SVM gestational diabetes risk prediction model under mixed kernel function

By using an SVM model with hybrid kernel functions to predict the risk of GDM using data from pregnant women in early pregnancy, this method solves the problems of late detection window and uncertain diagnostic criteria in existing technologies, enabling early and accurate screening and intervention, reducing the incidence of GDM, and protecting maternal and infant health.

CN118942706APending Publication Date: 2024-11-12TONGJI HOSPITAL ATTACHED TO TONGJI MEDICAL COLLEGE HUAZHONG SCI TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411164448.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-23
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Current technologies for screening gestational diabetes mellitus (GDM) have problems such as a late detection window, difficulty for pregnant women to adhere to testing, and a lack of research support for standardized screening thresholds, which makes it impossible to detect high-risk groups early.

Method used

A hybrid kernel function-based SVM model for predicting gestational diabetes mellitus risk was adopted. The model uses basic information and data from the early pregnancy of pregnant women to predict the risk of gestational diabetes mellitus, including basic information, medical history and physiological test data of pregnant women. The data is then classified using a support vector machine algorithm.

Benefits of technology

Accurately predicting the risk of GDM in early pregnancy can reduce unnecessary oral glucose tolerance tests, provide opportunities for early intervention, reduce the incidence of GDM, and protect maternal and infant health.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118942706A_ABST
    Figure CN118942706A_ABST
Patent Text Reader

Abstract

The invention relates to a risk prediction model of SVM gestational diabetes mellitus under a mixed kernel function. The risk prediction model comprises a pregnant woman data set and a pregnant woman SVM model. Wherein the pregnant woman data set comprises a pregnant woman basic information set and a pregnant woman filing data set; the data in the pregnant woman data set is the data of the pregnant women in the early pregnancy period; the pregnant woman SVM model is used for predicting the GDM risk of the pregnant woman as a GDM high risk or a GDM low risk according to the pregnant woman data set. Invasive physical examination on the pregnant woman is not needed, and the pregnant woman does not need to take various medicines; reducing unnecessary oral glucose tolerance test (OGTT) applications; the time for discovering the GDM is greatly shortened, and a very large time window is provided; and the range of high-risk groups can be determined very early.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of clinical prediction, and in particular to a risk prediction model of gestational diabetes mellitus using an SVM under a mixed kernel function. Background Art

[0002] Gestational diabetes mellitus (GDM) is one of the common complications of pregnancy. In the middle and late stages of pregnancy, pregnant women's sensitivity to insulin decreases. If the compensatory secretion of insulin is insufficient at this time, the disease is prone to occur. Most GDM patients have no obvious clinical manifestations, and the 75g glucose tolerance test is the main diagnostic method. Pregnancy can cause GDM in pregnant women without diabetes, and also make the condition of patients with prediabetes worse. High blood sugar can cause abnormal fetal development or even death. The possibility of gestational hypertension is 2-4 times higher than that of non-diabetic pregnant women. Pregnant women who fail to control blood sugar well are prone to infection, repeated vaginal inflammation, and polyhydramnios. The recurrence rate of GDM pregnant women in the next pregnancy is as high as 33%-69%; the risk of fetal macrosomia, growth restriction (FGR), miscarriage and premature birth, fetal distress and intrauterine fetal death, and fetal malformations increases; the risk of respiratory distress syndrome and hypoglycemia in newborns increases.

[0003] The existing technology for GDM screening uses OGTT technology, namely glucose tolerance test. Specifically, the technical means used are:

[0004] A 75g glucose tolerance test (OGTT) is performed between 24 and 28 weeks of pregnancy. The diagnostic criteria are: fasting blood glucose levels, 1 hour and 2 hours after taking sugar are lower than 5.1mmol / L, 10.0mmol / L and 8.5mmol / L respectively; GDM is diagnosed if any blood glucose value reaches or exceeds the above criteria.

[0005] The advantage of the existing technology is accuracy: the accuracy is undoubtedly reliable based on the real-time collected human body data indicators; however, the defects of the existing technology are:

[0006] 1. Because pregnant women cannot eat before and during the OGTT test, it is difficult for them to adhere to the test;

[0007] 2. Since the window for using OGTT is 24 to 28 weeks of pregnancy, which is already between the second and third trimesters, it is actually too late to provide dietary or medical interventions for pregnant women diagnosed with GDM through screening.

