Multi-parameter clinical decision support device for gestational diabetes risk

Abstract

The application discloses a kind of multi-parameter clinical decision-making auxiliary devices for gestational diabetes risk, and it relates to medical auxiliary equipment field, it includes: acquisition module is the input device of intelligent terminal including keyboard, for obtaining gestational diabetes related to the index to be detected, at least including age, first weight, second weight, weight gain, glycosylated hemoglobin, uric acid, triglyceride, total cholesterol and high-density lipoprotein cholesterol;Creation module is used to create index screening model and risk decision-making model;Calculation module is used to calculate the index to be detected obtained by using index screening model, and screening decision-making auxiliary index;Evaluation module is used to evaluate the risk of decision-making auxiliary index by using risk decision-making model;Feedback module is the display screen of intelligent terminal, for returning risk assessment result.The application can effectively evaluate gestational diabetes risk, achieve the purpose of early prediction, and provide effective assistance based on multiple parameters for clinical decision-making.

Description

Technical Field

[0001] This invention relates to the field of medical assistive devices, specifically to a multi-parameter clinical decision support device for gestational diabetes risk. Background Technology

[0002] Gestational diabetes mellitus (GDM) is a common complication of pregnancy, caused by a combination of environmental and genetic factors. According to data from the International Diabetes Federation (IDF), the global prevalence of GDM was 16.7% in 2021. In 2021, Maglano DJ et al. suggested that the prevalence of GDM is increasing with rapid socioeconomic development and improved living standards. Also in 2021, Wu et al. found that high-risk individuals for GDM tend to have elevated blood glucose levels before diagnosis, and a hyperglycemic environment has adverse effects on the fetus. It has become a major cause of maternal and child mortality worldwide. GDM has received increasing attention due to its dangerous consequences and long-term adverse effects on mothers and offspring. In 2015, Gabbay-Benziv R et al. proposed that early detection and standardized management of GDM are crucial for improving maternal and fetal health. In 2017, Umesawa M et al. suggested that GDM increases the risk of maternal infections and preeclampsia, and can also lead to premature birth, fetal malformations, and macrosomia, thus significantly increasing the risk of type 2 diabetes mellitus (T2DM) and metabolic diseases in both mother and offspring. In 2020, Li et al. found that advanced maternal age, pre-pregnancy overweight, and a first-degree relative's history of diabetes were associated with an increased risk of gestational diabetes mellitus (GDM). Also in 2020, YongHY et al. proposed that in early pregnancy, age and pre-pregnancy BMI are independent risk factors for GDM, with overweight or obese women aged 35 years and older having a 2.45 times higher risk of developing GDM than women of normal weight. Li et al.'s 2020 study showed a linear relationship between the risk of GDM and maternal age. Sartayeva A et al.'s 2022 study indicated that the prevalence of GDM increases with maternal age, and early diagnosis and treatment of GDM are crucial for reducing short-term and long-term complications for both mother and child.

[0003] Therefore, designing an auxiliary device to identify whether a pregnant woman has GDM in early pregnancy to assist clinical decision-making, carry out targeted interventions, and reduce maternal and fetal complications has become a research direction. Summary of the Invention

[0004] To address the shortcomings of the existing technology, the present invention aims to provide a multi-parameter clinical decision support device for gestational diabetes risk. This device analyzes the acquired indicators to be tested using an indicator screening model created by a module. An improved regression model is used to calculate and solve the model parameters through a proximal gradient inertia algorithm that combines simulated physical mechanisms of dry friction with Hessian driven damping. Decision support indicators and their corresponding weights are selected from the indicators to be tested. Finally, combined with a risk decision model and risk discrimination threshold, a gestational diabetes risk assessment result is given, achieving the goal of early prediction.

