Method and system for predicting preeclampsia risk of pregnant woman based on computer implementation

Abstract

The invention discloses a method and a system for predicting preeclampsia risk of a pregnant woman based on computer implementation. The method integrates the concentration level of vitamin D (especially 25-hydroxyvitamin D) in serum of pregnant women, angiogenesis biomarkers (such as sFlt-1 / PlGF ratio), conventional clinical indexes (such as mean arterial pressure and body mass index) and personal medical history data. A machine learning algorithm (such as a random forest or a gradient lifting model) is adopted to carry out weighted analysis on the multi-dimensional data, and a dynamic and high-precision prediction model is constructed. The system comprises a data acquisition module, an intelligent prediction module and a clinical decision support module, and can provide real-time risk assessment and personalized intervention suggestions (such as a vitamin D supplement scheme) for medical staff, so that early warning and effective management of preeclampsia are realized, and the risk of pregnancy outcome caused by maternal and infant poor pregnancy is reduced.

Description

Technical Field

[0001] This invention relates to the field of medical diagnostic technology, particularly predictive medicine related to gestational hypertension. More specifically, it relates to a method and system for predicting the risk of preeclampsia (PE) using biomarkers, clinical data, and artificial intelligence algorithms. Background Technology

[0002] Preeclampsia is a serious complication specific to pregnancy, usually occurring after 20 weeks of gestation. It is characterized by new-onset hypertension and proteinuria, and can progress to multiple organ dysfunction. It is one of the leading causes of maternal and fetal morbidity and mortality worldwide. Especially in regions with limited medical resources, the diagnosis and management of preeclampsia face significant challenges due to low patient health literacy, a shortage of professional medical personnel, and weak medical infrastructure.

[0003] Currently, the prediction and diagnosis of preeclampsia rely on clinical symptoms, blood pressure monitoring, and several biomarkers. Angiogenic factors, such as soluble fms-like tyrosine kinase-1 (sFlt-1) and placental growth factor (PlGF), have been shown to play important roles in the pathophysiology of preeclampsia. An elevated sFlt-1 / PlGF ratio is a hallmark of preeclampsia, and a ratio ≤38 is used to rule out preeclampsia in the short term. Some predictive algorithms, such as the Fetal Medicine Foundation (FMF) algorithm, have attempted to combine maternal characteristics, medical history, biophysical indicators (such as mean arterial pressure and uterine artery pulsatility index), and biomarkers (such as PlGF) for early risk screening. However, the sensitivity and specificity of these methods still have room for improvement, especially given their inconsistent performance across different populations and stages of pregnancy. Figure 1 As shown, the mechanism of action of the sFlt-1 / PlGF ratio in normal pregnancy and preeclampsia is that in the preeclamptic state, high levels of sFlt-1 neutralize circulating PlGF and VEGF, leading to endothelial dysfunction.

[0004] In recent years, an increasing number of studies have focused on vitamin D. Vitamin D is not only crucial for calcium and phosphorus metabolism, but also influences the normal development and function of the placenta through immune regulation, angiogenesis, and antioxidant pathways. Numerous studies have confirmed that maternal vitamin D deficiency during pregnancy is significantly associated with an increased risk of preeclampsia, while vitamin D supplementation can significantly reduce the risk of preeclampsia. One meta-analysis even indicated that for every 50 nmol / L decrease in serum 25-hydroxyvitamin D [25(OH)D] concentration, the risk of preeclampsia doubles. Figure 2As shown in the summary of clinical and epidemiological studies on vitamin D status and preeclampsia risk, multiple meta-analyses and case-control studies have shown that low vitamin D levels are associated with an increased risk of preeclampsia.

