JP2019041754A - Prediction device and prediction method - Google Patents

Abstract

To provide prediction devices and a prediction methods for achieving HDP onset prediction from oral bacterial flora.SOLUTION: One aspect of the present invention relates to a prediction device having: a data acquisition function to acquire bacterial flora data on bacterial flora in the oral cavity of pregnant women; a step of generating a prediction model of hypertensive disorders of pregnancy (HDP) based on the bacterial flora data; an HDP prediction function of carrying out the HDP onset prediction of the pregnant woman from the acquired bacterial flora data based on the HDP prediction model; and a result notifying function to notify the result of the HDP prediction, the data acquisition function acquiring the bacterial flora data at a plurality of time points in a data acquisition period from any time before or after pregnancy of the pregnant woman to any time before or after delivery, and the HDP prediction function carrying out the HDP onset prediction based on the bacterial flora data acquired at the plurality of time points.SELECTED DRAWING: Figure 3

Description

  TECHNICAL FIELD The present invention relates generally to medical diagnostic technology, and more particularly to a prediction device and prediction method for predicting the onset or absence of pregnancy / pregnancy-hypertensive syndrome for non-invasively predicting the onset of pregnancy / hypertension syndrome or detecting an abnormality.

  As a pregnancy complication, hypertensive disorders (HDP) is a typical perinatal medical treatment related to the life of mothers and infants causing maternal death, fetal growth failure, fetal death, permanent placenta early detachment, etc. It is a multifactorial disease that develops due to complex interactions. In addition, it is known that pregnant women suffering from HDP have a high risk of developing lifestyle-related diseases thereafter. Although the association between the oral flora of pregnant women and pregnancy complications has been reported, the detailed mechanism is still unclear.

  According to Non-Patent Document 1, for 124 pregnant women, clinical attachment level (CAL) and probing depth (PD) were measured during pregnancy or within 3 days after delivery, and cases with PLBW (Preterm Low Birth Weight) and When evaluated in the control group that did not, the CAL mean value of the case group (31 people) was significantly larger than that of the control group. In addition, in pregnant women in which teeth with 3 mm or more of CAL accounted for 60% or more, the PLBW onset risk rate was 5.9 times overall and 6.7 times at first birth.

  In addition, according to Non-Patent Document 2, when the dental plaque microflora of the 18 second trimesters who became premature from imminent premature birth and 40 other term second trimesters were compared, it was one of the periodontal disease related bacteria. The proportion of certain Tannerella forsythenesis was significantly higher. This Tannerella forsythenesis is one of three red complexes called bacteria associated with periodontal disease.

  In addition, according to Non-Patent Document 3, it is known that intestinal and vaginal flora fluctuate significantly during pregnancy, and it is assumed that oral flora affected by diet and tooth brushing is more volatile. Be done.

Offenbacher S, et al, Periodontal Infection as a Possible Risk Factor for Preterm Low Birth Weight, J Peridontol, 67, 1103-1113, 1996 Hasegawa K, et al, Associations between systematic status, oeruidibtak statysm seryn cytokine levels and delivery outcomes in pregnant women with a diagnosis of threatened premature labor (TPL), J Peridontol, 74, 1764-1770, 2003 Daniel B. DiGiulio, et al, Temporal and spatial variation of the human microbiota during pregnancy, PNAS, 112, 35, 11060-11065, 2015

  Thus, although the association between the oral flora and pregnancy complications has been reported, there has been little report on the association between the oral flora and one of pregnancy complications, HDP. In addition, although there is a racial difference in the relationship between oral flora and pregnancy complications, there are few studies targeting Japanese pregnant women. In addition, we can not take into consideration the effects of changes in oral flora over time on pregnancy complications. Furthermore, when predicting pregnancy complications from the oral flora, fluctuations in the bacterial flora within an individual may not be successful.

  In view of the above-mentioned problems, it is an object of the present invention to provide a prediction device and a prediction method for realizing the onset prediction of HDP from the intraoral bacterial flora.

