Bone trabecula feature construction method and system

Abstract

The invention relates to the field of medical imaging analysis, and discloses a bone trabecula feature construction method and system, and the method comprises the steps: preferentially extracting a fragile region from an obtained vertebral image sequence, and generating an initial feature set representing the stability of a fine-grained structure; performing coherence comparison on the initial feature set in a cross-slice dimension to obtain a composite feature vector capable of reflecting the vulnerability level of the structure; based on the composite feature vector, a connectivity change curve of the bone trabecula in a loading environment is extracted by adopting a multi-layer progressive clustering mode, and candidate bone trabecula units with potential micro-damage trends are marked; classifying the candidate bone trabecula units according to risk gradients, and reconstructing a bone trabecula local topology network with weight levels; and according to the local topology network, identifying high-risk nodes showing an imbalance tendency in a specific mechanical simulation period, and generating a feature construction path finally used for micro-fracture risk prediction. The method has the advantage of improving the micro-fracture risk prediction sensitivity.

Description

Technical Field

[0001] The present invention relates to the field of medical imaging analysis, and in particular to a method and system for constructing trabecular features. Background Art

[0002] In existing medical imaging analysis, characterization of trabecular bone structure primarily relies on bone density measurements or threshold-based segmentation-based 3D reconstruction methods. The former quantitatively assesses bone mass using DXA or CBCT, while the latter utilizes image segmentation and reconstruction algorithms to model trabecular bone morphology holistically. However, these methods generally focus on macroscopic parameters or global statistical features, insufficiently capturing the fine-grained structural characteristics of trabecular bone in localized regions. This can lead to discrepancies between predicted and actual outcomes in certain clinical scenarios. For example, in the context of osteoporotic vertebral microfracture prediction, existing risk assessments primarily rely on DXA-derived bone density. However, numerous studies have shown that some patients may have normal or only slightly decreased bone density, but localized vertebral trabeculae exhibit rarefaction, disorientation, or decreased connectivity, leading to occult microfractures despite minimal external forces. Because existing analysis methods often employ global statistical or averaging methods, they neglect the fine-grained characteristics of trabecular bone in stress-concentrated areas of the vertebral body, resulting in insufficient prediction sensitivity and difficulty in identifying high-risk individuals. Therefore, it is necessary to design a method and system for constructing trabecular features to improve the sensitivity of microfracture risk prediction. Summary of the Invention

[0003] In response to the deficiencies of the prior art, the present invention provides a method and system for constructing trabecular features, which has the advantage of improving the sensitivity of microfracture risk prediction and solves the problems in the above-mentioned background technology.

[0004] To achieve the above-mentioned purpose of improving the sensitivity of microfracture risk prediction, the present invention provides the following technical solution: a method for constructing trabecular bone characteristics, comprising the following steps: In the acquired vertebral image sequence, fragile areas are preferentially extracted, and the microscopic orientation, pore density, and local linkage rate of trabecular arrangement are captured through a directional consistency filter, thereby generating an initial feature set that characterizes fine-grained structural stability. The initial feature set is compared for consistency across slices, and the morphological evolution clues of the historical fracture area, the branching and degeneration pattern of the neighboring trabeculae, and the probability of local artifact disturbance are superimposed to obtain a composite feature vector that can reflect the structural vulnerability level. Based on the composite eigenvector, a multi-layer progressive clustering approach was used to extract the connectivity change curve of trabecular bone under loading conditions. By analyzing the degree of deviation between the curve and the baseline structure, candidate trabecular bone units with potential micro-damage trends were marked. Candidate trabecular bone units are classified according to the risk gradient, and a local topological network of trabecular bone with a weighted hierarchy is reconstructed by combining key support points, subtle branch nodes, and critical fracture nodes in high-load areas. Based on the local topological network, high-risk nodes that show imbalance tendency within a specific mechanical simulation cycle are identified, and combined with the dynamic fluctuations of trabecular connectivity under real-time imaging, a characteristic construction path is generated that is ultimately used for microfracture risk prediction.

[0005] Preferably, the process of generating an initial feature set characterizing the stability of fine-grained structures is: In the vertebral image sequence, the target bone tissue and surrounding soft tissue are segmented based on the region growing algorithm to extract the target trabecular bone area with potential microfracture risk; Applying a direction consistency filter to the target trabecular region to capture the main direction of trabecular arrangement, local fiber continuity, and pore spacing distribution; Based on the captured direction and density parameters, the node connectivity and local linkage rate of trabecular bone are further calculated to form a multi-dimensional structural parameter set; Through parameter standardization and normalization, the multidimensional structural parameter set is mapped into an initial feature set that characterizes the stability of fine-grained structures.

[0006] Preferably, the process of comparing the initial feature set for coherence across slice dimensions is as follows: In continuous image slices, the trabecular orientation and pore distribution of the initial feature set are tracked slice by slice to identify structural continuity and orientation deviation; Dynamically align the characteristic parameters between slices and eliminate local artifact errors between slices through morphological methods; Combining the morphological evolution clues of the historical fracture area with the branching and degeneration patterns of the adjacent trabecular bone, a time series comparison matrix was constructed; Based on the time series control matrix, the feature consistency index across slices is calculated to form a comparison result that can characterize the structural integrity and stability.