[0005] Although the association between vitamin D and preeclampsia has been widely recognized, there is currently a lack of a standardized, high-precision model and system that uses advanced algorithms to comprehensively predict vitamin D levels as a core indicator, combined with other key biomarkers and clinical data. The existing diagnostic process often considers each risk factor in isolation, failing to fully exploit the synergistic predictive value of multidimensional data. In addition, the potential of artificial intelligence and machine learning technology in precision medicine has not been fully applied and integrated in preeclampsia prediction. Therefore, the development of an innovative, multi-dimensional information integrated prediction tool has important clinical value and market prospects. SUMMARY

[0006] The purpose of the present application is to overcome the shortcomings of the prior art and provide a more accurate and earlier preeclampsia risk prediction method and supporting system. The present application aims to solve the following technical problems: How to use vitamin D levels as a core biomarker and effectively integrate other angiogenic factors, clinical indicators, and medical history information to improve prediction accuracy; How to use machine learning and other advanced algorithms to identify key risk patterns from complex multidimensional data and build a dynamic and personalized prediction model; How to translate the prediction results into intuitive and actionable clinical decision support to assist medical personnel in early intervention and pregnancy management.

[0007] To achieve the above-mentioned purpose, the present application provides the following technical solutions: A computer-implemented method for predicting the risk of preeclampsia in pregnant women, characterized by the following steps: a) Acquire multidimensional data of the pregnant woman via a data input interface, the multidimensional data including at least serum 25-hydroxyvitamin D concentration, at least one angiogenic biomarker, at least one clinical physiological indicator, and personal health history data; b) Input the multidimensional data into a pre-trained machine learning prediction model, and perform weighted calculation on the multidimensional data by the processing core of the model to generate a probability value representing the risk of preeclampsia in the pregnant woman; c) Based on the probability value, use the risk matrix method to stratify the risk of the pregnant woman, and output a prediction report containing risk level and personalized intervention suggestions through a user interface.

[0008] In preferred embodiments, the angiogenic biomarker is the ratio of soluble fms-like tyrosine kinase-1 to placental growth factor (sFlt-1 / PlGF).

[0009] In preferred embodiments, the clinical physiological indicators include mean arterial pressure (MAP) and / or body mass index (BMI).

[0010] In preferred embodiments, the method for measuring the concentration of 25-hydroxyvitamin D comprises the following steps: a) fasting blood sampling; b) separating serum by centrifugation at 3000 rpm for 10 minutes; c) detecting the concentration of 25-hydroxyvitamin D in serum by chemiluminescence immunoassay.

[0011] In preferred embodiments, the machine learning prediction model is selected from a combination of one or more of a random forest model, a gradient boosting model, a support vector machine model, or a deep neural network model.

[0012] In preferred embodiments, the personalized intervention recommendation includes a vitamin D supplementation regimen for pregnant women with vitamin D levels below a pre-set threshold.

[0013] The present application further provides a computer-implemented system for predicting the risk of preeclampsia in pregnant women, characterized by comprising: a) a data acquisition and interface module configured to obtain multi-dimensional data of pregnant women, the multi-dimensional data comprising at least serum 25-hydroxyvitamin D concentration, angiogenic biomarkers, clinical physiological indicators, and personal health history; b) an intelligent prediction engine module internally deploying a machine learning prediction model, configured to receive the multi-dimensional data and calculate a probability value representing the risk of preeclampsia; c) a clinical decision support module configured to generate a prediction report containing risk level and personalized intervention recommendation based on the probability value, and display it through a user interaction interface.

[0014] In preferred embodiments, the user interaction interface is integrated into a telemedicine platform to support remote pregnancy health management.

[0015] In preferred embodiments, the clinical decision support module adopts a user-centered design principle and provides visual risk attribution analysis.

