence distribution map, and extract the edge contour of the target lesion region based on the gradient amplitude range, generating binary mask data that retains the microfracture texture features. The gradient amplitude range is dynamically adjusted based on the local structural characteristics of the birefringence distribution map; A construction module 430, configured to convert the pixel points in the binary mask data into nodes, and construct a topological network representing the fissure branches and osteophyte morphology through the connection relationship between the nodes. The node attributes of the nodes include spatial coordinates, gradient direction, and birefringence. The connection relationship between the nodes is determined by the Euclidean distance of adjacent pixel points in the binary mask data; The extraction module 420 is further configured to determine the node weight of a node based on the connection relationship between nodes, according to the convergence density and birefringence of the nodes, and perform multi-layer feature aggregation using a graph attention network to extract the lesion topological features of the target lesion area; The generation module 440 is configured to map the lesion topological features to the spatial coordinate system of the three-dimensional birefringence distribution map to generate a lesion probability distribution field, and generate an interactive annotation result that supports three-dimensional space rotation and multi-plane slicing display based on the lesion probability distribution field through isosurface extraction and transparency blending rendering.

[0120] In one embodiment, the extraction module 420 is specifically configured to determine the node weight of a node based on the connection relationship between nodes, according to the convergence density, birefringence and gradient direction consistency of the nodes, and use a two-channel graph attention network to aggregate the node attribute and geometric relationship features respectively, and extract the lesion topological features of the target lesion area after fusion. The convergence density is the ratio of the number of adjacent edges of the node to the preset maximum number of connection edges.

[0121] In one embodiment, the extraction module 420 is specifically configured to calculate the gradient direction consistency parameter according to the included angle between the gradient direction of the node and the gradient direction of the adjacent node; based on the convergence density, birefringence and gradient direction consistency parameter of the node, perform weighted summation to generate the initial node weight of the node; perform weighted aggregation based on the spatial coordinates, birefringence and gradient direction of the initial node weight node to generate the node attribute features representing the biochemical characteristics of the lesion; construct the geometric relationship strength of the node according to the ratio of the Euclidean distance of the connection relationship between nodes to the initial node weight, and generate the geometric relationship features representing the morphological distribution of the lesion; input the node attribute features and geometric relationship features into the two-channel graph attention network, and perform weighted fusion on the node attribute features and geometric relationship features according to the initial node weight to generate the lesion topological features including the fissure branch topological structure and osteophyte morphology.

[0122] In one embodiment, the emission module 410 is specifically configured to emit multi-angle polarized light signals to the surface layer and deep layer of the cartilage in the osteoarticular region by adjusting the wavelength and polarization direction parameters of the polarization state tunable laser light source. The multi-angle polarized light signals cover the arrangement directions of the collagen fibers in the surface layer and deep layer of the cartilage; receive the interference signals reflected by the multi-angle polarized light signals at different depth layers in the osteoarticular region, and analyze the local birefringence value of the cartilage according to the phase delay amount and polarization state change of the interference signals. The phase delay amount and polarization state change are generated by the difference in the arrangement directions of the collagen fibers of the cartilage; map the local birefringence value to the three-dimensional spatial coordinate system of the osteoarticular region, and generate a birefringence gradient distribution map associated with the collagen fiber orientation to obtain a three-dimensional birefringence distribution map. The gradient direction in the birefringence gradient distribution map is orthogonal to the arrangement direction of the collagen fibers.

[0123] In one embodiment, the extraction module 420 is specifically configured to calculate the local structural uniformity parameter of the birefringence distribution map according to the consistency of the collagen fiber arrangement directions in adjacent regions of the birefringence distribution map, where the local structural uniformity parameter is used to characterize the fiber continuity of the bone joint region; adjust the gradient amplitude range based on the local structural uniformity parameter, expand the gradient amplitude range in the region where the local structural uniformity parameter is lower than the set threshold, and narrow the gradient amplitude range in the region where the local structural uniformity parameter is higher than the set threshold; traverse the birefringence distribution map according to the adjusted gradient amplitude range, and mark the pixel points within the gradient amplitude range and matching the gradient direction as edge contours to generate binary mask data including the topological connection relationship of microcracks.