  In order to solve the above-mentioned subject, one mode of the present invention generates a pregnancy hypertension syndrome (HDP) prediction model based on the data acquisition function which acquires bacterial flora data about bacterial flora in the oral cavity of a pregnant woman, and the bacterial flora data. An HDP prediction model generation function, an HDP prediction function that executes HDP onset prediction of the pregnant woman from the acquired bacterial flora data based on the HDP prediction model, and a result notification function that notifies the result of the HDP onset prediction The data acquisition function acquires the bacterial flora data at a plurality of time points in a data acquisition period from any time before or after pregnancy of the pregnant woman to any time before or after delivery, and the HDP prediction The function relates to a prediction device that executes the HDP prediction based on the bacterial flora data acquired at the plurality of time points.

  According to the present invention, it is possible to provide a prediction device and a prediction method for realizing the prediction of the onset of HDP from the oral flora.

FIG. 1 is a schematic diagram showing the configuration of a prediction system according to an embodiment of the present invention. FIG. 2 is a block diagram showing a functional configuration of a prediction device according to one embodiment of the present invention. FIG. 3 is a flow chart showing a prediction process according to one embodiment of the present invention. FIG. 4 is a flowchart showing feature quantity extraction / model creation processing according to an embodiment of the present invention. FIG. 5 is a flowchart showing feature quantity extraction / model creation processing according to another embodiment of the present invention. FIG. 6 is a flowchart showing feature quantity extraction / model creation processing according to another embodiment of the present invention.

  Hereinafter, embodiments of the present invention will be described based on the drawings.

  In the embodiments described below, a prediction device for predicting the onset of HDP is disclosed. According to the example described later, when bacterial flora data on the bacterial flora in the oral cavity of a pregnant woman collected in multiple pregnancy phases is acquired, the prediction device extracts a bacterial species having a correlation with HDP, and extracts the bacterial species Based on the HDP prediction model generated using, the prediction of HDP onset in pregnant women is performed from the acquired bacterial flora data. This makes it possible to generate an HDP prediction model suitable for a specific race against HDP prediction from bacterial flora data that is considered to change with time as well as differences between races.

  First, with reference to FIG. 1, a prediction system according to an embodiment of the present invention will be described. FIG. 1 is a schematic diagram showing the configuration of a prediction system according to an embodiment of the present invention.

  As shown in FIG. 1, the prediction system 10 includes a bacterial flora measurement device 50 and a prediction device 100.

  The bacterial flora measuring device 50 measures the bacterial flora in the oral cavity from the collected dental plaque sample, and provides the generated bacterial flora data to the prediction device 100. Specifically, a laboratory technician collects a dental plaque sample of a pregnant woman, acquires bacterial flora data indicating bacterial species present in the dental plaque specimen collected using the bacterial flora measuring device 50, and acquires the acquired bacterial flora data It is provided to the prediction device 100.

  More than 200 species of bacteria including, for example, Porphyromonas gingivalis, Treponema denticola, Tabberella forsythensis, Bergeyella, Kingella denitrificans, Scardovia wiggsiae, Streptococcus mutans, Selenomonas, Capnocytophaga, Treponema vincentii etc. as bacteria species present in human oral cavity In particular, Porphyromonas gingivalis, Treponema denticola, and Tabberella forsythensis are classified as a red complex involved in periodontal disease, and a correlation with pregnancy complications has been pointed out. The bacterial flora measuring device 50 may be any known device used to measure bacterial species in the oral cavity.

  The prediction device 100 predicts HDP from the provided bacterial flora data. Specifically, as described in detail below, the prediction device 100 creates an HDP prediction model by creating an HDP prediction model from known data including the bacterial flora data of the HDP group and the bacterial flora data of the non-HDP group. By inputting the bacterial flora data of the pregnant woman to be predicted in the prediction model, it is predicted whether or not the onset of HDP of the pregnant woman is concerned. The prediction of the onset of HDP according to the present invention is not limited to the presence or absence of the onset of HDP, and may include the prediction of the date of onset of HDP, the calculation of the HDP risk, and the like as described later.

  Here, the prediction device 100 may be typically realized by an information processing device having a communication function such as a computer, a server, a smartphone, or a tablet. For example, a processor mounted on an information processing apparatus executes various functions and processes to be described later by processing and executing data and programs stored in a memory device. However, the prediction device 100 is not limited to any specific hardware configuration, and may be realized by an appropriate hardware configuration.

  Next, with reference to FIG. 2, a prediction apparatus according to an embodiment of the present invention will be described. FIG. 2 is a block diagram showing a functional configuration of a prediction device according to one embodiment of the present invention.