[0005] Specifically, the present invention provides a multi-parameter clinical decision support device for gestational diabetes risk, comprising: The acquisition module is used to acquire the indicators to be tested related to gestational diabetes. The indicators to be tested include at least age, first weight, second weight, weight gain, glycated hemoglobin, uric acid, triglycerides, total cholesterol, and high-density lipoprotein cholesterol. The creation module is used to create indicator screening models and risk decision-making models; The indicator selection model includes: in, The selected decision support indicators and their corresponding weights are as follows: For operators that return non-zero model parameters, Filter vectors for updated metrics. Select vectors for indicators before updating. Step size, To obtain the indicators to be detected, Pseudo-continuous response: The risk decision-making model is as follows: ; ; ; in, For the decision outcome, For 1 and The union set, *for The union set is the set of weights corresponding to each indicator in the set s. The number of items, Risk assessment threshold; The calculation module is used to calculate the acquired indicators to be detected using the indicator screening model, and to screen out decision support indicators. The assessment module is used to assess the risks of decision support indicators using a risk decision-making model; The feedback module is used to return the risk assessment results.

[0006] Furthermore, the risk discrimination threshold is 0.509.

[0007] Furthermore, the index selection model is created based on an improved regression model, which calculates and solves the model parameters using a proximal gradient inertia algorithm that combines dry friction, a physical mechanism, with Hessian-driven damping.

[0008] Furthermore, in the indicator selection model The update rule in the improved regression model is obtained by solving the Lasso problem: The Lasso problem is: Where n is the number of samples, It is a pseudo-continuous response, and r1 is the regularization parameter.

[0009] Furthermore, the update rule is as follows: in, This represents the current iteration number. For step size parameters, For inertial parameters, For viscous damping parameters, For time step, It is a positive damping parameter. express The near-end mapping function, For auxiliary functions, It is a data consistency item right The gradient.

[0010] Furthermore, the selected decision support indicators included age, first weight, second weight, weight gain, glycated hemoglobin, uric acid, triglycerides, total cholesterol, and high-density lipoprotein cholesterol.

[0011] Furthermore, the first weight is the weight in the first trimester, and the second weight is the weight at 28 weeks of pregnancy.

[0012] Furthermore, the acquisition module can be an input device for a smart terminal, including but not limited to a keyboard, scanner, and camera; smart terminals include but are not limited to computers, mobile phones, and tablets.

[0013] Furthermore, the feedback module can be used for the display screen or sound broadcasting system of a smart terminal.

[0014] Furthermore, there are electrical or communication connections between the various modules of the device.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The device of this invention provides evaluation results based on a model constructed by combining detection index data with an improved regression model. It has the characteristics of multiple parameters, which is conducive to objective evaluation and prediction. This invention's device solves for model parameters by creating an improved Logistic-Lasso regression model and employing a proximal gradient inertia algorithm that combines simulated physical mechanisms of dry friction with Hessian-driven damping. The simulated dry friction provides a non-smooth, "hard" damping mechanism that effectively suppresses oscillations and prevents parameters from repeatedly oscillating around the optimal solution. Hessian-driven damping utilizes the curvature information (Hessian matrix) of the objective function for adaptive damping, automatically adjusting the step size of the inertia term (momentum). This results in a more stable algorithm, less prone to numerical divergence or oscillations, and insensitive to initial value selection, enabling more reliable finding of the optimal solution. It offers significant advantages in convergence speed, stability, and the ability to find sparse solutions.

[0016] 3. The device of the present invention is highly operable and can be well applied in clinical practice. It can also effectively assess the risk of gestational diabetes in pregnant patients, achieving the goal of early prediction. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the multi-parameter clinical decision support device for gestational diabetes risk of the present invention; Figure 2 This is an AUC-ROC curve diagram of one embodiment of the device of the present invention; Figure 3 This is an AUC-PR curve diagram of one embodiment of the device of the present invention; Figure 4 This is a Log Loss curve diagram of one embodiment of the device of the present invention; Figure 5 This is a diagram illustrating part of the multi-parameter clinical decision data for gestational diabetes risk in one embodiment of the device of the present invention. Detailed Implementation

[0018] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0019] The purpose of this invention is to provide a multi-parameter clinical decision support device for gestational diabetes risk. Based on multiple parameters including age, first body weight, second body weight, glycated hemoglobin, alanine aminotransferase, and high-density lipoprotein cholesterol, it uses a creation module to create an indicator screening model and a risk decision model, and then uses a calculation module and an evaluation module to perform risk assessment, providing clinical decision opinions based on multi-parameter gestational diabetes risk assessment, so as to accurately identify early and adjust the adjuvant treatment strategy in a timely manner and improve patient prognosis.

