Neurosyphilis detection and prediction model based on clinical diagnostic variables, and construction method therefor and application thereof

Abstract

A neurosyphilis detection and prediction model based on clinical diagnostic variables, and a construction method therefor and an application thereof. The model is constructed on the basis of the following five key clinical variables: cerebral ischemia or infarction, serum Treponema pallidum-specific antibody (anti-TP), TRUST titer, ataxia and decreased vision. Numerical analysis is performed on the variables, LASSO regression is performed to screen variables and multivariate logistic regression analysis is performed to construct a risk prediction model in the form of a nomogram for calculating a neurosyphilis risk probability of patients. The model integrates serological, imaging and clinical manifestation data, and thus can help clinicians identify high-risk patients on a non-invasive basis. This innovation fills a gap in the prior art, and also significantly optimizes the clinical diagnosis process, thereby providing strong support for the early screening and intervention of neurosyphilis.

Description

A neurosyphilis detection and prediction model based on clinical diagnostic variables, its construction method, and its application. Technical Field

[0001] This application relates to the field of medical diagnostics, and more specifically, it relates to a neurosyphilis risk prediction model based on clinical data and mathematical modeling techniques, as well as the model construction method and application. Background Technology

[0002] Neurosyphilis (NS) is a central nervous system complication caused by infection with Treponema pallidum (TP). It can occur at any stage of syphilis infection and its course is complex. Based on clinical manifestations and pathological features, neurosyphilis can be classified into asymptomatic, symptomatic, and late-onset types. Asymptomatic patients usually have no obvious neurological symptoms and are diagnosed only through cerebrospinal fluid examination; symptomatic neurosyphilis presents with diverse neurological lesions such as meningitis, cerebral vasculitis, or spinal cord damage; late-onset neurosyphilis usually occurs more than 10 years after infection, is more severe, and is often accompanied by irreversible neurological damage. These diverse manifestations make neurosyphilis a major challenge in clinical diagnosis.

[0003] Currently, the diagnosis of neurosyphilis relies on cerebrospinal fluid (CSF) examination, particularly abnormalities in white blood cell count and protein levels. Specific tests such as a positive anti-TP antibody test and non-specific tests such as TRUST titer are also important reference indicators. However, all these methods require lumbar puncture to obtain samples, which is an invasive procedure that not only carries potential risks but also causes psychological burden for some patients. Furthermore, CSF test results vary significantly among different patients, meaning that the sensitivity and specificity of the diagnosis cannot fully meet clinical needs.

[0004] The clinical manifestations of neurosyphilis are complex and diverse, closely related to the time, location, and treatment of infection. Some patients may only present with mild symptoms such as headache and nausea, while others may experience ischemic stroke caused by cerebrovascular inflammation, ataxia, or mental disorders. These symptoms largely overlap with various central nervous system diseases (such as encephalitis, stroke, and Alzheimer's disease), making neurosyphilis easily misdiagnosed or missed. Furthermore, due to the slow progression of symptoms, many patients are not diagnosed until serious complications develop, missing the optimal window for early intervention.

[0005] Epidemiological studies show that the incidence of neurosyphilis varies significantly among different populations. HIV-positive patients have a significantly higher incidence of neurosyphilis due to impaired immune function, and their condition is often more complex. Furthermore, elderly patients are more susceptible to central nervous system infections due to a decline in immune system function. In the eastern coastal areas of my country, the number of neurosyphilis cases is relatively high, which may be related to the prevalence of syphilis in that region.

[0006] Currently, diagnostic techniques for neurosyphilis still have many shortcomings. On the one hand, traditional diagnostic methods mainly rely on cerebrospinal fluid (CSF) testing, lacking non-invasive diagnostic tools based on serological, imaging, and clinical manifestations. On the other hand, existing research often focuses on the analysis of single variables (such as TRUST titer or CSF leukocyte levels), failing to integrate multiple clinical indicators. Although some studies have indicated that specific variables (such as high TRUST titer) are associated with the risk of neurosyphilis, a standardized risk prediction model has not yet been developed.

