Cervical vertebra postoperative settlement risk assessment method, device and equipment and storage medium

Abstract

The invention discloses a cervical vertebra postoperative settlement risk assessment method, device and equipment and a storage medium, and the method comprises the steps: obtaining clinical and image data sets of a settlement group and a non-settlement group of an interbody fusion cage; carrying out single-factor analysis on the data set by utilizing a statistical difference analysis method, and screening out characteristics related to settlement; then based on the settlement characteristics, independent risk factors are identified through a multi-factor regression model; constructing a settlement risk prediction model according to the independent risk factors; and verifying the performance of the model through the area under the curve and the calibration index, optimizing the parameters of the model, and outputting a settlement risk probability value by using the optimized model. Through systematic statistical analysis and model verification, quantitative evaluation of the settlement risk of the fusion cage after the cervical vertebra operation is realized, and an objective basis is provided for clinical intervention.

Description

Technical Field

[0001] The present invention belongs to the field of medical diagnosis technology, and in particular relates to a method, device, equipment and storage medium for assessing the risk of cervical vertebrae postoperative settlement. Background Art

[0002] With the advancement of spinal surgical techniques, anterior cervical discectomy and fusion (ACDF) has been widely used to treat cervical degenerative diseases. This technique achieves bony fusion between vertebrae through the implantation of an intervertebral fusion cage, offering the advantages of minimal surgical trauma and rapid recovery. Traditionally, the assessment of the risk of postoperative cage subsidence has relied primarily on the following methods: clinical empirical assessment, where physicians subjectively determine risk based on single factors such as patient age and osteoporosis; qualitative imaging analysis, where postoperative observation of cage position changes via X-ray or CT scans is performed, but this can only provide a passive diagnosis after subsidence has occurred; and postoperative follow-up, where multiple imaging reviews are performed to compare cage height changes, which is subject to lag and lacks early warning. However, current assessment methods present significant challenges: a lack of quantitative predictive models, making it difficult to integrate multidimensional clinical and imaging data; reliance on postoperative diagnosis, which prevents risk prediction before or during surgery; and a high degree of subjectivity and the lack of a statistically validated risk factor screening mechanism, leading to delayed intervention. Summary of the Invention

[0003] Based on this, it is necessary to provide a method, device, equipment and storage medium for assessing the risk of postoperative cervical spine subsidence that can solve the above problems.

[0004] In a first aspect, the present application provides a method for assessing the risk of postoperative cervical spine subsidence, comprising:

[0005] Clinical and imaging data sets were obtained for the interbody fusion cage settlement group and the non-settlement group;

[0006] Univariate analysis of clinical and imaging data sets was performed using statistical variance analysis to obtain sedimentation characteristics;

[0007] Based on the sedimentation characteristics, the multivariate regression model was used to screen the independent risk factors;

[0008] Construct a subsidence risk prediction model based on independent risk factors;

[0009] Based on the verification results of the area under the curve and the calibration index, the parameters of the settlement risk prediction model are optimized, and the settlement risk probability value is output using the optimized settlement risk prediction model.

[0010] In one embodiment, obtaining clinical and imaging data sets of an intervertebral fusion cage settlement group and a non-settlement group includes:

[0011] The patients' age, gender, BMI, disease duration, disease type, smoking history, history of hypertension and diabetes, lesion segment, operation time, intraoperative blood loss, and JOA scores at preoperative and postoperative time points were obtained as clinical data sets;

[0012] The patient's cervical spine X-ray, CT and MRI were obtained as the imaging dataset.

[0013] In one embodiment, the image dataset also includes the following feature indicators calculated based on CT and MRI:

[0014] EBQ: Based on the midsagittal image of the patient's cervical spine MRI T1-weighted image, the signal intensity of the upper and lower endplates was measured within the 0.6 mm thick endplate ROI of the surgical segment excluding Schmorl's tubercle. upper +endplates lower ) / 2(C2 CSF ) EBQ is calculated, where endplates upper is the signal intensity of the upper endplate, lower is the signal intensity of the lower end plate, C2 CSF is the cerebrospinal fluid signal intensity at the level of the C2 vertebra;

[0015] VBQ: Based on the midsagittal image of the patient's cervical spine MRI T1-weighted image, within the vertebral cancellous bone ROI of C3 to C6, avoiding the hemangioma and posterior venous plexus, according to the formula VBQ = M(C3 to C6) / C2 CSF The VBQ was calculated, where M(C3–C6) was the median signal intensity of the C3–C6 vertebrae;

[0016] CT-HU: Based on the CT image of the patient's cervical spine, the HU values ​​of the upper and lower end plates were measured in the bone area ROI 0.6 mm below the fusion device contact surface to obtain CT-HU.

[0017] In one embodiment, a univariate analysis of clinical and imaging data sets is performed using statistical variance analysis to obtain sedimentation characteristics, including:

[0018] Define the data types of clinical and imaging datasets and classify them into measurement data and counting data based on the data type;

[0019] Perform the Shapiro-Wilk normality test on the measurement data and select the statistical method based on the test results: if the measurement data conforms to the normal distribution, use the independent sample t test; otherwise, use the Mann-Whitney U rank sum test; combine the independent sample t test results and the Mann-Whitney U rank sum test results to generate the measurement test results;

[0020] Chi-square test was used for count data to generate count test results;

[0021] Based on the results of measurement and counting tests, factors with a significance level of P < 0.05 were selected as sedimentation characteristics.

