JP2018140172A - Data collection device and data collection method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a data collection technique for improving accuracy of predicting a premature delivery.SOLUTION: One embodiment of this invention relates to a data collection device which has: a data acquisition function of acquiring data on a pregnant woman from plural data source devices; a premature delivery prediction function of executing prediction of a premature delivery by the pregnant woman from the acquired data on the basis of a premature delivery prediction model; and a result report function of reporting a result of the premature delivery prediction. The data acquisition function acquires the data at plural points of time in a data acquisition period from any time around the pregnancy of the pregnant woman to any time around a delivery.SELECTED DRAWING: Figure 1

Description

  The present invention generally relates to medical diagnostic techniques, and more particularly to a data collection apparatus and method for predicting the onset of premature birth for noninvasively predicting the onset or detecting abnormalities of premature birth.

  Pregnancy-related diseases have been difficult to predict. In other words, the onset of maternally related diseases is not only genetic factors but also environmental factors are involved in a complicated manner, but it is difficult to grasp environmental factors. For this reason, there are many pregnant-related diseases for which no cause has yet been identified and no preventive method has been established.

  In the present obstetric practice, the onset prediction is mainly based on examination information, medical history and / or examination information obtained at the time of consultation, and it is difficult to obtain continuous living environment information such as at home. In the research on the onset of disease, the main method for obtaining the living environment information of pregnant women has been to conduct surveys using questionnaires. However, this method has problems with accuracy due to inaccurate memory of pregnant women and incorrect answers (recall bias), and there is a problem with the frequency of acquisition, such as once every six months, because the implementation interval is early. It was. Therefore, it was an inadequate method for capturing the actual life of pregnant women and clarifying the causal relationship with the disease. In the future, in order to accurately predict the onset of disease, it will be necessary to acquire continuous environmental information other than at the time of consultation and to analyze the information at the time of consultation in an integrated manner. Because it was difficult to obtain, it was not used to predict the onset of pregnancy related diseases.

  Premature birth is a typical perinatal medical disease whose cause is unknown by about one third, and is a multifactorial disease that develops through a complex interaction of genetic and environmental factors.

  As an epidemiological study of premature birth based on the living environment of pregnant women, for example, there is a report that BMI (Body Mass Index) before pregnancy is measured, and that the low BMI increases the risk of preterm birth (prior art documents). 1). However, in this study, since the prediction accuracy is 2.91 in terms of the adjusted odds ratio because it is based on a single parameter and the measurement result at one time point, it cannot be said that the accuracy is clinically sufficient. In addition, focusing on the quality of sleep of pregnant women, there is a report that if the quality of sleep is poor in early pregnancy (14-16 weeks of pregnancy) and late pregnancy (30-32 weeks), the risk of preterm birth is increased (prior art) Reference 2). However, this study is also a questionnaire survey, which includes recall bias, and the prediction accuracy is 1.25 for the odds of early pregnancy and 1.18 for the odds of late pregnancy. Is insufficient.

  Prior art reference 1 and prior art reference 2 both compare only odds ratios and can discuss the risk of premature birth, but can predict the premature birth of the pregnant woman. is not.

"Prepregnancy body mass index is an independent risk factor for gestational hypertension, gestational diabetes, preterm labor, and small- and large-for-gestational-age infants", Dayeon Shin and Won O. Song, The Journal of Maternal-Fetal & Neonatal Medicine, Volume 28, 2015 (http://www.tandfonline.com/doi/abs/10.3109/14767058.2014.964675) "" "Poor Sleep Quality is Associated with Preterm Birth", Michele L. Okun and et al. (Https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3198204/)

  In view of the above-described problems, an object of the present invention is to provide a data collection technique for improving the accuracy of prediction of premature birth.

  In order to solve the above problems, according to one aspect of the present invention, based on a data acquisition function for acquiring data on a pregnant woman from a plurality of data source devices and a preterm birth prediction model, the preterm birth prediction of the pregnant woman is executed from the acquired data. A premature birth prediction function and a result notification function for notifying the result of the preterm birth prediction, wherein the data acquisition function acquires data from any time before and after pregnancy of the pregnant woman to any time before and after childbirth The present invention relates to a data collection device that acquires the data at a plurality of points in time.

  ADVANTAGE OF THE INVENTION According to this invention, the data collection technique for improving the prediction precision of premature birth can be provided.

FIG. 1 is a block diagram showing a system configuration according to an embodiment of the present invention. FIG. 2 is a block diagram showing a functional configuration of the measurement value transfer apparatus according to one embodiment of the present invention. FIG. 3 is a block diagram showing a functional configuration of the physical condition registration device according to one embodiment of the present invention. FIG. 4 is a block diagram showing a functional configuration of a weather information acquisition apparatus according to an embodiment of the present invention. FIG. 5 is a block diagram showing a functional configuration of the survey form registration apparatus according to an embodiment of the present invention. FIG. 6 is a block diagram showing a functional configuration of a medical record registration apparatus according to an embodiment of the present invention. FIG. 7 is a block diagram showing a functional configuration of the data collection device according to one embodiment of the present invention. FIG. 8 is a flowchart showing data collection processing according to an embodiment of the present invention. FIG. 9 is a flowchart showing premature birth prediction model generation processing according to an embodiment of the present invention. FIG. 10 is a graph showing the results of predicting the onset of preterm birth at 24 weeks of gestation. FIG. 11 is a graph showing the results of predicting the onset of preterm birth at 28 weeks of gestation. FIG. 12 is a graph showing the results of predicting the onset of preterm birth at 32 weeks of gestation. FIG. 13 is a flowchart showing a feature amount extraction / classification model creation process according to an embodiment of the present invention. FIG. 14 is a flowchart showing a feature amount extraction / regression model creation process according to an embodiment of the present invention. FIG. 15 is a flowchart showing a feature amount extraction / risk prediction model creation process according to an embodiment of the present invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  In an embodiment described later, a data collection device for predicting preterm birth is disclosed. Since the prediction accuracy is low in the conventional preterm birth prediction based on a single index such as BMI and sleep quality, the data collection device according to the following embodiment predicts the accuracy by combining a plurality of data among the data on the living environment of the pregnant woman. To improve. In addition, the data collection device acquires the data to be acquired at a plurality of time points, for example, continuously or periodically, instead of the measurement value at one time point, and objectively eliminates the recall bias from the acquired data as much as possible. Use simple data.