[0007] In recent years, the application of machine learning technology in medical diagnosis has provided a new direction for predicting the risk of neurosyphilis. By integrating various clinical data and statistical analysis methods, efficient and accurate risk assessment tools can be constructed. For example, the LASSO regression method filters key variables in high-dimensional data to avoid overfitting; logistic regression models can quantitatively predict the probability of a patient developing the disease. Meanwhile, the visualization of nomograms facilitates rapid interpretation of risk scores by clinicians, improving the convenience of practical operation.

[0008] Therefore, if an efficient detection and prediction model for neurosyphilis can be developed, the clinical diagnostic process for neurosyphilis can be optimized, providing strong support for early screening and intervention of neurosyphilis. Summary of the Invention

[0009] To address the shortcomings of the aforementioned background technology, this invention proposes a neurosyphilis risk prediction model based on retrospective data and multivariate analysis, which can efficiently assess the risk of patients and provide decision support for clinicians.

[0010] To achieve the above objectives, this application adopts the following technical solution:

[0011] This invention discloses a neurosyphilis risk prediction model based on multivariate analysis, aiming to rapidly quantify the risk probability of a patient developing neurosyphilis by constructing a non-invasive prediction tool. This technical solution screens multiple clinical variables significantly associated with neurosyphilis, combines machine learning and statistical modeling techniques to complete model construction and evaluation, ultimately realizing a clinically operable risk assessment system.

[0012] In the first aspect, this application discloses a neurosyphilis detection and prediction model based on clinical diagnostic variables. The model is constructed based on the following five key clinical variables: cerebral ischemia or infarction, serum syphilis-specific antibody (anti-TP), TRUST titer, ataxia, and visual impairment. Through numerical analysis of the above variables, LASSO regression is used to screen variables and multivariate logistic regression analysis to construct a risk prediction model in the form of a nomogram, which is used to calculate the probability of neurosyphilis risk for patients.

[0013] It should be noted that the core variables of the model were selected based on clinical practice and statistical analysis results. The risk factors include cerebral ischemic infarction, serum syphilis-specific antibodies, serum TRUST titer, ataxia, and visual impairment.

[0014] The "cerebral ischemia or infarction" described in this invention is a marker of central nervous system injury identified through cranial MRI. This variable is a relatively common imaging abnormality in patients with neurosyphilis and is closely related to central nervous system infection and cerebral vasculitis. Treponema pallidum infection can cause meningovasculitis, leading to intimal fibrosis, luminal narrowing, and reduced blood flow, thereby causing insufficient cerebral blood supply and ischemic lesions. This variable has a high detection rate in patients with neurosyphilis and is an important indicator for assessing central nervous system involvement.

[0015] The serum syphilis-specific antibody described in this invention, also known as "Serum anti-TP(S / CO)", is a specific serological marker of syphilis infection and can indicate the status of syphilis infection. Although its lifelong positivity may interfere with the dynamic assessment of some patients, its numerical changes are still an important reference for judging central nervous system involvement. As a highly specific indicator, anti-TP provides basic support for neurosyphilis risk assessment.

[0016] The "serum TRUST titer" described in this invention is the same as "Serum TRUST titer," which is a non-specific antibody test for detecting syphilis activity. The result is expressed as a titer, reflecting the activity and degree of infection. In neurosyphilis risk assessment, TRUST titer is an important parameter for measuring the patient's disease activity and can significantly improve the predictive ability of the model.

[0017] The "ataxia" described in this invention is a neurological disorder characterized by unsteady gait, dysarthria, limb weakness, and dizziness. In neurosyphilis, damage to the cerebellum or spinal cord often causes ataxia, manifesting as incoordination or impaired balance. These symptoms may gradually worsen, severely impacting the patient's daily activities. The presentation of ataxia is diverse; some patients may only exhibit mild instability, while others experience significant dizziness and limb weakness. The presence of ataxia usually indicates deep involvement of the nervous system and is an important clinical feature in assessing neurosyphilis.

[0018] The term "decreased vision" as used in this invention refers to blurred vision, narrowed visual field, or even blindness, typically associated with optic nerve damage. Decreased vision is a common neurological manifestation in patients with neurosyphilis, usually caused by optic neuritis or cerebrovascular disease. Patients may describe blurred vision, shadows obscuring vision, or visual field defects, which can lead to unilateral or bilateral vision loss in severe cases. Decreased vision not only affects the patient's quality of life but may also indicate the progression of neurosyphilis. Early identification and intervention are crucial for improving patient prognosis.

