Intelligent osteoporosis diagnosis method based on centrum microenvironment and pathology co-evolution

Abstract

The invention provides an intelligent osteoporosis diagnosis method based on centrum microenvironment and pathology co-evolution, which comprises the following steps: extracting centrum three-dimensional bone trabecula structural characteristics, and generating a quantitative characteristic spectrum containing bone trabecula density, trend and connectivity; on the basis of the structural features, recognizing a vulnerable region in combination with a centrum function partition model; key search is carried out on the recognized vulnerable area, pathological changes are detected and quantified, and spatial relevance between the pathological changes and structural degradation is analyzed; integrating information, constructing a pathology-mechanics propagation network between vertebral bodies, and predicting a pathology diffusion path; establishing an osteoporosis progress prediction model by comparing the time sequence change of each parameter; generating a structured diagnosis report; and performing quality inspection on intermediate results of the steps. According to the osteoporosis diagnosis method and system, the progress from single index evaluation to multi-dimensional comprehensive analysis of osteoporosis diagnosis and from static diagnosis to dynamic prediction is realized, and powerful technical support is provided for accurate diagnosis and treatment of osteoporosis.

Description

Technical Field

[0001] This invention relates to the field of intelligent medical imaging diagnostic technology, specifically to an intelligent diagnostic method for osteoporosis based on the co-evolution of vertebral microenvironment and pathology. Background Technology

[0002] Osteoporosis is a systemic bone disease characterized by decreased bone mass and destruction of bone microstructure, seriously threatening the health of middle-aged and elderly people. Statistics show that the prevalence of osteoporosis in people over 60 years of age in my country reaches 36%, with vertebral compression fractures being the most common complication, resulting in high disability rates and severely impacting patients' quality of life. Early and accurate diagnosis of osteoporosis is of significant clinical importance for fracture prevention and treatment planning.

[0003] Currently, the gold standard for diagnosing osteoporosis in clinical practice is dual-energy X-ray absorptiometry (DXA), which assesses the degree of osteoporosis by measuring bone mineral density (BMD). However, DXA has significant limitations: it only provides area bone density in a two-dimensional projection and cannot reflect the three-dimensional structural information of bone trabeculae; the measurement results are easily affected by degenerative changes in the spine, such as osteophytes, vascular calcification, and osteosclerosis, which can lead to elevated BMD values ​​and missed diagnoses; and DXA cannot assess bone quality, with approximately 50% of fractures occurring in patients with normal BMD or only mildly reduced BMD.

[0004] With the development of CT technology, quantitative CT (QCT) can provide true volumetric bone mineral density measurements, avoiding interference from surrounding tissues. In recent years, CT image-based trabecular bone analysis technology has gradually emerged, assessing bone quality by extracting parameters such as the number, thickness, and spacing of trabeculae. However, existing CT methods for diagnosing osteoporosis still have shortcomings.

[0005] Most methods focus solely on bone mineral density or trabecular bone parameters, neglecting the various pathological changes often associated with osteoporosis. Clinical practice shows that over 80% of patients over 60 years of age exhibit varying degrees of spinal degenerative changes, including vertebral osteophyte formation, endplate sclerosis, and intervertebral disc degeneration. These pathological changes not only affect the accuracy of bone mineral density measurements but, more importantly, have complex interactions with osteoporosis. For example, osteophytes often form in high-stress areas, reflecting local biomechanical abnormalities; compression fractures alter the load distribution of adjacent vertebrae, accelerating their degenerative process. Existing methods typically exclude these pathological changes as confounding factors, failing to recognize that they may be crucial components of the pathophysiological process of osteoporosis.

[0006] Current methods lack consideration for the vertebral biomechanical environment. As the central weight-bearing structure of the human body, the spine's biomechanical properties significantly influence bone remodeling. Different functional areas bear vastly different loads; the anterior column bears approximately 60% of the axial load, while the posterior column primarily plays a stabilizing role. When bone density in a particular area decreases and becomes mismatched with its load-bearing capacity, the risk of fracture increases significantly. However, current methods typically perform homogenized analysis of the entire vertebral body, failing to identify such localized biomechanical-structural imbalances. Furthermore, current methods primarily perform cross-sectional analyses, lacking assessment of the dynamic evolution of the disease. Osteoporosis is a progressive process, with varying rates of progression among patients. Assessments based solely on a single time point are insufficient for accurately predicting future risks and cannot provide adequate evidence for personalized treatment plans. While some studies have attempted longitudinal analyses, they are mostly limited to simple density changes and fail to comprehensively consider multi-dimensional information such as structural degeneration and pathological evolution.

[0007] Therefore, there is an urgent need to develop an intelligent diagnostic method that can comprehensively analyze vertebral microstructure, mechanical environment, pathological changes and their interactions, in order to improve the accuracy and clinical applicability of osteoporosis diagnosis. Summary of the Invention

[0008] To overcome the shortcomings of existing technologies, this invention proposes an intelligent diagnostic method for osteoporosis based on the co-evolution of the vertebral microenvironment and pathology. This method comprehensively analyzes trabecular microstructure, mechanical load distribution, and pathological changes, overcoming the limitations of traditional methods that rely solely on a single bone mineral density index. By establishing a trabecular structural complexity index and a load-density mismatch model, structural degradation and mechanical imbalance can be identified even before a significant decrease in bone mineral density, achieving early warning of osteoporosis. Particularly for complex cases with multiple degenerative changes, this method no longer excludes pathological factors such as osteophytes and sclerosis as confounding factors but incorporates them into the comprehensive evaluation system, significantly improving the reliability of the diagnosis.

