Method and device for predicting risk of preeclampsia by utilizing artificial intelligence model

Abstract

The present disclosure relates to a method and device for predicting the risk of preeclampsia by utilizing an artificial intelligence model. One embodiment of the present disclosure may provide a method for predicting the risk of preeclampsia by utilizing an artificial intelligence model, the method comprising the steps of: inputting data about an expectant mother to a preeclampsia prediction model generated on the basis of a training data set; and acquiring an incidence rate of the preeclampsia from the preeclampsia prediction model, wherein the data about the expectant mother includes final variables determined from a plurality of initial variables of the training data set.

Description

Method and device for predicting the risk of preeclampsia using an artificial intelligence model

[0001] The present disclosure provides a method and device for predicting the risk of preeclampsia using an artificial intelligence model.

[0002] The three most common high-risk pregnancy conditions in obstetrics are preeclampsia, gestational diabetes, and premature birth. All three conditions not only negatively impact the health of the fetus and mother, but can also affect the survival and development of the fetus.

[0003] For these high-risk diseases, early diagnosis, effective prevention, and management are crucial. Therefore, while active research is currently underway into technologies that comprehensively predict the risk of various obstetric diseases, there is currently no technology capable of accurately predicting the risk of obstetric diseases based solely on early pregnancy data.

[0004] The background technology described above is technical information that the inventor possessed for the purpose of deriving the present invention or acquired in the process of deriving the present invention, and cannot necessarily be considered as publicly known technology disclosed to the general public prior to the application for the present invention.

[0005] The purpose of the present disclosure is to provide a method and device for predicting the risk of preeclampsia using an artificial intelligence model. The problems addressed by the present disclosure are not limited to those mentioned above. Other problems and advantages of the present disclosure not mentioned above can be understood through the following description and will be more clearly understood through the embodiments of the present disclosure. Furthermore, it will be appreciated that the problems and advantages addressed by the present disclosure can be realized by the means and combinations thereof set forth in the claims.

[0006] A first aspect of the present disclosure provides a method for predicting the risk of preeclampsia using an artificial intelligence model, comprising: a step of inputting data about a pregnant woman into a preeclampsia prediction model generated based on a learning data set; and a step of obtaining an incidence of preeclampsia from the preeclampsia prediction model; wherein the data about the mother includes a final variable determined from a plurality of initial variables of the learning data set.

[0007] A second aspect of the present disclosure includes a memory having at least one program stored therein; and a processor performing a calculation by executing the at least one program; wherein the processor inputs data about a pregnant woman into a pregnancy-related toxemia prediction model generated based on a learning data set, and obtains a pregnancy-related toxemia incidence rate from the pregnancy-related toxemia prediction model, wherein the data about the mother includes a final variable determined from a plurality of initial variables of the learning data set, and may provide a device for predicting the risk of pregnancy-related toxemia using an artificial intelligence model.

[0008] A third aspect of the present disclosure can provide a computer-readable recording medium having recorded thereon a program for executing the method of the first aspect on a computer.

[0009] In addition, other methods for implementing the present invention, other devices, and computer-readable recording media recording a program for executing the method may be further provided.

[0010] Other aspects, features and advantages other than those described above will become apparent from the following drawings, patent claims and detailed description of the invention.

[0011] According to the problem solving means of the present disclosure described above, a service for early prediction of obstetric diseases can be installed in the obstetrics internal network of each hospital to support decision-making by medical staff.

[0012] In addition, according to the problem solving means of the present disclosure, a service for early prediction of obstetric diseases can be provided to pregnant women through a web server, thereby contributing to the preservation and promotion of national health.

[0013] Figure 1 is an example of an environment for providing a service for early prediction of obstetric diseases.

[0014] FIG. 2 is a block diagram of a system for early prediction of obstetric diseases according to one embodiment of the present invention.

[0015] Figure 3 is an exemplary diagram of an input interface according to one embodiment of the present invention.

[0016] Figures 4 to 6 are exemplary diagrams of an output interface according to one embodiment of the present invention.

[0017] Figure 7 is a flowchart of a method for early prediction of obstetric diseases according to one embodiment of the present invention.

[0018] Figure 8 is a conceptual diagram of a method for generating an obstetric disease prediction model according to one embodiment of the present invention.

[0019] FIG. 9 is a diagram for explaining a method for preprocessing an acid and acid data set according to one embodiment of the present invention.

[0020] FIG. 10 and FIG. 11 are graphs of the performance evaluation results and the Hosmer-Lemshaw test results of a pregnancy-induced hypertension prediction model according to one embodiment of the present invention.

[0021] Figures 12 to 14 are graphs of performance evaluation results and Hosmer-Lemshaw test results of a gestational diabetes prediction model according to one embodiment of the present invention.

[0022] Figures 15 and 16 are graphs of the performance evaluation results and Hosmer-Lemshaw test results of a preterm birth prediction model according to one embodiment of the present invention.

[0023] Figure 17 is a flowchart of a method for predicting the risk of preeclampsia using an artificial intelligence model according to one embodiment of the present invention.

[0024] Figure 18 is a flowchart of a method for predicting the risk of gestational diabetes using an artificial intelligence model according to one embodiment of the present invention.

[0025] Figure 19 is a flowchart of a method for predicting the risk of premature birth using an artificial intelligence model according to one embodiment of the present invention.

[0026] Figure 20 is a block diagram of a device for early prediction of obstetric diseases according to one embodiment of the present invention.

[0027] The present disclosure relates to a method and device for predicting the risk of preeclampsia using an artificial intelligence model. One embodiment of the present disclosure may provide a method for predicting the risk of preeclampsia using an artificial intelligence model, comprising the steps of: inputting data about a pregnant woman into a preeclampsia prediction model generated based on a learning data set; and obtaining an incidence of preeclampsia from the preeclampsia prediction model; wherein the data about the woman includes a final variable determined from a plurality of initial variables of the learning data set.

[0028] The advantages and features of the present invention, and the methods for achieving them, will become clearer with reference to the embodiments described in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments presented below, but can be implemented in various different forms, and it should be understood that it includes all transformations, equivalents, and substitutes included in the spirit and technical scope of the present invention. The embodiments presented below are provided to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the invention of the scope of the invention. In describing the present invention, if a detailed description of a related known technology is judged to obscure the gist of the present invention, the detailed description thereof will be omitted.

[0029] The terminology used in this application is only used to describe specific embodiments and is not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this application, it should be understood that the terms "comprise" or "have" indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the specification, but do not preclude the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0030] Some embodiments of the present disclosure may be represented by functional block configurations and various processing steps. Some or all of these functional blocks may be implemented by various hardware and / or software configurations that perform specific functions. For example, the functional blocks of the present disclosure may be implemented by one or more microprocessors or by circuit configurations for a given function. Furthermore, for example, the functional blocks of the present disclosure may be implemented by various programming or scripting languages. The functional blocks may be implemented by algorithms that execute on one or more processors. Furthermore, the present disclosure may employ conventional techniques for electronic configuration, signal processing, and / or data processing. Terms such as "mechanism," "element," "means," and "configuration" may be used broadly and are not limited to mechanical and physical configurations.

[0031] Additionally, the connecting lines or connecting members between components depicted in the drawings are merely exemplary representations of functional connections and / or physical or circuit connections. In an actual device, connections between components may be represented by various functional connections, physical connections, or circuit connections that may be replaced or added.