[0022] In one embodiment, a settlement risk prediction model is constructed based on independent risk factors, including:

[0023] Based on the independent risk factors, the regression coefficient of each independent risk factor was calculated using the multivariate logistic regression model;

[0024] Construct a nomogram prediction model based on the regression coefficients;

[0025] The nomogram prediction model is used to output a visual nomogram containing factor weight scores as a settlement risk prediction model.

[0026] In one embodiment, the parameters of the subsidence risk prediction model are optimized according to the verification results based on the area under the curve and the calibration index, including:

[0027] When the AUC calculated by the validation set and the predicted probability output by the sedimentation risk prediction model is ≤ 0.75 or the AUC calculated based on the sedimentation feature grouping is When the subsidence risk prediction model parameters are adjusted, the following verification operations are repeated based on the adjusted model:

[0028] Based on the prediction probability of the current sedimentation risk prediction model for the validation set, the ROC curve is drawn and the AUC is calculated;

[0029] generating a calibration curve based on the sedimentation characteristic grouping and calculating the slope of the calibration curve;

[0030] Until AUC>0.75 and the slope of the calibration curve∈[0.8,1.2], the optimized sedimentation risk prediction model is output.

[0031] In one embodiment, adjusting the parameters of the subsidence risk prediction model includes any of the following operations:

[0032] Based on the results of univariate analysis, independent risk factor variables were added or deleted in the multivariate logistic regression model;

[0033] By modifying the score weight of each independent risk factor variable in the visual nomogram, the regression coefficient weight of the independent risk factor variable is adjusted.

[0034] In a second aspect, the present application further provides a device for assessing the risk of cervical spine postoperative subsidence, comprising:

[0035] A data acquisition module is used to obtain clinical and imaging data sets of the intervertebral fusion cage settlement group and the non-settlement group;

[0036] Univariate analysis module, used to perform univariate analysis on clinical and imaging data sets using statistical difference analysis to obtain sedimentation characteristics;

[0037] Multivariate regression module, used to screen independent risk factors based on sedimentation characteristics using a multivariate regression model;

[0038] Prediction model building module, used to build a settlement risk prediction model based on independent risk factors;

[0039] The model optimization output module is used to optimize the parameters of the settlement risk prediction model based on the verification results of the area under the curve and the calibration index, and output the settlement risk probability value using the optimized settlement risk prediction model.

[0040] In a third aspect, the present application also provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned method for assessing the risk of postoperative settlement of the cervical spine are implemented.

[0041] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned method for assessing the risk of postoperative cervical subsidence.

[0042] The above-mentioned method, device, computer equipment and storage medium for assessing the risk of postoperative settlement of the cervical spine collect clinical and imaging data sets of the intervertebral fusion device settlement group and the non-settlement group, use statistical difference analysis to screen out settlement features with significant correlation, and identify independent risk features from these features through a multivariate regression model; the settlement risk prediction model constructed in this way outputs a quantitative settlement risk probability value after parameter optimization based on the area under the curve and calibration index, thereby achieving accurate prediction of the patient's postoperative fusion device settlement risk, solving the technical problem that traditional clinical assessment methods are difficult to quantify and predict settlement risk, and providing an objective decision-making basis for preoperative intervention and postoperative management. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0044] Figure 1 This is a flow chart of a method for assessing the risk of cervical vertebrae postoperative settlement;

[0045] Figure 2 This is a structural diagram of a cervical vertebra postoperative settlement risk assessment device according to the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0047] The implementation environment of the present invention is based on a distributed architecture consisting of terminals and servers. The terminals include hospital imaging workstations and clinical data terminals, which are capable of reading and entering medical images. The servers are deployed in a data center and integrate data storage, model calculation, and management modules. The two realize data interaction through a medical communication network. In the application scenario, after surgery, the patient's clinical and imaging data are input and transmitted to the server. After model calculation, the risk prediction results are returned for terminal visualization. In multi-center clinical studies, the terminals upload anonymous case data to the server to iteratively optimize the prediction model. Through the collaborative interaction between the terminal and the server, the risk of cervical spine subsidence after surgery can be quantitatively assessed and clinical decision support can be achieved.

[0048] In one embodiment, Figure 1 As shown, a method for assessing the risk of postoperative cervical spine subsidence is provided. This embodiment uses the method applied to a terminal and server collaborative system as an example for explanation. It is understood that the method can also be applied alone to the server to implement batch data processing, or a lightweight model can be deployed locally on the terminal to complete real-time risk calculation. In this embodiment, the method includes the following steps:

[0049] S01, obtain clinical and imaging data sets of the intervertebral fusion cage settlement group and non-settlement group.

[0050] Among them, the collected clinical and imaging data sets include basic physiological indicators such as patient age, gender, BMI, etc. from the demographic and medical history dimensions, combined with disease type, smoking history and chronic disease history (such as hypertension, diabetes); from the surgical and efficacy dimensions, surgical-related parameters such as lesion segment, operation time, intraoperative blood loss, and functional evaluation indicators such as preoperative / postoperative JOA score are obtained to achieve quantitative records of surgical intervention and efficacy. A combination of clinical follow-up and imaging evaluation can be used to divide patients into fusion cage settlement group and non-sedimentation group: the settlement group, based on postoperative imaging review (such as X-ray, CT), confirms that the fusion cage position or height change meets the settlement standard (such as height loss ≥3mm); the non-sedimentation group, a control group that did not show settlement according to the same standard. Through standardized grouping process, the clinical characteristics of the two groups of samples are ensured to be comparable, providing a reliable control sample set for subsequent statistical analysis.

