System and method for predicting premature child delivery on basis of machine-learning utilizing clinical index and dental index

Abstract

Disclosed are a system and method for predicting premature child delivery, which predict the probability of premature child delivery on the basis of a machine-learning analysis model by utilizing clinical indices and periodontal indices of prenatal mothers. A device for predicting a premature child delivery by a mother according to the present invention comprises: a data input unit for receiving clinical index data and dental index data of a mother; and a premature child delivery prediction unit for predicting a premature child delivery by the mother by means of a machine-learning model on the basis of the clinical index data and the dental index data, wherein the machine-learning model has been trained to predict premature child delivery on the basis of clinical index data collected for mothers and periodontal index data collected through dental examination and image capturing of the mothers.

Description

A machine learning-based system and method for predicting premature delivery using clinical and dental indicators.

[0001] The present invention relates to a system and method for predicting premature birth based on machine learning utilizing clinical and dental indicators, and more particularly, to a system and method for predicting premature birth that predicts the probability of premature birth using clinical and periodontal indicators of a pregnant woman before birth based on a machine learning analysis model.

[0002] The incidence of premature birth is increasing worldwide, making it crucial to predict premature births before birth, establish preventive measures, and develop postpartum care plans. Risk factors for premature birth can be broadly categorized into two groups. First, maternal / environmental factors, including older age, low socioeconomic status, smoking, drinking, drugs, depression, nutrition, and body mass index. Second, medical / obstetric factors, including previous premature births, multiple pregnancies, short cervical length, urinary tract infections, diabetes, hypertension, sacral edema, premature rupture of membranes, and fetal malformations.

[0003] Despite various risk factors, the accuracy of models predicting prematurity remains low, and the predictive value for prematurity remains low. This is because existing models primarily assess maternal clinical factors. While risk factor assessment, cervical measurement, and biochemical biomarkers are used to predict prematurity, studies present varying criteria and predictive values, and no generalized model exists. While some previous studies have suggested a link between maternal periodontitis and prematurity, most studies are retrospective, resulting in limited accuracy and unclear mechanisms.

[0004] This invention was derived from research conducted as part of the Creative Challenge Research Foundation Support Project of the National Research Foundation of Korea (Project Unique Number: 1345364829, Subproject Number: 2020R1I1A1A01073697, Research Project Title: Analysis of the Relationship Between Maternal Oral Bacterial Imbalance and Long-Term Neonatal Prognosis, Host Organization: Korea University Industry-Academic Cooperation Foundation, Research Period: 2020.06.01 ~ 2023.05.31).

[0005] Meanwhile, the Korean government, which is the subject of the task, has no property interest in any aspect of the present invention.

[0006] The present invention provides a system and method for predicting premature birth by utilizing clinical and periodontal indicators of a pregnant woman before birth based on a machine learning analysis model.

[0007] A device for predicting premature delivery of a mother according to one aspect of the present invention includes: a data input unit for receiving clinical indicator data and dental indicator data of the mother; and a premature delivery prediction unit for predicting premature delivery of the mother by a machine learning model based on the clinical indicator data and the dental indicator data, the machine learning model being trained to predict premature delivery based on clinical indicator data collected for the mothers and periodontal indicator data collected through dental examinations and imaging of the mothers.

[0008] The above clinical indicator data may include at least two of maternal age, body mass index before pregnancy, whether or not artificial insemination was performed, history of previous premature birth, preeclampsia, chronic hypertension, gestational diabetes, premature rupture of membranes, and stage of chorioamnionitis.

[0009] The above dental indicator data may include at least two of periodontitis stage, altered gingival index, plaque index, and caries experience permanent tooth index.

[0010] The above machine learning model may be a model trained to predict variable importance of the clinical indicator data and the dental indicator data based on a random forest model.

[0011] According to another aspect of the present invention, a method for predicting premature delivery of a mother includes the steps of: receiving clinical indicator data and dental indicator data of the mother by a data input unit; and predicting premature delivery of the mother by a machine learning model based on the clinical indicator data and the dental indicator data by a premature delivery prediction unit, wherein the machine learning model is trained to predict premature delivery based on clinical indicator data collected for the mothers and periodontal indicator data collected through dental examinations and imaging of the mothers.