[0124] In one embodiment, the construction module 430 is specifically configured to traverse the pixel points in the binary mask data, mark the pixel points that satisfy the preset gradient direction continuity as initial nodes, where the preset gradient direction continuity is determined by the included angle between the gradient direction of the pixel point and the gradient direction of the adjacent pixel point and the preset angle threshold; calculate the spatial distribution density and birefringence difference degree of the initial nodes based on the spatial coordinates, gradient directions, and birefringence of the initial nodes, and screen out the initial nodes with a birefringence difference degree greater than the preset difference threshold as target nodes; take the target nodes as the center, generate connection edges according to the Euclidean distance between adjacent pixel points in the binary mask data, and construct a topological network representing the crack branches and osteophyte morphology, where the maximum connection distance of the connection edges is determined according to the birefringence difference degree.

[0125] In one embodiment, the generation module 440 is specifically configured to establish a probability mapping relationship between the target node weights and the spatial coordinates according to the target node weights in the lesion topological features and the spatial coordinates of the three-dimensional birefringence distribution map, and generate a three-dimensional probability distribution field including the lesion probability gradient; determine the isosurface boundary of the lesion probability distribution field based on the lesion probability gradient direction of the three-dimensional probability distribution field and the collagen fiber orientation information of the birefringence distribution map, where the extraction range of the isosurface boundary is determined by the product of the lesion probability gradient and the birefringence difference degree; align the isosurface boundary with the spatial coordinates of the three-dimensional birefringence distribution map, allocate transparency parameters according to the birefringence difference degree, and perform transparency blending rendering on the area covered by the isosurface boundary to generate an interactive annotation result supporting multi-plane slice display.

[0126] Figure 4 Each module in the system shown has the function of implementing Figures 1 to 3 each step therein and can achieve its corresponding technical effects. For the sake of brevity, it will not be elaborated here.

[0127] Figure 5 FIG. shows a schematic hardware structure diagram of an electronic device provided by an embodiment of the present application.

[0128] The electronic device may include a processor 510 and a memory 520 storing computer program instructions.

[0129] Specifically, the processor 510 may include a central processing unit (CPU), or an application specific integrated circuit (ASIC), or may be configured as one or more integrated circuits for implementing the embodiments of the present application.

[0130] The memory 520 may include a mass storage for data or instructions. By way of example and not limitation, the memory 520 may include a hard disk drive (HDD), a floppy disk drive, a flash memory, an optical disk, a magneto-optical disk, a magnetic tape, or a universal serial bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 520 may include a removable or non-removable (or fixed) medium. Where appropriate, the memory 520 may be internal or external to the integrated gateway disaster recovery device. In a particular embodiment, the memory 520 is a non-volatile solid state memory.

[0131] The memory may include a read-only memory (ROM), a random access memory (RAM), a magnetic disk storage media device, an optical storage media device, a flash memory device, an electrical, optical, or other physical / tangible memory storage device. Thus, in general, the memory includes one or more tangible (non-transitory) computer-readable storage media (e.g., memory devices) encoded with software including computer-executable instructions, and when the software is executed (e.g., by one or more processors), it is operable to perform the operations described with reference to the method according to the first aspect of the present disclosure.

[0132] The processor 510 reads and executes the computer program instructions stored in the memory 520 to implement any of the automatic annotation methods for bone and joint lesion regions based on image processing in the above embodiments.

[0133] In one example, the electronic device may further include a communication interface 530 and a bus 540. Among them, as Figure 5 shown, the processor 510, the memory 520, and the communication interface 530 are connected through the bus 540 to complete communication with each other.

[0134] The communication interface 530 is mainly used to implement communication between the various modules, devices, units, and / or devices in the embodiments of the present application.

[0135] The bus 540 includes hardware, software, or both, and couples the components of the online data flow metering device to each other. By way of example and not limitation, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a MicroChannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses or a combination of two or more of these. Where appropriate, the bus 540 may include one or more buses. Although the embodiments of the present application describe and illustrate specific buses, the present application contemplates any suitable bus or interconnect.