[0009] To achieve the above objectives, this invention proposes an intelligent diagnostic method for osteoporosis based on the co-evolution of the vertebral microenvironment and pathology, comprising the following steps:

[0010] Step S1: Extract the three-dimensional trabecular structure features of the vertebral body and generate a quantitative feature map containing trabecular density, orientation, and connectivity, which serves as the structural basis for subsequent mechanical analysis and pathological identification.

[0011] Step S2: Based on the trabecular bone structure characteristics obtained in Step S1, a mechanical-structural coupling model is established in conjunction with the vertebral body functional zoning to identify vulnerable areas with high load and low density.

[0012] Step S3: Focus on searching the vulnerable areas identified in step S2, detect and quantify pathological changes such as osteophytes, compression, and sclerosis, and analyze their spatial correlation with structural degradation.

[0013] Step S4: Integrate the structural features of Step S1, the mechanical distribution of Step S2, and the pathological information of Step S3 to construct a pathological-mechanical propagation network between vertebral bodies and predict the pathological diffusion path.

[0014] Step S5: Apply the analysis process of steps S1-S4 to multi-time point images, and establish an osteoporosis progression prediction model by comparing the temporal changes of each parameter.

[0015] Step S6: Combine all the analysis results from steps S1 to S5 to generate a structured diagnostic report that includes current status assessment, risk prediction, and personalized recommendations.

[0016] Step S7: Perform quality checks on the intermediate results of each step from S1 to S6 to ensure the reliability of the diagnostic process and the traceability of the results.

[0017] Furthermore, step S1 is detailed as follows:

[0018] S11: Identify the cancellous bone region of each vertebra in CT images, divide it into cubic units with a side length of 2mm, and assign a unique spatial location identifier to each unit;

[0019] S12: Within each cubic cell, bone trabeculae and medullary cavity are identified by threshold segmentation. The threshold is set to the 75th percentile of the HU value within that cell, and a bone trabeculae distribution map is generated.

[0020] S13: Based on the segmentation results of S12, the skeleton extraction algorithm is used to trace the centerline of each trabecular bone, record the three-dimensional coordinates of its start point, end point and bifurcation point, and construct the trabecular bone connection network.

[0021] S14: Along the trabecular centerline extracted from S13, measure the equivalent diameter of the vertical section every 0.5mm, and calculate the ratio of the actual path length to the straight distance of the endpoint as the curvature.

[0022] S15: Based on the connection network of S13 and the measurement results of S14, the number of trabeculae, average spacing, and connection point density in each cubic unit are statistically analyzed to form the local structural features of the unit.

[0023] S16: Combine the local structural features of all units according to their spatial positions to form a three-dimensional structural map that reflects the distribution of trabecular bone in the entire vertebral body.

[0024] Furthermore, step S2 is detailed as follows:

[0025] S21: Based on vertebral anatomical landmarks, the vertebral body is divided into five functional zones. The average trabecular density of each functional zone is extracted from the structural atlas of S16 as a structural strength index.

[0026] S22: Set 100 evenly distributed sampling points in each functional area, and calculate the effective bone mineral density value considering the influence of microstructure by combining the local trabecular bone features obtained in step S1.

[0027] S23: Based on the principles of spinal biomechanics and the structural strength distribution of S21, the load ratio borne by each functional area is dynamically adjusted, with areas of higher strength bearing more load;

[0028] S24: Using the effective density of S22 and the load distribution of S23, calculate the degree of mismatch between load and density. When the proportion of low-density points in a high-load area exceeds 30%, it is marked as a high-risk area.

[0029] S25: Analyze the spatial distribution of high-risk areas identified in S24, and determine the stress concentration locations through mechanical simulation. These locations will be the focus of subsequent pathological searches.

[0030] Furthermore, step S3 is as follows:

[0031] S31: Prioritize scanning within 5mm of the stress concentration location identified in S25 to detect high-density areas protruding outside the normal bone cortex and mark them as potential osteophytes;

[0032] S32: For each osteophyte detected in S31, combined with the mechanical load information from step S2 at its location, measure the base width, protrusion height, and growth angle, and analyze the correlation between osteophyte formation and local stress.

[0033] S33: Focus on searching for abnormally high-density areas within the high-risk area marked by S24, and distinguish bone islands, sclerotic foci, or calcifications based on morphological characteristics and location information.

[0034] S34: Based on the degree of trabecular bone structure degeneration in step S1, detect changes in vertebral body height. When height loss occurs in sparse areas of the trabecular bone, it is determined to be structural compression.

[0035] S35: In the stress concentration areas identified in S25, examine the integrity of the endplate, identify degenerative changes such as endplate depression and Schmorl's nodes, and record their correlation with local trabecular bone loss.

[0036] S36: Integrate all pathological information from S31 to S35, and combine it with the mechanical distribution from step S2 to establish a spatial distribution map showing the interrelationship between pathological and mechanical factors.