[0012] The step of predicting the premature delivery of the mother may include a step of predicting the probability of premature birth by determining the importance ranking of variables of pre-pregnancy body mass index, modified gingival index, pre-eclampsia, caries-experienced permanent tooth index, artificial insemination, history of previous premature birth, maternal age, chronic hypertension, gestational diabetes, premature rupture of membranes, and chorioamnionitis stage in relation to premature birth by the random forest model.

[0013] According to another aspect of the present invention, a computer program recorded on a computer-readable recording medium is provided to execute the method for predicting premature delivery of a mother.

[0014] According to an embodiment of the present invention, a system and method for predicting premature birth are provided, which predicts the probability of premature birth by utilizing clinical indicators and periodontal indicators of a pregnant woman based on a machine learning analysis model.

[0015] In addition, according to an embodiment of the present invention, by adding an indicator related to the maternal periodontal condition in addition to the maternal side as a predictive indicator and introducing a machine learning model based on the same, it is possible to predict premature birth with high predictive accuracy.

[0016] In addition, according to an embodiment of the present invention, it is expected that the accuracy of predicting premature birth will be increased, and that individualized management and prevention strategies will be applied to mothers expected to have premature births, thereby reducing premature births.

[0017] Figure 1 is a configuration diagram of a device for predicting premature delivery of a mother according to an embodiment of the present invention.

[0018] Figure 2 is a flowchart of a method for predicting premature delivery of a mother according to an embodiment of the present invention.

[0019] Figure 3 shows the results of predicting the probability of premature birth (total premature birth, natural premature birth) and full-term birth using various machine learning models.

[0020] Figure 4 shows the results of calculating variable importance using a random forest machine learning model.

[0021]

[0022] The purposes and effects of the present invention, as well as the technical configurations for achieving them, will become clearer with reference to the embodiments described below in detail, along with the accompanying drawings. In describing the present invention, detailed descriptions of known functions or configurations will be omitted if they are deemed to unnecessarily obscure the gist of the invention. Furthermore, the terms described below are defined in light of their functions in the present invention and may vary depending on the intentions or practices of the user or operator.

[0023] However, the present invention is not limited to the embodiments disclosed below and can be implemented in various other forms. These embodiments are provided solely to ensure complete disclosure of the present invention and to fully inform those skilled in the art of the invention of the scope of the invention. The present invention is defined solely by the scope of the claims. Therefore, such definitions should be based on the contents of this specification.

[0024] The term "~unit" used in this specification refers to a unit that processes at least one function or operation, and may refer to, for example, software, an FPGA, or a hardware component. The function provided by the "~unit" may be performed separately by multiple components, or may be integrated with other additional components. The "~unit" in this specification is not necessarily limited to software or hardware, and may be configured to be located in an addressable storage medium, or may be configured to reproduce one or more processors. Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

[0025] The present invention relates to a model for predicting prematurity by developing a model for predicting prematurity through accurate collection and machine learning analysis of prenatal maternal risk factors and periodontal factors according to a prospective study design, and to a model for predicting premature delivery that combines prenatal maternal risk factors and periodontal factors. According to an embodiment of the present invention, a device and method for predicting premature delivery of a mother train a machine learning model to predict premature delivery based on clinical indicator data collected for mothers and periodontal indicator data collected through dental examinations and imaging of mothers, and predicts premature delivery of a mother based on clinical indicator data (e.g., maternal age, body mass index before pregnancy, whether or not artificial insemination was performed, history of past premature birth, pre-eclampsia, chronic hypertension, gestational diabetes, premature rupture of membranes, chorioamnionitis stage, etc.) and dental indicator data (e.g., periodontitis stage, altered gingival index, plaque index, caries-experienced permanent tooth index, etc.) of the mother by the trained machine learning model. In an embodiment, the machine learning model may be a model trained to predict variable importance of clinical indicator data and dental indicator data based on a random forest model.