[0019] Furthermore, the present invention presents the prediction model in the form of a nomogram, which includes five key risk factors: cerebral ischemic infarction, serum syphilis-specific antibodies, TRUST titer, ataxia, and visual impairment.

[0020] Furthermore, the prediction model generates a patient's total risk score by accumulating the individual scores of each variable, and predicts the probability of the patient developing neurosyphilis based on the total score.

[0021] Preferably, when the total neurosyphilis risk prediction score is less than 42, the patient is assessed as low risk;

[0022] Preferably, when the total score is greater than 46, the patient is assessed as high risk;

[0023] Preferably, for patients with a total score between 42 and 46, their risk level cannot be clearly determined and requires comprehensive evaluation in conjunction with other clinical factors or further examinations.

[0024] Secondly, this application provides a method for constructing the above-mentioned model, including the following steps:

[0025] Step 1: Collect syphilis patients, screen out patients who are positive for both anti-TP and TRUST serum to include in the study, and collect clinical diagnostic variable data;

[0026] Step 2: Randomly divide patients into training and testing sets according to a certain proportion. The training set is used for model development, and the validation set is used to evaluate the model's discriminative ability, predictive consistency, and clinical applicability.

[0027] Step 3: Perform univariate analysis and LASSO regression screening on the clinical diagnostic variable data in the training set to identify key variables that are significantly associated with neurosyphilis;

[0028] Step 4: Use a multi-factor logistic regression model to model the selected key variables, calculate the regression coefficient of each variable, and generate a risk prediction model;

[0029] Step 5: Model Validation.

[0030] Preferably, the construction method has at least one of the following features:

[0031] In step 1, clinical diagnostic variable data include age, sex, TRUST titer, anti-TP, cerebral ischemic infarction (MRI), ataxia, visual impairment, blood pressure, blood sugar, hearing impairment, syncope, headache, and history of stroke.

[0032] In step 2, patients are randomly divided into a training set and a test set at a ratio of 8:2;

[0033] In step 3, the key variables identified included cerebral ischemic infarction, anti-TP, TRUST titer, ataxia, and visual impairment.

[0034] Step 5 involves the following steps: First, the receiver operating characteristic (ROC) curve is used to assess the model's discriminative ability, and the area under the curve (AUC) of the training and validation sets is calculated separately. Second, the consistency between the model's predicted values ​​and actual observed values ​​is assessed through calibration curve analysis. Third, the net benefit of the model at different risk thresholds is assessed through clinical decision curve (DCA) analysis.

[0035] The performance of the model was evaluated using ROC curves, with an AUC of 0.816 for the training set and 0.843 for the validation set.

[0036] Thirdly, this application provides a computer device for constructing the above-described model, the computer device including a memory and a processor, the memory storing a program, and the processor executing the program to implement the above-described construction method.

[0037] Fourthly, this application provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program;

[0038] The computer program controls the computer-readable storage medium to implement the above-described construction method during runtime.

[0039] Fifthly, this application provides the application of the above-mentioned model and construction method in the preparation of neurosyphilis detection and prediction products.

[0040] Sixthly, this application provides a system for the detection and prediction of neurosyphilis, comprising:

[0041] The data acquisition module is used to collect patients' clinical diagnostic variable data;

[0042] The data processing module is used to randomly divide patients into training and test sets according to a certain proportion. It performs univariate analysis and LASSO regression screening on the clinical diagnostic variable data of the training set to identify key variables that are significantly related to neurosyphilis. It uses a multivariate logistic regression model to model the selected key variables, calculates the regression coefficient of each variable, generates a risk prediction model, and validates it in the test set.

[0043] The risk prediction module is used to stratify patients based on their total score and output the results.

[0044] Patients with a total score less than 42 are considered low-risk; patients with a total score greater than 46 are considered high-risk; and patients with a total score between 42 and 46 are considered medium-risk.

[0045] Seventhly, this application provides a clinical application method for the above-described model and construction method, including the following steps:

[0046] a) Collect patients’ clinical data, including cerebral ischemia or infarction, serum anti-TP, TRUST titer, ataxia and visual impairment;

[0047] b) Calculate the total risk score based on patient variable values ​​using a nomogram;

[0048] c) Predict the probability of neurosyphilis in patients based on the total score and perform risk stratification;

[0049] d) Provide further cerebrospinal fluid examination and treatment recommendations for patients at medium to high risk.