[0008] 3. Since there is no definitive research on the current standardized screening threshold for the diagnosis of GDM and its value and pregnancy outcomes in selected patients, it is unknown whether the actual use of OGTT for the diagnosis of GDM exceeds the necessary level. Summary of the invention

[0009] In view of the above problems, the present invention provides a risk prediction model for gestational diabetes mellitus using an SVM under a mixed kernel function, the purpose of which is to eliminate the need for invasive physical examinations of pregnant women and the need for pregnant women to take various medications; reduce unnecessary use of oral glucose tolerance tests (OGTTs); greatly improve the time for discovering GDM and provide a very large time window; and determine the scope of high-risk groups very early.

[0010] To solve the above problems, the technical solution provided by the present invention is:

[0011] A risk prediction model for gestational diabetes mellitus using an SVM under a mixed kernel function, comprising a pregnant woman data set and a pregnant woman SVM model; wherein:

[0012] The pregnant woman data set includes a pregnant woman basic information set and a pregnant woman file data set; the data in the pregnant woman data set is data of pregnant women in the early pregnancy period; the early pregnancy period is 6 weeks to 13 weeks + 6 days of pregnancy;

[0013] The pregnant woman SVM model is used to predict the GDM risk of the pregnant woman as a high risk of GDM or a low risk of GDM according to the pregnant woman data set.

[0014] Preferably, the basic information set of the pregnant woman includes age, BMI before pregnancy, a first medical history set, and a second medical history set; wherein:

[0015] The first medical history set includes a history of impaired glucose tolerance, a history of PCOS, and a family history of diabetes;

[0016] The second medical history set includes a history of unexplained fetal death, a history of stillbirth, a history of miscarriage, a history of delivery of a macrosomia, a history of fetal malformation, a history of polyhydramnios, and a history of GDM.

[0017] Preferably, the pregnant woman's file data set includes blood routine, urine routine, fasting blood sugar, and thyroid stimulating hormone; wherein:

[0018] The routine blood test includes white blood cell count and neutrophil count;

[0019] The urine routine test includes urine protein and urine bilirubin.

[0020] Preferably, the pregnant woman SVM model is expressed as follows:

[0021]

[0022] Where: f(x) is used to characterize the pregnant woman SVM model; N is the number of pregnant woman samples collected in the pregnant woman data set; x is the pregnant woman data set, expressed as follows:

[0023] x m =(xm1 ,x m2 ,x m3 ,x m4 ,x m5 ,x m6 ,x m7 ,x m8 )

[0024] Wherein, m is a counter of pregnant women in the real space, used to represent the mth pregnant woman in the real space, and i=1,2,…,m; x m1 represents the age of the mth pregnant woman; x m2 represents the pre-pregnancy BMI of the mth pregnant woman; m3 represents the first medical history set of the mth pregnant woman; m4 represents the second medical history set of the mth pregnant woman; m5 represents the blood routine test of the mth pregnant woman in the early pregnancy period; x m6 represents the urine routine test of the mth pregnant woman in the early pregnancy period; x m7 represents the fasting blood sugar of the mth pregnant woman in the early pregnancy period; x m8 represents the thyroid stimulating hormone of the mth pregnant woman in the early pregnancy;

[0025] φ(x) is the feature vector that maps the pregnant woman data set x to the feature space; κ(x,x i ) is the kernel function about x; b is the model parameter; w is the model parameter; i is the counter about pregnant women in the feature space, which is used to characterize the i-th pregnant woman in the feature space; y is used to characterize the GDM risk of pregnant women, and y∈-1,1}, when y m =1, it indicates that the GDM risk of the mth pregnant woman is the high risk of GDM. m = -1, indicating that the GDM risk of the m-th pregnant woman is the low risk of GDM; ... is a set of Lagrange multipliers, expressed as follows:

[0026] α=(α1,α2,…α n ,…,α N )

[0027] Where: α i is the Lagrange multiplier of pregnant woman i, and α i ≥0.

[0028] Preferably, the kernel function κ(x,x i ) is expressed as follows:

[0029]

[0030] Where: j is a counter of pregnant women in the feature space, used to represent the jth pregnant woman in the feature space; ||xi -x j || 2 is the vector x i -x j The corresponding L 2 norm;||x i -x j || is vector x i -x j The corresponding L 1 norm; σ is the bandwidth of the Gaussian kernel, and σ>0.