[0051] S02, univariate analysis of clinical and imaging data sets was performed using statistical difference analysis method to obtain sedimentation characteristics.

[0052] The data set can be categorized into continuous measurement data (e.g., numerical indicators) and discrete count data (e.g., categorical indicators). Appropriate statistical test methods are selected based on the distribution characteristics of the measurement data: parametric tests are used for data that conform to a normal distribution, and nonparametric tests are used for data that are non-normally distributed. Distribution morphology tests and adapted mean / median difference tests are used to quantify the statistical differences in continuous features between the sedimentation and non-settlement groups, generating an index of the degree of association between features and sedimentation. Frequency distribution difference tests are used to analyze the distribution significance of classification features in the two groups of samples, identify statistically significant classification-associated features, and form a sedimentation feature set using a preset significance threshold (e.g., P < 0.05) as a screening benchmark. By deeply integrating medical statistical theory with clinical data and using a standardized significance analysis framework, a universal method for identifying features associated with sedimentation risk is established, providing a scientific feature input foundation for the subsequent construction of prediction models.

[0053] S03, based on the sedimentation characteristics, independent risk factors were screened using a multivariate regression model.

[0054] Specifically, for the binary classification of cage subsidence (subsidence / non-subsidence), a multivariate regression model suitable for classification problems (e.g., logistic regression model) can be used to construct a quantitative association equation between the dependent variable (subsidence status) and the independent variable (subsidence characteristics). By estimating the regression model parameters, the independent contribution of each subsidence characteristic to subsidence risk is quantified. Meanwhile, the interaction between each variable is considered. Based on a preset statistical significance threshold (P < 0.05), stepwise regression (forward / backward) or full subset regression methods are used to iteratively eliminate variables without independent significant effects, retaining features with independent predictive value for subsidence risk. The independent influence strength of each feature is quantified using model parameters (e.g., regression coefficient). The positive or negative sign and absolute value of the regression coefficient respectively represent the direction of risk association (positive / negative correlation) and the degree of influence. This forms an importance ranking of independent risk factors and a standardized set of independent risk factors, including the regression coefficient and significance level of the factors. By deeply combining multivariate statistical analysis theory with clinical risk assessment, a quantifiable and verifiable independent risk factor identification mechanism is established to solve the technical problems of vague variable correlation and difficulty in defining independent influencing factors in traditional assessments.

[0055] S04, construct a subsidence risk prediction model based on independent risk factors.

[0056] Among them, independent risk factors are used as input variables, and a visual prediction model (such as a nomogram) containing the weight scores of each risk factor is constructed based on a multifactor regression model. The combined impact of multiple factors is converted into a risk probability value that can be intuitively interpreted through graphical mapping. The regression coefficient is mapped to the factor weight score in the visualization model, and a three-layer mapping system of risk factor value-score accumulation-risk probability is established to achieve quantitative superposition calculation of multi-factor risks. Statistical modeling is combined with visualization technology to construct a risk prediction tool that is both scientific and clinically practical, breaking through the technical bottleneck of traditional assessments that are mainly based on qualitative judgment and lack quantitative models.

[0057] S05: Optimize the parameters of the settlement risk prediction model based on the verification results of the area under the curve and the calibration index, and use the optimized settlement risk prediction model to output a settlement risk probability value.

[0058] The area under the curve (AUC) metric can be used to quantify the model's ability to distinguish between the sedimentation and non-settlement groups. The ROC curve is plotted using the predicted probabilities of the validation set samples; the closer the AUC value is to 1, the better the model's discriminatory effect. A calibration curve is generated based on the sedimentation feature groups, and the slope of the calibration curve quantifies the consistency between the model's predicted probability and the actual sedimentation probability. A slope close to 1 indicates a high degree of agreement between the predicted probability and the actual result, while a deviation from 1 indicates calibration bias. When the validation metric does not reach the preset threshold, the variable set can be adjusted based on the results of univariate analysis to add potential risk factors, remove redundant variables, or optimize the weight distribution of risk factors. After each parameter adjustment, the process of AUC calculation and calibration curve calibration assessment is repeated until both metrics meet the preset standards. Based on the optimized model, the combined impact of multiple factors is converted into a risk probability value in the range of 0-1. By deeply integrating medical statistical validation metrics with model optimization, a quantitative risk assessment system with self-iterative capabilities is established, breaking through the technical bottlenecks of difficult model performance verification and low risk prediction reliability in traditional assessments.

[0059] The above-mentioned method for assessing the risk of postoperative cervical subsidence obtains clinical and imaging data sets of the intervertebral fusion device subsidence group and the non-subsidence group, uses statistical difference analysis to perform univariate analysis to screen out subsidence characteristics, identifies independent risk factors based on a multivariate regression model, constructs a subsidence risk prediction model, optimizes model parameters through the area under the curve and calibration index, and outputs risk probability values. It systematically integrates multidimensional clinical and imaging data, screens risk factors through objective statistical analysis, and constructs a quantitative prediction model, breaking through the limitations of traditional assessment that relies on clinical experience, qualitative imaging analysis, and subsequent follow-up, and solving the technical problems of existing technologies that lack quantitative prediction models, cannot provide early warnings, and are highly subjective, thereby achieving a quantitative assessment of the risk of postoperative cervical fusion device subsidence and providing an objective decision-making basis for preoperative intervention and postoperative management.