[0050] This invention employs advanced machine learning algorithms and statistical methods to construct a neurosyphilis risk prediction model, specifically including the following steps:

[0051] a) Variable selection. Univariate logistic regression analysis was used to identify variables significantly associated with neurosyphilis. To avoid model overfitting and redundancy, LASSO regression was further used to screen key variables, ultimately determining five core variables: cerebral ischemic infarction, anti-TP, TRUST titer, ataxia, and visual impairment.

[0052] b) Multifactor Logistic Regression Modeling. Five key variables selected are introduced into a multifactor logistic regression model. By calculating the regression coefficients, the contribution of each variable to the risk of neurosyphilis is quantified, forming a complete mathematical model. Based on the logistic regression model, an intuitive noctilinear graph tool is generated.

[0053] c) Nonoplot construction. The nonoplot quickly calculates the patient's total risk score by summing the risk weight scores of each variable, and predicts the probability of developing neurosyphilis.

[0054] d) Model performance validation. The model's discriminative ability is evaluated using ROC curve analysis, with the AUC value reflecting the accuracy of the model's predictions. Calibration curves are used to verify the consistency between the model's predicted values ​​and actual observed values. Clinical decision curves (DCA) are used to evaluate the model's net benefit at different thresholds, thereby ensuring the model's practicality and reliability.

[0055] e) Clinical Applications. The model of this invention can achieve risk assessment of neurosyphilis without invasive procedures. Low-risk patients can avoid unnecessary cerebrospinal fluid examinations, while high-risk patients can obtain further diagnosis and timely intervention.

[0056] Due to the adoption of the above technical solutions, the beneficial effects of this application are as follows:

[0057] This invention integrates multi-dimensional information from imaging, serology, and neurological symptomatology, employs machine learning algorithms to optimize variable selection, and constructs a non-invasive, accurate, and efficient neurosyphilis risk prediction model. The model is visualized through nomograms, providing clinicians with an easy-to-use and highly reliable diagnostic support tool. Attached Figure Description

[0058] Figure 1 is a flowchart illustrating the recruitment process for patients with neurosyphilis and those without, detailing the specific steps involved in patient enrollment, as well as the inclusion and exclusion criteria. This flowchart clearly explains the case selection and data processing procedures, forming the basis for model development and analysis.

[0059] Figure 2 illustrates the LASSO feature selection process, showcasing the detailed steps of variable selection via LASSO regression. Figure 2A: LASSO regression coefficient path diagram. Each curve represents the regression coefficient trajectory of a variable. As the lambda value increases, the regression coefficients of some variables gradually decrease to zero, ultimately identifying the five key variables that contribute most to the model: cerebral ischemia or infarction, ataxia, visual impairment, serum anti-TP, and TRUST titer. Figure 2B: Binomial bias plot, with the vertical axis representing binomial bias and the horizontal axis representing log(λ). The optimal lambda value (lambda.1se) is selected through cross-validation, corresponding to the lowest model complexity and best model performance.

[0060] Figure 3 is an example of a nomogram for a neurosyphilis risk prediction model, showing the weights of each variable and their contribution to the total risk score. The nomogram assigns different weights to each variable, allowing for the calculation of a total risk score and prediction of the probability of developing the disease based on the variable values ​​for a specific patient.

[0061] Figure 4 shows the ROC curves, used to evaluate the discriminative ability of the predictive model, displaying the AUC values ​​for the training and validation sets. Figure 4A: ROC curve for the training set. The model's AUC on the training set is 0.816, indicating that the model has good discriminative ability in distinguishing between patients with neurosyphilis and those without. Figure 4B: ROC curve for the validation set. The model's AUC on the validation set is 0.843, further validating the model's discriminative ability and demonstrating its stability and reliability.

[0062] Figure 5 shows the calibration curves, evaluating the consistency between the model's predictions and actual observations. Figure 5A: Calibration curves for the training set. The model's predictions are close to the diagonal line with the actual observations, indicating that the model's prediction accuracy is high and its calibration performance is good in the training set. Figure 5B: Calibration curves for the validation set. The calibration curves for the model in the validation set are also close to the ideal line, indicating that the model's predictions have good calibrability in independent datasets, further demonstrating the model's reliability and practicality.