[0020] Optionally, in this embodiment, the multi-parameter clinical decision support device for gestational diabetes risk of the present invention can be a variety of smart terminals capable of realizing the functions of the present invention. Its acquisition module can be an input device such as a keyboard, scanner, or camera, including but not limited to a computer, mobile phone, or PAD. Its creation module, calculation module, and evaluation module are integrated into a computer CPU or mobile phone chip. The created decision model and evaluation model can be stored as a program in a readable storage medium of a computer or mobile phone. Its feedback module can be the display screen or sound broadcasting system of the smart terminal. The modules are electrically or communicatively connected to each other.

[0021] As attached Figure 1 As shown, the present invention provides a multi-parameter clinical decision support device for gestational diabetes risk, the clinical decision support device 10 comprising: The acquisition module 101 is used to acquire the indicators to be detected. Decision support indicators include age, height, initial weight, second weight, body mass index (BMI), weight gain, glycated hemoglobin (HbA1c), free triiodothyronine (TTI), free tetraiodothyronine (THT), thyroid-stimulating hormone (TSH), alanine aminotransferase (ALT), aspartate aminotransferase (AST), gamma-glutamyl transferase (GGT), uric acid, blood urea nitrogen (BUN), creatinine, triglycerides, total cholesterol, high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C), etc., which may be related to GDM. In this embodiment of the invention, the initial weight is the weight in early pregnancy, the second weight is the weight at 28 weeks of gestation, and weight gain = second weight - initial weight.

[0022] Create module 102 to create indicator screening models and risk decision-making models; The created indicator selection model is as follows: ; in, The selected decision support indicators and their corresponding weights are as follows: For operators that return non-zero model parameters, Filter vectors for updated metrics. Select vectors for indicators before updating. Step size, To obtain the indicators to be detected, It is a pseudo-continuous response.

[0023] The risk decision-making model created is as follows: ; ; ; in, For the decision outcome, For 1 and The union set, *for The union set is the set of weights corresponding to each indicator in the set s. The number of items, Risk assessment threshold; The calculation module 103, in response to the indicator to be detected acquired by the acquisition module 101, calls the indicator screening model created by the creation module 102 to calculate and filter out the input information items; The assessment module 104, in response to the decision support indicators selected by the calculation module 103, calls the risk decision model created by the creation module 102 to perform risk assessment. Feedback module 105 is used to return the gestational diabetes risk assessment results obtained by assessment module 104.

[0024] In practical use, the following common knowledge and parameters are involved: Gestational diabetes mellitus (GDM) is a common pregnancy complication that seriously affects the health of both the mother and fetus. Pregnant women with GDM may develop type 2 diabetes postpartum, and their lifelong risk is higher than the general population. Poor blood sugar control in GDM mothers may increase the risk of neural tube defects (such as spina bifida), congenital heart disease, and limb malformations in the fetus. High blood sugar can cross the placenta and cause fetal hyperinsulinemia, promoting protein and fat synthesis and leading to excessive fetal growth and macrosomia (large baby).

[0025] [Baseline] Indicators: Age, height, weight in early pregnancy, weight at 28 weeks of pregnancy, body mass index, weight gain, glycated hemoglobin, free triiodothyronine, free tetraiodothyronine, thyroid-stimulating hormone, alanine aminotransferase, aspartate aminotransferase, gamma-glutamyl transferase, uric acid, blood urea nitrogen, creatinine, triglycerides, total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and other indicators that may be related to GDM.

[0026] 【Output Results】 Predict whether the response variable meets the target, i.e., assess whether it is gestational diabetes.