[0060] In one embodiment, obtaining clinical and imaging data sets of an intervertebral fusion cage settlement group and a non-settlement group includes:

[0061] S11, obtain the patient's age, gender, BMI index, disease duration, disease type, smoking history, history of hypertension and diabetes, lesion segment, operation time, intraoperative blood loss, and JOA score at set time points before and after the operation as the clinical data set;

[0062] S12, obtaining the patient's cervical spine X-ray, CT, and MRI as an image dataset.

[0063] Specifically, the patient's age, gender, BMI index, disease duration, disease type, smoking history, history of hypertension and diabetes, lesion segment, operation time, intraoperative blood loss, and JOA scores at set time points before and after the operation were obtained as a clinical data set, covering the patient's demographic characteristics, underlying disease history, operation-related parameters and neurological function assessment indicators, providing a quantitative basis for analyzing the impact of systemic factors and surgical intervention on fusion cage settlement; the MRI of the imaging data set can quantitatively reflect the patient's cervical spine bone quality, and the X-ray lateral film and cervical spine CT provide intuitive imaging evidence of the position and morphology of the fusion cage, supplementing the clinical data from the perspective of bone biomechanical properties and anatomical structure changes. The combination of the two forms a multi-dimensional data system, which provides scientific and comprehensive data support for the subsequent screening of settlement-related features and the construction of a prediction model through statistical analysis.

[0064] In one embodiment, the image dataset also includes the following feature indicators calculated based on CT and MRI:

[0065] S21, EBQ: Based on the midsagittal image of the patient's cervical spine MRI T1-weighted image, the signal intensity of the upper and lower endplates was measured within the 0.6 mm thick endplate ROI of the surgical segment excluding Schmorl's tubercle. The signal intensity was calculated according to the formula EBQ = (endplates upper +endplates lower ) / 2(C2 CSF ) EBQ is calculated, where endplates upper is the signal intensity of the upper endplate, lower is the signal intensity of the lower end plate, C2 CSF is the cerebrospinal fluid signal intensity at the level of the C2 vertebra;

[0066] S22, VBQ: Based on the midsagittal image of the patient's cervical spine MRI T1-weighted image, within the vertebral cancellous bone ROI of C3-C6, avoiding the hemangioma and posterior venous plexus, according to the formula VBQ = M(C3-C6) / C2 CSFThe VBQ was calculated, where M(C3–C6) was the median signal intensity of the C3–C6 vertebrae;

[0067] S23, CT-HU: Based on the CT image of the patient's cervical spine, the HU values ​​of the upper and lower end plates were measured within the bone area ROI 0.6 mm below the fusion cage contact surface to obtain CT-HU.

[0068] For example, based on the midsagittal image of cervical spine MRI T1 weighted image, a 0.6 mm thick endplate ROI (region of interest) was delineated in the surgical segment, and the influence of Schmorl's node (protrusion of nucleus pulposus caused by damage to vertebral endplate cartilage) was excluded. The signal intensity values ​​of the upper and lower endplates were measured respectively, and the formula EBQ=(endplates upper +endplates lower ) / 2(C2 CSF ) was used to calculate the EBQ. The ROI was set based on a 0.6mm thickness focusing on the cancellous bone area of ​​the endplate to exclude interference from cortical bone; excluding Schmorl's nodes can avoid the influence of abnormal signals in the lesion area on the results. The signal intensity of C2 cerebrospinal fluid was used as a benchmark to eliminate the signal deviation caused by differences in imaging parameters of different devices and improve the comparability of indicators. EBQ reflects the MRI signal characteristics of the endplate bone tissue. Decreased signal intensity often indicates bone degeneration or osteoporosis, which is associated with a decrease in the support force of the fusion device and can be used as an imaging predictor of subsidence risk. In the midsagittal image of the cervical spine MRIT1-weighted image, the C3-C6 vertebral cancellous bone ROI was selected (avoiding hemangioma and posterior venous plexus to exclude interference from vascular structures), and the formula VBQ=M(C3-C6) / C2 was used. CSF VBQ was calculated. C3 to C6 are segments with greater cervical spine mobility, and the quality of their cancellous bone directly impacts spinal biomechanical stability. Avoiding hemangiomas and the posterior venous plexus in the ROI prevents interference of high vascular signals with bone signals. VBQ reflects trabecular bone density and the degree of bone marrow adiposity by measuring the signal intensity of vertebral cancellous bone. Decreased signal values ​​indicate bone loss or osteomalacia, potentially increasing the risk of cage subsidence. Based on cervical spine CT images, the CT-HU values ​​of the superior and inferior endplates were measured within a 0.6 mm bone region ROI below the cage contact surface to reflect the degree of tissue X-ray attenuation. The ROI focuses on the stress concentration area at the cage-bone interface at a depth of 0.6 mm. Bone density in this area directly impacts the initial stability and long-term risk of cage subsidence. CT-HU values ​​objectively reflect endplate bone density. Studies have shown that decreased CT-HU values ​​of the endplate are positively correlated with the incidence of cage subsidence, making it an important imaging indicator for assessing bone support.

[0069] In one embodiment, a univariate analysis of clinical and imaging data sets is performed using statistical variance analysis to obtain sedimentation characteristics, including:

[0070] S31, define the data types of clinical and imaging datasets, and divide them into measurement data and counting data according to the data types;

[0071] S32, perform a Shapiro-Wilk normality test on the measurement data and select a statistical method based on the test results: if the measurement data conforms to a normal distribution, use an independent sample t-test; otherwise, use a Mann-Whitney U rank sum test; combine the independent sample t-test results and the Mann-Whitney U rank sum test results to generate the measurement test results;

[0072] S33, applying a chi-square test to the count data to generate a count test result;

[0073] S34. Based on the measurement test results and the counting test results, factors with a significance level of P < 0.05 are selected as sedimentation characteristics.