[0037] Furthermore, step S4 is detailed as follows:

[0038] S41: Based on the pathological-mechanical correlation map of S36, the functional connection strength between vertebral bodies is defined, taking into account the anatomical adjacency relationship and the mechanical transmission path identified by S2.

[0039] S42: Analyze the spatial correspondence between adjacent vertebral bodies of the pathology identified in step S3, and combine it with the stress transmission path in S25 to assess the possibility of pathology propagation through mechanical pathways.

[0040] S43: When the superior vertebral body has the compressive changes identified in S34, the load distribution of the inferior vertebral body is recalculated based on the mechanical model in step S2, and its high-risk areas are updated.

[0041] S44: Track the growth direction of osteophytes recorded in S32, and combine it with the functional connectivity strength of S41 to predict the location of possible osteophytes in adjacent vertebrae;

[0042] S45: Based on the analysis results of S41-S44, a pathological propagation probability network is constructed to quantify the possibility of each pathological condition spreading from one vertebra to an adjacent vertebra.

[0043] Furthermore, step S5 is detailed as follows:

[0044] S51: Repeat the complete analysis process of steps S1-S4 for CT images at different time points to obtain data on trabecular bone structure, pathological distribution and propagation network at each time point;

[0045] S52: Compare the time series data obtained in S51, calculate the rate of change of trabecular parameters, and verify whether the location of structural degradation is consistent with the high-risk area predicted in S2.

[0046] S53: Analyze whether the location of the newly emerging pathology in S51 conforms to the propagation path predicted in S4, and adjust the propagation model parameters according to the actual evolution.

[0047] S54: Based on the degradation rate of S52 and the pathological evolution law of S53, and combined with the propagation network of step S4, a spatiotemporal evolution model is established to predict the future risk level.

[0048] S55: When S52 detects that the annual degeneration rate of a certain vertebra exceeds 5%, or S53 finds that new pathology appears in the high-probability location predicted in step S4, an early warning signal is generated.

[0049] Furthermore, step S6 is as follows:

[0050] S61: Based on the risk prediction results of step S5, draw a panoramic view of the spine and use color gradients to visually display the current status and future risks of each vertebra;

[0051] S62: Detailed analysis report of high-risk vertebrae identified in S55, including the degree of structural degradation in step S1, the mechanical mismatch in step S2, and the pathological burden in step S3.

[0052] S63: Compare the key indicators extracted in steps S1-S5 with the reference values ​​of age-matched normal populations, and calculate the relative positions of the patient's various indicators.

[0053] S64: Based on the risk prediction in S54 and the relative assessment in S63, and combined with the pathological transmission characteristics revealed in step S4, develop a targeted intervention plan;

[0054] S65: Based on the evolution trend analysis in step S5, mark the key node vertebrae in the network of step S4, and provide suggestions on key monitoring contents and optimal re-examination time.

[0055] Furthermore, step S7 is detailed as follows:

[0056] S71: Verify the coverage integrity of trabecular bone identification in step S1, the spatial consistency between pathological annotation in step S3 and high-risk areas in step S2, and ensure the reliability of the results of each step.

[0057] S72: When the prediction result of step S5 contradicts the mechanical analysis of step S2, or the propagation path of step S4 does not conform to the anatomical law, the reanalysis process is automatically triggered.

[0058] S73: Verify whether the risk scores of adjacent vertebrae in the report of step S6 conform to the propagation pattern established in step S4, and ensure the overall logical consistency of the diagnostic results;

[0059] S74: Based on the stability and consistency of the intermediate results of each step S1-S6, generate a diagnostic confidence score. If the score is below 85%, a manual review is required.

[0060] S75: Completely preserve the original feature data of step S1, the analysis process of steps S2-S4, and the time-series comparison results of step S5 to form a traceable chain of diagnostic evidence.

[0061] Furthermore, the construction of the fine trabecular bone structure atlas in step S1 also includes:

[0062] Perform 26-neighbor connectivity analysis on each 2mm cube unit to determine the integrity of the trabecular bone network;

[0063] Fractured trabeculae are defined as trabecular segments whose two ends are less than 0.5 mm apart but not connected.

[0064] The branch complexity of the trabecular network is calculated by the ratio of the number of bifurcation points to the total length of the trabecular bone.

[0065] Assess the orientation of trabeculae and statistically analyze the distribution of the angles between the main orientations and the long axis of the vertebral body;

[0066] A trabecular degeneration index is generated, taking into account factors such as reduced trabecular number, increased spacing, and decreased connectivity.

[0067] Furthermore, the identification of pathological changes in step S3 also includes the detection of vascular lesions:

[0068] Search for round or near-round low-density areas within the vertebral body, with HU values ​​ranging from -50 to 50; determine whether it is a hemangioma based on morphological characteristics, including typical features such as the "fence sign" and "dot sign"; measure the maximum diameter and volume percentage of the hemangioma to assess its impact on the structural integrity of the vertebral body; track the thickening of trabecular bone at the boundary of the hemangioma, which is a compensatory change of the body; when the volume of the hemangioma exceeds one-third of the vertebral body volume, it indicates that close follow-up is required.