[0016] Compared with the prior art, the present application has the following advantages: By integrating the key and intervenable risk factor of vitamin D and using the powerful fitting ability of machine learning algorithm, the accuracy and early prediction of preeclampsia are significantly improved. The clinical decision support function provided by the system can convert the prediction results into actual clinical actions, realize the closed-loop management from "prediction" to "prevention", and has great clinical application value. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 : Mechanism diagram of the role of sFlt-1 / PlGF ratio in normal pregnancy and preeclampsia; Figure 2 : Summary of clinical and epidemiological studies on vitamin D status and preeclampsia risk; Figure 3 : Method flowchart; Figure 4 : Immune regulation mechanism of vitamin D deficiency in preeclampsia; Figure 5 : Role of vitamin D deficiency in regulating placental angiogenesis. DETAILED DESCRIPTION

[0018] As shown in Figure 3 , the computer-implemented method for predicting the risk of preeclampsia in pregnant women specifically includes the following steps: a) Data collection: Obtain multi-dimensional data of pregnant women in early pregnancy (such as 11-16 weeks) or mid-pregnancy, including at least: (1) Core biomarker: serum 25-hydroxyvitamin D [25(OH)D] concentration; The measurement method of 25-hydroxyvitamin D concentration includes the following steps: fasting blood collection; centrifugation at 3000 rpm for 10 minutes to separate serum; detecting the concentration of 25-hydroxyvitamin D in serum by chemiluminescence immunoassay analyzer; By the above method, the accuracy of 25-hydroxyvitamin D concentration can be greatly improved, thereby improving the accuracy of preeclampsia risk prediction; (2) Auxiliary biomarker: angiogenic factor, especially sFlt-1 / PlGF ratio; (3) Clinical physiological indicators: body mass index (BMI), mean arterial pressure (MAP), etc. (4) Personal health history: history of preeclampsia, chronic hypertension, autoimmune disease, polycystic ovary syndrome, etc. b) Model prediction: input the collected multi-dimensional data into a pre-trained machine learning prediction model. The model uses algorithms such as Random Forest, Support Vector Machine (SVM), or Gradient Boosting to non-linearly fit the weights of each risk factor and calculate the risk probability value of the pregnant woman developing preeclampsia; c) Risk stratification and output: based on the calculated risk probability value, use the risk matrix method to divide the pregnant women into low-risk, medium-risk, or high-risk levels, and output visual risk reports and interpretations to medical staff or the pregnant woman herself through the user interface; d) Intervention recommendations: based on the risk stratification results, the system automatically generates personalized clinical management recommendations, such as recommending safe doses of vitamin D supplementation for pregnant women with high risk and low vitamin D levels, and recommending increasing the frequency of prenatal examinations and blood pressure monitoring for high-risk pregnant women.

[0019] The construction process of the above prediction model is as follows: First, collect a large-scale retrospective data of pregnant women, including patients eventually diagnosed with preeclampsia and normal pregnancy controls.

[0020] Second, the collected data strictly follows the requirements of the aforementioned "Data Collection" step, and the missing data is processed as follows: (1) Variable analysis and interpolation framework in the model construction stage: a) Interpolation variable selection: all prediction variables and outcome variables are included, with a focus on the following clinical indicators: Continuous variables: serum 25(OH)D concentration (significantly affected by season), sFlt-1 / PlGF ratio, BMI, MAP; additionally, for the sFlt-1 / PlGF ratio, check for extreme outliers, and if > 85%, perform log transformation; Binary variables: history of preeclampsia, chronic hypertension, autoimmune disease, polycystic ovary syndrome; b) Parameter setting: Number of iterations (maxit): default 10-20 times, for variables with complex correlation such as sFlt-1 / PlGF ratio and MAP, the number of iterations is 30 times; Interpolation model: logistic regression for categorical variables, and prediction mean matching (PMM) or linear regression for continuous variables (such as 25(OH)D, which needs to consider nonlinear trends); (2) Algorithm selection and clinical adaptation in the data interpolation stage: a) Number of interpolations (m): Base setting: m = 20 (satisfies general missing rate ≤ 30% data); High missing rate adjustment: if the missing rate of a key variable (such as 25(OH)D or sFlt-1 / PlGF) is > 40%, it needs to be increased to m = 30-50 times to ensure stability; b) Convergence algorithm: Metropolis-Hasting algorithm is used to drive Markov chain, which is particularly suitable for skewed distribution of biochemical indicators (such as sFlt-1 / PlGF ratio) and seasonally sensitive 25(OH)D data (high missing rate in winter); Convergence diagnosis: verified by trajectory diagram and Gelman-Rubin statistic (\hat{R} < 1.1); (3) Results integration stage, merging strategy and clinical significance verification: a) Step-by-step analysis: The specific code is as follows: markdown for i in 1 to m: Fit regression model (such as logistic / Cox model) in the imputed dataset D_i; Extract regression coefficients β_i (such as the effect of 25(OH)D on preeclampsia risk), standard error SE_i; b) Rubin's Rules: Merge coefficients: (\bar{\beta}=\frac{1}{m}\sum{\beta_i}) Merge variance: (T=\bar{U}+(1+\frac{1}{m})B), where (\bar{U}) is the mean of within-group variance, (B) is the between-group variance; c) Statistical test: Use Wald test to calculate the 95% confidence interval and P value of the merged β (such as to verify the significance of MAP or 25(OH)D); For categorical variables (such as autoimmune diseases), use likelihood ratio test to avoid small sample bias.