[0032] In the following, operations according to various embodiments may be understood as being performed by a processor included in a device for predicting obstetric diseases at an early stage (hereinafter, “obstetric disease prediction device”) or a prediction service providing device.

[0033] The present disclosure will be described in detail with reference to the attached drawings below.

[0034] Figure 1 is an example of an environment for providing a service for early prediction of obstetric diseases.

[0035] Referring to Fig. 1, an environment for providing a service for early prediction of obstetric diseases may include an obstetric disease prediction device (1), a pregnant woman (2), and an obstetric database (3). In this case, the obstetric database (3) may be included in the obstetric disease prediction device (1) or may be a separate configuration from the obstetric disease prediction device (1).

[0036] In one embodiment, the obstetric disease prediction device (1) can obtain input data related to an obstetric disease of a pregnant woman (2) and output the incidence rate of the obstetric disease of the pregnant woman (2).

[0037] For example, the mother's height, blood pressure, -Fetoprotein, etc. are significantly related to the incidence of obstetric diseases such as gestational diabetes, preeclampsia, and premature birth, and maternal age, weight, creatinine, etc. are significantly related to the incidence of gestational diabetes and preeclampsia. Therefore, the obstetric disease prediction device (1) can predict and output the incidence of each obstetric disease from the data of the mother (2) through a prediction model learned based on various obstetric data of existing mothers.

[0038] Meanwhile, in this detailed explanation, "significant" can be used to mean statistically significant. That is, it can mean that the statistical significance of a hypothesis about a population is significant enough to be considered unlikely to be due to mere chance.

[0039] In one embodiment, the obstetric disease prediction device (1) may be a server. As an example, the obstetric disease prediction device (1) may provide a web interface to the mother (2) so that the mother (2) can input weight, blood pressure, blood test result data, data related to previous pregnancy, etc. at an early stage of pregnancy. In other words, the obstetric disease prediction device (1) may construct a service providing system having a specific domain address as a website, and when electronic devices access the constructed website, the obstetric disease prediction service based on the service providing system may be provided to the electronic devices. However, the web interface described in the drawings below is merely an example, and the web interface displayed when the obstetric disease prediction device (1) provides the obstetric disease prediction service to the mother (2) is not limited thereto.

[0040] In one embodiment, the obstetric database (3) may store obstetric data of multiple existing pregnant women. In addition, the obstetric database (3) may store input data entered by the pregnant woman (2).

[0041] In one embodiment, the obstetric disease prediction device (1) or prediction model may employ a machine learning algorithm for predicting the incidence of obstetric diseases. That is, the obstetric disease prediction device (1) or prediction model may utilize a machine learning algorithm to predict the incidence of obstetric diseases.

[0042] In one embodiment, the obstetric disease prediction device (1) or the prediction model may be updated based on obstetric data of a plurality of existing pregnant women stored in the obstetric database (3). Furthermore, when new data, such as input data of a pregnant woman (2), is accumulated and stored, the obstetric disease prediction device (1) or the prediction model may also be updated based on the new data. Accordingly, the weights of the prediction model of the obstetric disease prediction device (1) may be updated based on the obstetric data stored in the obstetric database (3).

[0043] FIG. 2 is a block diagram of a device for early prediction of obstetric diseases according to one embodiment of the present invention.

[0044] In one embodiment, the obstetric disease prediction device may include an input unit (210), a prediction unit (220), an output unit (230), and a database management system (240). In this case, the input unit (210), the prediction unit (220), the output unit (230), and the database management system (240) are merely divided into functional units for convenience of explanation and may not mean physically separate configurations.

[0045] In one embodiment, the input unit (210) may receive input data related to an obstetric disease of the pregnant woman. Specifically, the input unit (210) may provide the pregnant woman with an input interface for receiving data, and may acquire data entered by the pregnant woman as input data. In this case, the obstetric disease may include at least one of preeclampsia, gestational diabetes, and premature birth. Furthermore, in this document, gestational diabetes mellitus (GDM) may be used to encompass both A1GDM and A2GDM.

[0046] In one embodiment, the prediction unit (220) can predict the incidence of obstetric diseases through a machine learning algorithm based on input data.

[0047] In one embodiment, the prediction unit (220) may select a prediction model corresponding to an obstetric disease. Specifically, the prediction unit (220) may select a prediction model corresponding to each of two or more obstetric diseases. For example, the prediction unit (220) may select a prediction model corresponding to preeclampsia to predict the incidence of preeclampsia and input at least some of the input data. Similarly, the prediction unit (220) may select a prediction model corresponding to gestational diabetes or premature birth to predict the incidence of gestational diabetes or premature birth and input at least some of the input data.

[0048] At this time, the variables required for input to the selected prediction model may be at least a portion of the input data. That is, since the variables required for input to the prediction models corresponding to two or more obstetric diseases may be different, only a portion, not all, of the acquired input data may be variables required for input to the selected prediction model.

[0049] In one embodiment, the prediction unit (220) can obtain the incidence of an obstetric disease predicted by the prediction model.

[0050] In one embodiment, the output unit (230) may output the incidence of obstetric diseases in pregnant women. In this case, when there are two or more obstetric diseases as described above, the output unit (230) may output the incidence rate for each obstetric disease.

[0051] In one embodiment, the database management system (240) may collect obstetric data from multiple existing pregnant women to build an obstetric database. Furthermore, the database management system (240) may store the maternal input data obtained through the input unit (210) as obstetric data in the obstetric database.

[0052] As an example, the database management system (240) can store the obstetric data of multiple existing pregnant women in an obstetric database in a form that allows the prediction unit (220) to perform machine learning. Furthermore, the database management system (240) can periodically or aperiodically update the prediction model of the prediction unit (220) based on the obstetric data.

[0053] Predictive models can predict the incidence of obstetric diseases even when some of the required input variables are missing. However, the accuracy of the incidence of obstetric diseases may be lower than when all required variables are included.

[0054] Figure 3 is an exemplary diagram of an input interface according to one embodiment of the present invention.

[0055] In one embodiment, the obstetric disease prediction device may provide an input interface (300) for receiving at least one of the following data: the mother's age, height, weight, blood pressure, expected date of delivery, method of current pregnancy, history of diabetes, history of hypertension, history of premature birth, family history of diabetes, family history of hypertension, number of pregnancies, number of fetuses, year of last delivery, number of weeks of last delivery, birth weight, history of gestational diabetes, history of gestational hypertension, and blood test result.

[0056] For example, the method of conception could be one of natural conception, ovulation induction, artificial insemination, or in vitro fertilization. Furthermore, the number of fetuses could be one of singletons, twins, or triplet or more. Furthermore, the past fertility could be one of "none," 1, 2, or 3 or more. Selecting any option other than "none" would activate an interface for entering the birth history for each birth.

[0057] At this time, the mother is not required to input data for all variables of the input interface, and the input interface may be configured to allow selective input for some variables that are difficult or impossible to measure. That is, in one embodiment, the obstetric disease prediction device may provide an interface that allows for input of alternative data for at least some of the mother's input data.

[0058] For example, an obstetric disease prediction device may provide an interface for inputting whether a patient has experienced premature birth (preterm birth history) as surrogate data for the previous week of delivery, or an interface for inputting whether a patient has given birth to a large baby (e.g., weighing more than 4 kg) as surrogate data for birth weight. In this case, the interface for inputting surrogate data may include an item for selecting yes / no or 0 / 1.