[0027] With the aid of decision-making by the device of this invention, the evaluation results are predicted based on whether the response variable meets the target: patients with a value less than or equal to 0.509 are evaluated as having no risk of gestational diabetes; patients with a value greater than 0.509 are evaluated as having gestational diabetes. Clinicians are requested to make comprehensive assessments and adjust treatment strategies based on the patient's condition.

[0028] Model expression source: This invention adopts the Logistic-Lasso model and solves the model parameters by combining the dry friction of the physical mechanism with the Hessian driven damping algorithm.

[0029] Based on the Logistic regression model, predictor variables Set as: (1) Where s is a positive integer.

[0030] Model parameters Set as: There is a linear relationship between the predictor variables and the model parameters: Among them, the error term It follows a standard Logistic distribution. A critical value can be chosen; this invention assumes it to be 0. When the response variable... A value higher than the threshold value is used to determine whether an event has occurred or whether the sample belongs to a certain category. When the response variable... If the value is lower than the critical value c, the critical value discrimination threshold of this invention is 0.509, indicating that the event has not occurred or the sample does not belong to a certain category. Note: in, For indicator functions, Let c be the response variable and c be the critical value.

[0031] The probability of the event occurring is: The probability that the event will not occur is: remember: The cross-entropy loss of logistic regression is: Where n is the number of samples.

[0032] The probability space [0, 1] is mapped to the real number space using the Logit operation. : Constructing a pseudo-continuous response: Will exist Second-order Taylor expansion: Then we have: Will exist Taylor's expansion: Then we have: (15) This invention introduces The penalties include: Where r1 is the regularization parameter.

[0033] Therefore, the Lasso problem that this invention needs to solve is: in, It is a pseudo-continuous response, and r1 is the regularization parameter.

[0034] The value of β is obtained by solving expression (17), which is to say, the effective feature is obtained. Furthermore, this invention uses a β value greater than 0.05 as the criterion for effective features, and values ​​below 0.05 are considered invalid, thus making the effective features more concise. The "Occam's Razor" principle states that, under the same explanatory conditions, the simpler the model, the better. The more features there are, the higher the model complexity, and the greater the risk of overfitting (i.e., the model performs well on training data but poorly on new data). By setting a coefficient threshold for secondary screening, a more rigorous feature selection is essentially being performed. This helps to remove features that may only have small coefficients due to data noise or chance factors, thus obtaining a simpler model that is more likely to maintain good predictive performance on new data.

[0035] To solve this Lasso problem, this invention considers it as... The solution is a composite optimization problem, in which... For data consistency items, As regularization terms, this invention defines them as follows: The dry friction damping function is: Among them, dry friction damping threshold Hessian drive damping is ,in It is a positive damping parameter.

[0036] Here, the dry friction damping function is given by equation (20). In the heavy ball system, dry friction damping provides constant resistance force opposite to the velocity direction, which can effectively consume energy, enabling the system to stabilize to the equilibrium point more quickly and improving the system convergence effect. Dry friction introduces a threshold mechanism in the algorithm. When the gradient magnitude of the iteration point is less than a certain preset threshold, the algorithm will stop directly or proceed to the next step, instead of just approaching the minimum value infinitely. This can effectively avoid repeated oscillations near the optimal point and achieve finite-step convergence. Hessian-driven damping automatically adjusts the step size of the inertia term (momentum) based on the local curvature of the objective function (described by the eigenvalues ​​of the Hessian matrix). The step size is reduced (increased damping) where the curvature is large, and the step size is increased (decreased damping) where the curvature is small, thereby adaptively accelerating convergence and avoiding oscillations.

[0037] To reduce the complexity of the solution, this invention defines an auxiliary function. And the Hessian driven damping is discretized as The update rule for the composite optimization problem is: in, This represents the current iteration number. The iteration number is set based on the premise that the loss remains essentially constant, and iteration will stop once this number is reached. For step size parameters, For inertial parameters, For viscous damping parameters, time step . express The proximal mapping function, where the proximal mapping is defined as . yes right The gradient.