  As shown in FIG. 2, the prediction device 100 has a data acquisition function 110, a data storage function 120, an HDP prediction model generation function 130, an HDP prediction function 140, and a result notification function 150.

  The data acquisition function 110 acquires bacterial flora data from the bacterial flora measurement device 50. Specifically, the data acquisition function 110 acquires bacterial flora data related to a pregnant woman at a plurality of time points in a data acquisition period from any time before or after pregnancy of the pregnant woman to any time before or after delivery. For example, the data acquisition function 110 may acquire bacterial flora data in the oral cavity of a pregnant woman collected at different pregnancy phases (for example, 12 to 23 weeks and 24 to 34 weeks) from the bacterial flora measuring device 50.

  The data storage function 120 stores the acquired bacterial flora data in the oral cavity of the pregnant woman. Specifically, the data storage function 120 stores the acquired bacterial flora data as independent variables for each of the bacterial species (eg, 200 bacterial species and the like) collected in each pregnancy phase. For example, when bacterial species of 200 bacterial species are collected in each of two pregnancy phases, 400 pieces (200 bacterial species × 2) can be prepared so that a model can be created that takes into account changes in plaque bacterial composition at multiple time points. Pregnancy phase) variables may be set.

  The HDP predictive model generation function 130 generates an HDP predictive model that predicts the onset of HDP based on the acquired bacterial flora data in the oral cavity of a pregnant woman. Specifically, HDP predictive model generation function 130 is data on bacterial flora in the oral cavity acquired from a plurality of pregnant women including pregnant women who did not develop HDP (non-HDP group) and pregnant women who developed HDP (HDP group) Based on, the HDP onset prediction model may be generated and updated. For example, as described in detail below, the HDP prediction model generation function 130 identifies two types of bacterial species from the acquired bacterial flora data so as to divide the HDP group and the non-HDP group, and to reduce variables. 10 to 20 bacterial species × pregnant phase may be extracted as a bacterial species that divides well. Porphyromonas gingivalis, Treponema denticola, Tabberella forsythensis classified as red complex may be extracted as a bacterial species that divides HDP group and non-HDP group well. In addition, HDP predictive model generation function 130 is a bacterial species that divides HDP group and non-HDP group well, such as Actinomyces sp., Staphylococcus sp., Capnocytophaga sp., Prevotella sp., Leptotrichia sp., GN02 sp., Streptococcus sp. , Haemophilus sp., Corynebacterium sp., Aggregatibacter sp., Rothia sp., Leptotrichia sp., Lautropia sp.

  Subsequently, the HDP prediction model generation function 130 generates an HDP prediction model using the extracted bacterial species. As a specific prediction model creation method, for example, the HDP prediction model generation function 130 may select a bacterial species to be used as a feature amount by forward stepwise selection using a linear SVM. For example, for the first 20 variables (bacterial species), the HDP prediction model generation function 130 generates HDP prediction models M (1) to M (20) using only one variable, and the M with the best performance. (A) (use variable A) is specified. Next, the HDP prediction model generation function 130 selects one variable from the remaining nineteen variables excluding the variable, and combines it with A, a two-variable prediction model M (A, 1) to M (A, 19) Create and select the best model including M (A). The HDP prediction model generation function 130 recursively executes the above-described series of processes, and finally identifies the best performing prediction model M (A, B, C,...).

  In addition, although the Example mentioned above demonstrated the HDP onset prediction model which shows the onset presence or absence of HDP, the HDP prediction model by this invention is not limited to this, HDP onset date prediction which predicts the onset date of HDP, and HDP It may be an onset risk model of HDP or an HDP onset risk model for predicting the onset probability.

  The HDP prediction function 140 executes HDP prediction of the pregnant woman from the acquired bacterial flora data of the pregnant woman based on the HDP prediction model. Specifically, the HDP prediction function 140 inputs bacterial flora data of pregnant women obtained at multiple time points into the HDP prediction model, and the HDP prediction function 140 is based on the output result of the HDP predictive model based on the bacterial flora data obtained at multiple time points. The prediction of the onset of HDP, the prediction of the date of onset of HDP, or the risk calculation for the onset of HDP may be performed on a pregnant woman. That is, the HDP prediction function 140 predicts whether there is a possibility of developing HDP in the pregnant woman (presence or absence of developing HDP), or predicts the date of onset of HDP in the pregnant woman (predicting the date of onset of HDP) Risk of developing HDP may be calculated (HDP risk calculation).