[0007] Preferably, the process of obtaining the composite feature vector that can reflect the structural vulnerability level is: The feature consistency index obtained by cross-slice comparison is weightedly fused with the local perturbation probability to obtain an intermediate feature set containing multiple parameters; Principal component analysis and feature selection algorithm were applied to the intermediate feature set to extract the core factors that contributed most to trabecular bone stability. Combined with the evolution trajectory of morphological features in the historical fracture patient database, the core factors are matched and compared, and risk markers are established to form a composite feature vector that can quantitatively reflect the vulnerability level of trabecular bone structure.

[0008] Preferably, the process of extracting the connectivity change curve of trabecular bone under loading environment by using multi-layer progressive clustering is as follows: The composite feature vector is mapped into a multidimensional feature space, and the trabecular connectivity index is constructed based on different loading levels. A hierarchical clustering approach was used to progressively separate the connectivity patterns of trabecular bone under different stress conditions; The clustering results at each level are tracked in time series to form the dynamic connectivity change curve of trabecular bone under loading environment.

[0009] Preferably, the process of marking candidate trabecular units with potential micro-damage tendency is: Based on the difference between the connectivity change curve and the benchmark structure curve, the node areas with the deviation degree exceeding the set threshold are extracted; A clustering anomaly detection algorithm is used to identify the key points of local structural deviation and locate their occurrence locations in combination with time series features; A candidate trabecular unit set is established for the identified offset area, and the coordinates and connection relationships of the candidate trabecular unit set in the overall trabecular network are marked to form candidate trabecular units with potential micro-damage trends.

[0010] Preferably, the process of classifying candidate trabecular bone units according to the risk gradient is: Quantify the connectivity decay rate, porosity growth rate, and strike deviation amplitude of candidate trabecular units; Map the quantitative indicators to the risk scoring function and divide them into low risk, medium risk and high risk gradients according to the set risk level range; Combined with the overall structural position of the trabecular bone where the candidate unit is located, higher weights are given to the candidate units in the high-load area to form the classification results of the candidate trabecular bone units divided according to the risk gradient.

[0011] Preferably, the process of reconstructing a trabecular bone local topological network with a weight hierarchy is: The candidate trabecular bone units under different risk gradients are used as nodes to establish a multi-attribute node set; Based on the spatial adjacency, connectivity and support between nodes, a set of edges with directionality and weight is constructed; Using graph modeling methods, a topological network reflecting the local structural mechanics of trabecular bone is formed; By superimposing node risk weights and structural positions, a local topological network with weight hierarchy is constructed.

[0012] The process of generating the final feature construction pathway for microfracture risk prediction is as follows: In the local topological network, mechanical simulation methods are applied to simulate the response of trabecular bone under different loads; Extract the displacement response, stress concentration and connectivity fluctuation of nodes during the simulation cycle; The extracted mechanical response parameters are combined with the risk gradient to identify high-risk nodes that show imbalance tendencies and mark these high-risk nodes in the simulation results. Based on the identified high-risk nodes, the transmission paths and neighborhood dependencies of the high-risk nodes in the local topological network are tracked; Combined with the dynamic fluctuations of trabecular connectivity under real-time imaging, the risk evolution trajectory of high-risk nodes in the time dimension is formed; Perform sequence modeling on the risk evolution trajectory to generate a complete feature construction path.

[0013] A trabecular bone feature construction system, comprising: Fragility Extraction Module: locates potential fragile areas in vertebral image sequences and extracts the microscopic arrangement and pore characteristics of trabeculae; Coherence comparison module: integrates initial features across slice dimensions and generates a composite feature vector based on historical fracture and artifact factors; Clustering curve module: Extracts the connectivity change curve of trabecular bone under loading environment and identifies the deviation trend through progressive clustering method; Risk classification module: classifies the candidate trabecular units according to risk gradient and reconstructs the local topological network; Dynamic prediction module: Detects imbalances in high-risk nodes during the simulation cycle and generates the final risk prediction path based on the real-time connectivity rate.

[0014] Compared with the prior art, the present invention provides a method and system for constructing trabecular bone features, which has the following beneficial effects: The present invention prioritizes identifying vulnerable areas in vertebral image sequences and uses a directional consistency filter to accurately extract the microscopic orientation, pore density, and local linkage rate of trabeculae. This method can comprehensively capture the fine-grained structural information of trabeculae and significantly improve sensitivity to signs of local micro-injury. By comparing coherence across slice dimensions and superimposing clues of historical fracture area evolution, a composite feature vector is constructed to achieve a quantitative assessment of the vulnerability level of trabecular bone structure, compensating for the neglect of local abnormalities by traditional bone density measurement or threshold segmentation methods. By using multi-layer progressive clustering and connectivity change analysis under loading environments, the stability and potential damage trend of trabeculae under different mechanical effects can be dynamically characterized, and candidate micro-injury units can be accurately labeled. Through risk gradient classification and weighted hierarchical local topological network reconstruction, the risk distribution pattern of trabeculae in high-load areas and key support nodes can be revealed, and the potential location of microfractures can be identified. Finally, the dynamic fluctuations of trabecular connectivity under mechanical simulation and real-time imaging are combined to generate feature construction paths, providing a spatiotemporally continuous, structured, and quantifiable data foundation for microfracture risk prediction. The present invention can significantly improve the sensitivity and accuracy of microfracture risk prediction, provide a scientific basis for early clinical intervention and personalized osteoporosis protection, and has the potential for scalability and adaptability to different image resolutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Schematic diagram of the method of the present invention; Figure 2 Schematic diagram of the system of the present invention. DETAILED DESCRIPTION