[0021] Third, the dataset is randomly divided into training set and validation set in the ratio of 7:3. In the training set, various machine learning algorithms are used for model training, such as decision tree, random forest, support vector machine and logistic regression. Through cross-validation, the model with the best comprehensive performance in terms of area under the receiver operating characteristic curve (AUROC), accuracy, recall rate, etc. is selected, such as the random forest model, as the final prediction model.

[0022] The core scientific basis of this model is that vitamin D deficiency can promote the occurrence of preeclampsia through multiple mechanisms. For example, vitamin D deficiency can disrupt immune tolerance at the maternal-fetal interface, leading to a predominance of pro-inflammatory Th1 cells and suppression of immunosuppressive regulatory T cells (Tregs), thereby exacerbating placental inflammation. In addition, vitamin D deficiency can also lead to poor remodeling of uterine spiral arteries and increased vascular constriction by suppressing the expression of vascular endothelial growth factor (VEGF) and activating the renin-angiotensin system (RAAS), ultimately triggering placental ischemia. As shown in Figure 4 , the immune regulation mechanism of vitamin D deficiency in preeclampsia, vitamin D deficiency leads to a decrease in regulatory T cells, Th1 / Th2 imbalance, increased secretion of pro-inflammatory cytokines, and disruption of maternal-fetal immune tolerance. As shown in Figure 5 , the role of vitamin D deficiency in regulating placental angiogenesis, vitamin D deficiency suppresses angiogenesis and promotes vasoconstriction by down-regulating VEGF and up-regulating the renin-angiotensin system, respectively.

[0023] The present embodiment also provides a system (preeclampsia dynamic prediction system) for predicting the risk of preeclampsia in pregnant women, which implements the above method, comprising: a data acquisition and interface module for receiving and integrating various data from hospital information systems (HIS), laboratory information systems (LIS), and user manual input, and being configured to obtain multi-dimensional data of pregnant women, including at least serum 25-hydroxyvitamin D [25(OH)D] concentration, sFlt-1 / PlGF ratio, body mass index (BMI), mean arterial pressure (MAP), history of preeclampsia, chronic hypertension, autoimmune disease, polycystic ovary syndrome, etc.; an intelligent prediction engine module embedded with the above machine learning prediction model, responsible for performing risk calculation; a clinical decision support system (CDSS) module designed with the user in mind, which converts complex prediction results into clear charts, risk levels, and specific action recommendations, providing support for clinical decision-making; a user interaction interface module providing different access ports and interfaces for doctors and pregnant women, which can be integrated into desktop software, mobile applications, or telemedicine platforms.