  The result notification function 150 notifies pregnant women, medical personnel, etc. of the result of the HDP prediction performed by the HDP prediction function 140.

  Next, prediction processing according to an embodiment of the present invention will be described with reference to FIGS. The said prediction process is performed by the prediction apparatus 100 mentioned above. FIG. 3 is a flow chart showing a prediction process according to one embodiment of the present invention.

  As shown in FIG. 3, in step S100, the prediction device 100 acquires bacterial flora data from the bacterial flora measurement device 50, and stores the acquired bacterial flora data. Specifically, the bacterial flora measuring device 50 measures a dental plaque sample collected from a pregnant woman, and provides the bacterial flora data indicating the amount of bacterial species contained in the dental plaque specimen to the prediction device 100. The prediction device 100 stores the bacterial flora data acquired from the bacterial flora measurement device 50 for each pregnant woman. According to this example, the bacterial flora data is collected multiple times for each pregnant woman in a data acquisition period from any time before and after pregnancy to any time before and after delivery, for example, every pregnancy phase.

In step s101, the prediction device 100 detects and removes outliers from the acquired bacterial flora data, and complements missing values. For example, when the amount of bacterial species outside the threshold range is measured from the reference value, the prediction device 100 regards the measured value as not a real value but a measurement error, removes the measured value, and removes other bacterial species. Or random noise is added to the measurement value of another plaque sample having a similar value of quantity and diverted, or another plaque sample having similar values of other bacterial species is extracted and those measurements are made The measured value may be complemented by any method such as calculating one estimated value so as to follow the distribution of the values. For example, when the bacterial species spA-2 of the sample S is out of the threshold range or is a missing value, the variance V of the bacterial species amount of spA-2 in the sample group other than the sample S is first calculated. Next, a sample similar to the sample S in the amount of each bacterial species of the bacterial species group B _SUB obtained by removing spA-2 from the bacterial species group B _ALL = {spA-1, spB-1, ..., spA-2, spB-2} There is a method of extracting one sample and adding a random one point according to normal distribution N (0, V) to the amount of bacterial species of spA-2 as the estimated amount of spA-2 of the sample S. At this time, for the extraction of a sample close to the sample S, the Yougrid distance may be calculated using a standardized value obtained by dividing each bacterial species amount by the standard deviation of each bacterial species amount, or MSE (mean square error Or R ² (coefficient of determination) may be used. As another complementary method, only 10 samples, for example, where the bacterial species amount of the bacterial species group B _SUB is close to the sample S are extracted, the average value E2 and the variance V2 are determined, and random according to the normal distribution N (E2, V2) There is a method of using a single point as an estimate of spA-2 of the sample S. Also, the distribution is not necessarily a normal distribution, and may be estimated as an arbitrary distribution.

  In step s102, the prediction apparatus 100 extracts, as a feature value, a variable that well divides the non-HDP group and the HDP group, using data of a pregnant woman who has already given birth and whose birthing status has been determined. These include methods using Student's t test, Welch's t test, methods using Lasso regression, Ridge regression, Elastic Net, etc. to limit useful variables, methods using principal component analysis to limit variables, etc. But there is no limitation to them.

  In steps s103, s105 and s107, the prediction device 100 generates an HDP onset prediction model using the extracted feature quantities. Specifically, in step s103, the prediction device 100 generates an HDP onset prediction model as a classification model for predicting whether to become an HDP. Further, in step s105, the prediction device 100 generates an HDP onset date prediction model as a regression model for predicting the HDP onset date of each pregnant woman. The prediction device 100 also generates an HDP onset risk calculation model for calculating the risk of HDP onset in s107. The regression model includes, but is not limited to, linear regression model, logistic regression model, regression model by SVM, regression model by multi-layered neural network, and the like. Further, in the classification model, there are linear classification, logistic regression, Bayesian network, SVM, k-neighbor method, random forest, classification model by multilayer neural network, etc., but it is not limited thereto. The HDP risk calculation model includes, but is not limited to, odds ratio, contribution risk, relative risk, and the like. Further, steps s103, s105, s107 may be processed integrally without being separated from s102.