[0063] Figure 6 shows the Clinical Decision Achievement (DCA) curves, illustrating the model's net benefit at different decision thresholds. Figure 6A: Decision curve for the training set. It shows that the model delivers high net benefits across most risk thresholds, outperforming strategies such as "all patients receive intervention" or "no intervention for all patients." Figure 6B: Decision curve for the validation set. Consistent with the training set results, the model also exhibits high net benefits in the validation set, indicating good clinical applicability and the ability to provide effective decision support in actual diagnosis and treatment. Detailed Implementation

[0064] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that the following description is merely a preferred embodiment of the present invention, used to better understand and implement the technical solution of the present invention. For those skilled in the art, various improvements and additions can be made without departing from the core method of the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention.

[0065] Example: Constructing a clinical diagnostic model for neurosyphilis based on a retrospective study.

[0066] 1. Data source and queue information.

[0067] This study was based on syphilis patients admitted to the Shanghai Public Health Clinical Center from January 2021 to December 2023. A total of 702 patients with positive serum anti-TP and TRUST results were selected from 1078 suspected syphilis cases (see Figure 1 for details). Inclusion criteria included: age ≥18 years, positive serum Treponema pallidum-specific antibody (anti-TP), and complete clinical and imaging data, including TRUST titer, cranial MRI, and neurological examination results. Exclusion criteria included patients with other neurological diseases (such as encephalitis or brain tumors) or those lacking key diagnostic data. All patient data were desensitized before enrollment and approved by the ethics committee.

[0068] The baseline characteristics of the patient cohort were as follows: 586 males (83.5%) and 116 females (16.5%); median age was 50 years (IQR: 34-62). In terms of age distribution, the largest group of patients was over 60 years old (199 cases, 28.3%), followed by 147 patients aged 51-60 (20.9%), 101 patients aged 41-50 (14.4%), 146 patients aged 31-40 (20.8%), and 109 patients under 30 (15.5%). Based on serum TRUST titers, 397 patients (56.6%) had low titers (≤1:16), 167 patients (23.8%) had medium titers (1:32 and 1:64), and 136 patients (19.4%) had high titers (≥1:128). Of the neurological findings, 531 patients (75.6%) were normal, while 171 patients (24.4%) had neurological abnormalities. Specifically, 67 patients (9.5%) presented with ataxia, including speech disorders, dysarthria, unsteady gait, limb weakness, disorientation, dizziness, and headache; 48 patients (6.8%) presented with mental and behavioral abnormalities, including delirium, depression, dementia, and apathy; 31 patients (4.4%) presented with decreased vision; 22 patients (3.1%) presented with impaired memory; and 3 patients (0.4%) presented with decreased hearing. Imaging studies showed that the most common abnormality on cranial MRI was cerebral ischemic infarction (120 cases, 17.1%), with other abnormalities including leukoencephalopathy (13 cases, 1.9%), cerebral atrophy (11 cases, 1.6%), and demyelinating lesions (2 cases, 0.3%) (see Table 1 for details).

[0069] Table 1. Baseline characteristics of the 702 syphilis patients included in the study.

[0070] Based on cerebrospinal fluid (CSF) examination results, patients were divided into a neurosyphilis group and a non-neurosyphilis group. The neurosyphilis group (NS) comprised 246 patients who exhibited typical neurosyphilis characteristics in their CSF examination, including elevated white blood cell count, abnormal protein levels, and a positive CSF TRUST result. The non-neurosyphilis group (NNS) consisted of 456 patients with normal CSF examination results. Comparison of clinical and laboratory characteristics between the two groups revealed no statistically significant differences in age, sex, or HIV infection status between the non-neurosyphilis and neurosyphilis groups. However, significant differences were found between the two groups in the following indicators: cerebral ischemic infarction (P<0.0001), memory loss (P=0.0221), visual impairment (P<0.0001), ataxia (P<0.0001), abnormal mental behavior (P=0.0071), serum specific antibody (anti-TP, P<0.0001), serum TRUST titer (P<0.0001), whole blood red blood cell count (P=0.0002), platelet count (P=0.0222), and prothrombin time (PT, P=0.0474) (see Table 2 for details).