Specifically, according to the data collection system according to the following embodiment, in order to clarify the environmental factors of pregnancy-related diseases by acquiring the living environment of pregnant women more objectively and more frequently, The following various data are acquired from before.
1) Daily data of blood pressure meters, weight scales, activity meters, thermometers and other measuring devices 2) Physical condition data of daily sleep, bowel movement, nausea, pain, uterine contraction, palpitation, fetal movement interval, meal content, medication content 3 ) Daily living environment data such as temperature, weather, humidity, barometric pressure, tide, etc. of the pregnant woman's residence 4) Regular questionnaire data on lifestyle habits such as eating habits, sleeping conditions, smoking 5) Regular pregnancy checkups, etc. Medical record data By acquiring such data, it is possible to use many kinds of data, and the prediction accuracy can be improved compared to the conventional preterm birth prediction based on a single index. In addition, during 1) to 3) above, during and after the pregnancy period, data is acquired frequently, such as several times a day to several times a week, or continuously during the period. It is possible to eliminate or reduce the recall bias by shortening the interval that a pregnant woman should memorize in objective data acquisition by the above and daily registration in 2). Of the above data, the data items 1) to 3) are defined as life logs.

  Thereby, in this invention, in order to create a preterm birth prediction model, these life logs, survey forms, and medical records can be acquired, and environmental information of pregnant women can be enriched. That is, the present invention is a non-invasive method for predicting the onset of premature birth or detecting abnormalities, such as a blood pressure meter, a body composition monitor, an activity meter, and a thermometer as daily life logs during pregnancy and after delivery of pregnant women. Objective data measured by, etc., daily sleep quality, bowel movement, nausea, pain, uterine contraction, palpitations, fetal movement, diet, medication, etc. By acquiring data related to one or more data items from objective data such as humidity, atmospheric pressure, and tide, and analyzing them in combination, it is possible to realize more accurate preterm birth prediction.

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

  As shown in FIG. 1, the data collection system 10 includes a measurement device 20, a measurement value transfer device 30, a physical condition registration device 40, a weather information acquisition device 50, a survey form registration device 60, a medical record registration device 70, and a data collection device. 100. The illustrated system configuration is merely an example, and some devices may be omitted, or other devices may be added. For example, the data collection system 10 includes one or both of the measurement device 20 and the physical condition registration device 40, the medical record registration device 70, and the data collection device 100. Optionally, the measurement value transfer device 30, the weather information acquisition You may have the apparatus 50 and / or the survey form registration apparatus 60. FIG.

  The measuring device 20 includes a sphygmomanometer, a weight scale, an activity meter, a thermometer, and the like, and measures physiological data such as blood pressure of a pregnant woman.

  The measurement value transfer device 30 transfers data between the measurement device 20 and the data collection device 100. Note that the measurement device 20 itself may include the function of the measurement value transfer device 30, and the measurement value may be directly transmitted to the data collection device 100.

  The physical condition registration device 40 is a device for a pregnant woman to register daily physical condition information, and can be implemented in, for example, a smartphone and an application. Specifically, the pregnant woman inputs her physical condition information into the physical condition registration apparatus 40. Here, the physical condition information may include, for example, data items such as sleep, bowel movement, nausea, pain, uterine contraction, palpitation, fetal movement interval, meal content, and medication content, but is not limited thereto.

  The meteorological information acquisition device 50 acquires meteorological information on a pregnant woman's residence from a publicly available weather information database (not shown). Here, the weather information includes, for example, data items such as weather, precipitation probability, temperature, humidity, wind speed, wind direction, atmospheric pressure, tide, etc. every predetermined period (one hour unit, one day unit, etc.). However, it is not limited to these.

  The survey form registration device 60 is a device for registering information on regular survey forms from pregnant women. Specifically, the person in charge registers information on the survey form acquired from the pregnant woman in the survey form registration device 60. Here, the information on the questionnaire includes, for example, the family composition of pregnant women and their families, the history of pregnant women and their families, pregnancy and childbirth history, dietary habits, sleep status, medication status, smoking and drinking status, exercise status, work status, mental status. It may include data items such as a target situation, but is not limited thereto.

  The medical record registration device 70 is a device for registering a medical record at the time of delivery of a pregnant woman. In addition, medical information such as a regular pregnant medical checkup of a pregnant woman may optionally be registered in the medical record registration apparatus 70.

  As will be described in detail below, the data collection device 100 is one of the measurement device 20, the measurement value transfer device 30, the physical condition registration device 40, the weather information acquisition device 50, the survey form registration device 60, and the medical record registration device 70. Data on the above-described data items is collected from the network 200 through the network 200, a premature birth onset prediction model is generated, and prediction is performed. The network 200 may be composed of a plurality of networks instead of a single network.

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

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

  As shown in FIG. 2, the measurement value transfer device 30 includes a data acquisition function 31, a data transfer function 32, a measurement device authentication function 33, and a user authentication function 34.

  The data acquisition function 31 communicates with one or more measurement devices 20 and acquires data and measurement values from the measurement devices 20. The measurement value transfer device 30 and the measurement device 20 are not limited to the following, but are not limited to Bluetooth (registered trademark), Wi-Fi, NFC, ZigBee, USB connection, mobile communication (3G, LTE, 4G, 5G, etc.) It may be realized by a communication method such as For example, the data items acquired from the measurement device 20 are not limited to the following, and may be physiological data related to pregnant women such as a blood pressure monitor, a weight scale, an activity meter, and a thermometer.

  The data transfer function 32 transfers the acquired data and measurement values to the data collection device 100 via the network 200. The data may be acquired at a high frequency such as several times a day to several times a week during or around the pregnancy period, or continuously during the period, and may be transferred to the data collection device 100.

  The measurement device authentication function 33 authenticates the communication device 20 as a communication destination.

  The user authentication function 34 authenticates a user of the measurement value transfer device 30 such as a pregnant woman.

  Next, a physical condition registration device according to an embodiment of the present invention will be described with reference to FIG. FIG. 3 is a block diagram showing a functional configuration of the physical condition registration device according to one embodiment of the present invention.

  As shown in FIG. 3, the physical condition registration device 40 includes a data registration function 41, a data transfer function 42, and a user authentication function 43.

  The data registration function 41 registers input contents from a user such as a pregnant woman as data. For example, data items input from pregnant women are not limited to the following, but are physical condition data of pregnant women such as daily sleep, bowel movements, nausea, pain, uterine contraction, palpitation, fetal movement interval, meal content, medication content, etc. May be.

  The data transfer function 42 transfers the registered data to the data collection device 100. The data may be acquired at a high frequency such as several times a day to several times a week during or around the pregnancy period, or continuously during the period, and may be transferred to the data collection device 100.

  The user authentication function 43 authenticates a user of a physical condition registration device such as a pregnant woman.

  Next, a weather information acquisition apparatus according to an embodiment of the present invention will be described with reference to FIG. FIG. 4 is a block diagram showing a functional configuration of a weather information acquisition apparatus according to an embodiment of the present invention.

  As shown in FIG. 4, the weather information acquisition apparatus 50 has a data acquisition function 51 and a data transfer function 52.

  The data acquisition function 51 acquires weather information from the weather information database. As the weather information database, for example, there is a publicly available database. For example, the weather information is not limited to the following, and may be living environment data such as the temperature, weather, humidity, atmospheric pressure, and tide of a pregnant woman's residence.