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0017] Example 1: Please refer to Figure 1 As shown, a method for constructing trabecular features according to an embodiment of the present invention includes the following steps: S1: In the acquired vertebral image sequence, fragile areas are preferentially extracted, and the microscopic orientation, pore density, and local linkage rate of trabecular arrangement are captured through a directional consistency filter, thereby generating an initial feature set that characterizes fine-grained structural stability.

[0018] The process of generating the initial feature set characterizing the stability of fine-grained structures in S1 is as follows: In vertebral image sequences, the target bone tissue is segmented from the surrounding soft tissue using a region growing algorithm to extract target trabecular bone regions with potential microfracture risk. By setting grayscale thresholds and connectivity conditions, pixels in the vertebral images close to the trabecular bone are used as initial seed points. The region growing algorithm is then expanded layer by layer to automatically segment bone tissue regions with significant grayscale differences from the surrounding soft tissue. Morphological operations (such as erosion and expansion) are used to remove artifacts and isolated points, ultimately obtaining candidate trabecular bone regions with potential microfracture risks. A directional consistency filter is applied to the target trabecular region to capture the principal direction of trabecular arrangement, local fiber continuity, and pore spacing distribution. A structural tensor-based directional consistency filter is used to perform a convolution operation on the pixel gradient field within the target region to calculate the principal direction vector and directional variance of each point. Based on this, the principal direction distribution of trabecular fibers is extracted, and the directional consistency within the neighborhood is statistically analyzed using a sliding window to obtain local fiber continuity. At the same time, the pore spacing is estimated based on the periodic distribution of pixel grayscale gaps. Based on the captured direction and density parameters, the node connectivity and local linkage rate of trabecular bone are further calculated to form a multidimensional structural parameter set. The trabecular bone structure is abstracted into a graph structure model of nodes and edges, where nodes represent intersections or branching points and edges represent fiber paths. The degree of each node in this graph model is counted to reflect the connectivity, and the local linkage rate is obtained by calculating the ratio of the actual number of edges in the subgraph to the theoretically possible number of edges, thus forming a multidimensional structural parameter set including direction, continuity, porosity, node connectivity and linkage rate. Through parameter standardization and normalization, the multidimensional structural parameter set is mapped into an initial feature set that characterizes the stability of fine-grained structures. Structural parameters of different dimensions (such as directional consistency, pore spacing, and node connectivity) are standardized (zero mean, variance normalization), and their numerical ranges are linearly normalized to the interval [0,1] to eliminate dimensional differences between different parameters. Finally, all types of parameters are uniformly mapped into the feature space to form an initial feature set that can comprehensively characterize the stability of fine-grained trabecular bone structures.

[0019] S2: The initial feature set is compared for consistency across slice dimensions, and the morphological evolution clues of the historical fracture area, the branching degeneration pattern of the neighboring trabeculae, and the probability of local artifact perturbation are superimposed to obtain a composite feature vector that can reflect the structural vulnerability level.

[0020] The process of comparing the initial feature set for consistency across slice dimensions in S2 is as follows: In continuous image slices, the orientation and pore distribution of the trabeculae of the initial feature set are tracked slice by slice to identify structural continuity and orientation deviation; continuous vertebral image slices are input into the 3D reconstruction module to establish a spatial index relationship between slices; principal component analysis or structural tensor method is then used to calculate the main direction of the trabecular fibers on each slice, and the cosine similarity of the direction vector field is compared in adjacent slices to achieve slice-by-slice orientation tracking. At the same time, Fourier spectrum analysis is performed on the pore distribution between the trabeculae in each slice to calculate the center spacing and distribution density of the pores. Combined with the differential changes between adjacent slices, the continuity, breakpoints and orientation deviation of the trabecular structure in the continuous space are identified to further locate potential structural weaknesses. Dynamically align the characteristic parameters between slices, and use morphological methods to eliminate local artifact errors between slices; extract the key points of the trabeculae of each slice (such as node intersections, branch starting points, and boundary points), and use an affine transformation algorithm or elastic registration algorithm based on feature points to map the feature points of adjacent slices to a unified reference coordinate system; introduce a dynamic weight adjustment mechanism in the registration process to achieve higher fitting accuracy in high-variability areas while maintaining low computational overhead in stable areas. After completing the registration, use morphological filtering (such as opening and closing operations, connected region analysis) to remove isolated pixels and non-structural edges caused by motion artifacts and scanning noise, thereby improving the stability and reliability of feature alignment between slices and ensuring the accuracy of subsequent comparison results; Combining the morphological evolution clues of the historical fracture area with the branching degeneration pattern of the neighboring trabeculae, a time series comparison matrix is ​​constructed; the morphological evolution trajectory of the confirmed fracture area at different time points is extracted from the previous fracture case database, including typical change patterns such as a decrease in the number of trabeculae, sparse branches, and disordered directions; these evolutionary features are then encoded into a priori pattern curves and stored in a feature library. Subsequently, the cross-slice trabecular characteristic parameters (including branch number, branch length, thickness, porosity, etc.) of the current target patient are input into the comparison system slice by slice, and a dynamic time-warped comparison is performed with the prior pattern, and the degree of deviation is calculated. Finally, these deviation parameters are stored in matrix form to construct a time series comparison matrix, which can reflect whether the target trabeculae have a degeneration trend similar to the fracture evolution in the time and space dimensions; Based on the time series control matrix, a cross-slice feature consistency index is calculated to form a comparison result that can characterize structural integrity and stability. Based on the various parameters in the time series control matrix, a comprehensive evaluation function including directional offset, pore distribution difference, and branch degeneration rate is established, and these indicators are weighted and summarized to obtain the cross-slice consistency index. The index value range is set to 0 to 1, where a value close to 1 indicates high consistency of trabecular structure in consecutive slices, indicating that the overall structure remains stable; while a value close to 0 indicates significant cross-slice feature differences, suggesting the risk of local structural fracture or abnormal degeneration. Finally, the system compares the consistency index with the clinical threshold. If the index is lower than the set value, a high-risk warning is output, which serves as the input basis for subsequent risk prediction and intervention recommendations.