[0059] In one embodiment, the blood test result data may include at least one of numerical data for Hemoglobin, White blood cell count, Platelet Count, Total cholesterol, Triglyceride, Albumin, AST (GOT), ALT (GPT), Creatinine, HgbA1C, PAPP-A, uE3, a-fetoprotein, AntiTPO, FT4 (Free thyroxine), and Inhibin-A.

[0060] In one embodiment, an obstetric disease prediction device can predict the incidence of obstetric diseases early based on input data from the mother's early pregnancy. In one embodiment, the obstetric disease prediction device may display a message (310) recommending the input of input data corresponding to a gestational age of 14 weeks or less. Alternatively, the obstetric disease prediction device may only allow the input of input data corresponding to a gestational age of 14 weeks or less.

[0061] Figures 4 to 6 are exemplary diagrams of an output interface according to one embodiment of the present invention.

[0062] Referring to FIG. 4, an output interface (400) including an incidence rate (410) by obstetric disease is illustrated.

[0063] In one embodiment, the obstetric disease prediction device can predict the incidence of two or more obstetric diseases and output an incidence rate (410) for each obstetric disease. At this time, the incidence rate (410) for each obstetric disease can be displayed as a probability on the output interface (400). Accordingly, the expectant mother can comprehensively receive the incidence rate (410) for each obstetric disease corresponding to the input data.

[0064] Referring to FIG. 5, an output interface (500) is illustrated that requests input for essential data that is not normally input to the input interface.

[0065] The aforementioned input data may include information essential for predicting the incidence of obstetric diseases. In this case, if the essential data is not input from the pregnant woman, the obstetric disease prediction device may provide an output interface (500) that requests the input of the essential data. While Fig. 5 illustrates an output interface (500) that requests the input of "weight," the essential data is not limited thereto and may not include "weight."

[0066] Referring to FIG. 6, an output interface (600) is illustrated that recommends input for optional data that is not normally input to the input interface.

[0067] The predictive model of the obstetric disease prediction device can predict the incidence of obstetric diseases even when some of the required input variables are missing. However, the accuracy of the incidence of obstetric diseases may be lower than when all required variables are included.

[0068] In one embodiment, the obstetric disease prediction device may compare the accuracy of the acquired obstetric disease incidence rate with a preset expected value to determine whether the acquired obstetric disease incidence rate is lower than the preset expected value. Furthermore, in one embodiment, if the accuracy of the acquired obstetric disease incidence rate is determined to be lower than the preset expected value, the obstetric disease prediction device may provide an output interface (600) that recommends input of optional data that is not normally input to the input interface. For example, among blood test result data, -Fetoprotien is an optional data as an input variable for predicting the incidence of gestational diabetes, but if it is not included in the input data and the accuracy of the incidence of gestational diabetes is determined to be lower than the preset expected value, the obstetric disease prediction device - An output interface (600) recommending input for fetoprotien may be provided. At this time, the obstetric disease prediction device may provide the input interface again if the user or the pregnant woman selects 'Back', and may provide the incidence of gestational diabetes with an accuracy lower than the preset expected value through the output interface if the user or the pregnant woman selects 'Ignore'. Meanwhile, the accuracy of the predicted incidence of obstetric diseases may be calculated as the performance according to the final variable of each obstetric disease prediction model. In one embodiment, the accuracy of the incidence of obstetric diseases may be calculated as the p-value of the Hosmer-Lemshaw test result. Alternatively, the obstetric disease prediction device may simulate a case where some of the final variables are not input and calculate the accuracy according to the simulation result and missing values. The performance of each obstetric disease prediction model will be described later. However, the output interface (600) regarding gestational diabetes is only one embodiment, and it is obvious that it can be applied to embodiments related to various obstetric diseases.

[0069] Figure 7 is a flowchart of a method for early prediction of the risk of obstetric diseases according to one embodiment of the present invention.

[0070] Referring to FIG. 7, in step 710, the obstetric disease prediction device can obtain input data related to the obstetric disease of the mother.

[0071] In one embodiment, the obstetric disease prediction device may provide an input interface for receiving at least one of the following data: the mother's age, height, weight, blood pressure, expected date of delivery, method of pregnancy, history of diabetes, history of hypertension, history of premature birth, family history of diabetes, family history of hypertension, number of pregnancies, year of last delivery, number of weeks of last delivery, birth weight, history of gestational diabetes, history of gestational hypertension, and blood test result.

[0072] In one embodiment, the blood test result data includes Hemoglobin, White blood cell count, Platelet Count, Total cholesterol, Triglyceride, Albumin, AST(GOT), ALT(GPT), Creatinine, HgbA1C, PAPP-A, uE3, - May include at least one of the numerical data for fetoprotein, AntiTPO, FT4 (Free thyroxine), and Inhibin-A.

[0073] In one embodiment, the obstetric disease prediction device may provide an interface for inputting alternative data for at least some of the maternal input data.

[0074] In one embodiment, the input interface may include a message recommending entry of said input data corresponding to a gestational age of 14 weeks or less.

[0075] In step 720, the obstetric disease prediction device can obtain the incidence rate of the obstetric disease by inputting input data into a prediction model corresponding to the obstetric disease.

[0076] In one embodiment, the predictive model may be a machine learning algorithm trained based on obstetric data from multiple existing mothers.

[0077] In one embodiment, the obstetric disease prediction device can select a prediction model corresponding to the obstetric disease.

[0078] In one embodiment, the obstetric disease prediction device can select a prediction model corresponding to each of two or more obstetric diseases.

[0079] In one embodiment, the variables required for input to the selected predictive model may be at least a portion of the input data.

[0080] In step 730, the obstetric disease prediction device can output the incidence of obstetric diseases of the pregnant woman.

[0081] In one embodiment, the obstetrics and gynecology disease prediction device may provide an output interface that requests input of essential data that is not normally input to the input interface.

[0082] In one embodiment, if it is determined that the accuracy of the acquired incidence of obstetric diseases is lower than a preset expected value, the obstetric disease prediction device may provide an output interface that recommends input of optional data that is not normally input to the input interface.

[0083] In one embodiment, the obstetric disease prediction device can output an incidence rate for each obstetric disease.

[0084] In one embodiment, the obstetric disease prediction device can update the prediction model by storing obstetric data of a plurality of existing pregnant women in an obstetric database in a machine learning-enabled form.

[0085] In one embodiment, the acquired input data can be stored in the above mountain and database.

[0086] Figure 8 is a conceptual diagram of a method for generating an obstetric disease prediction model according to one embodiment of the present invention.

[0087] Referring to FIG. 8, a process of generating an obstetric disease prediction model (840) from data input from multiple pregnant women by an obstetric disease prediction device is illustrated. At this time, the data input from multiple pregnant women may be constructed as a raw data set (810). In addition, the obstetric disease prediction model (840) may include a preeclampsia prediction model, a gestational diabetes prediction model, and / or a premature birth prediction model. In addition, the 'obstetric disease' described below may include preeclampsia, gestational diabetes, and / or premature birth.

[0088] In addition, the obstetric disease prediction model (840) generated by the obstetric disease prediction device may refer to the prediction model described above through FIGS. 1 to 7.

[0089] In one embodiment, the obstetric disease prediction device can generate an obstetric data set (820) by setting a plurality of initial variables related to obstetric diseases in the raw data set (810).