[0031] Preferably, in the feature space, the corresponding hyperplane that correctly classifies the sample is (w, b), and the relationship is expressed as follows:

[0032]

[0033] Where: w T The transposed matrix representing w; in:

[0034] D is the sample set of pregnant women, expressed as follows:

[0035] D={(x1,y1), (x2,y2),…, (x n ,y n ),…,(x N ,y N )}

[0036] Where: n represents any natural number from 1 to N; (x i ,y i ) is any sample point in D.

[0037] Preferably, in the feature space, the distance between the two closest sample points that do not belong to the same set of pregnant women samples is expressed as follows:

[0038]

[0039] Where: γ is the distance between the two closest sample points that do not belong to the same type of the pregnant sample set, and the larger the γ is, the larger the distance between the two corresponding classes is, and the better the classification effect is; ||w|| is the L corresponding to the vector w 1 Norm.

[0040] Preferably, the optimization equation that maximizes γ is expressed as follows:

[0041]

[0042] αi i ≥0,i=1,2,…,m

[0043] Where: the Lagrange multiplier α i The relationship between the model parameters b and w is expressed as follows:

[0044]

[0045] Where: L(w,b,α) is the Lagrangian function of α, b, and w.

[0046] Preferably, the basic information set of pregnant women includes:

[0047] If the age of the mth pregnant woman is less than 35 years old, then x m1 = -1; if the age of the pregnant woman is ≥ 35 years old, then x m1 =1;

[0048] If the pre-pregnancy BMI of the mth pregnant woman is 18≤BMI≤24, then x m2 = -1, otherwise x m2 =1;

[0049] In the first medical history set, if the mth pregnant woman has any one of the history of impaired glucose tolerance, the history of PCOS, and the family history of diabetes, then x m3 =1, otherwise x m3 = -1;

[0050] In the second medical history set, if the mth pregnant woman has any one of the history of unexplained fetal death, the history of stillbirth, the history of miscarriage, the history of delivery of a macrosomia, the history of fetal malformation, the history of polyhydramnios, and the history of GDM, then x m4 =1, otherwise x m4 =-1.

[0051] Preferably, the pregnant woman profile data set includes:

[0052] If the white blood cell count in the blood routine test of the mth pregnant woman falls within the manually preset normal range of white blood cells, and the neutrophil count falls within the manually preset normal range of neutrophils, then x m5 = -1; otherwise, x m5 =1;

[0053] If the urine protein in the urine routine test of the mth pregnant woman is negative and the urine bilirubin is negative, then x m6 = -1, otherwise, x m6 =1;

[0054] If the fasting blood glucose of the mth pregnant woman in the early pregnancy period falls within the manually preset normal range of fasting blood glucose in the early pregnancy period, then x m7= -1, otherwise x m7 =1;

[0055] If the thyroid stimulating hormone of the mth pregnant woman in the early pregnancy falls within the manually preset normal range of thyroid stimulating hormone in the early pregnancy, then x m8 = -1, otherwise x m8 =1.

[0056] Compared with the prior art, the present invention has the following advantages:

[0057] 1. Since the SVM model used in the present invention is based on the historical data of pregnant women who have experienced GDM, it predicts GDM in a statistical way, so there is no need to conduct invasive physical examinations on pregnant women, nor do they require pregnant women to take various drugs. Instead, the prospect of pregnant women developing GDM can be accurately predicted by only collecting routine basic information and archiving data;

[0058] 2. Since the model of the present invention is based on historical data of pregnant women, unnecessary oral glucose tolerance tests (OGTT) are reduced, saving costs and time;

[0059] 3. Since the present invention can be implemented in the early pregnancy period, there is no need to wait until the 24th to 28th week of pregnancy, thus greatly improving the time for discovering GDM, giving doctors a very large time window and providing doctors with a buffer period for early intervention prevention and treatment;

[0060] 4. Due to the high prediction accuracy of the present invention, the adverse effects that may be caused by preventing unnecessary OGTT in patients who are not in the high-risk group for GDM are greatly reduced;

[0061] 5. Since the present invention can accurately assess the risk of GDM in early pregnancy, it is possible to determine the scope of high-risk groups very early, and then formulate accurate prevention plans to ensure maternal and fetal safety, reduce the incidence of GDM, improve eugenics, and ensure the health of mothers and babies. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of the logical structure of the GMD risk prediction model of a specific embodiment of the present invention. DETAILED DESCRIPTION

[0063] The present invention is further explained below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0064] It should be noted in advance that this specific embodiment uses 236 pieces of real obstetrics and gynecology data managed by the applicant as training sets, and predicts early pregnancy diabetes based on this; among these 236 pregnant women, 54 are at high risk of gestational diabetes (GDM) and 182 are at low risk of GMD. The training set contains a set of basic information of each pregnant woman and a set of pregnant woman file data, which, after packaging, is the pregnant woman data set of the present invention.