[0026] Fig. 1 is a configuration diagram of a device for predicting premature delivery of a pregnant woman according to an embodiment of the present invention. Referring to Fig. 1, a device for predicting premature delivery of a pregnant woman (100) according to an embodiment of the present invention may include a learning unit (110), a clinical database (DB; database, 120), a data input unit (140), and a premature delivery prediction unit (150).

[0027] The learning unit (110) can train a machine learning model (130) to predict premature birth based on clinical indicator data collected about pregnant women and periodontal indicator data collected through dental examinations and imaging of pregnant women. In an embodiment, the machine learning model (130) can be trained by the learning unit (110) to predict variable importance of clinical indicator data and dental indicator data based on a random forest model.

[0028] The clinical DB (120) can store various clinical indicator data and periodontal indicator data for training the machine learning model (130). Such clinical indicator data and periodontal indicator data can include clinical indicators (e.g., maternal age, pre-pregnancy body mass index, artificial insemination, history of past premature birth, preeclampsia, chronic hypertension, gestational diabetes, premature rupture of membranes, chorioamnionitis stage, etc.) and periodontal indicators (e.g., periodontitis stage, altered gingival index, plaque index, caries experience permanent tooth index, etc.) collected for pregnant women, particularly high-risk pregnant women and non-high-risk pregnant women.

[0029] The data input unit (140) can receive clinical indicator data and dental indicator data related to a pregnant mother who is the target of prediction of premature delivery. In an embodiment, the clinical indicator data input by the data input unit (140) can include at least two or more of maternal age, body mass index before pregnancy, whether or not artificial insemination was performed, history of past premature birth, preeclampsia, chronic hypertension, gestational diabetes, premature rupture of membranes, and chorioamnionitis stage. In an embodiment, the dental indicator data input by the data input unit (140) can include at least two or more of periodontitis stage, altered gingival index, dental plaque index, and caries experience permanent tooth index.

[0030] The premature delivery prediction unit (150) can predict the premature delivery of a mother by using a learned machine learning model (130) based on the clinical indicator data and dental indicator data of the mother input by the data input unit (140). In an embodiment, the premature delivery prediction unit (150) can predict the premature delivery by using the variable importance of the clinical indicator data and dental indicator data predicted by the random forest model.

[0031] Fig. 2 is a flowchart of a method for predicting premature delivery of a pregnant woman according to an embodiment of the present invention. Referring to Figs. 1 and 2, the method for predicting premature delivery of a pregnant woman according to an embodiment of the present invention may include a step (S10) of learning a machine learning model (130) to predict premature delivery based on clinical indicator data collected for pregnant women by a learning unit (110) and periodontal indicator data collected through dental examinations and imaging of pregnant women, a step (S20) of receiving clinical indicator data and dental indicator data related to a pregnant mother who is a target of premature delivery prediction by a data input unit (140), and a step (S30) of predicting premature delivery of the pregnant woman by the machine learning model (130) based on the clinical indicator data and dental indicator data of the pregnant woman by a premature delivery prediction unit (150).

[0032] Hereinafter, the experimental results for verifying the performance of a device and method for predicting premature delivery according to an embodiment of the present invention are described. High-risk pregnant women admitted to the intensive care unit were prospectively enrolled, and not only various obstetric clinical indicator data but also periodontal indicators were collected through dental examinations and panoramic X-ray imaging. In addition, various clinical indicator data related to newborns born from the mothers were collected. Infants were classified into full-term and premature groups, and the differences in clinical and dental indicators between the two groups were analyzed. Based on the results, a model was constructed through machine learning, and a model for predicting premature delivery was implemented, and its performance and indicator importance were measured.

[0033] Figure 3 shows the results of predicting the probability of preterm birth (total preterm birth, natural preterm birth) and full-term birth using various machine learning models. In Figure 3, 'LR' stands for logistic regression model, 'DT' stands for decision tree model, 'NB' stands for naive Bayes model, 'RF' stands for random forest model, 'SVM' stands for support vector machine, 'ANN' stands for artificial neural network model, and 'AUC' stands for area under the curve such as ROC (receiver operating characteristic curve). When the random forest model was used, it showed a prediction accuracy of 83%, which is higher than other machine learning models, and the AUC characteristic was also confirmed to be the highest among various machine learning models.