[0069] Furthermore, step S4 also includes constructing a biomechanical coupling model:

[0070] Identify abnormalities in the physiological curvature of the spine, including decreased lumbar lordosis and increased thoracic kyphosis; analyze the impact of curvature changes on the load distribution of each vertebra, with decreased lordosis leading to an increase of more than 50% in the load on the posterior column; detect the degree of intervertebral disc degeneration by assessing changes in intervertebral disc height and signal; when the intervertebral disc is severely degenerated, the stress on adjacent vertebrae increases, accelerating the progression of osteoporosis; comprehensively assess the biomechanical transmission chain of the entire spine and identify weak points.

[0071] Furthermore, it also includes a deep learning-based automatic feature extraction step:

[0072] A convolutional neural network specifically designed for vertebral body analysis was constructed, consisting of 5 convolutional layers and 3 fully connected layers. Batch normalization and ReLU activation were added after each convolutional layer to prevent gradient vanishing. Multi-scale feature extraction branches were designed to process features from receptive fields of 4mm, 8mm, and 16mm respectively. Information from different scales was fused through a feature pyramid to capture pathological features from microscopic to macroscopic. A 256-dimensional feature vector was output, which was then concatenated with manually designed features and fed into the final classifier.

[0073] Furthermore, a continuous learning and optimization step is included after step S6:

[0074] Collect feedback from doctors on diagnostic reports, including evaluations of diagnostic accuracy and omissions in pathological annotations; add the doctors' revised results to the training set, and update the model every 100 cases; adopt an incremental learning strategy to retain existing knowledge while learning new pathological patterns; regularly analyze misdiagnosed cases to identify weaknesses in the system and make targeted improvements; maintain a knowledge base update log to record the content and effectiveness evaluation of each update.

[0075] Compared with the prior art, the beneficial effects of the present invention are:

[0076] 1. This invention provides an intelligent diagnostic method for osteoporosis based on the co-evolution of the vertebral microenvironment and pathology. It visually displays the risk level of each vertebra, and the detailed analysis card provides quantitative indicators such as trabecular structure score, biomechanical mismatch index, and pathological burden, enabling doctors to comprehensively understand the patient's bone status. More importantly, it provides clear traceability of diagnostic evidence, allowing doctors to view the specific pathological features and analysis process behind each diagnostic conclusion, greatly enhancing the credibility and clinical acceptance of the diagnostic results.

[0077] 2. This invention provides an intelligent diagnostic method for osteoporosis based on the co-evolution of the vertebral microenvironment and pathology. Through comprehensive analysis of multi-timepoint data, a spatiotemporal evolution model of osteoporosis progression is established. The system can not only assess the current state but also predict risk changes at different future time periods and identify key warning signals. This prospective assessment provides a scientific basis for developing personalized prevention and treatment strategies and helps to take effective intervention measures before fractures occur.

[0078] 3. This invention provides an intelligent diagnostic method for osteoporosis based on the co-evolution of vertebral microenvironment and pathology. The trabecular structure complexity index comprehensively considers the characteristics of quantity, length and direction, and more comprehensively reflects the integrity of the trabecular network. The pathological spatial distribution entropy quantifies the degree of pathological aggregation, providing a new tool for assessing the impact of pathology on vertebral stability. The pathological propagation intensity model combines the severity of pathology and spatial relationship, realizing the quantitative prediction of pathological evolution.

[0079] 4. This invention provides an intelligent diagnostic method for osteoporosis based on the co-evolution of the vertebral microenvironment and pathology. By setting multiple quality checkpoints, it automatically verifies the rationality and consistency of the results at each step, and automatically triggers re-analysis when abnormalities occur. The diagnostic confidence scoring mechanism ensures that only high-quality results are directly output, and cases requiring manual review are clearly marked. This human-machine collaborative mode ensures both diagnostic efficiency and diagnostic safety. Attached Figure Description

[0080] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0081] Figure 1 This is a flowchart of the invention.

[0082] Figure 2 Vertebral body localization diagram Detailed Implementation

[0083] The technical solution of the present invention will be more clearly and completely explained below with reference to the accompanying drawings and through the description of preferred embodiments of the present invention.

[0084] like Figure 1 As shown, the present invention is:

[0085] Step S1: Extract the three-dimensional trabecular structure features of the vertebral body and generate a quantitative feature map containing trabecular density, orientation, and connectivity, which serves as the structural basis for subsequent mechanical analysis and pathological identification;

[0086] Step S2: Based on the trabecular bone structure features obtained in Step S1, a mechanical-structural coupling model is established in conjunction with the functional zonation of the vertebral body to identify vulnerable areas with high load and low density.

[0087] Step S3: Focus on searching the vulnerable areas identified in step S2, detect and quantify pathological changes, and analyze their spatial correlation with structural degradation;

[0088] Step S4: Integrate the structural features of Step S1, the mechanical distribution of Step S2, and the pathological information of Step S3 to construct a pathological-mechanical propagation network between vertebral bodies and predict the pathological diffusion path.

[0089] Step S5: Apply the analysis process of steps S1-S4 to multi-time point images, and establish an osteoporosis progression prediction model by comparing the temporal changes of each parameter.