[0065] like Figure 1 As shown, a risk prediction model for gestational diabetes mellitus under a mixed kernel function SVM includes a pregnant woman data set and a pregnant woman SVM model; wherein:

[0066] The pregnant women data set includes a pregnant women basic information set and a pregnant women file data set; the data in the pregnant women data set is the data of pregnant women in the early pregnancy period; the early pregnancy period is from 6 weeks to 13 weeks + 6 days of pregnancy.

[0067] In this specific embodiment, the basic information set of the pregnant woman includes age, BMI before pregnancy, a first medical history set, and a second medical history set; wherein:

[0068] The first medical history set includes history of impaired glucose tolerance, history of PCOS, and family history of diabetes.

[0069] The second medical history set included history of unexplained fetal death, stillbirth, miscarriage, delivery of a macrosomia, fetal malformation, polyhydramnios, and GDM.

[0070] In this specific embodiment, the data set for creating a pregnant woman's profile includes blood routine, urine routine, fasting blood sugar, and thyroid stimulating hormone; wherein:

[0071] Routine blood test includes white blood cell count and neutrophil count.

[0072] Urinalysis includes urine protein and urine bilirubin.

[0073] In this specific embodiment, the basic information set of pregnant women specifically includes:

[0074] If the age of the mth pregnant woman is less than 35 years old, then x m1 =-1; if the pregnant woman is ≥35 years old, then x m1 =1.

[0075] If the pre-pregnancy BMI of the mth pregnant woman is 18≤BMI≤24, then x m2 = -1, otherwise x m2 =1.

[0076] In the first medical history set, if the mth pregnant woman has any of the following: history of impaired glucose tolerance, history of PCOS, or family history of diabetes, then x m3 =1, otherwise xm3 =-1.

[0077] In the second medical history set, if the mth pregnant woman has any of the following: history of unexplained fetal death, history of stillbirth, history of miscarriage, history of delivery of a macrosomia, history of fetal malformation, history of polyhydramnios, or history of GDM, then x m4 =1, otherwise x m4 =-1.

[0078] In this specific embodiment, the pregnant woman's file data set specifically includes:

[0079] If the white blood cell count in the blood routine of the mth pregnant woman falls within the manually preset normal range of white blood cells, and the neutrophil count falls within the manually preset normal range of neutrophils, then x m5 = -1; otherwise, x m5 =1.

[0080] In this specific embodiment, the normal range of white blood cells is 3.5*10 9 ~9.5*10 9 / L.

[0081] In this specific embodiment, the normal range of neutrophils is 1.8*10 9~ 6.3*10 9 / L.

[0082] If the urine protein in the urine routine of the mth pregnant woman is negative and the urine bilirubin is negative, then x m6 = -1, otherwise, x m6 =1.

[0083] If the fasting blood glucose of the mth pregnant woman in early pregnancy falls within the manually preset normal range of fasting blood glucose in early pregnancy, then x m7 = -1, otherwise x m7 =1.

[0084] In this specific embodiment, the normal range of fasting blood glucose in early pregnancy is 3.9mmol~5.1mmol / L.

[0085] If the TSH of the mth pregnant woman in the early pregnancy falls within the manually preset normal range of TSH in the early pregnancy, then x m8 = -1, otherwise x m8 =1.

[0086] In this specific embodiment, the normal range of thyroid stimulating hormone in early pregnancy is 0.27uIU to 4.20uIU / ml.

[0087] The pregnant woman SVM model is used to predict the GDM risk of the pregnant woman as a high risk of GDM or a low risk of GDM based on the pregnant woman data set.

[0088] As shown in Table 1, this is the data set of pregnant women in this specific embodiment:

[0089] Table 1. Partial data display table of pregnant women data set

[0090]

[0091]

[0092] It should be noted that the cause of gestational diabetes mellitus (GDM) has not yet been explored in the academic community, but the currently recognized, statistically based view is that inflammation and thyroid dysfunction are potential factors for the development of diabetes. Based on this, the present invention collects the above physiological test data of pregnant women to assess the risk of diagnosing GDM during pregnancy.