[0136] The electronic device can execute the automatic annotation method for bone and joint lesion regions based on image processing in the embodiments of the present application, so as to implement the combination with Figures 1 to 3 the described automatic annotation method for bone and joint lesion regions based on image processing.

[0137] In addition, in combination with the automatic annotation method for bone and joint lesion regions based on image processing in the above embodiments, the embodiments of the present application can be implemented by providing a computer-readable storage medium. Computer program instructions are stored on the computer-readable storage medium; when the computer program instructions are executed by a processor, any of the automatic annotation methods for bone and joint lesion regions based on image processing in the above embodiments is implemented.

[0138] It should be clear that the present application is not limited to the specific configurations and processes described above and illustrated in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and illustrated as examples. However, the method process of the present application is not limited to the specific steps described and illustrated, and those skilled in the art can make various changes, modifications, and additions, or change the order between steps after understanding the spirit of the present application.

[0139] The functional blocks shown in the above-described structural block diagrams can be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, it can be, for example, an electronic circuit, an application-specific integrated circuit (ASIC), appropriate firmware, a plug-in, a functional card, and so on. When implemented in software, the elements of the present application are programs or code segments for performing the required tasks. The program or code segment can be stored in a machine-readable medium or transmitted via a data signal carried in a carrier wave over a transmission medium or a communication link. "Machine-readable medium" can include any medium that can store or transmit information. Examples of machine-readable media include electronic circuits, semiconductor memory devices, ROM, flash memory, erasable ROM (EROM), floppy disks, CD-ROMs, optical discs, hard disks, fiber optic media, radio frequency (RF) links, and so on. The code segment can be downloaded via a computer network such as the Internet, an intranet, and so on.

[0140] It should also be noted that the exemplary embodiments mentioned in the present application describe some methods or systems based on a series of steps or devices. However, the present application is not limited to the order of the above steps, that is, the steps can be executed in the order mentioned in the embodiments, or different from the order in the embodiments, or several steps can be executed simultaneously.

[0141] Aspects of the present application have been described above with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present application. It should be understood that each block in the flowcharts and / or block diagrams, and the combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing devices to produce a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices enable the implementation of the functions / actions specified in one or more blocks of the flowcharts and / or block diagrams. Such a processor can be, but is not limited to, a general-purpose processor, a special-purpose processor, a special application processor, or a field-programmable logic circuit. It should also be understood that each block in the block diagrams and / or flowcharts, and the combinations of blocks in the block diagrams and / or flowcharts, can also be implemented by dedicated hardware for performing the specified functions or actions, or by a combination of dedicated hardware and computer instructions.

[0142] As described above, this is only the specific implementation manner of the present application. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, modules, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein. It should be understood that the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present application.

Claims

1. An automatic annotation method for osteoarticular lesion areas based on image processing, applicable to osteoarticular lesion areas with osteophytes or cartilage fissures, characterized in that, Including: Emitting an optical signal to the bone and joint area through a polarization-state tunable laser light source, receiving the reflected signal through polarization-sensitive optical coherence tomography, and calculating the birefringence distribution by matrix analysis to obtain a three-dimensional birefringence distribution map including collagen fiber orientation information; Performing non-local mean filtering on the birefringence distribution map, and extracting the edge contour of the target lesion area based on the gradient magnitude range to generate binary mask data retaining microcrack texture features, where the gradient magnitude range is dynamically adjusted based on the local structural characteristics of the birefringence distribution map; Converting the pixel points in the binary mask data into nodes, and constructing a topological network representing the morphology of crack branches and osteophytes through the connection relationship between nodes. The node attributes of the nodes include spatial coordinates, gradient direction, and birefringence, and the connection relationship between nodes is determined by the Euclidean distance of adjacent pixel points in the binary mask data; Based on the connection relationship between nodes, determining the node weights of the nodes according to the intersection density and birefringence of the nodes, and using a graph attention network for multi-layer feature aggregation to extract the lesion topological features of the target lesion area; Mapping the lesion topological features to the spatial coordinate system of the three-dimensional birefringence distribution map to generate a lesion probability distribution field, and generating an interactive annotation result supporting three-dimensional space rotation and multi-plane slice display based on the lesion probability distribution field through isosurface extraction and transparency blending rendering.