[0038] Efficient solutions to these types of complex optimization problems are crucial for related model prediction applications. For such complex optimization problems, the proximal gradient method and its variants have become one of the effective numerical solutions. The function value convergence rate of the proximal gradient algorithm is sublinear, and the function value convergence rate of the inertial proximal gradient algorithm is also sublinear. Under the regulation of dry friction simulating the physical mechanism, this invention progresses from fast strong convergence with finite length properties to linear convergence, and may eventually achieve finite-step convergence.

[0039] According to the definition of proximal mapping, we have: ; in, The solution to the compound proximal mapping can be decomposed into multiple one-dimensional subproblems as shown in expression (25): in, It is an intermediate input variable scalar form, It is the current model parameter vector The corresponding scalar form. In expression (25) The closed-form solution is: when hour, The closed-form solution is shown as a piecewise function in expression (27).

[0040] By orthogonalizing the residuals of the non-zero model parameters, the update expression for the model parameters can be obtained using the gradient descent method: Expression (28) is the solution to the Lasso problem of expression (17), yielding non-zero parameters. Among them, Let S(·) be the step size, S(·) be the operator that returns non-zero model parameters, and r be the residual between the response variable and the pseudo-continuous response. according to , For r about The gradient, i.e.: in, This is the final output one-dimensional model parameter vector. When the value of an element in this vector is 0 or close to 0, it means that the feature corresponding to that element is not relevant to GDM risk prediction or has a very low correlation. These corresponding features will be considered as features to be removed.

[0041] Furthermore, after the model is built, it is trained. The specific training settings are as follows: the collected dataset, including the diabetes detection indicators, is randomly divided into two groups at a ratio of 7:3. The group with a 7:3 ratio serves as the training set, and the group with a 3:3 ratio serves as the test set. The iteration rule is the update rule described above. The parameters used in the iteration are configured as follows: L1 regularization parameter r1 = 0.01, dry friction damping parameter r2 = 0.05, time step h = 0.01, and viscous damping parameter γ is adaptively adjusted according to the convergence condition, time step h, and Lipschitz constant L. The number of iterations k = 5000. In addition, to select more concise and highly relevant effective features, the threshold for selecting effective features is set to 0.05. Ten-fold cross-validation is used on the training set to select the optimal model parameters. Then, the model is built using all training data and the optimal parameters, and the obtained model is validated on the test set to obtain experimental results.

[0042] Furthermore, the calculation module also includes calculations to standardize the input indicators to be detected, including: ; Among them, X std The data represents the standardized indicator data, while X represents the original indicator data. It is the mean of the corresponding indicators for all samples in the dataset. It is the standard deviation of the corresponding index for all samples in the dataset.

[0043] This invention employs the Logistic-Lasso model and solves for the model parameters using a proximal gradient inertia algorithm that combines dry friction with Hessian driven damping. The objective function of the Logistic-Lasso method is a unified optimization of "cross-entropy loss (classification objective) + L1 regularization (feature selection)". Integrating feature selection and the classifier into a single model reduces computational cost. However, the Logistic-Lasso method suffers from low accuracy, slow speed, poor convergence, and high objective function value. Traditional proximal gradient algorithms have slow convergence speeds. To accelerate convergence and improve the optimization trajectory, an inertia term is introduced into the dynamic system. However, simple inertial acceleration may induce oscillations or even cause the algorithm to diverge in highly nonlinear or non-convex composite functions. To suppress the instability that inertia may cause and to more finely control the optimization dynamics, this invention introduces dry friction damping and Hessian driven damping. This method effectively improves the convergence of the Logistic-Lasso method, thereby enhancing its classification accuracy.

[0044] like Figure 2As shown, the AUC-ROC curve of the device model of the present invention is 0.7979. The ROC curve is clearly above the random line and far away from the random line, which indicates that the model has a strong ability to distinguish between positive and negative samples.

[0045] like Figure 3 The figure shows the AUC-PR curve of the device model of the present invention. It shows that the AUC-PR is 0.6673. The PR curve of the model is always above the random line, indicating that the model's ability to identify positive GDM samples is significantly better than random guessing.