[0034] Figure 4 shows the results of calculating variable importance using a random forest machine learning model. The random forest model is an ensemble method that uses decision trees as a base model. It creates multiple decision trees and comprehensively considers their results to derive a conclusion. A decision tree is an artificial intelligence model with a tree structure in which the middle point represents the test condition for the independent variable that serves as the basis for branching, the branches represent the test results, and the final point represents the dependent variable value according to the test results. The random forest model is an artificial intelligence model that trains a large number of decision trees and then predicts the dependent variable value by having them vote in a majority vote. For example, premature birth or birth weight can be considered as dependent variables in random forest analysis.

[0035] For example, if the data size is 100 and the number of trees is 100, the random forest training process is as follows. (1) Random selection with replacement is performed on the training data (size 100) with 100 participants to create a subset of data (size 100) with 100 participants. (2) Some or all of the independent variables are randomly selected without replacement and trees are created. (3) The above process is executed 1,000 times to create 1,000 trees, and the majority rule is applied to each observation value that is not included in the subset of data (out of bag) to derive the predicted value.

[0036] We tested the association between clinical and dental predictors and preterm birth using machine learning and prospective cohort data. Prospective cohort data were obtained from 60 women who delivered singletons by cesarean section. Thirty women were classified as preterm births (PTB) if they delivered before 37 weeks of gestation. The remainder were classified as full-term births (after 37 weeks of gestation). The dependent variables were preterm birth (PTB), spontaneous preterm birth (SPTB), and full-term birth. Of the 20 variables considered in clinical examinations and records, 15 independent variables (10 medical factors and 5 dental factors) were selected for inclusion in the machine learning analysis. Random Forest (RF) variable importance was used to identify key predictors of PTB / SPTB. Shapley Additive Explained (SHAP) values ​​were calculated to analyze the direction of the association between predictors and PTB / SPTB.

[0037] Random forest analysis revealed that prepregnancy body mass index (BMI), modified gingival index (MGI), pre-eclampsia, decayed missing filled teeth (DMFT) index, and maternal age were ranked in the top five RF variables of importance for preterm birth (PTB). For spontaneous preterm birth (SPTB), pre-labor premature rupture of membranes (PROM), prepregnancy body mass index (BMI), maternal age, decayed missing filled teeth (DMFT) index, and chorioamnionitis (CAM) stage were the top five factors. In terms of SHAP values, some positive correlations were found between PTB / SPTB and key predictors such as premature rupture of membranes (PROM), prepregnancy body mass index (BMI), maternal age, modified gingival index (MGI), pre-eclampsia, and chorioamnionitis (CAM) stage. Modified Gingival Index (MGI), along with its clinical counterpart, was found to be one of the major predictors of PTB / SPTB.

[0038] Preterm birth (PTB), defined as birth before 37 weeks of gestation, is one of the most common and serious complications of pregnancy, causing significant morbidity and mortality in newborns, infants, and children under five years of age. Infants born prematurely due to PTB can suffer from a variety of serious perinatal complications, including intraventricular hemorrhage, retinopathy of prematurity, necrotizing enterocolitis, bronchopulmonary dysplasia, visual and hearing impairments, motor disabilities, and neurodevelopmental disorders. Some of these complications persist beyond the neonatal period and are lifelong complications.

[0039] Preterm birth (PTB) can be broadly categorized into two subsets. First, there are intentional preterm births, in which labor is induced or the infant is delivered by cesarean section without prior analgesia due to maternal or fetal indications such as preeclampsia or fetal instability. Second, there are spontaneous PTBs (SPTBs), in which preterm labor occurs with or without prelabor rupture of membranes (PROM). Spontaneous PTBs account for approximately two-thirds of all PTBs and are difficult to predict. Frequently discussed risk factors for preterm or spontaneous PTB can be broadly categorized into two categories. First, maternal / environmental factors, including older age, lower socioeconomic status, smoking, alcohol use, drug use, mood disorders, or psychological stress, nutritional factors, and prepregnancy body mass index (BMI). Second, there are medical / obstetric factors, including previous preterm birth, multiple gestations, short cervical length, urinary tract infections, diabetes, hypertension, thyroid disorders, preterm premature rupture of membranes (PROM), and fetal malformations.