2. The method according to claim 1, wherein The determining the node weights of the nodes according to the intersection density and birefringence of the nodes based on the connection relationship between nodes, and using a graph attention network for multi-layer feature aggregation to extract the lesion topological features of the target lesion area includes: Based on the connection relationship between nodes, determining the node weights of the nodes according to the intersection density, birefringence, and gradient direction consistency of the nodes, and using a two-channel graph attention network to aggregate node attribute and geometric relationship features respectively, and extracting the lesion topological features of the target lesion area after fusion. The intersection density is the ratio of the number of adjacent edges of the node to the preset maximum number of connection edges.

3. The method according to claim 2, wherein The determining the node weights of the nodes according to the intersection density, birefringence, and gradient direction consistency of the nodes based on the connection relationship between nodes, and using a two-channel graph attention network to aggregate node attribute and geometric relationship features respectively, and extracting the lesion topological features of the target lesion area after fusion includes: Calculating the gradient direction consistency parameter of the node according to the included angle between the gradient direction of the node and the gradient direction of adjacent nodes; Based on the intersection density, birefringence, and gradient direction consistency parameter of the node, generating the initial node weight of the node by weighted summation; Performing weighted aggregation based on the spatial coordinates, birefringence, and gradient direction of the node with the initial node weight to generate a node attribute feature representing the biochemical characteristics of the lesion; Constructing the geometric relationship strength of the node according to the ratio of the Euclidean distance of the connection relationship between nodes to the initial node weight, and generating a geometric relationship feature representing the morphological distribution of the lesion. Input the node attribute features and geometric relationship features into a dual-channel graph attention network, and perform weighted fusion on the node attribute features and geometric relationship features according to the initial node weights to generate lesion topological features including fissure branch topological structures and osteophyte morphologies.

4. The method according to claim 1, characterized in that, Emitting an optical signal to the osteoarticular region through a polarization-state tunable laser light source, receiving the reflected signal through polarization-sensitive optical coherence tomography, and calculating the birefringence distribution using matrix analysis to obtain a three-dimensional birefringence distribution map including collagen fiber orientation information, including: By adjusting the wavelength and polarization direction parameters of the polarization-state tunable laser light source, emitting multi-angle polarized light signals to the superficial and deep layers of the cartilage in the osteoarticular region, and the multi-angle polarized light signals cover the arrangement directions of the collagen fibers in the superficial and deep layers of the cartilage; Receiving the interference signals reflected by the multi-angle polarized light signals at different depth layers in the osteoarticular region, and analyzing the local birefringence values of the cartilage according to the phase delay amount and polarization-state change of the interference signals, and the phase delay amount and the polarization-state change are generated by the difference in the arrangement directions of the collagen fibers in the cartilage; Mapping the local birefringence values to the three-dimensional space coordinate system of the osteoarticular region, generating a birefringence gradient distribution map associated with the collagen fiber orientation to obtain the three-dimensional birefringence distribution map, and the gradient direction in the birefringence gradient distribution map is orthogonal to the collagen fiber arrangement direction.

5. The method according to claim 4, characterized in that, Performing non-local mean filtering on the birefringence distribution map, and extracting the edge contour of the target lesion area based on the gradient amplitude range to generate binary mask data retaining microfissure texture features, including: Calculating the local structural uniformity parameter of the birefringence distribution map according to the consistency of the collagen fiber arrangement directions in adjacent regions of the birefringence distribution map, and the local structural uniformity parameter is used to characterize the fiber continuity in the osteoarticular region; Adjusting the gradient amplitude range based on the local structural uniformity parameter, expanding the gradient amplitude range in the region where the local structural uniformity parameter is lower than the set threshold, and shrinking the gradient amplitude range in the region where the local structural uniformity parameter is higher than the set threshold; Traversing the birefringence distribution map according to the adjusted gradient amplitude range, and marking the pixel points within the gradient amplitude range and matching the gradient direction as edge contours to generate binary mask data including microfissure topological connection relationships.