[0074] For example, the data types of clinical and imaging datasets are defined as follows: measurement data, quantitative data with continuous numerical features, such as patient age (years), BMI (kg / m 2), operation time (minutes), etc., can reflect the strength of the feature through the size of the value; counting data, qualitative data presented in the form of classification or enumeration, such as gender (male / female), smoking history (yes / no), lesion segment (single segment / multiple segments), etc., reflect the category distribution of the feature. Through the classification of data types, the foundation is laid for the subsequent selection of appropriate statistical test methods. Measurement data focuses on numerical differences, and counting data focuses on frequency distribution differences. The two require different statistical models for significance analysis to ensure the scientific nature of the statistical methods and the reliability of the results. By calculating the degree of fit between the sample data and the normal distribution, the value range of the test statistic W is [0,1]. When P>0.05, the data is considered to conform to the normal distribution, otherwise it is non-normally distributed. Normal distribution is the premise assumption of parametric tests (such as t-test). If the data does not meet the normality, non-parametric test methods must be used to avoid statistical deviations due to invalid assumptions. The independent sample t-test is applicable to normally distributed measurement data. By comparing the mean difference between the sedimentation group and the non-subsidence group, the t-value and P-value are calculated to quantify the significant association between the two groups. The Mann-Whitney U rank sum test is applicable to non-normally distributed measurement data. By sorting the data and comparing the rank sum difference, the rank sum difference is calculated to mitigate the influence of abnormal data distribution on the test results. The results of the two tests are integrated to generate a measurement test report containing statistics (t-value / U-value) and P-values, which intuitively presents the degree of association between each measurement feature and sedimentation. The chi-square test measures the distributional significance of count data between two sample groups by calculating the difference between the actual observed frequency and the theoretical expected frequency. It is used to analyze the association between categorical variables such as gender, disease type, and smoking history and cage subsidence. For example, the proportion of smoking history in the sedimentation group is significantly higher than that in the non-subsidence group, thereby identifying potential risk factors. A P < 0.05 threshold is used as the statistical significance threshold. When the test result for a feature is P < 0.05, it is considered to be significantly associated with cage subsidence and included in the sedimentation feature set. Through statistical methods, clinical and imaging data are converted into significant settlement characteristics, which solves the technical problems of the lack of quantitative models and objective standards for risk factor screening in traditional evaluations, and lays a data foundation for the accurate assessment of the risk of fusion device settlement after cervical surgery.

[0075] In one embodiment, a settlement risk prediction model is constructed based on independent risk factors, including:

[0076] S41, based on independent risk factors, the regression coefficient of each independent risk factor was calculated using the multivariate logistic regression model;

[0077] S42, constructing a nomogram prediction model based on regression coefficients;

[0078] S43, using the nomogram prediction model to output a visual nomogram containing factor weight scores as a settlement risk prediction model.

[0079] Specifically, the multivariate logistic regression model is applicable to binary outcomes (such as cage subsidence / non-subsidence). It can simultaneously incorporate multiple independent risk factors, control for interactions between variables, and quantify the independent contribution of each factor to subsidence risk using regression coefficients, addressing the inability of traditional univariate analysis to assess the synergistic effects of multiple variables. When the regression coefficient of an independent risk factor is greater than 0, it indicates an increased risk of subsidence; conversely, it indicates a decreased risk. The larger the absolute value of the regression coefficient, the greater the risk factor's impact. A nomogram prediction model is constructed based on the regression coefficients. By converting the regression coefficients of the multivariate logistic regression model into a visual score scale, a three-level mapping system is established: risk factor value selection, score calculation, and risk probability. For each independent risk factor, a scale interval is divided according to its value range, and each interval corresponds to a score calculated based on the regression coefficient (e.g., each 10-year increase in age corresponds to X points). The scores of each factor are accumulated to obtain a total score. The final subsidence probability is generated by mapping the total score to the risk probability (based on the model fitting results). By converting the abstract statistical model into a graphical tool, clinicians can quickly accumulate scores using a ruler scale, avoiding complex formula calculations and improving risk assessment efficiency. The output is a visual nomogram, whose components include: a horizontal scale, the value range and corresponding score of each independent risk factor; a total score scale, a summary area for accumulating the scores of each factor; and a risk probability axis, which is the probability of settlement (0-100%) based on the total score mapping. Through the technical path of statistical modeling-visualization conversion-clinical adaptation, a settlement risk prediction model that is both scientific and practical has been constructed, filling the technical gaps in traditional assessments that lack quantitative tools and are difficult to integrate multiple factors, and providing a standardized solution for the precise prevention of settlement of fusion devices after cervical surgery.

[0080] In one embodiment, the parameters of the subsidence risk prediction model are optimized according to the verification results based on the area under the curve and the calibration index, including:

[0081] S51, when the AUC calculated by the validation set and the predicted probability output by the settlement risk prediction model is ≤ 0.75 or the AUC calculated based on the settlement feature grouping When the subsidence risk prediction model parameters are adjusted, the following verification operations are repeated based on the adjusted model:

[0082] S51.1, based on the prediction probability of the current subsidence risk prediction model for the validation set, draw the ROC curve and calculate the AUC;

[0083] S51.2, generating a calibration curve based on the sedimentation characteristic grouping and calculating the slope of the calibration curve;

[0084] S52, until AUC>0.75 and the slope of the calibration curve ∈[0.8,1.2], output the optimized sedimentation risk prediction model at this time.