  The data transfer function 52 transfers the acquired data to the data collection device 100. The data may be acquired at a high frequency such as several times a day to several times a week during or around the pregnancy period, or continuously during the period, and may be transferred to the data collection device 100.

  Next, a survey form registration apparatus according to an embodiment of the present invention will be described with reference to FIG. FIG. 5 is a block diagram showing a functional configuration of the survey form registration apparatus according to an embodiment of the present invention.

  As shown in FIG. 5, the survey form registration device 60 has a data registration function 61 and a data transfer function 62.

  The data registration function 61 registers data input by the person in charge of survey form information. For example, questionnaire information on pregnant women includes, but is not limited to, the family composition of pregnant women and their families, the history of pregnant women and their families, pregnancy and childbirth history, dietary habits, sleeping status, medication status, smoking and drinking status, exercise It may be a data item regarding the lifestyle of pregnant women and their families, such as status, work status, and mental status.

  The data transfer function 62 transfers the registered data to the data collection device 100.

  Next, a medical record registration apparatus according to an embodiment of the present invention will be described with reference to FIG. FIG. 6 is a block diagram showing a functional configuration of a medical record registration apparatus according to an embodiment of the present invention.

  As shown in FIG. 6, the medical record registration device 70 has a data registration function 71 and a data transfer function 72.

  The data registration function 71 registers data input by the person in charge of medical record input. For example, the medical record is not limited to the following, and may be a medical record such as a pregnant woman medical checkup recorded by a doctor or the like.

  The data transfer function 72 transfers the registered data to the data collection device 100.

  Next, a data collection apparatus according to an embodiment of the present invention will be described with reference to FIG. FIG. 7 is a block diagram showing a functional configuration of the data collection device according to one embodiment of the present invention.

  As illustrated in FIG. 7, the data collection device 100 includes a data acquisition function 110, a data storage function 120, a preterm birth prediction model generation function 130, a preterm birth prediction function 140, and a result notification function 150.

  The data acquisition function 110 acquires data regarding pregnant women from a plurality of data source devices. In addition, the data acquisition function 110 acquires data on a pregnant woman at a plurality of points in a data acquisition period from any time before and after pregnancy of the pregnant woman to any time before and after childbirth. That is, the data acquisition function 110 is received from a data source device such as the measurement device 20, the measurement value transfer device 30, the physical condition registration device 40, the weather information acquisition device 50, the survey form registration device 60, and / or the medical record registration device 70 described above. In the data acquisition period from any time before and after pregnancy of the pregnant woman to any time before and after childbirth, data of one or more data items related to the pregnant woman are acquired at a plurality of times such as regularly.

  Here, data on pregnant women include, for example, pregnancy age, blood pressure, pulse rate, body weight, BMI, number of steps, calories, body temperature, sleep time, number of waking up during sleep, quality of sleep, number of bowel movements, softness of bowel movements Uterine contraction, head pain, tooth pain, upper abdominal pain, lower abdominal pain, lower back pain, nausea, vomiting, palpitations, fetal interval, meal content, medication content, past premature birth history, pregnancy status, temperature , Room temperature, humidity, barometric pressure, weather conditions, tides, and age.

  The data storage function 120 stores the acquired data regarding pregnant women.

  The preterm birth prediction model generation function 130 generates and holds a preterm birth prediction model. For example, the preterm birth prediction model generation function 130 may generate and update the preterm birth prediction model based on data on pregnant women acquired from a plurality of pregnant women including a pregnant woman who delivered by term birth and a pregnant woman who delivered by premature birth. Specifically, the preterm birth prediction model may be generated according to a model derivation method as described in detail below.

  The preterm birth prediction function 140 executes the preterm birth prediction of the pregnant woman from the acquired data on the pregnant woman based on the preterm birth prediction model. Specifically, the preterm birth prediction function 140 inputs the acquired data on the pregnant woman into the preterm birth prediction model, and executes the preterm birth onset prediction, the birth date prediction, or the preterm birth risk calculation based on the output result of the preterm birth prediction model. May be. That is, in preterm birth prediction, the preterm birth prediction function 140 predicts whether or not the pregnant woman has the possibility of premature birth (predicting the onset of premature birth), predicts the date of birth of the pregnant woman (prediction of birth date), or the pregnant woman The risk of preterm birth may be calculated (risk calculation of preterm birth). The preterm birth prediction function 140 may execute premature birth onset prediction, birth date prediction, or premature birth risk calculation based on different data items depending on the number of weeks of the pregnant woman.

  The result notification function 150 notifies a pregnant woman, a medical staff, and the like of the result of the preterm birth prediction executed by the preterm birth prediction function 140.

Note that each of the above-described devices may be a single device, or a certain device may have some or all of the functions of another device.
Example 1
Next, a data collection process in the data collection apparatus 100 according to an embodiment of the present invention will be described with reference to FIGS. FIG. 8 is a flowchart showing data collection processing according to an embodiment of the present invention.

  As shown in FIG. 8, in step S <b> 100, the data collection device 100 performs the above-described measurement device 20, measurement value transfer device 30, physical condition registration device 40, weather information acquisition device 50, survey form registration device 60, and / or medical care. Data is acquired from all or part of the data source device such as the record registration device 70, and the acquired data is stored. Specifically, the data collection device 100 periodically acquires data in a data acquisition period from any time before and after pregnancy of the pregnant woman to any time before and after childbirth.

  First, the pregnant woman uses the measuring device 20 from an arbitrary time point before or during pregnancy to an arbitrary time point after giving birth or after giving birth. The measurement device 20 includes, for example, a sphygmomanometer, a weight scale, an activity meter, a thermometer, and the like. The sphygmomanometer measures systolic blood pressure, diastolic blood pressure, pulse rate, room temperature, and the like. The weight scale measures body weight, BMI, body fat percentage, skeletal muscle percentage, and the like. The activity meter measures the number of steps, calorie consumption, and the like. A thermometer measures body temperature and the like. The measurement by the measuring device 20 is assumed to be twice a day for morning and evening for the sphygmomanometer and once a morning for the other, but not limited thereto, and can be measured at an arbitrary interval. The measurement value transfer device 30 acquires measurement data from each measurement device 20 and transfers the measurement data to the data collection device 100.

  In addition, the pregnant woman registers her physical condition in the physical condition registration device 40. Although it is assumed that the physical condition registration device 40 uses a physical condition registration application or a web page on a smartphone, the present invention is not limited thereto. Physical condition information includes bedtime, wakeup time, sleep time, number of times you wake up during sleep, quality of sleep, number of bowel movements, softness of fecal pain, total time of nausea, number of vomiting, number of vomiting, One of head pain, tooth pain, lower back pain, upper abdominal pain, lower abdominal pain, painless uterine contractions, painful uterine contractions, palpitation, fetal interval, meal content, medication content It is assumed that the above data items are registered, but the present invention is not limited to these. The registration period of physical condition information is assumed to be from any time point before or during pregnancy to any time point after childbirth or postpartum. As for the physical condition registration interval, main data items are assumed to be once a day, but some data items such as meal contents and medication contents are assumed to be input multiple times a day. Further, these data items can be measured at arbitrary intervals instead of every day.