[0021] The process of obtaining the composite feature vector that can reflect the structural vulnerability level in S2 is as follows: The feature consistency index obtained by cross-slice comparison is weightedly fused with the local perturbation probability to obtain an intermediate feature set containing multiple parameters; the obtained cross-slice feature consistency index is used as the main input parameter, which reflects the overall stability of trabeculae in continuous slices; at the same time, the perturbation probability is extracted from the images of the local area, including the probability value of abnormal fluctuations caused by noise, artifacts or trabecular fracture points. Subsequently, the fusion weight coefficient is set, and the consistency index and the perturbation probability are fused through linear weighting or multi-objective optimization methods. A dynamic adjustment mechanism is introduced in the weighting process, that is, a higher weight is given to the perturbation probability in high-risk areas, while the dominance of the consistency index is maintained in low-risk areas. The final intermediate feature set not only contains parameters reflecting global stability, but also incorporates factors that characterize the sensitivity to local abnormalities; Principal component analysis and feature selection algorithms are applied to the intermediate feature set to extract the core factors that contribute most to trabecular stability; the intermediate feature set is input into the principal component analysis module, and the covariance matrix of the multidimensional parameters is decomposed to obtain the main component directions and the corresponding variance contribution rates. By screening the top principal components with a cumulative contribution rate greater than 85%, key variables with high explanatory power are retained. At the same time, LASSO regression and recursive feature elimination feature selection algorithms are combined to perform sparse processing on the parameters to remove features that have a weak effect on trabecular stability or are redundant. The core factors that are finally retained usually include trabecular directional stability indicators, pore distribution differences, node connectivity, and local disturbance intensity. These factors have been clinically proven to be highly correlated with fracture risk and can represent the core stability characteristics of trabecular structure; By combining the evolutionary trajectories of morphological features from a historical fracture patient database, core factors were matched and compared to establish risk markers, forming a composite feature vector that quantitatively reflects the vulnerability of trabecular bone structure. The historical fracture patient database, which contains a large number of confirmed cases' trabecular structural evolutionary data, includes dynamic curves of directional shifts, increased porosity, and decreased connectivity at different disease stages. The extracted core factors were individually matched with the evolutionary trajectories in the database, and similarity scores were calculated using the cosine similarity method. When a core factor highly matches the historical evolutionary pattern, the factor is assigned a corresponding risk label, such as "high vulnerability," "medium vulnerability," or "low vulnerability." Furthermore, multiple risk-labeled core factors are recombined and encoded to form a multidimensional vector. This vector not only retains the quantitative characteristics of the parameters but also incorporates clinical risk information, forming a composite feature vector that quantitatively characterizes the vulnerability of trabecular bone and provides input for subsequent microfracture risk prediction.

[0022] S3: Based on the composite eigenvector, a multi-layer progressive clustering method is used to extract the connectivity change curve of trabecular bone under loading environment. By analyzing the degree of deviation between the curve and the reference structure, candidate trabecular bone units with potential micro-damage trends are marked.