[0085] For example, the area under the curve (AUC) is used to draw the ROC curve based on the predicted probability of the validation set. The AUC quantifies the model's ability to distinguish between the sedimentation group and the non-sedimentation group. AUC = 1 indicates complete distinction, and AUC = 0.5 indicates random prediction. AUC>0.75 is used as the standard for effective distinction, and reference is made to the conventional requirements of medical prediction models (AUC>0.7 indicates clinical application value) to ensure that the model has sufficient risk stratification capabilities. The slope of the calibration curve groups the validation set according to the predicted probability (such as each 10% probability as a group), and calculates the linear regression slope of the actual sedimentation rate and the predicted probability of each group. Slope = 1 means that the predicted probability is completely consistent with the actual result, and a slope deviation from 1 indicates that there is a calibration bias (such as a slope <0.8 indicates that the model over-predicts the risk, and a slope >1.2 indicates that the risk is underestimated). This ensures the consistency of the model's predicted probability with the actual clinical results and avoids misjudgment of risk due to calibration bias. Model adjustment was initiated when any of the following situations occurred: AUC ≤ 0.75: the model had insufficient ability to distinguish between the sedimentation and non-settlement groups, and high-risk patients might be missed or low-risk populations might be misjudged; The deviation between the model's predicted probability and the actual sedimentation rate was too large, reducing its clinical guidance value. The adjusted model was applied to the validation set to generate a predicted sedimentation probability for each sample. Using the predicted probability as the threshold, a receiver operating characteristic (ROC) curve was plotted, and the area under the curve (AUC) was calculated to assess whether the model's discriminatory ability had improved. The validation set was grouped by predicted probability, and the actual sedimentation rate for each group was calculated. A calibration curve of predicted probability versus actual sedimentation rate was plotted. The slope of the calibration curve was obtained through linear regression fitting to assess the reliability of the model's predictions. When AUC > 0.75 and the slope of the calibration curve ∈ [0.8, 1.2], optimization was stopped and the model was output to ensure that the model met the dual requirements of discriminatory ability and calibration accuracy. By deeply coupling medical statistical validation indicators with model parameter optimization, a risk prediction system with self-iterative capabilities was established, breaking through the technical bottleneck of the lack of validation and optimization after traditional model construction, and achieving a technological leap from empirical judgment to precise quantitative prediction in the risk assessment of cervical spine surgery.

[0086] In one embodiment, adjusting the parameters of the subsidence risk prediction model includes any of the following operations:

[0087] S61, based on the results of univariate analysis, add or delete independent risk factor variables in the multivariate logistic regression model;

[0088] S62, by modifying the score weight of each independent risk factor variable in the visual nomogram, adjusting the regression coefficient weight of the independent risk factor variable.

[0089] Specifically, the parameters of the sedimentation risk prediction model were adjusted. Using sedimentation characteristics identified in univariate analysis (P < 0.05) as a benchmark, the following steps were performed in the multivariate logistic regression model: adding variables. When a characteristic was significantly associated with sedimentation in univariate analysis (e.g., the newly discovered imaging indicator CT-HU) but not included in the current model, it was introduced as a candidate variable into the regression equation. Deleting variables: If a variable in the model had a multivariate regression P value ≥ 0.05, indicating insufficient independent predictive value, it was removed from the model. By dynamically adjusting the variable set, only independently statistically significant risk factors were retained in the model, avoiding overfitting or omission of key factors. The score weights of each factor in the nomogram directly correspond to the regression coefficients of the multivariate logistic regression model. Adjusting the score weights is equivalent to proportionally modifying the regression coefficients of the regression model, thus changing the contribution of the factor to the risk probability. This can be achieved by rescaling the factor scales in the nomogram or modifying the score calculation rules (e.g., adjusting the score for each decade of age from 20 to 25 points). Regression coefficient weights can be optimized through graphical manipulation. Through a systematic parameter adjustment mechanism, the cervical spine postoperative subsidence risk prediction model can be dynamically optimized according to the characteristics of clinical data, thereby improving prediction accuracy and clinical practicality, and providing a standardized solution for postoperative risk assessment in the context of precision medicine.

[0090] The above-mentioned method for assessing the risk of postoperative cervical spine subsidence obtains a clinical data set including age, gender, BMI, disease course, disease type, smoking history, history of hypertension and diabetes, lesion segment, operation time, intraoperative blood loss, preoperative and postoperative JOA scores, as well as X-ray lateral film, CT, and MRI imaging data sets (including characteristic indicators such as EBQ and VBQ calculated based on MRI and CT-HU calculated based on CT). The clinical and imaging data are classified as measurement data and counting data. After the Shapiro-Wilk normality test, univariate analysis is performed using the independent sample t-test or Mann-Whitney U rank sum test for measurement data and the chi-square test for counting data. Subsidence characteristics with P < 0.05 are screened out. Then, independent risk factors are identified based on the multivariate logistic regression model. A nomogram prediction model including regression coefficients is constructed and a visual nomogram is output. The AUC (the model is adjusted when AUC ≤ 0.75) and the slope of the calibration curve (when AUC ≤ 0.75) are calculated using the validation set. The model is adjusted in real time to optimize parameters and output the subsidence risk probability value. Through the systematic integration of multi-dimensional data, statistical analysis and model verification, quantitative assessment is achieved, which solves the problems of traditional assessment such as lack of quantitative models, reliance on post-diagnosis and high subjectivity, and provides an objective basis for clinical intervention.