  Moreover, the weather information acquisition apparatus 50 collects daily weather information. For example, it can be obtained from the website of the Japan Meteorological Agency. The data items to be acquired assume date, precipitation, average temperature, maximum temperature, minimum temperature, total solar radiation, average cloud cover, snow depth, maximum snow depth, maximum instantaneous wind speed, tide and / or age, Furthermore, although hourly temperature, precipitation, wind direction, wind speed, sunshine duration, humidity, atmospheric pressure, and weather conditions are assumed, the present invention is not limited to these. Further, the data unit is not limited to one day or time, and may be a week unit, a month unit, or a year unit.

  In addition, the survey form registration device 60 acquires information on a regular survey form entered by a pregnant woman. The questionnaire information includes data such as the family structure of pregnant women and their families, the history of pregnant women and their families, pregnancy and childbirth history, dietary habits, sleeping status, medication status, smoking and drinking status, exercise status, working status, and mental status. Including but not limited to items. As for the frequency of acquisition, the frequency of survey forms is assumed.

  In addition, the medical record registration device 70 acquires records such as test values and findings when a pregnant woman has received a medical checkup for pregnant women. As for the frequency of acquisition, the frequency of consultation is assumed.

  In step s101, the data collection device 100 detects and removes abnormal values from the acquired data, and interpolates missing values. For example, when a pregnant woman whose daily systolic blood pressure is 120 mmHg is 60 mmHg for one day, the data collection device 100 regards the measurement value as a measurement error instead of an actual value, and removes the measurement value. To do. As an example of this removal criterion, the data collection device 100 calculates the mean μ and standard deviation σ of the systolic blood pressure of the pregnant woman so far, and when a certain measured value is dissociated more than nσ from the average μ, There is a method of removing a measured value (n is an arbitrary numerical value), but is not limited thereto. Further, regarding missing values, there are linear interpolation and spline interpolation using front and rear points, and a method of complementing using a model fitted to measurement points other than missing values, but it is not limited thereto.

  In step s102, the data collection apparatus 100 extracts data items having a large difference in data distribution between the premature birth group and the other groups as feature amounts using the data of pregnant women who have already given birth and the delivery status is determined. . To define the magnitude of the difference in data distribution, for example, indicators such as “the probability that the average value of the data cannot be said to be the same” or “the probability that the data distribution cannot be said to be the same” are available. There are, but are not limited to them. This is useful by methods using Student's t-test, Welch's t-test, multivariate regression, multivariate logistic regression, chi-square test, Fisher's exact test, Lasso regression, Ridge regression, and Elastic Net. There are a method of limiting data items and a method of limiting data items based on the contribution ratio of principal component analysis, but the method is not limited to these.

  In steps s103, s105, and s107, the data collection device 100 creates a preterm birth prediction model using the extracted feature amount. Specifically, in step s103, the data collection device 100 generates a preterm birth prediction model as a classification model for predicting whether or not preterm birth occurs. In step s105, the data collection device 100 generates a birth date prediction model as a regression model for predicting the birth date of each pregnant woman. In s107, the data collection device 100 generates a premature labor risk calculation model for calculating the risk of premature labor. Examples of the regression model include, but are not limited to, a linear regression model, a logistic regression model, a regression model using SVM, and a regression model using a multilayer neural network. The classification model includes linear classification, logistic regression, Bayesian network, SVM, k-nearest neighbor method, random forest, classification model by multilayer neural network, and the like, but is not limited thereto. The preterm birth risk calculation model includes, but is not limited to, odds ratio, contribution risk, relative risk. In addition, steps s103, s105, and s107 may be integrally processed without being separated from s102.

  In steps s104, s106, and s108, the data collection device 100 inputs the data of the pregnant woman to be predicted into the generated preterm birth prediction model, birth date prediction model, and preterm birth risk calculation model, and executes preterm birth prediction. Specifically, in step s104, the data collection device 100 uses the preterm birth prediction model to predict the presence or absence of future preterm birth from the data of the pregnant woman to be predicted. In step s106, the data collection apparatus 100 predicts the date of birth from the data of the pregnant woman to be predicted using the birth date prediction model. Further, in step s108, the data collection device 100 uses the preterm birth risk calculation model to calculate the preterm birth risk from the predicted pregnant woman data. Prediction of premature birth, prediction of final premature birth, prediction that premature birth will not occur in the next n1 weeks (n1 is an arbitrary value), prediction of premature birth in the next n2 weeks (n2 is an arbitrary value), etc. Including, but not limited to.

In step s109, the data collection device 100 notifies the pregnant woman or the medical institution of the premature birth prediction result, the birth date prediction result, and / or the premature birth risk calculation result, and ends the data collection process.
(Example 2)
In addition, in Example 1 mentioned above, the data items used to generate and / or input the preterm birth prediction model, the birth date prediction model or the preterm birth risk calculation model are the age at pregnancy, blood pressure, pulse, body weight, BMI. , Steps, calories burned, body temperature, sleeping time, number of times you wake up during sleep, quality of sleep, number of bowel movements, softness of bowel movement, uterine contraction, head pain, tooth pain, upper abdominal pain, lower abdominal pain Pain, low back pain, nausea, vomiting, palpitation, fetal interval, meal content, medication content, past premature birth history, pregnancy status (multiple births, etc.), temperature, room temperature, humidity, barometric pressure, weather conditions, tide and age It may be limited to the above. This makes it possible to efficiently create and use a model with a smaller number of data.
(Example 3)
FIG. 9 is a flowchart showing premature birth prediction model generation processing according to an embodiment of the present invention. With regard to the above-described second embodiment, the preterm birth prediction model generation process described later uses the magnitude of the difference in data distribution between preterm birth and full term birth, and further reduces the number of data items, depending on the number of gestation weeks. The data items necessary for this are changed and a preterm birth prediction model is generated. The premature birth prediction model generation process can be executed in steps s102 and s103 of FIG. 8 described above. Steps s102-1 to s6 to be described later correspond to the feature amount extraction process of step s102, and step s103-1 ˜11 corresponds to the premature birth prediction model generation process in step s103. In the third embodiment, the preterm birth prediction model generation function 130 divides the acquired data into the preterm birth group and the non-preterm birth group, and there is a large difference in data distribution between the two groups of each data item (that is, distinguish, discriminate, separate) Extract data items, and then select one of the highly correlated data items in each data item to reduce the number of data items, thereby extracting features used to generate the preterm birth prediction model To do. Further, the preterm birth prediction model generation function 130 selects a preterm birth prediction model with high prediction accuracy by sequentially increasing the number of feature amounts from the feature amount having a large difference in data distribution between the preterm birth group and the non-preterm birth group. Furthermore, the preterm birth prediction model generation function 130 predicts from the preterm birth prediction model created using the feature amount extracted from the data of a certain period and the preterm birth prediction model group created using the feature amount extracted from the data of another period. Select a model for predicting preterm birth with high accuracy. Note that the above is an example of a feature quantity extraction method and a preterm birth prediction model selection method used for the preterm birth prediction model, and the preterm birth prediction generation model 130 is not limited thereto. For example, there is a method in which extraction of feature amounts is extracted by a multilayer neural network or Lasso regression to create a preterm birth prediction model.