[0090] In one embodiment, the obstetric disease prediction device may select at least two variables related to obstetric diseases through univariate analysis of multiple variables in a raw data set (810). The raw data set (810) may include various data with a very low correlation with obstetric diseases. Accordingly, the obstetric disease prediction device may select two or more variables related to obstetric diseases, such as physical information of the pregnant woman, such as height, weight, and blood pressure, and data related to her medical history, obstetric history, and blood test results.

[0091] As an example, the results of univariate analysis on multiple variables in the raw data set (810) with respect to gestational diabetes may be as shown in Tables 1 and 2 below.

[0092]

[0093]

[0094]

[0095]

[0096]

[0097]

[0098] As another example, the results of a univariate analysis on multiple variables in the raw data set (810) with respect to preeclampsia may be as shown in Table 3 below.

[0099]

[0100]

[0101]

[0102]

[0103] As another example, the results of a univariate analysis on multiple variables in the raw data set (810) with respect to premature birth may be as shown in Table 4 below.

[0104]

[0105]

[0106]

[0107]

[0108] In one embodiment, the obstetric disease prediction device may select two or more variables with a p-value less than a predetermined value based on univariate analysis results. For example, the obstetric disease prediction device may select two or more variables with a p-value less than 0.05 based on univariate analysis results.

[0109] In one embodiment, the obstetrics and gynecology disease prediction device can determine multiple initial variables by standardizing two or more selected variables. For example, the maternal height can be in cm or m and have a value between 100 and 200 or 10 and 20, while the BMI index can be in kg / As a unit, the numerical value can have a value between 12 and 42, and in the case of disease history, it can have a value of 0 or 1. Therefore, the unit of each variable must be removed, normalized, and converted to a standardized constant before reflecting the weight of each variable according to importance. Accordingly, the obstetric disease prediction device can standardize two or more selected variables and determine the standardized variables as initial variables.

[0110] In this way, the obstetric disease prediction device can select and standardize variables from the raw data set (810) to generate an obstetric data set (820) having only a plurality of initial variables.

[0111] Meanwhile, the raw data set (810) may be a data set constructed in a prospective or retrospective cohort manner in a database management system. For example, a raw data set (810) for learning a pregnancy-induced hypertension prediction model or a premature birth prediction model may be a data set constructed in a retrospective cohort manner, and a raw data set (810) for learning a gestational diabetes prediction model may be a data set constructed in a prospective cohort manner.

[0112] In one embodiment, the obstetric disease prediction device may preprocess the obstetric data set (820) to generate a learning data set (830). Specifically, the obstetric disease prediction device may generate the learning data set (830) by dividing the obstetric data set (820) into an obstetric disease data set and a control data set based on at least one of a plurality of initial variables.

[0113] FIG. 9 is a diagram for explaining a method for preprocessing an acid and acid data set according to one embodiment of the present invention.

[0114] Referring to FIG. 9, a process in which an obstetric disease prediction device preprocesses an obstetric data set (910) to generate a learning data set is illustrated.

[0115] For example, a process of generating a learning data set for training a prediction model for pregnancy-induced hypertension by preprocessing an obstetric data set (910) on pregnancy-induced hypertension is described.

[0116] In one embodiment, the obstetric disease prediction device can separate the obstetric data set (910) into a preeclampsia data set (930) and a control data set (911, 921) based on blood pressure values ​​and proteinuria values ​​among a plurality of initial variables.

[0117] The obstetric disease prediction device can generate a sub-data set (920) by excluding data (911) in which blood pressure values ​​satisfy the first blood pressure condition but urine test values ​​do not satisfy the first proteinuria condition from the obstetric data set (910). At this time, the first blood pressure condition may be systolic blood pressure of 140 mmHG or higher or diastolic blood pressure of 90 mmHG or higher, and the first proteinuria condition may be 24-hour urine of 300 mg or higher or dipstick 1+ or higher.

[0118] The obstetric disease prediction device may determine data (921) that does not satisfy at least one of the second blood pressure condition, the first disease condition, the second disease condition, and the newborn condition in the sub data set (920) as the pregnancy toxemia data set (930). At this time, the second blood pressure condition is systolic blood pressure of 160 mmHG or higher or diastolic blood pressure of 90 mmHG or higher, the first disease condition is pulmonary edema, newly developed headache, and visual impairment even without proteinuria according to the ISSHP standard, and the second disease condition is thrombocytopenia (platelet count 150,000 / L), renal failure (serum creatinine 1 mg / dL or higher), liver dysfunction (AST and ALT 40 U / L or higher), and the fetus may be SGA (small for gestational age).

[0119] At this time, the obstetric disease prediction device can determine data not included in the pregnancy-induced hypertension data set (930) from the obstetric data set (910) as the control data set (911, 921). In this way, the obstetric disease prediction device can separate the obstetric data set (910) into the pregnancy-induced hypertension data set (930) and the control data set (911, 921) based on the blood pressure level and the proteinuria level. The separated pregnancy-induced hypertension data set (930) and the control data set (911, 921) can be used as data for learning the pregnancy-induced hypertension prediction model described later as a learning data set (930).

[0120] As another example, a process of an obstetric disease prediction device preprocessing an obstetric data set (910) regarding gestational diabetes to generate a learning data set (930) for training a gestational diabetes prediction model is described.

[0121] In one embodiment, the obstetrics and gynecology disease prediction device can separate the obstetrics and gynecology data set (910) into a gestational diabetes data set (930) and a control data set (911, 921) based on a proteinuria level among a plurality of initial variables.

[0122] The obstetric disease prediction device can generate a sub-data set (920) from the remaining data, excluding the first criterion-deficient data (911) from the obstetric data set (910). As an example, the first criterion-deficient data may be foreign mother data, data with missing important variables, withdrawn data, etc. In this case, the obstetric data set (910) may be generated from a raw data set constructed in a prospective cohort manner, and the sub-data set (920) may be composed of data corresponding to the first trimester of pregnancy, the second trimester of pregnancy, the third trimester of pregnancy, the last trimester of pregnancy, and 4 to 6 weeks after birth.

[0123] The obstetric disease prediction device may determine the remaining data, excluding the second standard below-standard data (921) from the sub-data set (920), as the gestational diabetes data set (930). At this time, the second standard below-standard data (921) may be transfer data, tracking failure data, twin abnormality data, miscarriage data, and overt diabetes data. The tracking failure data may refer to data that failed to be tracked up to 4 to 6 weeks after birth, and the overt diabetes data may refer to data corresponding to general diabetes (diabetes mellitus) rather than gestational diabetes mellitus.

[0124] At this time, the obstetric disease prediction device can determine data not included in the gestational diabetes data set (930) in the obstetric data set (910) as the control data set (911, 921). In this way, the obstetric disease prediction device can separate the obstetric data set (910) into the gestational diabetes data set (930) and the control data set (911, 921) based on the blood pressure level and the proteinuria level. The separated gestational diabetes data set (930) and the control data set (911, 921) can be used as data for learning the gestational diabetes prediction model described later as the learning data set (930).

[0125] As another example, a process of an obstetric disease prediction device preprocessing an obstetric data set (910) regarding premature birth to generate a learning data set (930) for training a premature birth prediction model is described.

[0126] In one embodiment, the obstetric disease prediction device may separate the obstetric data set (910) into a premature birth data set (930) and a control data set (911, 921) based on the timing of delivery among a plurality of initial variables.