[0040] Despite the numerous risk factors identified over decades of research, the predictive value for PTB remains low. Commonly considered PTB prediction strategies include risk assessment, serial measurement of cervical length (CL), and assessment of biochemical biomarkers, including fetal fibronectin and inflammatory cytokines. Although CL measurement has become a routine obstetrical procedure from the second trimester, its clinical utility, particularly in low-risk populations, remains controversial. Recommendations for fibronectin screening are inconsistent, and its low cost-effectiveness and patient burden must also be considered.

[0041] Periodontitis is a chronic inflammatory disease of the oral cavity and is widely associated with preterm birth (PTB) and pregnancy complications, including preeclampsia and low birth weight. Periodontitis and preterm birth (PTB) share risk factors, including low socioeconomic status, high BMI, and medical conditions such as smoking, alcohol use, diabetes, and hypertension. Theoretically, periodontitis may directly or indirectly contribute to the inappropriate activation of the labor stage leading to PTB by elevating the inflammatory state of the placental tissue.

[0042] The present invention utilizes prospectively collected data and machine learning methods to predict the importance of periodontitis-related parameters as predictors of preterm birth (PTB), thereby increasing the accuracy of predicting preterm birth. Unlike traditional linear regression models, machine learning-based data analysis calculates the difference between predictors and outcomes. It is relatively free from complex, nonlinear relationships and can model them "ceteris paribus" (all other variables held constant), demonstrating its superiority over existing approaches in risk factor assessment. Therefore, utilizing machine learning methods in risk factor analysis can fill the gaps in existing approaches, enabling the identification of additional screening tools for the disease. Therefore, the present invention tested the association between clinical and dental predictors of preterm birth using machine learning and prospective cohort data.

[0043] This prospective cohort study was approved by the Institutional Review Board (IRB) of Korea University Anam Hospital. It was conducted as part of a study on the oral microbiome in preterm mothers. Pregnant women admitted for cesarean section for singletons were recruited and divided into two groups: 30 patients who delivered before 37 weeks of gestation and 30 patients who delivered at term. Medical and obstetric data were collected before and after delivery. These included maternal age, prepregnancy body mass index (BMI), changes in BMI during pregnancy, in vitro fertilization history, previous preterm birth (PTB), preeclampsia, chronic hypertension, gestational diabetes mellitus (GDM), preterm rupture of membranes (PROM), histologic chorioamnionitis (CAM), and enterocolitis. Preterm labor was defined as cervical dilation of 1 cm or more and regular uterine contractions.

[0044] Panoramic radiographs were taken before a full-oral periodontal examination of the participants during their postpartum hospitalization. Measurements included periodontal probing depth (PPD), clinical attachment level (CAL), modified gingival index (MGI), and plaque index (PI). Based on PPD and CAL measurements, in addition to radiographic examinations that revealed treatment complexity factors such as furcation involvement, the periodontitis stage was determined for each patient. Periodontitis severity was graded from stage 0 to 3 according to the periodontitis classification system. Stage 0 represents a healthy periodontium with no signs of periodontitis.

[0045] Prospective cohort data were obtained from 60 women who gave birth to unmarried mothers. The dependent variable was PTB / SPTB. The analysis included 15 independent variables. Logistic regression (LR), decision trees (DT), naive Bayes (NB), random forests (RF), support vector machines (SVM), and artificial neural networks (ANN) were used to predict PTB / SPTB. Random forests (RF) are decision tree models that perform majority voting on the dependent variable ("bootstrap aggregation"). Assuming the original data contains 42 participants, we will use a random forest with 1,000 decision trees as an example. Random forest training and testing involves two steps. First, new data is generated from the 42 participants using random sampling and replacement, and decision trees are generated based on these new data. Some participants in the original data are excluded from the new data, which is referred to as out-of-bag data. This process is repeated 1,000 times.