  As illustrated in FIG. 9, in step s102-1, the data collection device 100 processes some data items of the acquired data items. Specifically, the data collection device 100 is not limited to the following, for some data items, the difference from the data of several days ago, the average and variance of several days, and the derivative of the approximation model of the measured value Deriving second derivative etc.

  First, feature amount extraction processing will be described.

  In step s102-2, the data collection device 100 combines the original data and the data created in s102-1, and extracts the data of pregnant women who have already given birth for model creation.

  In step s102-3, the data collection device 100 assumes that the data items useful for preterm birth change for each pregnancy week, and performs a processing loop for extracting data items (features) useful for each week number. Start. For example, when extracting feature quantities in each week from week 16 to week 32, the in-loop processing is executed with t1 = 16 × 7 (days) as the parameter t1 indicating the number of days of pregnancy, and then t1 = 17 × 7. Finally, t1 = 32 × 7 is executed.

  In step s102-4, the data collection device 100 selects data whose period is from (t1-a) day to (t1 + b) day from the data extracted in s102-2 at the beginning of the in-loop processing, The data is divided into a group in which the pregnant women associated with the data had a premature birth and a group that did not. a and b are arbitrary numerical values. For example, when a model is created in units of one week, a = 0 and b = 6. Further, by increasing a, such as a = 28, it is possible to extract data items after that by using long-term data after b days from data 28 days before t1. Thereafter, the data collection apparatus 100 extracts data items having a large difference in data distribution between the two groups, using the method described in s102. Here, for example, data items whose p value is less than c are extracted by Welch's t-test. Here, c is an arbitrary numerical value, for example, c = 0.05. However, the present invention is not limited to this, and any appropriate discrimination model (for example, a statistical model, a learning model, or the like) that can separate the preterm birth group and the normal birth group may be applied.

  In step s102-5, the data collection device 100 narrows down the number of data items by collecting the extracted data items having high correlation. Here, the correlation coefficient of each data item is calculated, and those having d or more are classified as the same cluster, and the items having the lowest p-value are left in each cluster and the others are excluded. Is reduced to the number of clusters. Here, d is an arbitrary numerical value, for example, d = 0.75. As a cluster configuration method, it is assumed that the mutual data items in the same cluster are all d or more, but is not limited thereto. Note that s102-4 may be executed after executing s102-5, or s102-4 and s102-5 may be executed repeatedly.

  In step s102-6, the data collection device 100 rearranges the remaining data items in ascending order of the p value, and sets these as feature amounts created from the data at the time point t1.

  Next, a method for creating a preterm birth prediction model will be described. Since the model using all of the feature quantities created above is not necessarily optimal, consider creating a model by gradually increasing the number of feature quantities and selecting a model with high prediction accuracy.

  In step s103-1, the data collection device 100 changes the number of feature values for model creation from 1, stores a list of feature value names in the parameter f, and executes the subsequent in-loop processing.

  In step s103-2, the data collection device 100 generates a model using data from the time point (t1−g × 7) to (t1 + h × 7) using the feature amount calculated at the time point t1. Here, g and h are arbitrary numerical values, and may be g = 4 and h = 4, for example. This is because, for example, when making predictions using 25-week data, the model created using the feature values selected in the 24th week (eg “weight, nocturnal blood pressure”) was selected in the 25th week. This is because the prediction accuracy may be higher than a model created using feature quantities (eg, “morning blood pressure, sleep time”). In s103-2, the in-loop processing is executed every week by shifting the time t2 as a parameter and shifting from (t1−g × 7) to (t1 + h × 7) by 7 days.

  In step s103-3, the data collection device 100 creates learning data and evaluation data as preparation for model creation. Specifically, the data collection device 100 extracts data from the period (t2-a) to (t2 + b), and first divides the data into the preterm birth group D1 and the group D2 that did not. The data collection apparatus 100 further divides D1 into D1a and D1b, D2 into D2a and D2b, and the learning data as D1a + D2a, where e is the ratio of dividing the learning data and the evaluation data into evaluative data. The data is D1b + D2b. Here, e is an arbitrary numerical value, and may be e = 0.5, for example. Further, the number of pregnant women used as D2 can be adjusted. For example, the number of pregnant women of D2 can be up to twice the number of pregnant women of D1.

  In step s103-4, the data collection device 100 learns the model by limiting the feature quantity to the feature quantity name list f. For example, a support vector machine (SVM) is used as a prediction model at this time, and learning is performed using learning data. Since a large number of prediction models can finally be formed, this prediction model is set to M (t1, f, t2) for distinction.

  In step s103-5, the data collection device 100 evaluates the accuracy with which the model can predict premature birth using the evaluation data. In the evaluation method, for example, an ROC (Receiver Operating Characteristic) curve having a false positive rate on the horizontal axis and sensitivity on the vertical axis is created, and AUC (Area Under the Curve) which is an area under the curve is calculated. The higher the AUC, the better the prediction accuracy. The data collection device 100 records this evaluation result AUC together with the prediction model M (t1, f, t2). Note that the evaluation method is not limited to this, and it is possible to combine premature birth risk values calculated in the same manner as s102-1 to s103-11 in s107 of FIG. In the above description, f is one feature quantity name list, but a prediction model can be created by combining a plurality of feature quantity name lists. For example, when t1 is 24 weeks (t1 = 24 × 7), the feature quantity name list f1 extracted from the 23-week data and the feature quantity name list f2 extracted from the 24-week data are used, and different feature quantities are used. It is also possible to create and evaluate a prediction model M (t1, f1 + f2, t2) in which the 23-week data is learned from the feature quantity in f1, and the 24-week data is learned from the feature quantity in f2.

  In steps s103-6, s103-7, and s103-8, the data collection device 100 creates models for t2, f, and t1, and ends each loop.

  In step s103-9, the data collection device 100 selects the best model for each number of weeks from among the created many models. For this reason, the data collection apparatus 100 first performs processing in the loop for each number of weeks using t2 as a parameter, taking values such as 16 × 7 (days), 17 × 7, and the like.