  In steps s104, s106 and s108, the prediction device 100 inputs the bacterial flora data of the pregnant woman to be predicted into the generated HDP onset prediction model, the HDP onset prediction model and the HDP onset risk calculation model, respectively, and predicts the HDP onset. Run. Specifically, in step s104, the prediction device 100 predicts the presence or absence of future HDP onset from bacterial flora data of a pregnant woman to be predicted using the HDP onset prediction model. Further, in step s106, the prediction device 100 predicts the date of HDP onset from the bacterial flora data of the pregnant woman to be predicted using the HDP onset day prediction model. Further, in step s108, the prediction device 100 calculates the HDP risk from the data of the pregnant woman to be predicted using the HDP onset risk calculation model. In addition, the HDP presence / absence prediction is a final prediction of HDP onset / presence, prediction of not developing HDP in the next n1 weeks (n1 is any numerical value), prediction of developing HDP in the next n2 weeks (n2 is arbitrary numerical value), Including, but not limited to.

  In step s109, the prediction device 100 notifies the pregnant woman or medical institution of the HDP onset prediction result, the HDP onset date prediction result, and / or the HDP onset risk calculation result, and ends the prediction processing.

  FIG. 4 is a flowchart showing feature quantity extraction / model creation processing according to an embodiment of the present invention. The feature quantity extraction / model creation process can be executed in steps s102 and s103 of FIG. 3 described above, and steps s102-1 to s5 described later correspond to the feature quantity extraction process of step s102, and step s103 -1 to 2 correspond to the generation process of the HDP presence / absence prediction model in step s103. In the feature amount extraction / model creation processing, the HDP prediction model generation function 130 divides the acquired data into an HDP group and a non-HDP group, and divides two groups out of each variable (that is, distinguishes, discriminates, A variable, etc.) is extracted, for example, the upper predetermined number of variables that divides the two groups well, the extracted variables are used to create an HDP prediction model, and the HDP prediction model with the best performing variables is specified. Next, the HDP prediction model generation function 130 creates a two-variable prediction model combining the remaining variables excluding the variable and the specified variables, and selects an optimal model including these models. The HDP prediction model generation function 130 recursively executes the above-described series of processes, and finally identifies the HDP prediction model. The above is an example of a method of extracting feature quantities used in the HDP onset prediction model and a selection method of the HDP onset prediction model, and the HDP onset prediction model is not limited thereto. For example, there is also a method of extracting feature quantities by a multi-layered neural network or Lasso regression to generate an HDP onset prediction model.

  In step s102-2, the prediction device 100 extracts bacterial flora data of pregnant women who have already given birth, for model generation.

  In step s102-3, the prediction apparatus 100 stores bacterial flora data of dental plaque samples collected in each pregnancy phase as individual variables of each bacterial species collected in each pregnancy phase. For example, when 200 bacterial species are collected in each of two pregnancy phases, 400 (200 bacterial species × 2 phases) variables are set. For example, 400 pieces of data are represented by a variable string of "spA-1, spB-1, ..., spA-2, spB-2, ..." in association with the ID (userid) of the pregnant woman. May be

  In step S102-4, the prediction device 100 divides the bacterial flora data into the HDP group and the non-HDP group, and the bacterial species that divides the two groups with ap value less than a (for example, a = 0.01) in the test. Extract

  In step S103-1, the prediction device 100 divides the bacterial flora data into the HDP group D1 and the non-HDP group D2, and uses the HDP group D1 for learning data D1a and evaluation at each ratio e (for example, e = 0.9). It is divided into data D1b, and the non-HDP group D2 is divided into learning data D2a and evaluation data D2b.

  In step S103-2, the prediction device 100 selects feature amounts by forward stepwise selection using the learning data D1a and D2a, and creates an HDP prediction model M. For example, for the first 20 feature quantities (bacterial species), the prediction device 100 creates HDP prediction models M (1) to M (20) using only one variable. Then, the prediction device 100 specifies M (A) (using the variable A) with the highest performance, using the evaluation data D1b and D2b. As the performance score, for example, AUC or f-score may be used. Next, using the learning data D1a and D2a, the prediction device 100 selects one variable from the remaining 19 variables excluding the variable, and combines it with A to create a two-variable HDP prediction model M (A , 1) to M (A, 19), and using the evaluation data D1b and D2b, an optimal model including M (A) is selected. The prediction device 100 recursively executes the above-described series of processes, and finally identifies the HDP prediction model M (A, B, C,...).