[0046] like Figure 4 As shown, the Log Loss curve of the device model of the present invention shows that both the training set loss and the validation set loss decrease rapidly with the increase of the number of iterations and eventually tend to stabilize, indicating that the model gradually fits the data during the training process and has good generalization ability.

[0047] The evaluation process is shown in detail below as an example of the practical application of the device of the present invention to a single patient: The acquisition module 101 of the auxiliary device of the present invention acquires a total of 21 detectable indicators that may be related to GDM in a certain patient, as shown below: Patient's original predictor variables: =[Maternal age 40, height 156.0, weight in early pregnancy 66.0, weight at 28 weeks of pregnancy 70.0, body mass index (BMI) 27.12, weight gain 4.0, glycated hemoglobin 5.6, free triiodothyronine (FT3) 4.01, free tetraiodothyronine (FT4) 13.9, thyroid-stimulating hormone (TSH) 1.49, alanine aminotransferase (ALT) 14.2, aspartate aminotransferase (AST) 13.1, gamma-glutamyl transferase (GTT) 9.6, uric acid (UA) 226.9, blood urea nitrogen (Urea) 4.35, creatinine (CREA) 47.0, triglycerides (TG) 1.54, total cholesterol (TC) 6.02, high-density lipoprotein cholesterol (HDL-C) 2.15, low-density lipoprotein cholesterol (LDL-C) 3.17, TG / HDL-C 0.72].

[0048] Patient standardized predictor variables: _scaled=[np.float64(2.539074),np.float64(-1.556355),np.float64(0.715082),np.float64(0.48039),np.float64(1.386959), np.float64(-0.510047),np.float64(1.957516),np.float64(-1.031043),np.float64(-0.368695),np.float64(-0.117457),np.fl oat64(-0.061394),np.float64(-0.556025),np.float64(-0.435143),np.float64(0.252678),np.float64(1.492826),np.float64( 0.341204),np.float64(-0.08589),np.float64(1.573753),np.float64(0.42846),np.float64(1.923889),np.float64(-0.30442)].

[0049] Nine decision support indicators were selected from the 21 indicators to be tested. The decision support indicators and their corresponding weights are as follows: 1. Age (β=0.099352): Contribution value = 0.252261 (positive (increases the risk of disease)).

[0050] 2. Early pregnancy weight (β=0.128911): Contribution value = 0.092182 (positive (increases the risk of disease)).

[0051] 3. Weight at 28 weeks of gestation (β=-0.077076): Contribution value = -0.037026 (negative (reduced risk of disease)).

[0052] 4. Weight gain (β=0.082291): Contribution value = -0.041972 (negative (reduced risk of disease)).

[0053] 5. Glycated hemoglobin (β=0.266636): Contribution value = 0.521944 (positive (increases the risk of disease)).

[0054] 6. Uric acid UA (β=0.256316): Contribution value = 0.064765 (positive (increases the risk of disease)).

[0055] 7. Triglycerides (TG) (β=0.120047): Contribution value = -0.010311 (negative (reduced risk of disease)).

[0056] 8. Total cholesterol TC (β=-0.070733): Contribution value = -0.111316 (negative (reducing the risk of disease)).

[0057] 9. High-density lipoprotein cholesterol (β=-0.059371): Contribution value = -0.025438 (negative (reducing the risk of disease)).

[0058] Linear response variable: * = 0.68197.

[0059] Probability of disease after sigmoid conversion: Sigmoid( *) = 0.664178.

[0060] Judgment logic: 0.664178 > threshold (0.509).

[0061] Judgment result: =1, the patient was analyzed by a multi-parameter clinical decision support device for gestational diabetes risk according to the present invention, and the feedback result was that the patient had GDM.

[0062] Based on clinical trials, multi-parameter feature values ​​from several thousand actual patients were obtained and input into a multi-parameter clinical decision support device for gestational diabetes risk of this invention for evaluation, analysis and verification. Figure 5 (This is a partial display of the data.)