[0093] It should be further explained that maternal personal information and physiological test data are represented in the form of vectors as feature vectors corresponding to the support vector machine algorithm.

[0094] Let φ(x) be the feature vector that maps the pregnant women data set x to the feature space, κ(x,x i ) is the kernel function about x, b is the model parameter, w is the model parameter, so the model corresponding to the hyperplane divided in the feature space is expressed according to formula (1):

[0095] f(x)=w T φ(x)+b(1)

[0096] Among them: f(x) is used to characterize the pregnant woman SVM model; N is the number of pregnant women samples collected in the pregnant women data set; x is the pregnant women data set, expressed according to formula (2):

[0097] x m =(x m1 ,x m2 ,x m3 ,x m4 ,x m5 ,x m6 ,x m7 ,x m8 )(2)

[0098] Wherein, m is a counter of pregnant women in the real space, used to represent the mth pregnant woman in the real space, and i=1,2,…,m; x m1 represents the age of the mth pregnant woman; x m2 represents the pre-pregnancy BMI of the mth pregnant woman; x m3 represents the first medical history set of the mth pregnant woman; x m4 represents the second medical history set of the mth pregnant woman; m5 represents the blood routine test of the mth pregnant woman in early pregnancy; xm6 represents the urine routine test of the mth pregnant woman in early pregnancy; x m7 represents the fasting blood sugar of the mth pregnant woman in early pregnancy; x m8 represents the thyroid stimulating hormone of the mth pregnant woman in the early pregnancy.

[0099] It should be noted that support vector machine (SVM) is one of the models and algorithms in machine learning, mainly used to process binary variables; SVM is a class classifier, which is a hyperplane that can separate samples of different classes in the sample space.

[0100] In other words, given some labeled training samples (supervised learning), the SVM algorithm can output an optimal segmentation hyperplane; SVM is developed from the optimal classification surface in the linearly separable case.

[0101] It should be further explained that the optimal classification surface requires that the classification line can not only correctly separate the two categories, that is, the so-called training error rate is 0, but also maximize the classification interval; SVM considers finding a hyperplane that meets the classification requirements and makes the points in the training set as far away from the classification surface as possible, that is, to find a classification surface with the largest blank area (margin) on both sides of it.

[0102] It should be further explained that the present invention adopts a support vector machine method with a mixed kernel function to perform binary classification on samples of whether pregnant women suffer from GDM. Since it is generally difficult to have a hyperplane that can correctly divide the two types of samples in the actual sample space, the samples are mapped from the original space to a high-dimensional feature space so that the samples are linearly separable in the feature space.

[0103] i is a counter of pregnant women in the feature space, which is used to characterize the i-th pregnant woman in the feature space; y is used to characterize the GDM risk of pregnant women, and y∈{-1,1}, when y m =1, it indicates that the GDM risk of the mth pregnant woman is high risk of GDM. m = -1 indicates that the GDM risk of the mth pregnant woman is low risk of GDM; α is the set of Lagrange multipliers, expressed as formula (3):

[0104] α=(α1,α2,…α n ,…,α N ) (3)

[0105] Where: α i is the Lagrange multiplier of pregnant woman i, and α i ≥0.

[0106] In this specific embodiment, in the feature space, the corresponding hyperplane that correctly classifies the sample is (w, b), and the relationship is expressed by formula (4):

[0107]

[0108] Where: w T The transposed matrix representing w; in:

[0109] D is the sample set of pregnant women, expressed as follows:

[0110]

[0111] Where: n represents any natural number from 1 to N; (x i ,y i ) is any sample point in D.

[0112] It should be noted that, considering that the sample points closest to the hyperplane are , the above formula (4) holds true. Then, according to the geometric mapping relationship, in this specific embodiment, in the feature space, the distance between the two sample points that are closest to each other and do not belong to the same set of pregnant women samples is expressed by formula (6):

[0113]

[0114] Among them: γ is the distance between the two closest sample points that do not belong to the same set of pregnant women samples, and the larger the γ, the larger the distance between the two different classes, and the better the classification effect; ||w|| is the L corresponding to the vector w 1 Norm.