  In step s103-10, the data collection device 100 selects models M (t1, f, t2) that have been created at all t1 and all f in order to select the optimum model at time t2, and most of them. A model having a high AUC is selected as an optimum model at time t2. Note that this selection method is not limited to this, and it is also possible to combine the value of the risk of premature birth as calculated by the same method as s102-1 to s103-11 in s107 of FIG. 8 with AUC. is there.

  In step s103-11, the data collection device 100 executes processing at all points in time for t2, and ends the loop.

  As described above, it is possible to create an optimal preterm birth prediction model for each number of weeks. For example, in the preterm birth prediction model for data for 32 weeks, calorie consumption, room temperature, head pain, room temperature difference between morning and evening, number of bowel movements, body temperature in the morning are used as characteristic quantities. In the 28-week premature birth prediction model, calorie consumption, room temperature, softness of bowel movement, number of bowel movements, head pain, number of contractions without pain, room temperature difference between morning and evening, number of times that occurred during bedtime, morning and evening The difference in systolic blood pressure is used. In the 24-week preterm birth prediction model, the number of vomiting, calorie consumption, body temperature, and lower back pain are used as characteristic quantities. In the 20-week preterm birth prediction model, pulse, calorie consumption, and temperature are used. In the 16-week preterm birth prediction model, temperature, age at delivery, and diastolic blood pressure are used.

  In the above-described embodiment, the process of generating the preterm birth prediction model based on the data items that differ according to the number of weeks has been described. However, the birth date prediction model and the premature birth risk calculation model can be generated by the same process. It will be clear.

Further, as can be understood from the above-described embodiment, the premature birth prediction model generation function 130, in creating a model from the feature amount, sequentially determines the number of feature amounts from the feature amount having a large difference in data distribution between the preterm birth group and the non-preterm birth group. A model with high prediction accuracy may be selected by increasing. Further, when creating a model for predicting premature birth from data in a certain period, the premature birth prediction model generation function 130 uses a premature birth prediction model created using the feature amount extracted from the data in the period and data from another period. A model with high prediction accuracy may be selected from the premature birth prediction model group created using the extracted feature quantity. In these, when the number of feature quantities used in the model is set on the horizontal axis and the value of AUC is confirmed, the results shown in FIGS. 10 to 12 are obtained. Looking at the results for the 28th week, it can be seen that the number of features increases, the AUC becomes higher, and the accuracy of predicting preterm birth is increasing. Thereby, it can confirm that prediction accuracy improves by combining a plurality of data. Looking at the results for 32 weeks, even though six feature values were extracted in step s102-6, the one with the highest prediction accuracy was when the top five were used, and step s103. It can be understood that there is a loop effect of the parameter f at -1. Similarly, the prediction accuracy is highest when the top four feature values are used in the results for 24 weeks. In addition, for the 32 week prediction model, it is better to create a model using a feature quantity group extracted from 28 week data than to create a model using a feature quantity group extracted from 32 week data. It can be seen that good prediction accuracy can be realized, and it can be seen that there is an effect of the loop of the parameter t2 in step s103-2. As a result, the AUC indicated by the prediction model for each week's data is 0.82 in the 16-week data, 0.77 in the 20th week, 0.83 in the 24th week, 0.81 in the 28th week, 0.83 in the 32nd week. The prediction accuracy was approximately 0.8. The odds ratio for preterm birth shown by the prediction model for each week's data is 3.1 for the prediction from the 16-week data, 8.6 for the 20th week, 26.6 for the 24th week, 7.3 for the 28th week, 32 It is 8.7 per week, which greatly exceeds the odds ratio of 2.91 shown in the prior art document 1 and the odds ratios of 1.25 and 1.18 shown in the prior art document 2, and uses the data collection system of the present invention. This shows that the prediction of the onset of preterm birth has improved.
(Example 4)
The preterm birth prediction model generation function 130 and the preterm birth prediction function 140 of the data collection device 10 according to the present invention can set the target disease other than premature birth. For example, gestational hypertension syndrome, gestational diabetes, SGA (Small-for-gestational age), FGR (Fetal growth restriction), stillbirth, imminent miscarriage, imminent premature birth, early rupture, pPROM (preterm premature rapture of the membranes), excessive amniotic fluid, One or more of amniotic fluid deficiency, eclampsia, acute fatty liver of pregnancy, HELLP syndrome, infant atopy, infant mental related disorders, infant cardiovascular disease, infant metabolic disorders, and other maternal and fetal disorders It is possible to predict the presence / absence of onset, the onset date, or the risk of onset. It is also possible to predict full-term production.

As described above, according to the present invention, in addition to the conventional survey form and medical records, daily life logs can be collected non-invasively during and before the pregnancy period. It is possible to diversify data items, increase the data acquisition period and the number of acquisition points, minimize data objectivity and recall bias. Furthermore, by combining and analyzing these various data obtained, it is possible to predict the onset of premature birth, leading to advance prevention and quick response. In addition, prediction from daily life logs and detection of abnormalities are possible, and early detection of symptoms and prevention of seriousness are promoted by prompting consultations as needed without waiting for a pregnant woman medical checkup once every several weeks. , Leading to quick action.
(Example 5)
FIG. 13 is a flowchart showing the feature amount extraction / premature birth occurrence prediction model generation process according to an embodiment of the present invention. Compared to Example 3 described above, in the processing described later, by using data of a plurality of weeks, not only the magnitude of the difference in data distribution between premature delivery and full-term delivery but also the time-series change of the data distribution. A prediction model with higher performance can be generated by using the magnitude of the difference. The definition of the magnitude of the difference in time-series changes in the data distribution includes, for example, the fact that the data distribution of a certain data item in preterm birth and full-term birth changes according to the number of gestation weeks. However, this is not a limitation. For example, a certain data item for full-term birth declines according to the number of gestational weeks, whereas no decline is seen for preterm birth. This is useful by methods using Student's t-test, Welch's t-test, multivariate regression, multivariate logistic regression, chi-square test, Fisher's exact test, Lasso regression, Ridge regression, and Elastic Net. There are a method of limiting data items and a method of limiting data items based on the contribution ratio of principal component analysis, but the method is not limited to these.

  The feature quantity extraction / premature birth occurrence prediction model creation process can be executed in steps s102 and s103 in FIG. 8 described above, and steps s202-1 to s7-7 described later correspond to the feature quantity extraction process in step s102. Steps s203-1 to s <b> 203-6 correspond to the generation process of the preterm birth onset prediction model in step s <b> 103.