[0071] Table 2. Baseline characteristics and laboratory parameters of the non-neurosyphilis group and the neurosyphilis group.

[0072] 2. Variable selection.

[0073] After data collection, the 702 patients were randomly divided into a training set and a validation set at an 8:2 ratio. The training set contained 577 patients for model development, while the validation set contained 125 patients to evaluate model performance. Before modeling, 13 candidate variables were identified from the patient data, including age, sex, TRUST titer, anti-TP, cerebral ischemic infarction (MRI), ataxia, visual impairment, blood pressure, blood glucose, hearing loss, syncope, headache, and history of stroke. Univariate analysis and LASSO regression were used to screen the candidate variables, ultimately identifying five key variables significantly associated with neurosyphilis: cerebral ischemic infarction, anti-TP, TRUST titer, ataxia, and visual impairment (see Table 3). The screening process and results for these variables are shown in Figure 2.

[0074] Table 3. Logistic Regression Analysis of Clinical Diagnostic Models

[0075] 3. Model building.

[0076] After variable screening, a multivariate logistic regression model was used to model the selected variables, calculate the regression coefficient of each variable, and generate a risk prediction formula. The core of the model is to quantify the contribution weight of each variable to the risk of neurosyphilis, sum the scores of each variable to generate a total score for the patient, and further calculate the probability of developing neurosyphilis based on the total score. For ease of use by clinicians, the model is visualized using a nomogram, which clearly shows the weight of each variable and its contribution to the total risk score. A specific example of a nomogram is shown in Figure 3. Clinicians can determine the corresponding variable scores in the nomogram one by one based on the patient's specific data, sum the scores of each variable to obtain the total score, and predict the patient's risk level based on the total score.

[0077] 4. Model validation.

[0078] The model's performance was comprehensively evaluated using multiple metrics. First, the model's discriminative ability was assessed using receiver operating characteristic (ROC) curves, with the area under the curve (AUC) calculated for both the training and validation sets. The AUC for the training set was 0.816, and the AUC for the validation set was 0.843 (see Table 4), indicating that the model has high discriminative ability. The ROC curves are shown in Figure 4. Second, calibration curve analysis was used to evaluate the consistency between the model's predicted values ​​and actual observed values. The results showed that the calibration curves for both the training and validation sets were close to the ideal line, indicating that the model has good predictive stability. The calibration curves are shown in Figure 5. Furthermore, clinical decision curve (DCA) analysis was used to evaluate the model's net benefit at different risk thresholds. The results showed that within the medium-to-high risk threshold range (0.2 to 0.6), the model's net benefit was significantly higher than the "full intervention" or "no intervention" strategies, validating the model's clinical applicability. See Figure 6 for details.

[0079] Table 4 shows the ROC curve analysis results of the model on the training and validation sets.

[0080] 5. Specific application scenarios.

[0081] In practical applications, this invention proposes a risk score-based stratification scheme to guide clinical decision-making. Based on the model scores, patients are categorized into three risk levels. Patients with a total score less than 42 are assessed as low-risk and recommended for continued follow-up observation, without immediate cerebrospinal fluid examination. Patients with a total score greater than 46 are assessed as high-risk and recommended for timely cerebrospinal fluid examination for diagnosis and prompt intervention. Patients with a total score between 42 and 46 are considered intermediate-risk, and their risk level cannot be definitively determined, requiring comprehensive evaluation based on other clinical factors or supplementary examinations (see Table 5 for details). This stratification scheme effectively optimizes the allocation of clinical resources, reduces unnecessary invasive examinations, and ensures that high-risk patients receive timely diagnosis and treatment.

[0082] Table 5. Stratification of Neurosyphilis Risk Scores and Their Clinical Significance

[0083] This specific embodiment is merely an explanation of this application and is not intended to limit it. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they fall within the scope of the claims of this application.

Claims

1. A neurosyphilis detection and prediction model based on clinical diagnostic variables, characterized in that, The model is constructed based on the following five key clinical variables: cerebral ischemia or infarction, serum syphilis-specific antibody (anti-TP), TRUST titer, ataxia, and visual impairment. Through numerical analysis of the above variables, LASSO regression was used to screen variables and multivariate logistic regression analysis was performed to construct a risk prediction model in the form of a nomogram to calculate the probability of neurosyphilis risk in patients.