[0046] Second, 1,000 decision trees predict the dependent variable for all participants in the out-of-bag data. The majority of these votes is taken as the final prediction for that participant, and the out-of-bag error is calculated as the proportion of incorrect votes in the out-of-bag data. The data for 60 cases containing complete information were split into training and validation sets in a 75:25 ratio (42 cases vs. 14 cases). The validation criteria for the trained model were accuracy (correct prediction rate among 14 cases) and the area under the receiver operating characteristic curve (AUC).

[0047] Random Forest permutation importance was used to identify key predictors of PTB / SPTB. Shapley Additive Explained (SHAP) values ​​were calculated to analyze the direction of association between predictors and PTB / SPTB. RF permutation importance was chosen to measure the overall decrease in accuracy when predictor data are randomly shuffled. This decrease in accuracy indicates the degree to which the model relies on the predictor. SHAP values ​​were calculated by measuring the difference between the machine learning-predicted probabilities of PTB / SPTB with and without the predictor. RF permutation importance was used to derive the rank order of predictors. SHAP plots were generated to evaluate the direction of association between each predictor and the dependent variable, PTB / SPTB.

[0048] Among various medical variables, the incidence of preeclampsia, preterm rupture of membranes (PROM), preterm labor, and labor pains during delivery were significantly higher in the PTB group (P<0.05). Maternal age, incidence of in vitro fertilization, and gestational diabetes mellitus (GDM), known risk factors for PTB, did not show significant differences in the study group. Periodontitis stage, DMFT (caries-experienced permanent teeth) index, and plaque index did not differ between the two groups, but the modified gingival index (MGI) showed a significantly higher value in the PTB group (P<0.05).

[0049] The random forest (RF) model showed similar performance to the logistic regression (LR) model. Based on the random forest (RF) variable importance rankings, prepregnancy BMI, modified gingival index (MGI), pre-eclampsia, DMFT index, and maternal age were included in the top five items for PTB. Premature rupture of membranes (PROM), prepregnancy BMI, maternal age, DMFT index, and chorioamnionitis (CAM) stage were included in the top five items for SPTB.

[0050] Figure 4 shows the SHAP values ​​for each independent variable in the random forest. The SHAP values ​​for each independent variable indicate whether the probability of the dependent variable (PTB or SPTB) decreases or increases when a particular independent variable is included in the machine learning analysis. In general, an absolute value of the maximum SHAP (positive) that is greater than the absolute value of the minimum SHAP (negative) indicates a positive relationship between the predictor and the dependent variable. In particular, some positive associations were observed among variables ranked high in the random forest (RF) permutation importance rankings.

[0051] In the present invention, we suggest that dental factors such as the modified gingival index (MGI) and the dental caries-experienced permanent tooth (DMFT) index can act as major predictors in a preterm birth (PTB) prediction model constructed using machine learning techniques. What differentiates the present invention from previous studies is that in addition to well-known medical risk factors, dental factors such as various medical backgrounds and obstetric history are added. In existing prediction models using AI or multivariate regression analysis, the model performance calculated using the AUC was only in the range of 0.61 to 0.71. According to the present invention, among the five machine learning methods tested (DT, NB, RF, SVM, ANN), the random forest (RF) model showed the highest performance with an AUC of 0.73 in the preterm birth (PTB) prediction model and 0.86 in the spontaneous preterm birth (SPTB) prediction model.

[0052] The importance ranking of RF variables was used to verify the importance of dental factors in the random forest model. The modified gingival index (MGI) ranked second in the PTB model and sixth in the SPTB model. This ranking was higher than other well-known PTB risk factors from a medical perspective, such as maternal age (5th), previous PTB (14th), pre-eclampsia (3rd), chronic hypertension (15th), and gestational diabetes (10th). The MGI represents the level of gingival inflammation during the examination period and indicates susceptibility to chronic inflammation. A higher MGI significantly reflects the host's susceptibility to gingivitis and is strongly correlated with future periodontitis, especially in young adults. In this context, a high MGI, especially an average MGI ≥2 (Figure 2a), in prenatal screening or routine checkups during pregnancy indicates a higher maternal susceptibility to infection or a tendency toward its progression during pregnancy.