[0091] It should be understood that, although the steps in the flowcharts of the above embodiments are shown in sequence as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be performed in other orders. Moreover, at least a portion of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times. The execution order of these steps or stages is not necessarily to be performed in sequence, but can be performed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0092] Based on the same inventive concept, the embodiments of the present application also provide a cervical vertebrae postoperative settlement risk assessment device for implementing the aforementioned cervical vertebrae postoperative settlement risk assessment method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more embodiments of the cervical vertebrae postoperative settlement risk assessment device provided below can be found in the above-mentioned limitations of the cervical vertebrae postoperative settlement risk assessment method, and will not be repeated here.

[0093] In an exemplary embodiment, Figure 2 As shown, a device for assessing the risk of cervical spine postoperative settlement is provided, comprising:

[0094] The data acquisition module 101 is used to acquire clinical and imaging data sets of the intervertebral fusion cage settlement group and the non-settlement group;

[0095] A univariate analysis module 102 is used to perform univariate analysis on clinical and imaging data sets using statistical difference analysis to obtain sedimentation characteristics;

[0096] The multi-factor regression module 103 is used to screen and obtain independent risk factors based on the sedimentation characteristics using a multi-factor regression model;

[0097] The prediction model building module 104 is used to build a settlement risk prediction model based on independent risk factors;

[0098] The model optimization output module 105 is used to optimize the parameters of the settlement risk prediction model based on the verification results of the area under the curve and the calibration index, and output the settlement risk probability value using the optimized settlement risk prediction model.

[0099] In one embodiment, the data acquisition module 101 is further configured to:

[0100] The patients' age, gender, BMI, disease duration, disease type, smoking history, history of hypertension and diabetes, lesion segment, operation time, intraoperative blood loss, and JOA scores at preoperative and postoperative time points were obtained as clinical data sets;

[0101] The patient's cervical spine X-ray, CT and MRI were obtained as the imaging dataset.

[0102] In one embodiment, the image data set acquired by the data acquisition module 101 further includes the following characteristic indicators calculated based on CT and MRI:

[0103] The following characteristic indicators are calculated based on CT and MRI:

[0104] EBQ: Based on the midsagittal image of the patient's cervical spine MRI T1-weighted image, the signal intensity of the upper and lower endplates was measured within the 0.6 mm thick endplate ROI of the surgical segment excluding Schmorl's tubercle. upper +endplates lower ) / 2(C2 CSF ) EBQ is calculated, where endplates upper is the signal intensity of the upper endplate, lower is the signal intensity of the lower end plate, C2 CSF is the cerebrospinal fluid signal intensity at the level of the C2 vertebra;

[0105] VBQ: Based on the midsagittal image of the patient's cervical spine MRI T1-weighted image, within the vertebral cancellous bone ROI of C3 to C6, avoiding the hemangioma and posterior venous plexus, according to the formula VBQ = M(C3 to C6) / C2 CSF The VBQ was calculated, where M(C3–C6) was the median signal intensity of the C3–C6 vertebrae;

[0106] CT-HU: Based on the CT image of the patient's cervical spine, the HU values ​​of the upper and lower end plates were measured in the bone area ROI 0.6 mm below the fusion device contact surface to obtain CT-HU.

[0107] In one embodiment, the univariate analysis module 102 is further configured to:

[0108] Define the data types of clinical and imaging datasets and classify them into measurement data and counting data based on the data type;

[0109] Perform the Shapiro-Wilk normality test on the measurement data and select the statistical method based on the test results: if the measurement data conforms to the normal distribution, use the independent sample t test; otherwise, use the Mann-Whitney U rank sum test; combine the independent sample t test results and the Mann-Whitney U rank sum test results to generate the measurement test results;

[0110] Chi-square test was used for count data to generate count test results;

[0111] Based on the results of measurement and counting tests, factors with a significance level of P < 0.05 were selected as sedimentation characteristics.

[0112] In one embodiment, the prediction model building module 104 is further configured to:

[0113] Based on the independent risk factors, the regression coefficient of each independent risk factor was calculated using the multivariate logistic regression model;

[0114] Construct a nomogram prediction model based on the regression coefficients;

[0115] The nomogram prediction model is used to output a visual nomogram containing factor weight scores as a settlement risk prediction model.

[0116] In one embodiment, the model optimization output module 105 is further configured to:

[0117] When the AUC calculated by the validation set and the predicted probability output by the sedimentation risk prediction model is ≤ 0.75 or the AUC calculated based on the sedimentation feature grouping is When the subsidence risk prediction model parameters are adjusted, the following verification operations are repeated based on the adjusted model:

[0118] Based on the prediction probability of the current sedimentation risk prediction model for the validation set, the ROC curve is drawn and the AUC is calculated;

[0119] generating a calibration curve based on the sedimentation characteristic grouping and calculating the slope of the calibration curve;

[0120] Until AUC>0.75 and the slope of the calibration curve∈[0.8,1.2], the optimized sedimentation risk prediction model is output.

[0121] In one embodiment, the model optimization output module 105 is further configured to:

[0122] Adjust the parameters of the subsidence risk prediction model, including any of the following operations:

[0123] Based on the results of univariate analysis, independent risk factor variables were added or deleted in the multivariate logistic regression model;

[0124] By modifying the score weight of each independent risk factor variable in the visual nomogram, the regression coefficient weight of the independent risk factor variable is adjusted.