[0115] In this way, the optimization equation can be expressed as follows:

[0116]

[0117] Obviously, to maximize γ, we only need to maximize ||w|| -1 , which is equivalent to minimizing ||w|| 2 , so the above formula can be further expressed as formula (8):

[0118]

[0119] On this basis, since formula (8) itself is a convex quadratic optimization problem, the Lagrange multiplier method can be used to obtain the solution of its "dual problem"; applying Lagrange multipliers to each constraint condition in formula (8), the Lagrange function L(w, b, α) of the problem is expressed according to formula (9):

[0120]

[0121] Where: L(w,b,α) is the Lagrangian function of α, b, and w.

[0122] Furthermore, by taking L(w, b, α) and taking w and b as variables to find partial derivatives, we can obtain the following expression according to equations (10) and (11):

[0123]

[0124] Further simplification, the dual problem of equation (8) can be expressed as equation (12):

[0125]

[0126] Solving the above equation involves calculating φ(x i ) T φ(x j ), which is usually difficult to calculate, so φ(x i ) T φ(x j ) The calculation result is considered to be in the form of a kernel function and is expressed as follows:

[0127] φ(x i ) T φ(x j )=κ(x i ,x j ) (13)

[0128] Therefore, in this specific embodiment, the optimization equation for maximizing γ is expressed as follows:

[0129]

[0130] Through simulation verification, the kernel function is represented by a combination of Gaussian kernel and Laplace kernel, and the kernel function coefficients are half each. In this specific embodiment, the kernel function κ(x,x i ) is expressed according to formula (15):

[0131]

[0132] Where: j is a counter of pregnant women in the feature space, used to represent the jth pregnant woman in the feature space; ||x i -x j || 2 is the vector x i -x j The corresponding L 2 norm;||x i -x j || is vector x i -x j The corresponding L 1 norm; σ is the bandwidth of the Gaussian kernel, and σ>0.

[0133] It should be noted that when ym =1, it indicates that the GDM risk of the mth pregnant woman is high risk of GDM. m = -1 indicates that the GDM risk of the mth pregnant woman is low risk of GDM, so the information of the mth pregnant woman can be represented by a vector, denoted as d m =(x m ,y m ).

[0134] Further, formula (14) is a quadratic programming problem, and the SMO (Sequential minimal optimization) algorithm is used to calculate the α corresponding to this optimization problem. i (i=1,2,…,m) and the value of b.

[0135] So, in calculating α i (i=1,2,…,m) values, in this specific embodiment, the pregnant woman SVM model is finally expressed as formula (16):

[0136]

[0137] In order to verify the accuracy of the method of the present invention, the applicant performed OGTT routine on 236 pregnant women selected in this specific embodiment at 24 to 28 weeks of pregnancy, and collected the results of OGTT, which are as follows:

[0138] The applicant collected reports of OGTT tests performed on pregnant women between 22 and 28 weeks of pregnancy; the diagnostic criteria were: blood glucose levels when fasting, 1 hour and 2 hours after taking sugar were lower than 5.1mmol / L, 10.0mmol / L and 8.5mmol / L respectively; if any of the blood glucose levels reached or exceeded the above criteria, the pregnant woman was diagnosed with GDM.

[0139] Then, the applicant compared the OGTT test reports of 236 pregnant women with the results of SVM prediction under the mixed kernel function using the model of the present invention during the early pregnancy period for each of them. The results are shown in Table 2:

[0140] Table 2. Comparison of the prediction results of the model of the present invention and the actual results

[0141] SVM model predicts the number of people OGTT Number of people High risk for GDM 47 Abnormal (patient population) 45 Low risk of GDM 189 Normal (normal people) 191

[0142] It can be clearly seen from Table 2 that among the 236 pregnant women samples, the difference between the prediction using the model of the present invention and the final accurate result is 3 people, that is, the accuracy of the present invention in the application of this experiment is 98.72%.

[0143] Furthermore, the applicant is a large-scale comprehensive tertiary-level Class A hospital in the country. After experiments on multiple groups of samples by the applicant's obstetrics and gynecology department, the accuracy of the present invention is between 98.22% and 99.03%. It already has a de facto application basis and can be used to assist in decision-making on whether a pregnant woman suffers from gestational diabetes (GDM), thereby helping obstetricians to provide clinical advice to pregnant women in the early pregnancy period, start dietary adjustments in time, reduce medical interventions, reduce unnecessary OGTT applications, and reduce the possible adverse effects of OGTT on low-risk GDM patients, thereby reducing the incidence of GDM, improving eugenics, protecting the health of mothers and children, and improving the health quality of the people.