[0023] The process of extracting the connectivity change curve of trabecular bone under loading environment by using multi-layer progressive clustering in S3 is as follows: The composite feature vector is mapped into a multidimensional feature space, and the trabecular connectivity index is constructed based on different loading levels; the composite feature vector obtained in the previous step is input into the feature mapping module, and projected into the multidimensional feature space through kernel function mapping or dimensionality reduction methods (such as t-SNE or PCA) to more clearly reveal the nonlinear relationship between different parameters. Subsequently, multiple loading levels are set in this space, including static load, light load, medium load and high load. The stress field distribution is generated in each stage through finite element mechanics simulation or experimental data calibration. Based on the stress field and trabecular geometric parameters, connectivity indicators such as average node degree, number of trabecular support paths, and stability of key connection points are calculated. These connectivity indicators will show different change patterns under different loading levels; A hierarchical clustering method is used to progressively separate the connectivity patterns of trabecular bone under different stress conditions; in the multidimensional feature space, the Euclidean distance method is used to calculate the similarity measure between connectivity indicators with reference to the loading level. Then, a bottom-up hierarchical clustering algorithm (such as agglomerative clustering) is used to gradually merge the connectivity patterns with higher similarity into clusters. In each layer of clustering process, a threshold standard is set to ensure that patterns with high similarity are merged first, while patterns with low similarity are retained at a higher level. The results of progressive clustering can gradually reveal the evolution of trabecular bone structure from steady state to instability under different stress conditions, and distinguish connectivity patterns at multiple levels of "stable connection pattern", "potential weakening pattern" and "fracture precursor pattern"; The clustering results at each level are tracked in time series to form a dynamic connectivity change curve of trabecular bone under loading environment; the corresponding cluster center point is extracted as the representative connectivity state in each loading stage, and its position in the time series is recorded; the cluster center points of each stage are connected according to the loading sequence to form a dynamic curve, which can intuitively reflect the changing trend of the connectivity index of trabecular bone under gradually increasing stress environment, such as the gradual decrease from high stability to the critical fracture point. The critical turning point of structural vulnerability is identified through the curve slope and inflection point detection algorithm, thereby providing a basis for early warning of microfracture risk.

[0024] The process of marking the candidate trabecular units with potential micro-damage tendency in S3 is as follows: Based on the difference between the connectivity change curve and the benchmark structural curve, node regions with a degree of deviation exceeding a set threshold are extracted. Connectivity indicators of each trabecular node at different loading stages, such as node degree, number of paths, and stability of key connection points, are extracted from the obtained dynamic connectivity change curves. These curves are compared node by node with the established benchmark structural curve. The benchmark curve is generated by the average connectivity evolution of healthy vertebrae or low-risk samples. By calculating the degree of node deviation, including absolute difference, percentage change, or standardized residual, node regions with a deviation exceeding a pre-set threshold (such as 2 times the standard deviation of the mean deviation) are identified. The selection of the threshold takes into account image noise, measurement error, and clinically acceptable micro-damage sensitivity to ensure that the identified node regions truly reflect potential structural weakness. A clustering anomaly detection algorithm is used to identify key points of local structural offset and locate their occurrence locations in combination with time series features. Based on the selected offset nodes, an anomaly detection algorithm is used to cluster the offset patterns of local nodes. By analyzing the spatial proximity of nodes and the degree of anomaly in the changes in connectivity indicators, key points with the most significant offsets are identified. Furthermore, combined with time series features such as the loading stage, continuous offset trend, and peak point of node offset, the location and stage of possible microdamage occurrence are precisely located. This process not only considers spatial concentration but also uses temporal information to distinguish temporary fluctuations from true microdamage trends, thereby enhancing the accuracy of candidate node identification. A set of candidate trabecular units is established for the identified offset areas, and the coordinates and connection relationships of the candidate trabecular unit sets in the overall trabecular network are marked to form candidate trabecular units with potential micro-damage trends; the identified key nodes and their neighborhood nodes are spatially expanded to form a complete set of candidate trabecular units. Each unit contains node position, connectivity attributes and connection path information. The overall trabecular network topology is used to coordinately map the candidate units in the three-dimensional skeleton, and the connectivity relationships between nodes are recorded, including the number of directly connected trabeculae, support paths and key bridging points. Each candidate unit is uniquely identified in the network, and combined with its offset degree, critical weight and neighborhood influence to form a systematic and traceable candidate trabecular unit.

[0025] S4: Classify candidate trabecular units according to the risk gradient, and reconstruct a trabecular local topological network with a weighted hierarchy by combining key support points, subtle branch nodes, and critical fracture nodes in high-load areas.