[0090] Step S6: Combine all the analysis results from steps S1 to S5 to generate a structured diagnostic report that includes current status assessment, risk prediction, and personalized recommendations;

[0091] Step S7: Perform quality checks on the intermediate results of each step from S1 to S6 to ensure the reliability of the diagnostic process and the traceability of the results.

[0092] like Figure 2 As shown, the implementation process of multiplanar analysis of the vertebral body is demonstrated according to the diagnostic method proposed in this invention. The figure shows three orthogonal sections of a spinal CT image: the transverse, sagittal, and coronal planes. These form the basis for constructing a three-dimensional trabecular bone structure atlas of the vertebral body. The blue crosshairs precisely mark the analysis locations, ensuring accurate spatial correspondence between different sections.

[0093] The images clearly show the complete outline of a single vertebral body, which is a prerequisite for dividing the image into 2mm cubic units. The cross-sectional view is particularly suitable for identifying cancellous bone regions, laying the foundation for subsequent detailed trabecular analysis. The sagittal plane clearly shows the anterior and posterior columns of the vertebral body. The marked areas in the image (red, green, blue, etc.) represent the load distribution in different functional zones. The highlighted areas are stress concentration points identified through finite element analysis; these locations are high-risk areas for osteoporosis.

[0094] This embodiment uses spinal CT imaging data of 486 patients collected from a tertiary hospital between January 2022 and June 2024 for verification. All patients underwent standard DXA examination, considered the gold standard for diagnosing osteoporosis. CT scans were performed using a Siemens SOMATOM Force dual-source CT scanner with tube voltage of 120kVp, automatic tube current modulation, slice thickness of 0.625mm, and the acquisition range covering the T11 to L5 vertebral bodies. The following describes the implementation process of this invention through the complete diagnostic procedure of a 65-year-old female patient.

[0095] Image preprocessing employed conventional anisotropic diffusion filtering and trilinear interpolation resampling methods to unify all images to a voxel size of 1mm × 1mm × 1mm. Initial localization of the cone was performed using an improved 3D U-Net network; these are standard procedures in existing technologies.

[0096] The L1-L5 vertebral bodies of this patient were analyzed. First, the cancellous bone region of each vertebral body was divided into 2mm × 2mm × 2mm cubic units. A total of 1,250 units were divided in the L2 vertebral body, each assigned unique spatial coordinates (i, j, k). Voxels within each unit were statistically analyzed, and the 75th percentile of the HU value was calculated as an adaptive threshold. For example, in the unit with coordinates (10, 15, 8), the 75th percentile was HU = 125; voxels with values ​​higher than this were identified as trabecular bone. This adaptive thresholding method better accommodates individual differences in bone mineral density among patients.

[0097] Using a 3D skeleton extraction algorithm to trace the trabecular centerline, 3,472 trabecular segments were identified in the L2 vertebral body, and 8,156 connection nodes were recorded. Cross-sectional measurements were taken every 0.5 mm along the centerline of each trabecular segment. The results showed that the average diameter of the trabecular segments in this patient was 0.18 mm (normal range 0.20-0.25 mm), and the average curvature was 1.23 (normal range <1.15), indicating degenerative characteristics of thinning and increased curvature in the trabecular segments. A trabecular structural complexity index is proposed:

[0098]

[0099] Where N bNumber of trabecular branch points within a unit V represents the average length of the trabecular bone. unit Unit volume (fixed at 8mm) 3 ), σ θ The standard deviation of the trabecular orientation angle. The mean orientation angle is given. This indicator comprehensively reflects the number, length, and directional complexity of trabeculae. The mean C0 of the L2 vertebral body in this patient is... tb =0.42, significantly lower than the normal reference value for the same age (0.60-0.80), indicating severe degradation of the trabecular network structure. Finally, a three-dimensional structural map containing 1,250 unit features was generated, with each unit recording 7 structural parameters such as the number of trabeculae, average spacing, and connection density.

[0100] Based on the generated three-dimensional structural atlas, the L2 vertebral body was divided into five functional zones. Extraction from the structural atlas revealed that the average trabecular density in the anterior column weight-bearing zone was only 2.1 trabecular bones / mm². 3 (Normal value > 3.5 roots / mm) 3 This region bears 60% of the axial load. 100 sampling points were set in each functional area, and effective bone mineral density was calculated based on trabecular bone characteristics. To assess the mechanical performance degradation of the vertebral body, this invention established a load-density mismatch model:

[0101]

[0102] Where w i Let w1 be the load weight for the i-th functional area, with specific values ​​as follows: w1 = 0.6 (front column), w2 = 0.2 (rear column), w3 = w4 = w5 = 0.067 (other areas). N_{low,i} represents the number of low-density points (HU value below 100) in this area. low,i This represents the total number of sampling points in the area (set to 100). When M... Id A value >0.3 indicates a significant load-density mismatch in the vertebral body, suggesting a high risk of fracture. In this patient, the M value of the L2 vertebral body... Id =0.38, exceeding the high-risk threshold. Finite element analysis identified three main stress concentration points: the middle of the anterior edge (coordinates 125, 82, 45), the base of the left pedicle (coordinates 98, 65, 52), and the anterior third of the lower endplate (coordinates 115, 75, 38). The stress values ​​at these locations were more than 40% higher than those in the surrounding area.