  In the fifth embodiment, the preterm birth prediction model generation function 130 divides the acquired data into the preterm birth group and the full term birth group, and the data distribution is different (ie, distinguishes, discriminates, separates) in two groups among the respective data items. Extract data items, but extract confounding factors including multiple weekly data and static data (age, non-pregnant BMI, height, medical history, delivery history, etc.) and adjust the effects ,Extract. Furthermore, the feature-value utilized for the production | generation of a premature delivery prediction model is extracted by reducing the number of data items by selecting one among data items with high correlation in each data item. In addition, the preterm birth prediction model generation function 130 extracts a predetermined number of high-order feature quantities from among the feature quantities having different data distributions in the preterm birth group and the regular birth group, and creates a preterm birth prediction model using the extracted feature quantities. Then, the premature birth prediction model with the best performance feature quantity is specified. Next, the premature birth prediction model generation function 130 creates a two-variable prediction model that combines the remaining feature quantity excluding the feature quantity and the identified feature quantity, and selects an optimal model including these models. select. The preterm birth prediction model generation function 130 recursively executes the series of processes described above, and finally identifies the preterm birth prediction model. The above is an example of a feature quantity extraction method and a premature birth prediction model selection method used for the premature birth prediction model, and the premature birth prediction model generation function 130 is not limited thereto. For example, there is a method in which extraction of feature amounts is extracted by a multilayer neural network or Lasso regression to create a preterm birth prediction model.

  As illustrated in FIG. 13, in step s202-1, the data collection device 100 processes some data items of the acquired data items. Specifically, the data collection device 100 is not limited to the following, for some data items, the difference from the data of several days ago, the average and variance of several days, and the derivative of the approximation model of the measured value Deriving second derivative etc.

  First, feature amount extraction processing will be described.

  In step s202-2, the data collection device 100 combines the original data and the data created in s202-1 and extracts pregnancy data and time-lapse data of pregnant women who have already given birth for model creation. Pregnancy data is data of static data items that do not depend on the number of weeks of pregnancy, and includes gestational age, non-pregnant BMI, height, weight, birth history, past medical history, family history, and the like. The time-lapse data is an enumeration of data obtained by aggregating the measurement values of a specific period during pregnancy, for example, one week can be specified as a specific period, user ID, blood pressure of 16 weeks, pulse of 16 weeks, The total values of each pregnant woman are described in the time direction such as body temperature at 16 weeks,..., Blood pressure at 17 weeks, pulse at 17 weeks, body temperature at 17 weeks,. An arbitrary period can be specified as the specific period. Moreover, as a totaling method of the measured values in a specific period, an arbitrary method such as an average value or a median value can be designated.

  In step s202-3, with respect to the data at the time of pregnancy, data of a group in which the pregnant women associated with the data are prematurely delivered and a group in which the premature delivery is established are extracted. Thereafter, the data collection device 100 extracts data items having different data distributions between the two groups by using the method described in s102, and sets the data items as confounded data items. Here, for example, data items whose p-value is less than a (for example, a = 0.01) are extracted by multivariate logistic regression. Further, a seasonal term may be included in the confounding data item. The present invention is not limited to this, and any appropriate discrimination model (for example, a statistical model, a learning model, or the like) that can separate the preterm birth group and the full-term birth group may be applied.

  In step s202-4, the correlation coefficient between the confounded data items is calculated, and those having a correlation of b (for example, b = 0.75) or more are divided into the same cluster, and the item having the lowest p value is left in each cluster. By excluding others, the number of confounding data items is reduced to the number of clusters. As a cluster configuration method, it is assumed that all data items in the same cluster are b or more, but the method is not limited to this. Note that s202-3 may be executed after executing s202-4, or s202-3 and s202-4 may be executed repeatedly.

  Next, in step s202-5, the data collection device 100 calculates the p-value and odds ratio by logistic regression by adding confounding data items to each data item one by one with respect to the pregnancy data and the time-lapse data, and p By selecting the data item whose value is less than c (for example, c = 0.01), the data item that significantly divides the two groups is specified after removing the influence of the confounding factor.

  In step s202-6, a correlation coefficient between these data items is calculated, and those having a correlation of d (for example, d = 0.75) or more are divided into the same cluster, and the item having the lowest p-value is left in each cluster. The number of data items is reduced to the number of clusters by excluding other items. The cluster configuration method is the same as in s202-4.

  Next, in step s202-7, the data collection device 100 rearranges the remaining data items in ascending order of the p-value, and sets these as feature amounts for creating a preterm birth prediction model. In addition, information such as data item names, odds ratios, p values, and regression coefficients β are stored as feature quantity tables.

  Next, a method for creating a preterm birth prediction model will be described.

  Since the model using all of the feature quantities created above is not necessarily optimal, consider creating a model by gradually increasing the number of feature quantities and selecting a model with high prediction accuracy.

  In step s203-1, the data collection device 100 assumes that the risk of subsequent preterm onset changes according to the progress of pregnancy, and constructs a predictive model for the onset of premature birth every certain time unit. The certain time unit can be, for example, a week, but is not limited thereto. Here, it is assumed that a prediction model for each week (week t) from the 16th to 36th weeks of pregnancy is constructed.

  In step s203-2, the data collection device 100 extracts the pregnancy time data and the time data before t weeks from the feature amount specified in s202-7 from the pregnancy time data and time data of s202-2.

  In step s203-3, the data collection device 100 creates learning data and evaluation data as preparation for model creation. Specifically, the data collection device 100 first divides into the preterm birth group D1 and the full term birth group D2. The data collection apparatus 100 further divides D1 into D1a and D1b, D2 into D2a and D2b, and the learning data as D1a + D2a, where e is the ratio of dividing the learning data and the evaluation data into evaluative data. The data is D1b + D2b. Here, e is an arbitrary numerical value, and may be e = 0.5, for example. Further, the number of pregnant women used as D2 can be adjusted. For example, the number of pregnant women of D2 can be up to twice the number of pregnant women of D1.

  In step s203-4, the prediction apparatus 100 may oversample D1a and / or undersample D2a in the learning data D1a and D2a in order to improve prediction performance.

In step s203-5, the data collection device 100 selects feature amounts by forward stepwise selection using the learning data D1a and D2a, and learns the preterm birth prediction model M _t at the time point _t . For example, a support vector machine (SVM) is used as a prediction model at this time, and learning is performed using learning data. For example, for the first 100 feature quantities (pregnancy data and time-lapse data), the prediction apparatus 100 uses only one feature quantity to predict the premature birth prediction model M _t (1) to M _t (100) at time _t. Create Then, the prediction apparatus 100 identifies M _t (A) (using the feature amount A) with the best performance using the evaluation data D1b and D2b. As the performance score, for example, AUC (Area under the curve), f-score, harmonic average of sensitivity and specificity may be used. Next, the prediction device 100 selects one feature amount from the remaining 99 feature amounts excluding the feature amount using the learning data D1a and D2a, and uses a two-variable premature birth prediction model combined with A. create an _{M t (a, 1) ~M} t (a, 99), evaluation data D1b, with D2b, including M _t (a) selecting an optimum model M _t (a, B) . The prediction device 100 recursively executes the series of processes described above, and finally identifies the preterm birth prediction model M _t (A, B, C,...).