[0026] The process of classifying the candidate trabecular units according to the risk gradient in S4 is as follows: The connectivity decay rate, porosity growth rate, and direction deviation amplitude of the candidate trabecular bone unit are quantified. The node and connectivity information of each unit is extracted from the generated set of candidate trabecular bone units, including the node degree, the number of support paths, and the stability of key bridge points. The connectivity decay rate of the node is calculated under different loading stages. The structural weakening trend of the unit is quantified by differential analysis of the rate of change of the connectivity index with time or stress level. The porosity growth rate of the unit is calculated using the pore distribution obtained by image segmentation, and the expansion of micropores and the increase in the void ratio are quantified. The deviation amplitude of the main direction of the trabecular bone is also measured and expressed as the directional consistency angle difference or vector offset. The above three key indicators are normalized to form multidimensional quantitative parameters of the candidate unit. Quantitative indicators are mapped to a risk scoring function, and the risk level is divided into low-risk, medium-risk, and high-risk gradients according to the set risk level interval. A multi-parameter risk scoring function is designed, and the connectivity decay rate, porosity growth rate, and strike deviation amplitude are weighted and summed according to preset weights or mapped to a unified risk score interval using multi-objective optimization. For example, high decay rate, high porosity growth, and large strike deviation will jointly lead to a higher risk score. Based on statistical analysis and clinical experience, the risk score interval threshold is set, and candidate units with scores falling in different intervals are marked as low-risk, medium-risk, and high-risk respectively. In the risk classification process, confidence intervals or fuzzy logic processing can be introduced to deal with image noise and measurement errors, thereby improving the reliability and robustness of risk classification. Combined with the overall structural position of the trabeculae where the candidate units are located, higher weights are assigned to the candidate units in the high-load area to form a classification result of the candidate trabecular units divided according to the risk gradient; the spatial coordinates and topological position of each candidate unit are located in the overall trabecular network, and the high-load area where it is located is identified, such as the vertebral weight-bearing surface, stress concentration point or key support area. For the candidate units located in the high-load area, a weight factor is added to the original risk score, and the risk level is appropriately improved to reflect the greater impact it may have on structural instability. Through the weighted and corrected score, all candidate units are finally classified into low, medium and high risk gradients, and a structured output table or network diagram is formed, which not only retains the local micro-damage indicators but also reflects the mechanical importance of the global structure, thereby providing clear and traceable classification results for microfracture risk prediction.

[0027] The process of reconstructing a trabecular local topological network with a weight hierarchy in S4 is as follows: Candidate trabecular bone units under different risk gradients are used as nodes to establish a multi-attribute node set. The spatial coordinates, connectivity parameters, node degrees, and microdamage indicators of each unit are extracted from the candidate trabecular bone units divided by risk gradient obtained in the previous step. This information is encoded as node attributes. Each node not only contains its physical location but also records multi-dimensional characteristics such as its risk level, porosity, connectivity decay rate, and strike offset amplitude. In this way, a multi-attribute node set is formed so that each node in the topological network can reflect its microstructural state and potential mechanical impact. Based on the spatial adjacency, connectivity, and support between nodes, a set of edges with directionality and weight is constructed. The spatial proximity of candidate units in the overall trabecular network is analyzed, including the Euclidean distance, connectivity paths, and cross-support relationships between nodes. Based on the mechanical connection characteristics of trabecular bone, directed or undirected edges are established between nodes. The direction of the edge reflects the main stress conduction direction, and the weight of the edge is assigned based on the strength of node connectivity, the importance of local support, and the risk level. For example, edges located on high-load paths with higher risk levels are given larger weights to indicate their critical role in local structural stability. Through this method, a set of edges with directionality and weight is generated. A graph modeling approach is used to form a topological network that reflects the local structural mechanical relationships of trabecular bone. A multi-attribute node set and a weighted edge set are input into a graph modeling framework, such as an adjacency matrix or graph database representation, to form a complete local trabecular bone topological network. Each node in the network and its connected edges jointly describe the relationship between local trabecular bone in spatial distribution and mechanical conduction. By introducing graph metrics such as node centrality, edge weight distribution, and local clustering coefficient, the network's structural stability, key support paths, and potential vulnerable areas are analyzed to ensure that the topological network accurately reflects the microscopic mechanical properties of trabecular bone and potential damage risks. By superimposing node risk weights and structural positions, a local topological network with a weight hierarchy is constructed. Based on the generated topological network, the risk level of each node and its spatial importance in the overall trabecular structure are superimposed and calculated to form a comprehensive weight index. Nodes with high risk and located in critical support areas are given higher weights in the network, while low-risk or non-critical nodes are given relatively lower weights. By normalizing and hierarchically classifying node weights, the network weight hierarchy is achieved, so that the network not only reflects structural connectivity, but also reflects local mechanical importance and potential micro-damage risk distribution.

[0028] S5: Based on the local topological network, high-risk nodes that show imbalance tendency within a specific mechanical simulation cycle are identified, and combined with the dynamic fluctuations of trabecular connectivity under real-time imaging, a characteristic construction path is generated for the final prediction of microfracture risk.