[0103] Based on the identified stress concentration points, a fine scan was performed within a 5mm radius around each point. A 15mm × 8mm × 6mm osteophyte was detected near the anterior stress concentration point, its base precisely located in the predicted high-stress area. Measurements showed that the osteophyte grew at a 35° angle, pointing anteroinferiorly, consistent with the principal stress direction at that location, validating the stress-induced osteophyte formation mechanism. Within the high-risk area, 12 abnormally high-density lesions (HU>400) were identified, 8 of which were located around the stress concentration area, indicating that these sclerotic lesions are a compensatory response to localized high stress. Simultaneously, a slight wedge-shaped change was detected at the anterior edge of the L2 vertebral body (22mm), the middle (24mm), and the posterior edge (26mm). Combined with the sparse trabecular bone in this area (density <1.5 trabeculae / mm²), further evidence was found. 3 This was determined to be an early structural compression.

[0104] Regarding the quantification of pathological changes, this invention proposes the concept of spatial distribution entropy for osteophyte growth:

[0105]

[0106] Where p j H represents the volume percentage of the osteophyte in the j-th octet of the vertebral body, which is divided into 8 octets by three orthogonal planes at the center of the vertebral body. os The larger the value, the more dispersed the osteophyte distribution, and the more complex its impact on vertebral stability. The calculated H... os =1.82, indicating that the pathological changes are mainly concentrated in specific areas rather than randomly distributed.

[0107] By integrating structural features, biomechanical distribution, and pathological information, the pathological propagation pattern among the L1-L5 vertebral bodies was analyzed. It was found that the anterior osteophyte of L2 is only 3 mm away from the posterosuperior margin of the L3 vertebral body, forming a potential "kissing osteophyte." Based on the biomechanical transmission path, it was calculated that L2 compression leads to an 18% increase in the load on the anterior column of L3. To address the mutual influence between pathological conditions, this invention constructs a pathological propagation intensity matrix. The propagation intensity from pathological α\alphaα to pathological β\betaβ between adjacent vertebral bodies is defined as:

[0108]

[0109] Where S α The severity score of the source pathology is given (0-10 points), S0=5 is the transmission threshold, k=0.5 is the transmission sensitivity coefficient, and θ αβ The angle between the two pathological sites on the vertebral body is considered. When the pathological sites are close together (small angle) and the source pathology is severe, the propagation intensity is high. Calculations show that the probability of L2 osteophyte propagating to L3 is 0.73, and the probability of L2 compression leading to L3 compression is 0.65.

[0110] A complete analysis of the patient's CT data from 6 months prior was performed, with time-series comparisons. The number of trabeculae in the L2 vertebral body decreased from 3,856 to 3,472 (a decrease of 10.0%), with an average C... tb The value decreased from 0.51 to 0.42 (a decrease of 17.6%). The rate of change in vertebral bone mineral density is defined as follows:

[0111]

[0112] in and The values ​​are the average CT values ​​of the vertebral body from current and previous examinations, respectively, with Δ representing the time interval (years). The R value of the L2 vertebral body in this patient... bmd = -5.8% / year, exceeding the normal aging rate (<3% / year), suggesting the need for active intervention. More importantly, the newly emerging pathological location highly matched the predictions of the propagation model: early osteophytes appeared in the L3 vertebral body at the predicted high-probability location (anterosuperior margin), verifying the accuracy of the propagation model.

[0113] The system also calculates a comprehensive pathological burden score:

[0114]

[0115] Where n is the number of detected pathological types, v k Let s be the volume of the kth pathological type. k The severity coefficient is calculated as follows (0.5 for osteophytes, 2.0 for compression fractures, 0.3 for hemangiomas, etc.), I k This is the interaction coefficient, which is 0.2 when multiple pathologies coexist, and 0 otherwise. (This refers to the patient's...) total =186.4, which is considered a medium to high burden level.

[0116] Based on all the analysis results, the system generated a visual diagnostic report. In the panoramic spinal image, L2 is marked orange (moderate risk), L3 is yellow (mild risk), and L1, L4, and L5 are green (normal). Compared with the normal population of the same age group, this patient's trabecular bone density is at the 15th percentile, structural complexity is at the 20th percentile, and the overall risk score is at the 85th percentile (high risk). Based on the predictive model, the risk of L3 compression within 6 months is 45%, and the probability of pathological changes in L4 within 1 year is 62%. The system recommends immediately starting anti-osteoporosis drug treatment, focusing on protecting the L3 and L4 vertebral bodies, and a follow-up examination to evaluate the treatment effect after 3 months.

[0117] The system's automated quality check showed that the trabecular bone identification coverage reached 93.5% (>90% standard), the overlap between pathological locations and high-risk areas was 76% (>70% standard), and the overall confidence score was 91%, exceeding the 85% threshold, requiring no manual review. Validated on 486 patients, this method achieved a sensitivity of 92.3% and a specificity of 88.7% for diagnosing osteoporosis, with a Kappa value of 0.81 consistent with the DXA gold standard. Particularly in complex cases with multiple degenerative changes, the diagnostic accuracy of this method was 15.6% higher than traditional methods based solely on bone mineral density measurement. In terms of computational efficiency, the complete analysis process for a single patient took approximately 3.5 minutes on a workstation equipped with an NVIDIA RTX 3090 graphics card, representing a more than 20-fold improvement in efficiency compared to manual analysis.