6. The method according to claim 1, wherein Converting the pixel points in the binary mask data into nodes, and constructing a topological network representing fissure branches and osteophyte morphologies through the connection relationships between nodes, including: Traversing the pixel points in the binary mask data, and marking the pixel points satisfying the preset gradient direction continuity as initial nodes, and the preset gradient direction continuity is determined by the included angle between the gradient direction of the pixel point and the gradient direction of the adjacent pixel point and the preset angle threshold; Calculating the spatial distribution density and birefringence difference degree of the initial nodes based on the spatial coordinates, gradient directions and birefringence of the initial nodes, and screening out the initial nodes with the birefringence difference degree greater than the preset difference threshold as target nodes; Centering on the target node, connection edges are generated according to the Euclidean distances between adjacent pixel points in the binarized mask data, and a topological network representing the morphology of crack branches and osteophytes is constructed. The maximum connection distance of the connection edges is determined according to the birefringence difference degree.

7. The method according to claim 1, characterized in that, Mapping the lesion topological features to the spatial coordinate system of the three-dimensional birefringence distribution map to generate a lesion probability distribution field, and based on the lesion probability distribution field, generating an interactive annotation result supporting three-dimensional space rotation and multi-plane slice display through isosurface extraction and transparency blending rendering, including: According to the target node weight in the lesion topological features and the spatial coordinates of the three-dimensional birefringence distribution map, establishing a probability mapping relationship between the target node weight and the spatial coordinates, and generating a three-dimensional probability distribution field including lesion probability gradients; Based on the lesion probability gradient direction of the three-dimensional probability distribution field and the collagen fiber orientation information of the birefringence distribution map, determining the isosurface boundary of the lesion probability distribution field. The extraction range of the isosurface boundary is determined by the product of the lesion probability gradient and the birefringence difference degree; Aligning the isosurface boundary with the spatial coordinates of the three-dimensional birefringence distribution map, assigning transparency parameters according to the birefringence difference degree, and performing transparency blending rendering on the area covered by the isosurface boundary to generate an interactive annotation result supporting multi-plane slice display.

8. An automatic annotation system for bone and joint lesion regions based on image processing, applicable to bone and joint lesion regions with osteophytes or cartilage fissures, characterized in that, The system includes: A transmitting module for emitting an optical signal to the osteoarticular region through a polarization-state tunable laser light source, receiving a reflected signal through polarization-sensitive optical coherence tomography, and calculating the birefringence distribution by matrix analysis to obtain a three-dimensional birefringence distribution map including collagen fiber orientation information; An extraction module for performing non-local mean filtering on the birefringence distribution map, and extracting the edge contour of the target lesion region based on the gradient amplitude range to generate binarized mask data retaining the microcrack texture features. The gradient amplitude range is dynamically adjusted based on the local structural characteristics of the birefringence distribution map; A construction module for converting the pixel points in the binarized mask data into nodes, and constructing a topological network representing the morphology of crack branches and osteophytes through the connection relationship between the nodes. The node attributes of the nodes include spatial coordinates, gradient directions, and birefringence, and the connection relationship between the nodes is determined by the Euclidean distances between adjacent pixel points in the binarized mask data; The extraction module is further configured to determine the node weight of the nodes according to the intersection density and birefringence of the nodes based on the connection relationship between the nodes, and perform multi-layer feature aggregation using a graph attention network to extract the lesion topological features of the target lesion region; A generation module for mapping the lesion topological features to the spatial coordinate system of the three-dimensional birefringence distribution map to generate a lesion probability distribution field, and generating an interactive annotation result supporting three-dimensional space rotation and multi-plane slice display through isosurface extraction and transparency blending rendering based on the lesion probability distribution field.

9. An electronic device, characterized in that, The device includes: a processor and a memory storing computer program instructions; When the processor executes the computer program instructions, it implements the automatic annotation method for bone and joint lesion regions based on image processing as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, Computer program instructions are stored on the computer-readable storage medium, and when the computer program instructions are executed by the processor, they implement the automatic annotation method for bone and joint lesion regions based on image processing as described in any one of claims 1-7.

Images