[0144] In the above detailed description, various features are grouped together in a single embodiment to simplify the disclosure. This method of disclosure should not be interpreted as reflecting an intention that the embodiments of the claimed subject matter require more features than are clearly stated in each claim. On the contrary, as reflected in the appended claims, the invention is in a state of having less than all the features of the disclosed individual embodiments. Therefore, the appended claims are hereby expressly incorporated into the detailed description, with each claim standing on its own as a separate preferred embodiment of the invention.

[0145] The disclosed embodiments are described above to enable any person skilled in the art to implement or use the present invention. Various modifications of these embodiments are obvious to those skilled in the art, and the general principles defined herein may also be applied to other embodiments without departing from the spirit and scope of the present disclosure. Therefore, the present disclosure is not limited to the embodiments given herein, but is consistent with the broadest scope of the principles and novel features disclosed in this application.

[0146] The above description includes examples of one or more embodiments. Of course, it is not possible to describe all possible combinations of components or methods for the purpose of describing the above embodiments, but it should be recognized by those skilled in the art that the various embodiments may be further combined and arranged. Therefore, the embodiments described herein are intended to cover all such changes, modifications and variations that fall within the scope of protection of the appended claims. In addition, with respect to the term "comprising" used in the specification or claims, the word is covered in a manner similar to the term "including", just as "including," is explained as a transitional word in the claims. In addition, any term "or" used in the specification of the claims is intended to mean "non-exclusive or".

[0147] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A risk prediction model for gestational diabetes mellitus based on SVM under mixed kernel function. Features: Contains a pregnant woman data set and a pregnant woman SVM model; among which: The pregnant woman data set includes a pregnant woman basic information set and a pregnant woman file data set; the data in the pregnant woman data set is data of pregnant women in the early pregnancy period; the early pregnancy period is 6 weeks to 13 weeks + 6 days of pregnancy; The pregnant woman SVM model is used to predict the GDM risk of the pregnant woman as a high risk of GDM or a low risk of GDM according to the pregnant woman data set.

2. The risk prediction model for gestational diabetes mellitus using SVM under the mixed kernel function according to claim 1, Features: The basic information set of the pregnant woman includes age, BMI before pregnancy, a first medical history set, and a second medical history set; wherein: The first medical history set includes a history of impaired glucose tolerance, a history of PCOS, and a family history of diabetes; The second medical history set includes a history of unexplained fetal death, a history of stillbirth, a history of miscarriage, a history of delivery of a macrosomia, a history of fetal malformation, a history of polyhydramnios, and a history of GDM.

3. The risk prediction model for gestational diabetes mellitus using SVM under the mixed kernel function according to claim 2, Features: The data set for creating a file for pregnant women includes blood routine, urine routine, fasting blood sugar, and thyroid stimulating hormone; wherein: The routine blood test includes white blood cell count and neutrophil count; The urine routine test includes urine protein and urine bilirubin.

4. The risk prediction model for gestational diabetes mellitus using SVM under the mixed kernel function according to claim 3, Features: The pregnant woman SVM model is expressed as follows: Where: f(x) is used to characterize the pregnant woman SVM model; N is the number of pregnant woman samples collected in the pregnant woman data set; x is the pregnant woman data set, expressed as follows: x m =(x m1 ,x m2 ,x m3 ,x m4 ,x m5 ,x m6 ,x m7 ,x m8 ) Wherein, m is a counter of pregnant women in the real space, used to represent the mth pregnant woman in the real space, and i=1,2,…,m; x m1 represents the age of the mth pregnant woman; x m2 represents the pre-pregnancy BMI of the mth pregnant woman; m3 represents the first medical history set of the mth pregnant woman; m4 represents the second medical history set of the mth pregnant woman; m5 represents the blood routine test of the mth pregnant woman in the early pregnancy period; x m6 represents the urine routine test of the mth pregnant woman in the early pregnancy period; x m7 represents the fasting blood sugar of the mth pregnant woman in the early pregnancy period; x m8 represents the thyroid stimulating hormone of the mth pregnant woman in the early pregnancy; φ(x) is the feature vector that maps the pregnant woman data set x to the feature space; κ(x,x i ) is the kernel function about x; b is the model parameter; w is the model parameter; i is the counter about pregnant women in the feature space, which is used to characterize the i-th pregnant woman in the feature space; y is used to characterize the GDM risk of pregnant women, and y∈{-1,1}, when y m =1, it indicates that the GDM risk of the mth pregnant woman is the high risk of GDM. m = -1, indicating that the GDM risk of the mth pregnant woman is the low risk of GDM; α is a set of Lagrange multipliers, expressed as follows: α=(α 1 ,α 2 ,…α n ,…,α N ) Where: α i is the Lagrange multiplier of pregnant woman i, and α i ≥0.