[0029] The process of generating the feature construction path for microfracture risk prediction in S5 is as follows: In a local topological network, mechanical simulation methods are applied to simulate the response of trabecular bone under different loads. The constructed local topological network with a weighted hierarchy is input into a finite element analysis or multibody dynamics simulation platform, and different types of external load conditions are set, including static loads, cyclic loads, and sudden impact loads. By parametrically modeling the material properties of each node and its edges in the network, as well as the connection strength and weight between nodes, the deformation, stress distribution, and strain response of trabecular bone under different mechanical environments are simulated, thereby obtaining the mechanical behavior of each trabecular bone unit under realistic load conditions. Extract the displacement response, stress concentration, and connectivity fluctuations of the nodes during the simulation cycle. During the simulation, collect data on the displacement vector, stress tensor, and local connectivity over time for each node. The displacement response reflects the deformation amplitude of the node under load, the stress concentration indicates the risk area for micro-damage of the node and its adjacent elements, and the connectivity fluctuation reveals the potential trend of localized fracture or bridging loss of trabeculae. By performing time-series sampling and quantification on these indicators, the dynamic response characteristics of the node throughout the simulation cycle are formed into an analyzable data set. The extracted mechanical response parameters are combined with the risk gradient to identify high-risk nodes that show imbalance tendencies and mark these high-risk nodes in the simulation results. The displacement, stress, and connectivity fluctuation parameters of the nodes are comprehensively evaluated with the previously calculated risk levels. Nodes that show abnormal displacement, stress concentration, or obvious loss of connectivity during the simulation process are identified using weighted scores or multidimensional discriminant functions. These nodes are marked as high-risk nodes and their spatial position, weight level, and association with neighboring nodes in the local topological network are recorded so that they can be highlighted in the simulation visualization results, providing clear high-risk areas for microfracture prediction. Based on the identified high-risk nodes, the conduction paths and neighborhood dependencies of the high-risk nodes in the local topological network are tracked. Starting from the high-risk node, its connection relationship with adjacent nodes is traced along the topological network to analyze the paths along which mechanical stress or deformation may be transmitted. The strength of the dependencies between nodes is recorded, including the contribution of key supporting nodes to the stability of the high-risk node and the conductive effect of bridging nodes. Through this path tracing, a complete map of the potential impact range of the high-risk node in the network is formed, revealing the possible spread paths of micro-damage and the mechanical coupling characteristics with the surrounding trabecular bone units. By combining the dynamic fluctuations of trabecular connectivity under real-time imaging, the risk evolution trajectory of high-risk nodes in the time dimension is formed. The high-risk node response data obtained by simulation is compared and integrated with the actual changes in trabecular connectivity over time in the continuous image sequence. Through time alignment and interpolation methods, the structural connectivity, stress state, and displacement response of high-risk nodes at different time points are mapped to the time axis to form a node risk evolution curve. The curve not only reflects the risk change trend of a single node, but also shows its possible chain reaction in the local network, providing data support in the time dimension for dynamic monitoring and prediction of microfractures. Perform sequence modeling on the risk evolution trajectory to generate a complete feature construction path; input the time series data of each high-risk node into a sequence modeling algorithm, such as a long short-term memory network, a hidden Markov model, or an autoregressive model, to model and predict the risk evolution trajectory. By analyzing the dependencies, response patterns, and time correlations between nodes, a complete feature construction path is generated to comprehensively reflect the micro-damage risk distribution of trabeculae in space and time, providing systematic and quantifiable analysis results for microfracture risk prediction and providing a decision-making basis for clinical intervention.

[0030] Example 2: Figure 2 As shown, a trabecular bone feature construction system includes: Fragility Extraction Module: locates potential fragile areas in vertebral image sequences and extracts the microscopic arrangement and pore characteristics of trabeculae; Coherence comparison module: integrates initial features across slice dimensions and generates a composite feature vector based on historical fracture and artifact factors; Clustering curve module: Extracts the connectivity change curve of trabecular bone under loading environment and identifies the deviation trend through progressive clustering method; Risk classification module: classifies the candidate trabecular units according to risk gradient and reconstructs the local topological network; Dynamic prediction module: Detects imbalances in high-risk nodes during the simulation cycle and generates the final risk prediction path based on the real-time connectivity rate.

[0031] It should be noted that, in this document, relational terms such as first and second, etc., are used only 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 "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0032] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing trabecular bone characteristics, characterized in that: The following steps are involved: In the acquired vertebral image sequence, fragile areas are preferentially extracted, and the microscopic orientation, pore density, and local linkage rate of trabecular arrangement are captured through a directional consistency filter, thereby generating an initial feature set that characterizes fine-grained structural stability. The initial feature set is compared for consistency across slices, and the morphological evolution clues of the historical fracture area, the branching and degeneration pattern of the neighboring trabeculae, and the probability of local artifact disturbance are superimposed to obtain a composite feature vector that can reflect the structural vulnerability level. Based on the composite eigenvector, a multi-layer progressive clustering approach was used to extract the connectivity change curve of trabecular bone under loading conditions. By analyzing the degree of deviation between the curve and the baseline structure, candidate trabecular bone units with potential micro-damage trends were marked. Candidate trabecular bone units are classified according to the risk gradient, and a local topological network of trabecular bone with a weighted hierarchy is reconstructed by combining key support points, subtle branch nodes, and critical fracture nodes in high-load areas. Based on the local topological network, high-risk nodes that show imbalance tendency within a specific mechanical simulation cycle are identified, and combined with the dynamic fluctuations of trabecular connectivity under real-time imaging, a characteristic construction path is generated that is ultimately used for microfracture risk prediction.

2. The method for constructing trabecular bone characteristics according to claim 1, characterized in that: The process of generating the initial feature set that characterizes the stability of fine-grained structures is: In the vertebral image sequence, the target bone tissue and surrounding soft tissue are segmented based on the region growing algorithm to extract the target trabecular bone area with potential microfracture risk; Applying a direction consistency filter to the target trabecular region to capture the main direction of trabecular arrangement, local fiber continuity, and pore spacing distribution; Based on the captured direction and density parameters, the node connectivity and local linkage rate of trabecular bone are further calculated to form a multi-dimensional structural parameter set; Through parameter standardization and normalization, the multidimensional structural parameter set is mapped into an initial feature set that characterizes the stability of fine-grained structures.