2. The neurosyphilis detection and prediction model according to claim 1, characterized in that, The nomogram assigns a weight to each variable based on the regression coefficients of the variables, generates a total score by summing the scores of each variable, and predicts the probability of neurosyphilis in patients based on the total score.

3. The neurosyphilis detection and prediction model according to claim 2, characterized in that, The model stratifies patients by risk based on their total score, including: Patients with a total score of less than 42 are considered low-risk; Patients with a total score greater than 46 are considered high-risk; Patients with a total score between 42 and 46 are considered to be at medium risk.

4. The method for constructing the model according to any one of claims 1-3, characterized in that, Includes the following steps: Step 1: Collect syphilis patients, screen out patients who are positive for both anti-TP and TRUST serum to include in the study, and collect clinical diagnostic variable data; Step 2: Randomly divide patients into training and testing sets according to a certain proportion. The training set is used for model development, and the validation set is used to evaluate the model's discriminative ability, predictive consistency, and clinical applicability. Step 3: Perform univariate analysis and LASSO regression screening on the clinical diagnostic variable data in the training set to identify key variables that are significantly associated with neurosyphilis; Step 4: Use a multi-factor logistic regression model to model the selected key variables, calculate the regression coefficient of each variable, and generate a risk prediction model; Step 5: Model Validation.

5. The construction method according to claim 4, characterized in that, It has at least one of the following characteristics: In step 1, clinical diagnostic variable data include age, sex, TRUST titer, anti-TP, cerebral ischemic infarction (MRI), ataxia, visual impairment, blood pressure, blood sugar, hearing impairment, syncope, headache, and history of stroke. In step 2, patients are randomly divided into a training set and a test set at a ratio of 8:2; In step 3, the key variables identified included cerebral ischemic infarction, anti-TP, TRUST titer, ataxia, and visual impairment. Step 5 involves the following steps: First, the receiver operating characteristic (ROC) curve is used to assess the model's discriminative ability, and the area under the curve (AUC) of the training and validation sets is calculated separately. Second, the consistency between the model's predicted values ​​and actual observed values ​​is assessed through calibration curve analysis. Third, the net benefit of the model at different risk thresholds is assessed through clinical decision curve (DCA) analysis. The performance of the model was evaluated using ROC curves, with an AUC of 0.816 for the training set and 0.843 for the validation set.

6. A computer device for constructing the model according to any one of claims 1-3, characterized in that, The computer device includes a memory and a processor, the memory storing a program, and the processor executing the program to implement the construction method of claim 4 or 5.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored computer program; The computer program controls the computer-readable storage medium to implement the construction method of claim 4 or 5 during runtime.

8. The application of the model according to any one of claims 1-3 and the construction method according to claim 4 or 5 in the preparation of neurosyphilis detection and prediction products.

9. A system for detecting and predicting neurosyphilis, characterized in that, include: The data acquisition module is used to collect patients' clinical diagnostic variable data; The data processing module is used to randomly divide patients into training and test sets according to a certain proportion. It performs univariate analysis and LASSO regression screening on the clinical diagnostic variable data of the training set to identify key variables that are significantly related to neurosyphilis. It uses a multivariate logistic regression model to model the selected key variables, calculates the regression coefficient of each variable, generates a risk prediction model, and validates it in the test set. The risk prediction module is used to stratify patients based on their total score and output the results. Patients with a total score of less than 42 are considered low-risk; patients with a total score of more than 46 are considered high-risk. Patients with a total score between 42 and 46 are considered to be at medium risk.

10. A clinical application method of the model according to any one of claims 1-3 and the construction method according to claim 4 or 5, characterized in that, Includes the following steps: a) Collect patients’ clinical data, including cerebral ischemia or infarction, serum anti-TP, TRUST titer, ataxia and visual impairment; b) Calculate the total risk score based on patient variable values ​​using a nomogram; c) Predict the probability of neurosyphilis in patients based on the total score and perform risk stratification; d) Provide further cerebrospinal fluid examination and treatment recommendations for patients at medium to high risk.

Images