5. The risk prediction model for gestational diabetes mellitus using SVM under the mixed kernel function according to claim 4, Features: The kernel function κ(x,x i ) is expressed as follows: Where: j is a counter of pregnant women in the feature space, used to represent the jth pregnant woman in the feature space; ||x i -x j || 2 is the vector x i -x j The corresponding L 2 norm;||x i -x j || is vector x i -x j The corresponding L 1 norm; σ is the bandwidth of the Gaussian kernel, and σ>0.

6. The risk prediction model for gestational diabetes mellitus using SVM under the mixed kernel function according to claim 5, Features: In the feature space, the corresponding hyperplane that correctly classifies the sample is (w, b), and the relationship is expressed as follows: Where: w T The transposed matrix representing w; in: D is the sample set of pregnant women, expressed as follows: D={(x 1 ,and 1 ),(x 2 ,and 2 ),…,(x n ,and n ),…,(x N ,and N )} Where: n represents any natural number from 1 to N; (x i ,y i ) is any sample point in D.

7. The risk prediction model for gestational diabetes mellitus using SVM under the mixed kernel function according to claim 6, Features: In the feature space, the distance between the two closest sample points that do not belong to the same set of pregnant women samples is expressed as follows: Where: γ is the distance between the two closest sample points that do not belong to the same type of the pregnant sample set, and the larger the γ is, the larger the distance between the two corresponding classes is, and the better the classification effect is; ||w|| is the L corresponding to the vector w 1 Norm.

8. The risk prediction model for gestational diabetes mellitus using SVM under the mixed kernel function according to claim 7, Features: The optimization equation to maximize γ is expressed as follows: a i ≥0,i=1,2,…,m Where: the Lagrange multiplier α i The relationship between the model parameters b and w is expressed as follows: Where: L(w,b,α) is the Lagrangian function of α, b, and w.

9. The risk prediction model for gestational diabetes mellitus using SVM under the mixed kernel function according to claim 8, Features: The basic information set of pregnant women specifically includes: If the age of the mth pregnant woman is less than 35 years old, then x m1 = -1; if the age of the pregnant woman is ≥ 35 years old, then x m1 =1; If the pre-pregnancy BMI of the mth pregnant woman is 18≤BMI≤24, then x m2 = -1, otherwise x m2 =1; In the first medical history set, if the mth pregnant woman has any one of the history of impaired glucose tolerance, the history of PCOS, and the family history of diabetes, then x m3 =1, otherwise x m3 = -1; In the second medical history set, if the mth pregnant woman has any one of the history of unexplained fetal death, the history of stillbirth, the history of miscarriage, the history of delivery of a macrosomia, the history of fetal malformation, the history of polyhydramnios, and the history of GDM, then x m4 =1, otherwise x m4 =-1.

10. The risk prediction model for gestational diabetes mellitus using SVM under the mixed kernel function according to claim 9, Features: The pregnant woman file data set specifically includes: If the white blood cell count in the blood routine test of the mth pregnant woman falls within the manually preset normal range of white blood cells, and the neutrophil count falls within the manually preset normal range of neutrophils, then x m5 = -1; otherwise ,x m5 =1; If the urine protein in the urine routine test of the mth pregnant woman is negative and the urine bilirubin is negative, then x m6 = -1, otherwise, x m6 =1; If the fasting blood glucose of the mth pregnant woman in the early pregnancy period falls within the manually preset normal range of fasting blood glucose in the early pregnancy period, then x m7 = -1, otherwise x m7 =1; If the thyroid stimulating hormone of the mth pregnant woman in the early pregnancy falls within the manually preset normal range of thyroid stimulating hormone in the early pregnancy, then x m8 = -1, otherwise x m8 =1.