[0053] In this prospective study, data on several dental parameters, including periodontitis stage, modified gingival index (MGI), PI, and the DMFT index, were collected. Periodontitis staging indicates disease severity and is based on a newly established periodontitis classification. The periodontal status of the participants assessed using the staging system did not differ between the two groups. Cases of moderate / severe periodontitis, corresponding to stages 3 and 4, were rare, and the prevalence did not differ between the groups. No significant difference in prevalence was observed between periodontitis (stage 1 or higher) and healthy periodontium (stage 0). These results differ somewhat from previous data, which generally showed a higher incidence of periodontitis in premature infants.

[0054] Considering the high AUC (0.86) of dental parameters in the SPTB analysis model, preterm labor with or without premature rupture of membranes (SPTB) can be considered a major cause of PTB. It is considered a syndrome with multiple causes, including infection, vascular disorders, decidual aging, disruption of maternal-fetal tolerance, decreased progesterone function, and cervical disease. The delicate balance of maternal immunology is a prerequisite for a healthy pregnancy and labor. The transition from a quiescent to a proinflammatory state activates labor, and inflammatory cytokines such as interleukin (IL)-1, IL-6, and tumor necrosis factor (TNF)-α are involved in this process.

[0055] In this context, if pregnant women are susceptible to infections due to various factors, such as BMI, stress, behavioral factors, genetic factors, and nutritional deficiencies, the delivery process may be abnormally advanced. Our study suggests that a high modified gingival index (MGI) can serve as a screening tool for maternal susceptibility to infection. This may be a simple, noninvasive, and cost-effective test that can be incorporated into prenatal screening. Furthermore, pregnant women with high MGI scores are more likely to harbor pathogenic bacteria in their dental plaque, suggesting that periodontal treatment during the prenatal or mid-natal period is recommended to prevent hematogenous spread of oral pathogens.

[0056] The dependence plot revealed a negative correlation within the DMFT index range of ≤10, indicating that caries-susceptible patients are less likely to develop SPTB. Dental caries and periodontitis are two contrasting infections occurring in the oral cavity, as the bacterial species primarily associated with each disease have nearly opposite characteristics. In this context, patients susceptible to periodontitis have a lower caries propensity and therefore lower DMFT index and higher SHAP values. The present invention utilizes state-of-the-art AI approaches, such as random forest (RF) variable importance and SHAP summary / dependency plots, to identify key predictors and explain the direction of their associations. Traditional statistical approaches, such as linear or logistic regression, rely on unrealistic data assumptions, i.e., "all other variables held constant," as the underlying principle. In contrast, SHAP considers all realistic scenarios.

[0057] We assume three predictors of SPTB: preterm premature rupture of membranes (PROM), prepregnancy BMI, and modified gingival index (MGI). The SHAP value for the modified gingival index (MGI) for a participant is the average of four scenarios: (1) excluding PROM but excluding prepregnancy BMI; (2) including PROM but excluding prepregnancy BMI; (3) excluding PROM but including prepregnancy BMI; and (4) including PROM but including prepregnancy BMI. In other words, the SHAP value incorporates all possible subgroup analysis results that are ignored by existing methods. A limitation of this study is the small sample size. Machine learning techniques can generally structure data without explicit programming, and predictive performance and reliability improve with larger datasets. However, in this study, even with a relatively small dataset, we achieved robust performance of 73% (AUC) for PTB and 86% (AUC) for SPTB using the random forest (RF) method. Unlike previous studies, the dataset in the present invention was collected prospectively and included multifaceted parameters including dental, medical, and obstetric factors, which enabled the construction of a robust model due to the high quality of the data.

[0058] The embodiments described above may be implemented using hardware components, software components, and / or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose computers or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing instructions and responding.

[0059] A processing device can execute an operating system and one or more software applications running on the operating system. Furthermore, the processing device can access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing device is sometimes described as being used singly; however, those skilled in the art will understand that the processing device can include multiple processing elements and / or multiple types of processing elements.

[0060] For example, a processing unit may include multiple processors, or a processor and a controller. Other processing configurations, such as parallel processors, are also possible. Software may include a computer program, code, instructions, or a combination of one or more of these, and may configure a processing unit to perform a desired operation or may command the processing unit, either independently or collectively.