[0127] The obstetric disease prediction device can generate a sub-data set (920) by excluding data (911) that do not satisfy the delivery timing condition from the obstetric data set (910). At this time, the delivery timing condition may be delivery between 20 and 37 weeks, and the condition for determining the delivery timing may be one or more of dew, rupture of water, and contraction. That is, if any one of dew, rupture of water, and contraction occurs between 20 and 37 weeks, it can be considered that the delivery timing condition is satisfied.

[0128] The obstetrics and gynecology disease prediction device may determine data (921) that does not satisfy the natural premature birth condition from the sub-data set (920) as a premature birth data set (930). In this case, the data (921) that does not satisfy the natural premature birth condition may be data corresponding to medical premature birth. In other words, premature birth data due to medical treatment may be excluded from the premature birth data set (930).

[0129] At this time, the obstetric disease prediction device can determine data not included in the premature birth data set (930) from the obstetric data set (910) as the control data set (911, 921). In this way, the obstetric disease prediction device can separate the obstetric data set (910) into the premature birth data set (930) and the control data set (911, 921) based on the timing of delivery. The separated premature birth data set (930) and the control data set (911, 921) can be utilized as data for learning the premature birth prediction model described later as the learning data set (930).

[0130] Returning to FIG. 8, in one embodiment, the obstetric disease prediction device may generate an obstetric disease prediction model (840) based on a learning data set (830). At this time, the obstetric disease prediction model (840) may be a model that receives data regarding a pregnant woman (841) and outputs an obstetric disease incidence rate (842) of the pregnant woman (841).

[0131] In one embodiment, the obstetric disease prediction device may determine final variables to be applied to the obstetric disease prediction model (840) through univariate analysis of multiple initial variables of the learning data set (830). In this case, the obstetric disease prediction model (840) may be a logistic regression model.

[0132] In one embodiment, the obstetrics and gynecology disease prediction device can determine some of the initial variables as final variables through a backward stepwise elimination method. Stepwise elimination is a method of finding an optimal model by eliminating statistically insignificant variables one by one from multiple initial variables.

[0133] In one embodiment, the final variable determined may vary by acid and disease.

[0134] As an example, in the case of preeclampsia, the final variable may include at least one of systolic blood pressure, diastolic blood pressure, mean arterial pressure (MAP), body mass index (BMI), age, weight, history of pregnancy-induced hypertension, history of hypertension, previous childbirth, method of conception (method of conception), family history of hypertension, time interval from the last birth (interpregnancy interval), and blood test result data. In this case, the blood test result data may include inhibin-A, PAPP-A, HgbA1C, platelet count, unconjugated estriol (uE3), creatinine, - May include at least one of fetoprotein, glucose, hematocrit, white blood cell count (WBC), total cholesterol, total B-hCG, triglyceride, HgbA1C, ALT (GPT), hemoglobin, albumin, ALP, and T.bilirubin.

[0135] As another example, for gestational diabetes, the final variable may include at least one of systolic blood pressure, diastolic blood pressure, mean arterial pressure (MAP), age, height, weight, alcohol consumption, history of diabetes, family history of diabetes, history of hypertension, family history of hypertension, history of gestational hypertension, baby weight at previous birth, body mass index (BMI), history of gestational diabetes, interpregnancy interval, conception method (including in vitro fertilization), and blood test results. At this time, the blood test result data may include at least one of white blood cell count (WBC), creatinine, PAPP-A, total B-hCG, unconjugated estriol (uE3), inhibin-A, white blood cell count (WBC), hemoglobin, hematocrit, platelet count, creatinine, glucose, HgbA1C, T.bilirubin, AST (GOT), ALT (GPT), total cholesterol, triglyceride, and total cholesterol.

[0136] As another example, in the case of preterm birth, the final variable may include at least one of systolic blood pressure, diastolic blood pressure, mean arterial pressure (MAP), age, weight, height, body mass index (BMI), history of hypertension, history of alcohol consumption, history of diabetes, family history of hypertension, family history of diabetes, history of pregnancy-induced hypertension, method of conception (including in vitro fertilization), previous childbirth, baby weight at previous birth, and blood test results. In this case, the blood test results data may include: - It may include at least one of fetoprotein, PAPP-A, albumin, HgbA1C, triglyceride, unconjugated estriol (uE3), free thyroxine (FT4), total cholesterol, hemoglobin, anti-TPO, T.Protein, inhibin-A, glucose, ALT (GPT), hematocrit, and white blood cell count (WBC).

[0137] In one embodiment, the obstetrics and gynecology disease prediction device may weight the final variable based on the univariate analysis results, p-value.

[0138] According to Tables 1 to 3 above, the p-value of each variable for obstetric diseases is shown in the univariate analysis results. The obstetric disease prediction device can assign a higher weight to the final variable as the p-value is smaller. The weights of the final variables may refer to the parameter weights of the machine learning algorithm of the obstetric disease prediction model. As an example, the weight of age (p-value <0.001) among the final variables of the machine learning algorithm of the preeclampsia prediction model (840) may be higher than the weight of the current pregnancy method (p-value 0.015). The same example according to the p-value may also be applied to the gestational diabetes prediction model (840) and the premature birth prediction model (840).

[0139] In one embodiment, the obstetric disease prediction device may generate a preeclampsia prediction model (840) based on a learning data set (830). Similarly, the obstetric disease prediction device may generate a gestational diabetes prediction model (840) and / or a premature birth prediction model (840) based on the learning data set (830). In this case, each obstetric disease prediction model (840) may be a model that receives data regarding a pregnant woman (841) and outputs a preeclampsia incidence rate (842), a gestational diabetes incidence rate (842), and a premature birth incidence rate (842), respectively.

[0140] In one embodiment, the obstetric disease prediction device can obtain the incidence of preeclampsia (842) from data (841) about the pregnant woman by utilizing the preeclampsia prediction model (840) described above. Specifically, the obstetric disease prediction device can input data (841) about the pregnant woman into the preeclampsia prediction model (840) generated based on the learning data set (830). In addition, the obstetric disease prediction device can obtain the incidence of preeclampsia (842) from the preeclampsia prediction model (840).

[0141] Likewise, in one embodiment, the obstetric disease prediction device can obtain the incidence of gestational diabetes (842) from data (841) about the pregnant woman by utilizing the gestational diabetes prediction model (840) described above. Specifically, the obstetric disease prediction device can input data (841) about the pregnant woman into the gestational diabetes prediction model (840) generated based on the learning data set (830). In addition, the obstetric disease prediction device can obtain the incidence of gestational diabetes (842) from the gestational diabetes prediction model (840).

[0142] Likewise, in one embodiment, the obstetric disease prediction device can obtain the premature birth incidence rate (842) from data (841) about the mother by utilizing the premature birth prediction model (840) described above. Specifically, the obstetric disease prediction device can input data (841) about the mother into the premature birth prediction model (840) generated based on the learning data set (830). In addition, the obstetric disease prediction device can obtain the premature birth incidence rate (842) from the premature birth prediction model (840).

[0143] In one embodiment, the obstetric disease prediction device can verify an obstetric disease prediction model. Specifically, in one embodiment, the obstetric disease prediction device divides the output obstetric disease incidence rate by final variable into multiple intervals, maps the actual obstetric disease incidence rate in each of the divided intervals, and performs a statistical significance test between the output obstetric disease incidence rate and the actual obstetric disease incidence rate. As an example, the statistical significance test may be the Hosmer-Lemeshow test.