[0125] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the power supply safety management method as described above when executing the computer program.

[0126] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0127] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0128] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A method for assessing the risk of postoperative cervical spine subsidence, characterized in that: The method comprises: Clinical and imaging data sets were obtained for the interbody fusion cage settlement group and the non-settlement group; The clinical and imaging data sets were analyzed using univariate statistical analysis to obtain sedimentation characteristics; Based on the sedimentation characteristics, independent risk factors were screened using a multivariate regression model; Constructing a subsidence risk prediction model based on the independent risk factors; Based on the verification results of the area under the curve and the calibration index, the parameters of the settlement risk prediction model are optimized, and the settlement risk probability value is output using the optimized settlement risk prediction model.

2. The method according to claim 1, characterized in that The clinical and imaging data sets of the intervertebral fusion cage settlement group and the non-settlement group are obtained, including: The patient's age, gender, BMI index, disease duration, disease type, smoking history, history of hypertension and diabetes, lesion segment, operation time, intraoperative blood loss, and JOA scores at set time points before and after the operation were obtained as the clinical data set; The patient's cervical spine X-ray, CT and MRI are obtained as the image dataset.

3. The method according to claim 2, characterized in that The image dataset also includes the following feature indicators calculated based on the CT and MRI: EBQ: Based on the midsagittal image of the patient's cervical spine MRI T1-weighted image, the signal intensity of the upper and lower endplates was measured in the 0.6 mm thick endplate ROI of the surgical segment excluding Schmorl's tubercle. upper +endplates lower ) / 2(C2 CSF ) calculates the EBQ, where endplates upper is the signal intensity of the upper endplate, lower is the signal intensity of the lower end plate, C2 CSF is the cerebrospinal fluid signal intensity at the level of the C2 vertebra; VBQ: Based on the midsagittal image of the patient's cervical spine MRI T1-weighted image, within the vertebral cancellous bone ROI of C3-C6, avoiding the hemangioma and posterior venous plexus, according to the formula VBQ = M(C3-C6) / C2 CSF The VBQ was calculated, where M(C3–C6) was the median signal intensity of the C3–C6 vertebrae; CT-HU: Based on the CT image of the patient's cervical spine, the HU values ​​of the upper and lower end plates were measured within the bone region ROI 0.6 mm below the fusion cage contact surface to obtain the CT-HU.

4. The method according to claim 3, characterized in that The statistical difference analysis method is used to perform univariate analysis on the clinical and imaging data sets to obtain sedimentation characteristics, including: Defining the data types of the clinical and imaging data sets, and dividing them into measurement data and counting data according to the data types; Perform a Shapiro-Wilk normality test on the measurement data, and select a statistical method based on the test results: if the measurement data conforms to a normal distribution, use an independent sample t-test; otherwise, use a Mann-Whitney U rank sum test; combine the independent sample t-test results and the Mann-Whitney U rank sum test results to generate a measurement test result; Using a chi-square test on the counting data to generate a counting test result; Based on the measurement test results and the counting test results, factors with a significance level of P<0.05 were screened out as the sedimentation characteristics.

5. The method according to claim 1, wherein The constructing of a settlement risk prediction model based on the independent risk factors includes: Based on the independent risk factors, the regression coefficient of each independent risk factor was calculated using a multivariate logistic regression model; Constructing a nomogram prediction model based on the regression coefficients; The nomogram prediction model is used to output a visual nomogram containing factor weight scores as the subsidence risk prediction model.

6. The method according to claim 5, characterized in that Optimizing parameters of the settlement risk prediction model according to the verification results based on the area under the curve and the calibration index includes: When the AUC calculated by the prediction probability of the validation set and the output of the sedimentation risk prediction model is ≤ 0.75 or the slope of the calibration curve calculated based on the sedimentation feature grouping When the subsidence risk prediction model parameters are adjusted, the following verification operations are repeated based on the adjusted model: Based on the prediction probability of the current sedimentation risk prediction model for the validation set, the ROC curve is drawn and the AUC is calculated; generating a calibration curve based on the sedimentation characteristic grouping and calculating the slope of the calibration curve; Until the AUC>0.75 and the slope of the calibration curve∈[0.8,1.2], the optimized sedimentation risk prediction model is output.

7. The method according to claim 6, characterized in that The adjusting of the parameters of the subsidence risk prediction model includes any of the following operations: Based on the results of univariate analysis, add or delete independent risk factor variables in the multivariate logistic regression model; The regression coefficient weight of the independent risk factor variable is adjusted by modifying the score weight of each independent risk factor variable in the visual nomogram.

8. A device for assessing the risk of cervical vertebrae postoperative settlement, characterized in that: The device comprises: A data acquisition module is used to obtain clinical and imaging data sets of the intervertebral fusion cage settlement group and the non-settlement group; A univariate analysis module is used to perform univariate analysis on the clinical and imaging data sets using a statistical difference analysis method to obtain sedimentation characteristics; A multi-factor regression module is used to screen and obtain independent risk factors based on the sedimentation characteristics using a multi-factor regression model; A prediction model building module, used to build a settlement risk prediction model based on the independent risk factors; The model optimization output module is used to optimize the parameters of the settlement risk prediction model based on the verification results of the area under the curve and the calibration index, and output the settlement risk probability value using the optimized settlement risk prediction model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images