[0061] The software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal waves, for the purpose of being interpreted by a processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on one or more computer-readable recording media.

[0062] The method according to the embodiment may be implemented in the form of program commands that can be executed by various computer means and recorded on a computer-readable medium. The computer-readable medium may include program commands, data files, data structures, etc., either singly or in combination. The program commands recorded on the medium may be those specifically designed and configured for the embodiment or may be known and usable by those skilled in the art of computer software.

[0063] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical media such as CDROMs and DVDs; and hardware devices specifically configured to store and execute program instructions, such as ROMs, RAMs, and flash memories. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

[0064] The above description is merely an illustrative illustration of the technical idea of ​​the present invention, and those skilled in the art will appreciate that various modifications and variations can be made without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are intended to illustrate rather than limit the technical idea of ​​the present invention, and the scope of the technical idea of ​​the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted by the following claims, and all technical ideas within a scope equivalent thereto should be interpreted as being included in the scope of the rights of the present invention.

[0065]

[0066] [Explanation of symbols]

[0067] 100: Pregnant woman's premature delivery prediction device

[0068] 110: Learning Department

[0069] 120: Clinical Database

[0070] 130: Machine Learning Model

[0071] 140: Data input section

[0072] 150: Premature birth prediction department

Claims

1. Data input section for entering clinical indicator data and dental indicator data of the mother; and A device for predicting premature delivery of a mother, comprising a premature delivery prediction unit for predicting premature delivery of a mother by a machine learning model based on the clinical indicator data and the dental indicator data, the machine learning model being trained to predict premature delivery based on clinical indicator data collected about the mothers and periodontal indicator data collected through dental examinations and imaging of the mothers.

2. In claim 1, A device for predicting premature delivery of a mother, wherein the clinical indicator data includes at least two of maternal age, body mass index before pregnancy, whether or not artificial insemination was performed, history of previous premature birth, preeclampsia, chronic hypertension, gestational diabetes, premature rupture of membranes, and chorioamnionitis stage.

3. In claim 1, A device for predicting premature delivery of a mother, wherein the dental indicator data includes at least two of periodontal stage, altered gingival index, dental plaque index and caries experience permanent tooth index.

4. In claim 1, A device for predicting premature delivery of a pregnant woman, wherein the machine learning model is a model learned to predict the variable importance of the clinical indicator data and the dental indicator data based on a random forest model.

5. A step of entering the clinical indicator data and dental indicator data of the mother through the data input section; and A method for predicting premature delivery of a mother, comprising: a step of predicting premature delivery of a mother by a machine learning model based on the clinical indicator data and the dental indicator data, by a premature delivery prediction unit, wherein the machine learning model is trained to predict premature delivery based on clinical indicator data collected about the mothers and periodontal indicator data collected through dental examinations and imaging of the mothers.

6. In claim 5, A method for predicting premature delivery of a mother, wherein the clinical indicator data includes at least two of maternal age, body mass index before pregnancy, whether or not artificial insemination was performed, history of previous premature birth, pre-eclampsia, chronic hypertension, gestational diabetes, premature rupture of membranes, and chorioamnionitis stage.

7. In claim 5, A method for predicting premature delivery of a mother, wherein the dental indicator data includes at least two of periodontal stage, altered gingival index, dental plaque index and caries experience permanent tooth index.

8. In claim 5, A method for predicting premature delivery of a mother, wherein the machine learning model is a model learned to predict the variable importance of the clinical indicator data and the dental indicator data based on a random forest model.

9. In claim 8, The steps to predict premature delivery in the above mother are: A method for predicting premature delivery of a mother, comprising a step of predicting the probability of premature birth by determining the importance ranking of variables of pre-pregnancy body mass index, modified gingival index, pre-eclampsia, caries-experienced permanent tooth index, artificial insemination, past history of premature birth, maternal age, chronic hypertension, gestational diabetes, premature rupture of membranes, and chorioamnionitis stage, in relation to premature birth, by the random forest model.

10. A computer-readable, non-transitory recording medium having recorded thereon a program for executing the method for predicting premature delivery of claim 5.

Images