3. The method for constructing trabecular bone characteristics according to claim 2, characterized in that: The process of comparing the initial feature set for consistency across slice dimensions is: In continuous image slices, the trabecular orientation and pore distribution of the initial feature set are tracked slice by slice to identify structural continuity and orientation deviation; Dynamically align the characteristic parameters between slices and eliminate local artifact errors between slices through morphological methods; Combining the morphological evolution clues of the historical fracture area with the branching and degeneration patterns of the adjacent trabecular bone, a time series comparison matrix was constructed; Based on the time series control matrix, the feature consistency index across slices is calculated to form a comparison result that can characterize the structural integrity and stability.

4. The method for constructing trabecular bone characteristics according to claim 3, characterized in that: The process of obtaining the composite eigenvector that can reflect the structural vulnerability level is as follows: The feature consistency index obtained by cross-slice comparison is weightedly fused with the local perturbation probability to obtain an intermediate feature set containing multiple parameters; Principal component analysis and feature selection algorithm were applied to the intermediate feature set to extract the core factors that contributed most to trabecular bone stability. Combined with the evolution trajectory of morphological features in the historical fracture patient database, the core factors are matched and compared, and risk markers are established to form a composite feature vector that can quantitatively reflect the vulnerability level of trabecular bone structure.

5. The method for constructing trabecular bone characteristics according to claim 4, characterized in that: The process of extracting the connectivity change curve of trabecular bone under loading environment by using multi-layer progressive clustering is as follows: The composite feature vector is mapped into a multidimensional feature space, and the trabecular connectivity index is constructed based on different loading levels. A hierarchical clustering approach was used to progressively separate the connectivity patterns of trabecular bone under different stress conditions; The clustering results at each level are tracked in time series to form the dynamic connectivity change curve of trabecular bone under loading environment.

6. The method for constructing trabecular bone characteristics according to claim 5, characterized in that: The process of marking candidate trabecular bone units with potential micro-damage tendency is as follows: Based on the difference between the connectivity change curve and the benchmark structure curve, the node areas with the deviation degree exceeding the set threshold are extracted; A clustering anomaly detection algorithm is used to identify the key points of local structural deviation and locate their occurrence locations in combination with time series features; A candidate trabecular unit set is established for the identified offset area, and the coordinates and connection relationships of the candidate trabecular unit set in the overall trabecular network are marked to form candidate trabecular units with potential micro-damage trends.

7. The method for constructing trabecular bone characteristics according to claim 6, characterized in that: The process of classifying candidate trabecular bone units according to the risk gradient is as follows: Quantify the connectivity decay rate, porosity growth rate, and strike deviation amplitude of candidate trabecular units; Map the quantitative indicators to the risk scoring function and divide them into low risk, medium risk and high risk gradients according to the set risk level range; Combined with the overall structural position of the trabecular bone where the candidate unit is located, higher weights are given to the candidate units in the high-load area to form the classification results of the candidate trabecular bone units divided according to the risk gradient.

8. The method for constructing trabecular bone characteristics according to claim 7, characterized in that: The process of reconstructing a local topological network of trabecular bone with a weight hierarchy is as follows: The candidate trabecular bone units under different risk gradients are used as nodes to establish a multi-attribute node set; Based on the spatial adjacency, connectivity and support between nodes, a set of edges with directionality and weight is constructed; Using graph modeling methods, a topological network reflecting the local structural mechanics of trabecular bone is formed; By superimposing node risk weights and structural positions, a local topological network with weight hierarchy is constructed.

9. The method for constructing trabecular bone characteristics according to claim 8, characterized in that: The process of generating the final feature construction pathway for microfracture risk prediction is as follows: In the local topological network, mechanical simulation methods are applied to simulate the response of trabecular bone under different loads; Extract the displacement response, stress concentration and connectivity fluctuation of nodes during the simulation cycle; The extracted mechanical response parameters are combined with the risk gradient to identify high-risk nodes that show imbalance tendencies and mark these high-risk nodes in the simulation results. Based on the identified high-risk nodes, the transmission paths and neighborhood dependencies of the high-risk nodes in the local topological network are tracked; Combined with the dynamic fluctuations of trabecular connectivity under real-time imaging, the risk evolution trajectory of high-risk nodes in the time dimension is formed; Perform sequence modeling on the risk evolution trajectory to generate a complete feature construction path.

10. A trabecular bone feature construction system, applied to the method according to any one of claims 1 to 9, characterized in that: include: Fragility Extraction Module: locates potential fragile areas in vertebral image sequences and extracts the microscopic arrangement and pore characteristics of trabeculae; Coherence comparison module: integrates initial features across slice dimensions and generates a composite feature vector based on historical fracture and artifact factors; Clustering curve module: Extracts the connectivity change curve of trabecular bone under loading environment and identifies the deviation trend through progressive clustering method; Risk classification module: classifies the candidate trabecular units according to risk gradient and reconstructs the local topological network; Dynamic prediction module: Detects imbalances in high-risk nodes during the simulation cycle and generates the final risk prediction path based on the real-time connectivity rate.

Images