  Although the process for generating the HDP onset / prediction model is described in the above-described embodiment, it is apparent that the HDP onset day prediction model and the HDP onset risk calculation model can be generated by the same process. The HDP prediction model M may be generated based on SVM, linear logistic regression, neural network, random forest, and the like. Also, the performance score is not limited to AUC or f-score, and a harmonic mean of sensitivity and specificity may be used. Also, not limited to forward step selection, backward stepwise selection or bidirectional stepwise selection may be used. In addition, all bacterial species are used, and after dimensional reduction with PCA, MDS, NMDS, etc., they are converted into N feature quantities of PC1, PC2, ..., PCN, and these are forward / backward / bidirectional. You may apply stepwise selection.

  FIG. 5 is a flowchart showing feature quantity extraction / model creation processing according to another embodiment of the present invention. The feature quantity extraction / model creation process can be executed in steps s102 and s105 of FIG. 3 described above, and steps s102-1 to s5 described later correspond to the feature quantity extraction process of step s102, and step s105 -1 to 3 correspond to the process of generating a birth date prediction model in step s105. Note that steps S102-1 to 5 are the same as steps S102-1 to 5 in FIG.

  In step S105-1, the prediction device 100 divides the bacterial flora data into the HDP group D1 and the non-HDP group D2, and the HDP group D1 is used for learning data D1a and evaluation at each ratio e (for example, e = 0.9). It is divided into data D1b, and the non-HDP group D2 is divided into learning data D2a and evaluation data D2b.

  In step S105-2, in order to improve prediction performance, the prediction device 100 may oversample D1a in the learning data D1a and D2a, and / or undersample D2a.

In step S105-3, the prediction device 100 selects feature quantities by forward stepwise selection using the learning data D1a and D2a, and creates an HDP onset date prediction model M. For example, for the first 20 variables (bacterial species), the prediction device 100 creates HDP onset day prediction models M (1) to M (20) using only one variable. Then, the prediction device 100 specifies M (A) (using the variable A) with the highest performance, using the evaluation data D1b and D2b. As a performance score, for example, MSE (mean squared error) or R ² may be used. Next, using the learning data D1a and D2a, the prediction device 100 selects one variable from the remaining 19 variables excluding the variable, and combines it with A, a two-variable HDP onset day prediction model M (A, 1) to M (A, 19) are created, and the evaluation data D1b and D2b are used to select an optimal model including M (A). The prediction device 100 recursively executes the series of processes described above, and finally identifies the HDP onset day prediction model M (A, B, C,...).

The HDP onset day prediction model M may be generated based on SVM, linear logistic regression, neural network, or the like. Moreover, the performance score, MSE may not be limited to (mean square error) or R ^2. Also, not limited to forward step selection, backward stepwise selection or bidirectional stepwise selection may be used. In addition, after reducing the dimensions by PCA, MDS, NMDS, etc. using all bacterial species, apply forward / backward / bidirectional stepwise selection to N feature values of PC1, PC2, ..., PCN. It is also good.

  FIG. 6 is a flowchart showing feature quantity extraction / model creation processing according to another embodiment of the present invention. The feature quantity extraction / model creation process can be executed in steps s102 and s107 of FIG. 3 described above, and steps s102-1 to s5 described later correspond to the feature quantity extraction process of step s102, and step s107 -1 to 2 correspond to the generation process of the HDP risk model in step s107. Note that steps S102-1 to 5 are the same as steps S102-1 to 5 in FIG.

In step S107-1, the prediction device 100 sets the threshold range S _spA-1 = [quantity for each bacterial species (spA-1, spB-1, ..., spA-2, spB-2, ...). _Set A _spA-1 , B _spa-1 ], and calculate the risk ratio for HDP (RR _spA-1 ). Here, the risk ratio is RR = a (c + d) / c (a + b), and a is the number of dental plaque samples in the HDP group (the amount of bacterial species is within the threshold range), b is the number of plaque samples in the non-HDP group and within the threshold range, c is the number of plaque samples in the HDP group and outside the threshold range, and d is in the non-HDP group and outside the threshold range Is the number of plaque specimens.