[0144] FIG. 10 and FIG. 11 are graphs of the performance evaluation results and the Hosmer-Lemshaw test results of a pregnancy-induced hypertension prediction model according to one embodiment of the present invention.

[0145] In one embodiment, the obstetric disease prediction device can verify a prediction model for preeclampsia. Specifically, in one embodiment, the obstetric disease prediction device divides the outputted preeclampsia incidence rate by final variable into multiple intervals, maps the actual incidence rate of preeclampsia in each of the divided intervals, and performs a statistical significance test between the outputted preeclampsia incidence rate and the actual incidence rate of preeclampsia.

[0146] Referring to Figure 10, we can see the results of evaluating the performance of a pregnancy-induced hypertension prediction model by splitting the training and test data in a 9:1 ratio. The AUC (Area Under Receiver Operating Characteristic) measurement was 0.915. Meanwhile, among the 3,124 total test samples, only 998 samples with missing values ​​in two or fewer variables were included in the test.

[0147] Referring to Figure 11, the Hosmer-Remshaw test results confirm that the predicted incidence of preeclampsia by the preeclampsia prediction model is similar to the actual incidence of preeclampsia. The p-value is 0.14, which does not reject the null hypothesis that the predicted and actual incidence rates of preeclampsia are identical. Meanwhile, only those samples with missing values ​​in two or fewer variables were included in the test.

[0148] Table 5 below shows the results of mapping the actual incidence of preeclampsia in each interval of the segmented incidence of preeclampsia.

[0149]

[0150]

[0151]

[0152] Figures 12 to 14 are graphs of the performance evaluation results and the Hosmer-Lemshaw test results of a gestational diabetes prediction model according to one embodiment of the present invention. In one embodiment, the obstetric disease prediction device can verify the gestational diabetes prediction model. Specifically, in one embodiment, the obstetric disease prediction device can divide the gestational diabetes incidence rate output by final variable into a plurality of sections, map the actual gestational diabetes incidence rate in each of the divided sections, and perform a statistical significance test between the output gestational diabetes incidence rate and the actual gestational diabetes incidence rate.

[0153] Figure 12 illustrates the process of generating training and test data for 5-fold cross-validation, and Figure 13 illustrates the results of evaluating the performance of a gestational diabetes prediction model through 5-fold cross-validation. The AUC (Area Under Receiver Operating Characteristic) measurement is 0.757. Since the difference due to changes in the learning or training data is very small, it can be confirmed that the generated gestational diabetes model has stable performance.

[0154] Referring to Figure 14, the Hosmer-Remshaw test results confirm that the predicted incidence of gestational diabetes by the gestational diabetes prediction model is similar to the actual incidence of gestational diabetes. The p-value is 0.698, failing to reject the null hypothesis that the predicted incidence of gestational diabetes is identical to the actual incidence.

[0155] Table 6 below shows the results of mapping the actual incidence of gestational diabetes in each interval of the divided gestational diabetes incidence rate.

[0156]

[0157]

[0158]

[0159] Figures 15 and 16 are graphs of the performance evaluation results and the Hosmer-Lemshaw test results of a preterm birth prediction model according to one embodiment of the present invention. In one embodiment, the obstetric disease prediction device can verify the preterm birth prediction model. Specifically, in one embodiment, the obstetric disease prediction device can divide the preterm birth incidence rate output by final variable into multiple sections, map the actual preterm birth incidence rate in each of the divided sections, and perform a statistical significance test between the output preterm birth incidence rate and the actual preterm birth incidence rate.

[0160] Referring to Figure 15, we can see the results of evaluating the performance of a preterm birth prediction model by splitting the training and test data in a 9:1 ratio. The AUC (Area Under Receiver Operating Characteristic) measurement was 0.767. Meanwhile, among the 3,063 total test samples, only 185 samples with missing values ​​in two or fewer variables were included in the test.

[0161] Referring to Figure 16, the Hosmer-Remshaw test results confirm that the predicted incidence of preterm birth by the preterm birth prediction model is similar to the actual incidence of preterm birth. The p-value is 0.146, which does not reject the null hypothesis that the predicted and actual incidence rates are identical. Meanwhile, only those samples with missing values ​​in two or fewer variables were included in the test.

[0162] Table 7 below shows the results of mapping the actual incidence of preterm birth in each segment of the preterm birth incidence rate.

[0163]

[0164]

[0165]

[0166] FIG. 17 is a flowchart of a method for predicting the risk of preeclampsia using an artificial intelligence model according to one embodiment of the present invention. Referring to FIG. 17, in step 1710, an obstetric disease prediction device can input data about the pregnant woman into a preeclampsia prediction model generated based on a learning data set.

[0167] In one embodiment, the data about the mother may include a final variable determined from a plurality of initial variables of the learning data set.

[0168] In one embodiment, the final variable may include at least one of systolic blood pressure, diastolic blood pressure, mean arterial pressure (MAP), body mass index (BMI), age, weight, history of pregnancy-induced hypertension, history of hypertension, whether or not a child has been born, method of conception (method of conception), family history of hypertension, time interval from the last birth date (interpregnancy interval), and blood test result data. In this case, the blood test result data may include inhibin-A, PAPP-A, HgbA1C, platelet count, unconjugated estriol (uE3), creatinine, - It may include at least one of fetoprotein, glucose, hematocrit, white blood cell count (WBC), total cholesterol, total B-hCG, triglyceride, HgbA1C, ALT (GPT), hemoglobin, albumin, ALP, and T. bilirubin. In one embodiment, the obstetric disease prediction device may generate a learning data set for learning a preeclampsia prediction model before inputting data about the pregnant woman.

[0169] In one embodiment, the obstetric disease prediction device can generate an obstetric data set by setting a plurality of initial variables related to preeclampsia in the raw data set.

[0170] In one embodiment, the obstetric disease prediction device can select at least two variables associated with preeclampsia through univariate analysis of multiple variables of a raw data set.

[0171] In one embodiment, the obstetrics and gynecology disease prediction device can determine a plurality of initial variables by standardizing two or more selected variables.

[0172] In one embodiment, the obstetric disease prediction device can preprocess an obstetric data set to generate a learning data set.

[0173] In one embodiment, the obstetric disease prediction device can separate the obstetric data set into a preeclampsia data set and a control data set based on blood pressure levels and proteinuria levels among a plurality of initial variables.

[0174] In one embodiment, the obstetric disease prediction device may determine the final variable to be applied to the preeclampsia prediction model through stepwise backward elimination of multiple initial variables in the learning data set. That is, in one embodiment, the final variable may be determined through stepwise backward elimination of multiple initial variables in the learning data set.

[0175] In one embodiment, the obstetrics and gynecology disease prediction device may weight the final variable based on the univariate analysis result, p-value.

[0176] In step 1720, the obstetric disease prediction device can obtain the incidence of preeclampsia from the preeclampsia prediction model.

[0177] In one embodiment, the pregnancy toxemia prediction model may be a logistic regression model.

[0178] In one embodiment, the obstetric disease prediction device can verify a pregnancy toxemia prediction model.

[0179] In one embodiment, the obstetric disease prediction device can divide the incidence of preeclampsia output by final variable into multiple intervals.