[0063] This invention employs gradient descent, stochastic gradient descent, coordinate gradient descent, and the model of this device to solve for the parameters of the Logistic-Lasso regression model. The obtained models are then subjected to 10-fold cross-validation, and their AUCs are compared. The results are shown in Table 1 below: Table 1: Comparison of AUC-ROC between the device model and the traditional method model As can be seen from the experimental results in Table 1, the model constructed by the device of the present invention outperforms the gradient descent method, stochastic gradient descent method and coordinate gradient descent method in terms of AUC-ROC index. This indicates that the model constructed by the device of the present invention is superior to other traditional methods in terms of discrimination.

[0064] This invention demonstrates that the device can effectively assess patients at risk of gestational diabetes, achieving the goal of early prediction and providing effective multi-parameter-based assistance for clinical decision-making regarding gestational diabetes.

[0065] In the several embodiments provided in this application, it should be understood that the disclosed client can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the module division is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces, or indirect couplings or communication connections between modules, and may be electrical or other forms.

[0066] The modules described as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0067] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

[0068] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A multi-parameter clinical decision support device for gestational diabetes risk, characterized in that, It includes: The acquisition module is used to acquire the indicators to be tested related to gestational diabetes. The indicators to be tested include at least age, first weight, second weight, weight gain, glycated hemoglobin, uric acid, triglycerides, total cholesterol, and high-density lipoprotein cholesterol. The creation module is used to create indicator screening models and risk decision-making models; The indicator selection model includes: in, The selected decision support indicators and their corresponding weights are as follows: For operators that return non-zero model parameters, Filter vectors for updated metrics. Select vectors for indicators before updating. Step size, To obtain the indicators to be detected, Pseudo-continuous response: The risk decision-making model is as follows: ； ； ； in, For the decision outcome, For 1 and The union set, *for The union set is the set of weights corresponding to each indicator in the set s. The number of items, Risk assessment threshold; The calculation module is used to calculate the acquired indicators to be detected using the indicator screening model, and to screen out decision support indicators. The assessment module is used to assess the risks of decision support indicators using a risk decision-making model; The feedback module is used to return the risk assessment results.

2. The multi-parameter clinical decision support device for gestational diabetes risk according to claim 1, characterized in that, The risk assessment threshold is 0.

509.

3. The multi-parameter clinical decision support device for gestational diabetes risk according to claim 1, characterized in that, The index selection model is based on an improved regression model, which calculates and solves the model parameters using a proximal gradient inertia algorithm that combines dry friction, a physical mechanism, with Hessian-driven damping.

4. The multi-parameter clinical decision support device for gestational diabetes risk according to claim 3, characterized in that, Indicator screening model The update rule in the improved regression model is obtained by solving the Lasso problem: The Lasso problem is: Where n is the number of samples, It is a pseudo-continuous response, and r1 is the regularization parameter.

5. The multi-parameter clinical decision support device for gestational diabetes risk according to claim 4, characterized in that, The update rules are as follows: in, This represents the current iteration number. For step size parameters, For inertial parameters, For viscous damping parameters, For time step, It is a positive damping parameter. express The near-end mapping function, For auxiliary functions, It is a data consistency item right The gradient.

6. The multi-parameter clinical decision support device for gestational diabetes risk according to claim 1, characterized in that, The selected decision support indicators include age, first weight, second weight, weight gain, glycated hemoglobin, uric acid, triglycerides, total cholesterol, and high-density lipoprotein cholesterol.

7. The multi-parameter clinical decision support device for gestational diabetes risk according to claim 6, characterized in that, The first weight is the weight gained in the first trimester, and the second weight is the weight gained at 28 weeks of pregnancy.

8. The multi-parameter clinical decision support device for gestational diabetes risk according to claim 1, characterized in that, The acquisition module can be an input device for a smart terminal, including but not limited to keyboards, scanners, and cameras; smart terminals include but are not limited to computers, mobile phones, and tablets.

9. The multi-parameter clinical decision support device for gestational diabetes risk according to claim 1 or 4, characterized in that, The feedback module can provide feedback for the display screen or sound broadcasting system of a smart terminal.

10. The multi-parameter clinical decision support device for gestational diabetes risk according to claim 1, characterized in that, Electrical or communication connections between the various modules of the device.

Images