[0118] The above-described specific embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Various modifications, substitutions, and improvements made by those skilled in the art to the technical solutions of the present invention based on the provided textual description and drawings, without departing from the design concept and spirit of the present invention, should all fall within the scope of protection of the present invention. The scope of protection of the present invention is determined by the claims.

Claims

1. An intelligent diagnostic method for osteoporosis based on the co-evolution of vertebral microenvironment and pathology, characterized in that, Includes the following steps: Step S1: Extract the three-dimensional trabecular structure features of the vertebral body and generate a quantitative feature map containing trabecular density, orientation, and connectivity, which serves as the structural basis for subsequent mechanical analysis and pathological identification; Step S2: Based on the trabecular bone structure features obtained in Step S1, a mechanical-structural coupling model is established in conjunction with the functional zonation of the vertebral body to identify vulnerable areas with high load and low density. Step S3: Focus on searching the vulnerable areas identified in step S2, detect and quantify pathological changes, and analyze their spatial correlation with structural degradation; Step S4: Integrate the structural features of Step S1, the mechanical distribution of Step S2, and the pathological information of Step S3 to construct a pathological-mechanical propagation network between vertebral bodies and predict the pathological diffusion path. Step S5: Apply the analysis process of steps S1-S4 to multi-time point images, and establish an osteoporosis progression prediction model by comparing the temporal changes of each parameter. Step S6: Combine all the analysis results from steps S1 to S5 to generate a structured diagnostic report that includes current status assessment, risk prediction, and personalized recommendations; Step S7: Perform quality checks on the intermediate results of each step from S1 to S6 to ensure the reliability of the diagnostic process and the traceability of the results.

2. The intelligent diagnostic method for osteoporosis based on the co-evolution of vertebral microenvironment and pathology according to claim 1, characterized in that, Step S1 is as follows: S11: Identify the cancellous bone region of each vertebra in CT images, divide it into cubic units with a side length of 2mm, and assign a unique spatial location identifier to each unit; S12: Within each cubic cell, bone trabeculae and medullary cavity are identified by threshold segmentation. The threshold is set to the 75th percentile of the HU value within that cell, and a bone trabeculae distribution map is generated. S13: Based on the segmentation results of S12, the skeleton extraction algorithm is used to trace the centerline of each trabecular bone, record the three-dimensional coordinates of its start point, end point and bifurcation point, and construct the trabecular bone connection network. S14: Along the trabecular centerline extracted from S13, measure the equivalent diameter of the vertical section every 0.5mm, and calculate the ratio of the actual path length to the straight distance of the endpoint as the curvature. S15: Based on the connection network of S13 and the measurement results of S14, the number of trabeculae, average spacing, and connection point density in each cubic unit are statistically analyzed to form the local structural features of the unit. S16: Combine the local structural features of all units according to their spatial positions to form a three-dimensional structural map that reflects the distribution of trabecular bone in the entire vertebral body.

3. The intelligent diagnostic method for osteoporosis based on the co-evolution of vertebral microenvironment and pathology according to claim 1, characterized in that, Step S2 is as follows: S21: Based on vertebral anatomical landmarks, the vertebral body is divided into five functional zones. The average trabecular density of each functional zone is extracted from the structural atlas of S16 as a structural strength index. S22: Set 100 evenly distributed sampling points in each functional area, and calculate the effective bone mineral density value considering the influence of microstructure by combining the local trabecular bone features obtained in step S1. S23: Based on the principles of spinal biomechanics and the structural strength distribution of S21, the load ratio borne by each functional area is dynamically adjusted, with areas of higher strength bearing more load; S24: Using the effective density of S22 and the load distribution of S23, calculate the degree of mismatch between load and density. When the proportion of low-density points in a high-load area exceeds 30%, it is marked as a high-risk area. S25: Analyze the spatial distribution of high-risk areas identified in S24, and determine the stress concentration locations through mechanical simulation. These locations will be the focus of subsequent pathological searches.

4. The intelligent diagnostic method for osteoporosis based on the co-evolution of vertebral microenvironment and pathology according to claim 1, characterized in that, Step S3 is as follows: S31: Prioritize scanning within 5mm of the stress concentration location identified in S25 to detect high-density areas protruding outside the normal bone cortex and mark them as potential osteophytes; S32: For each osteophyte detected in S31, combined with the mechanical load information from step S2 at its location, measure the base width, protrusion height, and growth angle, and analyze the correlation between osteophyte formation and local stress. S33: Focus on searching for abnormally high-density areas within the high-risk area marked by S24, and distinguish bone islands, sclerotic foci, or calcifications based on morphological characteristics and location information. S34: Based on the degree of trabecular bone structure degeneration in step S1, detect changes in vertebral body height. When height loss occurs in sparse areas of the trabecular bone, it is determined to be structural compression. S35: In the stress concentration areas identified in S25, examine the integrity of the endplate, identify degenerative changes such as endplate depression and Schmorl's nodes, and record their correlation with local trabecular bone loss. S36: Integrate all pathological information from S31 to S35, and combine it with the mechanical distribution from step S2 to establish a spatial distribution map showing the interrelationship between pathological and mechanical factors.