[0180] In one embodiment, the obstetric disease prediction device can perform a statistical significance test between the outputted incidence of preeclampsia and the actual incidence of preeclampsia by mapping the actual incidence of preeclampsia in each of the segmented sections.

[0181] In one embodiment, the statistical significance test may be a hosmer-lemeshow test.

[0182] Figure 18 is a flowchart of a method for predicting the risk of gestational diabetes using an artificial intelligence model according to one embodiment of the present invention.

[0183] Referring to FIG. 18, in step 1810, the obstetric disease prediction device can input data about the pregnant woman into a gestational diabetes prediction model generated based on a learning data set.

[0184] In one embodiment, the data about the mother may include a final variable determined from a plurality of initial variables of the learning data set.

[0185] In one embodiment, the final variable may include at least one of systolic blood pressure, diastolic blood pressure, mean arterial pressure (MAP), age, height, weight, alcohol consumption, diabetes, family history of diabetes, hypertension, family history of hypertension, history of gestational hypertension, weight of a baby at a previous birth, body mass index (BMI), history of gestational diabetes, interpregnancy interval from the last birth date, conception method (including in vitro fertilization), and blood test result data. In this case, the blood test result data may include at least one of white blood cell count (WBC), creatinine, PAPP-A, total B-hCG, unconjugated estriol (uE3), inhibin-A, white blood cell count (WBC), hemoglobin, hematocrit, platelet count, creatinine, glucose, HgbA1C, T. bilirubin, AST(GOT), ALT(GPT), total cholesterol, triglyceride, and total cholesterol. In one embodiment, the obstetric disease prediction device can generate a learning data set for learning a gestational diabetes prediction model before inputting data about the pregnant woman.

[0186] In one embodiment, the obstetric disease prediction device can generate an obstetric data set by setting a plurality of initial variables related to gestational diabetes in a raw data set.

[0187] In one embodiment, the obstetric disease prediction device can select at least two variables related to gestational diabetes through univariate analysis of multiple variables of a raw data set.

[0188] In one embodiment, the obstetrics and gynecology disease prediction device can determine a plurality of initial variables by standardizing two or more selected variables.

[0189] In one embodiment, the obstetric disease prediction device can preprocess an obstetric data set to generate a learning data set.

[0190] In one embodiment, the obstetric disease prediction device can separate an obstetric data set into a gestational diabetes data set and a control data set based on a proteinuria level among a plurality of initial variables.

[0191] In one embodiment, the obstetrics and gynecology disease prediction device may determine the final variable to be applied to the gestational diabetes prediction model through stepwise backward elimination of multiple initial variables in the learning data set. That is, in one embodiment, the final variable may be determined through stepwise backward elimination of multiple initial variables in the learning data set.

[0192] In one embodiment, the obstetrics and gynecology disease prediction device may weight the final variable based on the univariate analysis result, p-value.

[0193] In step 1820, the obstetric disease prediction device can obtain the incidence of gestational diabetes from the gestational diabetes prediction model.

[0194] In one embodiment, the gestational diabetes prediction model may be a logistic regression model.

[0195] In one embodiment, the obstetric disease prediction device can verify a gestational diabetes prediction model.

[0196] In one embodiment, the obstetrics and gynecology disease prediction device can divide the gestational diabetes incidence rate output by final variable into multiple intervals.

[0197] In one embodiment, the obstetrics and gynecology disease prediction device can perform a statistical significance test between the output gestational diabetes incidence rate and the actual gestational diabetes incidence rate by mapping the actual gestational diabetes incidence rate in each segmented section.

[0198] In one embodiment, the statistical significance test may be a hosmer-lemeshow test.

[0199] Figure 19 is a flowchart of a method for predicting the risk of premature birth using an artificial intelligence model according to one embodiment of the present invention.

[0200] Referring to FIG. 19, in step 1910, the obstetric disease prediction device can input data about the mother into a premature birth prediction model generated based on a learning data set.

[0201] In one embodiment, the data about the mother may include a final variable determined from a plurality of initial variables of the learning data set.

[0202] In one embodiment, the final variable may include at least one of systolic blood pressure, diastolic blood pressure, mean arterial pressure (MAP), age, weight, height, body mass index (BMI), history of hypertension, history of drinking, history of diabetes, family history of hypertension, family history of diabetes, history of pregnancy-induced hypertension, conception method (including in vitro fertilization), whether or not there was childbirth, weight of a baby in a previous birth, and blood test result data. In this case, the blood test result data may include: - It may include at least one of fetoprotein, PAPP-A, albumin, HgbA1C, triglyceride, unconjugated estriol (uE3), free thyroxine (FT4), total cholesterol, hemoglobin, anti-TPO, T.Protein, inhibin-A, glucose, ALT (GPT), Hematocrit, and white blood cell count (WBC). In one embodiment, the obstetric disease prediction device may generate a learning data set for learning a premature birth prediction model before inputting data about the pregnant woman.

[0203] In one embodiment, the obstetric disease prediction device can generate an obstetric data set by setting a plurality of initial variables related to premature birth in the raw data set.

[0204] In one embodiment, the obstetrics and gynecology disease prediction device can select at least two variables associated with premature birth through univariate analysis of multiple variables of a raw data set.

[0205] In one embodiment, the obstetrics and gynecology disease prediction device can determine a plurality of initial variables by standardizing two or more selected variables.

[0206] In one embodiment, the obstetric disease prediction device can preprocess an obstetric data set to generate a learning data set.

[0207] In one embodiment, the obstetric disease prediction device can separate an obstetric data set into a preterm birth data set and a control data set based on the timing of delivery among a plurality of initial variables.

[0208] In one embodiment, the obstetrics and gynecology disease prediction device may determine final variables to be applied to the preterm birth prediction model through stepwise backward elimination of multiple initial variables in the learning data set. That is, in one embodiment, the final variables may be determined through stepwise backward elimination of multiple initial variables in the learning data set.

[0209] In one embodiment, the obstetrics and gynecology disease prediction device may weight the final variable based on the univariate analysis result, p-value.

[0210] In step 1920, the obstetric disease prediction device can obtain the incidence of premature birth from the premature birth prediction model.

[0211] In one embodiment, the premature birth prediction model may be a logistic regression model.

[0212] In one embodiment, the obstetrics and gynecology prediction device can verify a premature birth prediction model.

[0213] In one embodiment, the obstetrics and gynecology disease prediction device can divide the output premature birth incidence rate by final variable into multiple intervals.

[0214] In one embodiment, the obstetrics and gynecology disease prediction device can perform a statistical significance test between the output premature birth incidence rate and the actual premature birth incidence rate by mapping the actual premature birth incidence rate in each segmented interval.

[0215] In one embodiment, the statistical significance test may be a hosmer-lemeshow test.

[0216] Figure 20 is a block diagram of an obstetric disease prediction device according to one embodiment of the present invention.

[0217] Referring to FIG. 20, an obstetric disease prediction device (hereinafter referred to as "device") (2000) may include a communication unit (2010), a processor (2020), and a database (2030). Only components related to the embodiment are illustrated in the device (2000) of FIG. 20. Therefore, those skilled in the art will understand that other general components may be included in addition to the components illustrated in FIG. 20.

[0218] The communication unit (2010) may include one or more components that enable wired / wireless communication with an external server or external device. For example, the communication unit (2010) may include at least one of a short-range communication unit (not shown), a mobile communication unit (not shown), and a broadcast receiving unit (not shown).