In step S107-2, when the amount of bacterial species of the pregnant woman to be predicted is (x _spA-1 , x _spB-1 ,...), The prediction device 100:
RR = Π _sp RR _sp ^{f (x} _sp ⁾
f (x _sp ) = 1 if x _sp is in S _sp
f (x _sp ) = -1 if x _sp is not in S _sp
To generate an HDP risk prediction model.

In addition, not limited to forward step selection, backward stepwise selection or bidirectional stepwise selection may be used. In addition, after reducing the dimensions by PCA, MDS, NMDS, etc. using all bacterial species, apply forward / backward / bidirectional stepwise selection to N feature values of PC1, PC2, ..., PCN. It is also good.
As a result of carrying out the above example, Actinomyces sp., Staphylococcus sp., Prevotella sp., GN 02 sp. Staphylococcus ph. The performance of the HDP onset / prediction prediction model used had a sensitivity of 0.60 and a specificity of 0.91.

  There has been no model for predicting HDP from the dental plaque flora in Japan, but according to the present invention, it is suitable for Japanese by combining multiple bacterial species related to HDP from time-varying bacterial flora. It is possible to create an HDP prediction model, which enables HDP prior prevention and prompt response.

  Although the embodiments of the present invention have been described in detail, the present invention is not limited to the specific embodiments described above, and various modifications may be made within the scope of the subject matter of the present invention described in the claims.・ Change is possible.

10 Prediction System 50 Bacterial Flora Measurement Device 100 Prediction Device 110 Data Acquisition Function 120 Data Storage Function 130 HDP Prediction Model Generation Function 140 HDP Prediction Function 150 Result Notification Function

Claims (5)

Data acquisition function to acquire bacterial flora data on bacterial flora in the oral cavity of pregnant women,
An HDP predictive model generation function of generating a pregnancy hypertensive syndrome (HDP) predictive model based on the bacterial flora data;
An HDP prediction function that executes HDP prediction of the pregnant woman from the acquired bacterial flora data based on the HDP prediction model;
A result notification function for notifying the result of the HDP prediction;
Have
The data acquisition function acquires the bacterial flora data at a plurality of time points in a data acquisition period from any time before or after pregnancy of the pregnant woman to any time before or after delivery.
The prediction apparatus that executes the HDP onset prediction based on the bacterial flora data acquired at the plurality of time points. The HDP prediction function sets each bacterial species of the bacterial flora data acquired at the plurality of time points as an independent variable, and extracts a predetermined number of upper variables that are often used to divide the HDP group and the non-HDP group.
The prediction device according to claim 1, wherein the HDP prediction function generates the HDP prediction model from the extracted variables according to a forward stepwise selection scheme.   The HDP prediction model generation function is a bacterial species that divides HDP group and non-HDP group well, and includes Actinomyces sp., Staphylococcus sp., Capnocytophaga sp., Prevotella sp., Leptotrichia sp., GN02 sp., Streptococcus sp., The prediction device according to claim 1 or 2, wherein one or more of Haemophilus sp., Corynebacterium sp., Aggregatibacter sp., Rothia sp., Leptotrichia sp., Lautropia sp. Are extracted. The HDP prediction model generation function generates an HDP onset presence prediction model, an HDP onset date prediction model, and an HDP risk model,
The HDP prediction function executes HDP onset presence prediction, HDP onset date prediction, and HDP risk calculation based on the generated HDP onset presence prediction model, HDP onset date prediction model, and HDP risk model, respectively. The prediction apparatus as described in any one of 3). Obtaining bacterial flora data on the bacterial flora in the oral cavity of a pregnant woman;
Generating a pregnancy hypertensive syndrome (HDP) prediction model based on the flora data;
Performing HDP onset prediction of the pregnant woman from the acquired bacterial flora data based on the HDP prediction model;
Notifying the result of the HDP prediction.
A prediction method having
The acquiring includes acquiring the bacterial flora data at a plurality of time points in a data acquisition period from any time before or after pregnancy of the pregnant woman to any time before or after delivery.
The performing step is a prediction method of performing the HDP onset prediction based on the bacterial flora data acquired at the plurality of time points.

Images