[0024] Specifically, the system can be deployed on a server of a hospital, or provided as a cloud service. The data acquisition and interface module interfaces with the existing information system of the hospital through a standard interface (such as HL7), and automatically acquires the electronic medical record and test results of the patient. For data that cannot be automatically acquired, a simple form is provided for medical staff to manually enter. The intelligent prediction engine module is deployed on the backend server, and after receiving a front-end request, it calls the trained model for calculation and returns the result within a few seconds. The clinical decision support system (CDSS) module is the core interactive part of the system. Its interface shows the doctor: 1) the final risk score and risk level (such as represented by red, yellow, and green); 2) the contribution analysis of each input indicator in the model, helping the doctor understand the source of the prediction result; 3) personalized recommendations based on guidelines and the latest research, for example, when the system detects that the patient's vitamin D level is lower than 20 ng / mL and the risk score is high, the system will pop up a prompt: "The patient has severe vitamin D deficiency, which is a high-risk factor for preeclampsia. Consider supplementing 1000-2000 IU of vitamin D daily, and recheck in 4 weeks." In addition, in order to facilitate the implementation of the method of the present application, the present application can also provide a matching diagnostic kit. The kit is intended for performing the detection method for predicting preeclampsia, and can include the following inside: reagents for quantitatively detecting serum 25-hydroxyvitamin D; reagents for quantitatively detecting serum PlGF and / or sFlt-1; standard, quality control and operating instructions.

[0025] In addition, it should be noted that the specific embodiments described in the specification, the name of each part, etc. can be different, and any equivalent or simple change made according to the structure, features and principles of the patent concept of the present application is included in the protection scope of the present application. Those skilled in the art can make various modifications or supplements to the described specific embodiments or use similar ways to replace them, as long as they do not deviate from the structure of the present application or exceed the scope defined by the present claims, and should belong to the protection scope of the present application.

Claims

1. A computer-implemented based method for predicting the risk of a pregnant woman developing preeclampsia, characterized in that The method comprises the following steps: a) acquiring multi-dimensional data of the pregnant woman via a data input interface, the multi-dimensional data comprising at least serum 25-hydroxy vitamin D concentration, at least one angiogenesis biomarker, at least one clinical physiological indicator, and personal health history data; b) inputting the multi-dimensional data into a pre-trained machine learning prediction model, performing weighted calculation on the multi-dimensional data by a processing core of the model, and generating a probability value representing the risk of preeclampsia of the pregnant woman; c) based on the probability value, stratifying the risk of the pregnant woman by using a risk matrix method, and outputting a prediction report comprising risk level and personalized intervention suggestion through a user interface.

2. The method of claim 1, wherein: The angiogenesis biomarker is the ratio of soluble fms-like tyrosine kinase-1 to placental growth factor.

3. The method of claim 1, wherein: The clinical physiological indicator comprises mean arterial pressure and / or body mass index.

4. The method of claim 1, wherein: The method for measuring the concentration of 25-hydroxy vitamin D comprises the following steps: a) fasting blood sampling; b) separating serum by centrifugation at 3000 rpm for 10 minutes; c) detecting the concentration of 25-hydroxy vitamin D in serum by chemiluminescence immunoassay.

5. The method of claim 1, wherein: The machine learning prediction model is selected from one or more combinations of random forest model, gradient boosting model, support vector machine model, or deep neural network model.

6. The method of claim 1, wherein: The personalized intervention suggestion comprises a vitamin D supplementation scheme for pregnant women with vitamin D level below a preset threshold.

7. A computer-implemented based system for predicting the risk of a pregnant woman developing preeclampsia, characterized in that The method comprises: a) a data acquisition and interface module configured to acquire multi-dimensional data of the pregnant woman, the multi-dimensional data comprising at least serum 25-hydroxy vitamin D concentration, angiogenesis biomarker, clinical physiological indicator, and personal health history; b) an intelligent prediction engine module internally deploying a machine learning prediction model, configured to receive the multi-dimensional data and calculate a probability value representing the risk of preeclampsia; c) a clinical decision support module configured to generate a prediction report comprising risk level and personalized intervention suggestion according to the probability value, and display through a user interaction interface.

8. The system of claim 7, wherein: The user interaction interface is integrated into a remote medical platform to support remote pregnancy health management.

9. The system of claim 7, wherein: The clinical decision support module adopts a user-centered design principle to provide visual risk attribution analysis.

Images