5. The intelligent diagnostic method for osteoporosis based on the co-evolution of vertebral microenvironment and pathology according to claim 1, characterized in that, Step S4 is as follows: S41: Based on the pathological-mechanical correlation map of S36, the functional connection strength between vertebral bodies is defined, taking into account the anatomical adjacency relationship and the mechanical transmission path identified by S2. S42: Analyze the spatial correspondence between adjacent vertebral bodies of the pathology identified in step S3, and combine it with the stress transmission path in S25 to assess the possibility of pathology propagation through mechanical pathways. S43: When the superior vertebral body has the compressive changes identified in S34, the load distribution of the inferior vertebral body is recalculated based on the mechanical model in step S2, and its high-risk areas are updated. S44: Track the growth direction of osteophytes recorded in S32, and combine it with the functional connectivity strength of S41 to predict the location of possible osteophytes in adjacent vertebrae; S45: Based on the analysis results of S41-S44, a pathological propagation probability network is constructed to quantify the possibility of each pathological condition spreading from one vertebra to an adjacent vertebra.

6. The intelligent diagnostic method for osteoporosis based on the co-evolution of vertebral microenvironment and pathology according to claim 1, characterized in that, Step S5 is as follows: S51: Repeat the complete analysis process of steps S1-S4 for CT images at different time points to obtain data on trabecular bone structure, pathological distribution and propagation network at each time point; S52: Compare the time series data obtained in S51, calculate the rate of change of trabecular parameters, and verify whether the location of structural degradation is consistent with the high-risk area predicted in S2. S53: Analyze whether the location of the newly emerging pathology in S51 conforms to the propagation path predicted in step S4, and adjust the propagation model parameters according to the actual evolution. S54: Based on the degradation rate of S52 and the pathological evolution law of S53, and combined with the propagation network of step S4, a spatiotemporal evolution model is established to predict the future risk level. S55: When S52 detects that the annual degeneration rate of a certain vertebra exceeds 5%, or S53 finds that new pathology appears in the high-probability location predicted in step S4, an early warning signal is generated.

7. The intelligent diagnostic method for osteoporosis based on the co-evolution of vertebral microenvironment and pathology according to claim 1, characterized in that, Step S6 is as follows: S61: Based on the risk prediction results of step S5, draw a panoramic view of the spine and use color gradients to visually display the current status and future risks of each vertebra; S62: Detailed analysis report of high-risk vertebrae identified in S55, including the degree of structural degradation in step S1, the mechanical mismatch in step S2, and the pathological burden in step S3. S63: Compare the key indicators extracted in steps S1-S5 with the reference values ​​of age-matched normal populations, and calculate the relative positions of the patient's various indicators. S64: Based on the risk prediction in S54 and the relative assessment in S63, and combined with the pathological transmission characteristics revealed in step S4, develop a targeted intervention plan; S65: Based on the evolution trend analysis in step S5, mark the key node vertebrae in the network of step S4, and provide suggestions on key monitoring contents and optimal re-examination time.

8. The intelligent diagnostic method for osteoporosis based on the co-evolution of vertebral microenvironment and pathology according to claim 1, characterized in that, Step S7 is as follows: S71: Verify the coverage integrity of trabecular bone identification in step S1, the spatial consistency between pathological annotation in step S3 and high-risk areas in step S2, and ensure the reliability of the results of each step. S72: When the prediction result of step S5 contradicts the mechanical analysis of step S2, or the propagation path of step S4 does not conform to the anatomical law, the reanalysis process is automatically triggered. S73: Verify whether the risk scores of adjacent vertebrae in the report of step S6 conform to the propagation pattern established in step S4, and ensure the overall logical consistency of the diagnostic results; S74: Based on the stability and consistency of the intermediate results of each step S1-S6, generate a diagnostic confidence score. If the score is below 85%, a manual review is required. S75: Completely preserve the original feature data of step S1, the analysis process of steps S2-S4, and the time-series comparison results of step S5 to form a traceable chain of diagnostic evidence.

9. The method according to claim 2, characterized in that, The construction of the fine trabecular bone structure map in step S1 also includes: Perform 26-neighbor connectivity analysis on each 2mm cube unit to determine the integrity of the trabecular bone network; Fractured trabeculae are defined as trabecular segments whose two ends are less than 0.5 mm apart but not connected. The branch complexity of the trabecular network is calculated by the ratio of the number of bifurcation points to the total length of the trabecular bone. Assess the orientation of trabeculae and statistically analyze the distribution of the angles between the main orientations and the long axis of the vertebral body; A trabecular degeneration index is generated, taking into account factors such as reduced trabecular number, increased spacing, and decreased connectivity.

10. The method according to claim 4, characterized in that, The identification of pathological changes in step S3 also includes the detection of vascular lesions: Search for round or near-round low-density areas within the vertebral body, with HU values ​​ranging from -50 to 50; determine whether it is a hemangioma based on morphological characteristics; measure the maximum diameter and volume percentage of the hemangioma to assess its impact on the structural integrity of the vertebral body; track the thickening of trabecular bone at the boundary of the hemangioma, which is a compensatory change of the body; when the volume of the hemangioma exceeds one-third of the vertebral body volume, it indicates that close follow-up is required.

Images