[0219] DB (2030) is hardware that stores various data processed within the device (2000), and can store a program for processing and controlling the processor (2020).

[0220] DB (2030) may include random access memory (RAM) such as dynamic random access memory (DRAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), CD-ROM, Blu-ray or other optical disk storage, hard disk drive (HDD), solid state drive (SSD), or flash memory.

[0221] The processor (2020) controls the overall operation of the device (2000). For example, the processor (2020) can control the input unit (not shown), the display (not shown), the communication unit (2010), the DB (2030), etc., by executing programs stored in the DB (2030). The processor (2020) can control the operation of the device (2000) by executing programs stored in the DB (2030).

[0222] The processor (2020) can control at least some of the operations of the devices described above in FIGS. 1 to 19.

[0223] The processor (2020) may be implemented using at least one of application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors, and other electrical units for performing functions.

[0224] In one embodiment, the device (2000) may be a mobile electronic device. For example, the device (2000) may be implemented as a smartphone, tablet PC, PC, smart TV, personal digital assistant (PDA), laptop, media player, navigation device, camera-equipped device, or other mobile electronic device. Furthermore, the device (2000) may be implemented as a wearable device, such as a watch, glasses, hair band, or ring, equipped with communication and data processing capabilities.

[0225] In another embodiment, the device (2000) may be a server. The server may be implemented as a computer device or multiple computer devices that communicate over a network to provide commands, codes, files, content, services, etc. The server may receive data from an external source for early prediction of obstetric diseases or for generating an obstetric disease prediction model, and may predict obstetric diseases or generate an obstetric disease prediction model based on the received data. As an example, the server may be a web server and provide a web interface for a service for early prediction of obstetric diseases.

[0226] Embodiments according to the present invention may be implemented in the form of a computer program that can be executed through various components on a computer, and such a computer program may be recorded on a computer-readable medium. In this case, the medium may include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical recording media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specifically configured to store and execute program instructions, such as ROMs, RAMs, and flash memories.

[0227] Meanwhile, the computer program may be specifically designed and constructed for the present invention, or may be one known and available to those skilled in the computer software field. Examples of computer programs may include not only machine language code, such as that generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

[0228] According to one embodiment, the method according to various embodiments of the present disclosure may be provided as included in a computer program product. The computer program product may be traded as a product between a seller and a buyer. The computer program product may be distributed in the form of a machine-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or may be distributed online (e.g., downloaded or uploaded) through an application store (e.g., Play Store™) or directly between two user devices. In the case of online distribution, at least a portion of the computer program product may be temporarily stored or temporarily generated in a machine-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0229] Unless the steps constituting the method according to the present invention are explicitly described in a specific order or are otherwise described in a different order, the steps may be performed in any appropriate order. The present invention is not necessarily limited to the order in which the steps are described. The use of all examples or exemplary terms in the present invention is merely intended to illustrate the present invention in detail, and the scope of the present invention is not limited by the examples or exemplary terms unless otherwise defined by the claims. Furthermore, those skilled in the art will appreciate that various modifications, combinations, and variations can be configured according to design conditions and factors within the scope of the appended claims or their equivalents.

[0230] Therefore, the idea of ​​the present invention should not be limited to the embodiments described above, and not only the scope of the patent claims described below but also all scopes equivalent to or equivalently modified from the scope of the patent claims are considered to fall within the scope of the idea of ​​the present invention.

Claims

1. A step of inputting data about the pregnant woman into a pregnancy-induced hypertension prediction model generated based on a learning data set; and A step of obtaining the incidence of pregnancy-induced hypertension from the above pregnancy-induced hypertension prediction model; including; The data on the above mothers are: A method for predicting the risk of preeclampsia using an artificial intelligence model, the method comprising: determining a final variable from multiple initial variables of the above learning data set; 2. In paragraph 1, The final variable above is, A method for predicting the risk of preeclampsia using an artificial intelligence model, which includes at least one of systolic blood pressure, diastolic blood pressure, mean arterial pressure (MAP), body mass index (BMI), age, weight, history of gestational hypertension, history of hypertension, whether or not a woman has given birth, method of conception (method of conception), family history of hypertension, time interval from the last birth date (interpregnancy interval), and blood test result data.

3. In paragraph 1, The above blood test result data is, inhibin-A, PAPP-A, HgbA1C, platelet count, unconjugated estriol(uE3), creatinine, - A method for predicting the risk of preeclampsia using an artificial intelligence model, including at least one of fetoprotein, glucose, hematocrit, white blood cell count (WBC), total cholesterol, total B-hCG, triglyceride, HgbA1C, ALT (GPT), hemoglobin, albumin, ALP, and T.bilirubin.

4. In paragraph 1, further comprising a step of generating the above learning data set; The steps for generating the above learning data set are: A step of creating an obstetric data set by setting the above multiple initial variables related to pregnancy-induced hypertension in the raw data set; and A method for predicting the risk of preeclampsia using an artificial intelligence model, comprising: a step of preprocessing the above-mentioned obstetric data set to create a learning data set; 5. In paragraph 4, The steps for creating the above mountain and data set are: A step of selecting at least two variables related to preeclampsia through univariate analysis of multiple variables of a raw data set; and A method for predicting the risk of preeclampsia using an artificial intelligence model, comprising: a step of determining the plurality of initial variables by standardizing the two or more selected variables; 6. In paragraph 4, The steps for generating the above learning data set are: A method for predicting the risk of preeclampsia using an artificial intelligence model, comprising: a step of dividing the obstetric data set into a preeclampsia data set and a control data set based on blood pressure levels and proteinuria levels among the plurality of initial variables; 7. In paragraph 1, The final variable above is, A method for predicting the risk of preeclampsia using an artificial intelligence model, wherein the method is determined through backward stepwise elimination of the plurality of initial variables of the above learning data set.

8. In paragraph 5, A method for predicting the risk of preeclampsia using an artificial intelligence model, further comprising a step of assigning weights to the final variable based on the p-value of the univariate analysis results.

9. In paragraph 1, The above pregnancy toxemia prediction model is, A method for predicting the risk of preeclampsia using an artificial intelligence model, the logistic regression model.

10. In paragraph 1, A step of verifying the above pregnancy-induced hypertension prediction model; The above verification steps are: A step of dividing the outputted incidence of pregnancy-induced hypertension into multiple sections for each of the final variables; and A method for predicting the risk of preeclampsia using an artificial intelligence model, comprising: a step of mapping the actual incidence rate of preeclampsia in each of the above-described segments and performing a statistical significance test between the output incidence rate of preeclampsia and the actual incidence rate of preeclampsia.

11. In paragraph 10, The above statistical significance test is, A method for predicting the risk of preeclampsia using an artificial intelligence model, the Hosmer-Lemeshow test.

12. Memory in which at least one program is stored; and A processor that performs an operation by executing at least one program; The above processor, Input data about the mother into the pregnancy-induced hypertension prediction model generated based on the learning data set, Obtain the incidence of pregnancy-induced hypertension from the above pregnancy-induced hypertension prediction model, The data on the above mothers are: A device for predicting the risk of preeclampsia using an artificial intelligence model, the device including a final variable determined from multiple initial variables of the above learning data set.

13. A computer-readable recording medium recording a program for executing the method of Article 1 on a computer.

Images