  Note that the preterm birth prediction model M may be generated based on SVM, linear logistic regression, neural network, random forest, or the like. Further, the method for selecting the optimum feature quantity is not limited to forward step selection, and backward stepwise selection or bidirectional stepwise selection may be used. Also, after using all data items and reducing dimensions with PCA, MDS, NMDS, etc., apply forward / backward / bidirectional stepwise selection to N features of PC1, PC2, ..., PCN. Also good.

In step s203-6, the data collection device 100 creates models at all points in time for t, and ends the loop.
(Example 6)
FIG. 14 is a flowchart showing a feature quantity extraction / birth date prediction model creation process according to another embodiment of the present invention. By using the data of multiple weeks, it is possible to generate a prediction model with higher performance using not only the difference in data distribution between premature delivery and full-term delivery but also the difference in time-series changes in data distribution. The feature quantity extraction / model creation process can be executed in steps s102 and s105 of FIG. 8 described above, and the above-described steps s202-1 to s20-7 correspond to the feature quantity extraction process of step s102. Steps s205-1 to 6 correspond to the birth date prediction model generation process in step s105. 14 is the same as steps s202-1 to s72-1 to 7 in FIG. 13, and steps s205-1 to s205-6 are the same as steps s203-1 to s203-6 in FIG. The explanation is omitted because it is the same.

In step s205-5 in FIG. 14, the data collection device 100, the learning data D1a, select the feature quantity by forward stepwise selection with D2a, learns the birth date predictive model M _t at t weeks time. For example, a support vector machine (SVM) is used as a prediction model at this time, and learning is performed using learning data. For example, for the first 100 feature quantities (pregnancy data and time-lapse data), the data collection device 100 uses only one feature quantity to predict the premature birth prediction model M _t (1) to M _t (100 ). Then, the data collection device 100 identifies M _t (A) (using the feature amount A) with the best performance using the evaluation data D1b and D2b. As the performance score, for example, MSE (mean square error) or R ² (determination coefficient) may be used. Next, the data collection device 100 selects one feature amount from the remaining 99 feature amounts excluding the feature amount using the learning data D1a and D2a, and predicts premature birth of two variables combined with A Create models M _t (A, 1) to M _t (A, 99), and use evaluation data D1b and D2b to select the optimal model M _t (A, B) including M _t (A) To do. The data collection device 100 recursively executes the series of processes described above, and finally identifies the preterm birth prediction model M _t (A, B, C,...).

Note that the preterm birth prediction model M may be generated based on SVM, linear regression, neural network, or the like. Further, the method for selecting the optimum feature quantity is not limited to forward step selection, and backward stepwise selection or bidirectional stepwise selection may be used. Also, after using all data items and reducing dimensions with PCA, MDS, NMDS, etc., apply forward / backward / bidirectional stepwise selection to N features of PC1, PC2, ..., PCN. Also good.
(Example 7)
FIG. 15 is a flowchart showing a feature quantity extraction / premature birth risk prediction model creation process according to another embodiment of the present invention. By using the data of multiple weeks, it is possible to generate a prediction model with higher performance using not only the difference in data distribution between premature delivery and full-term delivery but also the difference in time-series changes in data distribution. The feature quantity extraction / model creation process can be executed in steps s102 and s107 of FIG. 8 described above, and the above-described steps s202-1 to s20-7 correspond to the feature quantity extraction process of step s102, and will be described later. Steps s207-1 to 6 correspond to the process for generating a preterm birth risk prediction model in step s107. Steps s202-1 to s7-7 in FIG. 15 are the same as steps s102-1 to s7-7 in FIG. 13, and steps s207-1 to s207-6 are steps s203-1 to s203-6 in FIG. The explanation is omitted because it is the same.

In step s207-5 of FIG. 15, the data collection device 100 creates a premature birth risk prediction model M _t for the number of weeks t by the same method as in s203-5. The model used at this time includes multivariate logistic regression. To do.

The data collection device 100 of FIG. 8 can calculate the probability p of premature birth by applying feature data to the preterm birth risk prediction model M _t1 in a pregnant woman who has not given birth in t1 gestation week in s108. . As mentioned above, although the Example of this invention was explained in full detail, this invention is not limited to the specific embodiment mentioned above, In the range of the summary of this invention described in the claim, various deformation | transformation・ Change is possible.

DESCRIPTION OF SYMBOLS 10 Data collection system 20 Measurement apparatus 30 Measurement value transfer apparatus 40 Physical condition registration apparatus 50 Weather information acquisition apparatus 60 Survey form registration apparatus 70 Medical record registration apparatus 100 Data collection apparatus 200 Network

Claims (8)

A data acquisition function for acquiring data about pregnant women from multiple data source devices;
Based on a preterm birth prediction model, a preterm birth prediction function for performing the preterm birth prediction of the pregnant woman from the acquired data;
A result notification function for notifying the result of the preterm birth prediction;
Have
The data acquisition function is a data collection device that acquires the data at a plurality of time points in a data acquisition period from any time before and after pregnancy of the pregnant woman to any time before and after childbirth.   Data on the pregnant woman include age at pregnancy, blood pressure, pulse rate, body weight, BMI, number of steps, calorie consumption, body temperature, sleep time, number of times awakened during sleep, quality of sleep, number of bowel movements, softness of bowel movements, uterine contraction , Head pain, tooth pain, upper abdominal pain, lower abdominal pain, lower back pain, nausea, vomiting, palpitation, fetal interval, meal content, medication content, past premature birth history, pregnancy status, temperature, room temperature, The data collection device according to claim 1, comprising one or more data items of humidity, atmospheric pressure, weather conditions, tide, and age.   Data on the pregnant woman include age at pregnancy, blood pressure, pulse rate, body weight, BMI, number of steps, calorie consumption, body temperature, sleep time, number of times awakened during sleep, quality of sleep, number of bowel movements, softness of bowel movements, uterine contraction , Head pain, tooth pain, upper abdominal pain, lower abdominal pain, lower back pain, nausea, vomiting, palpitation, fetal interval, meal content, medication content, past premature birth history, pregnancy status, temperature, room temperature, The data collection device according to claim 1 or 2, comprising a combination of two or more data items of humidity, atmospheric pressure, weather conditions, tide, and age.   The premature birth prediction function uses the acquired data, and uses the magnitude of the difference in data distribution between preterm birth and full-term birth or / and the magnitude of the difference in time series change in data distribution in the preterm birth prediction model. The data collection device according to any one of claims 1 to 3, wherein the premature birth occurrence prediction, birth date prediction, or preterm birth risk calculation of the pregnant woman is executed.   The data collection device according to claim 4, wherein the preterm birth predicting function executes premature birth occurrence prediction, birth date prediction, or preterm birth risk calculation based on data items that differ depending on the number of weeks of the pregnant woman.   Based on data on the pregnant women obtained from multiple pregnant women, including pregnant women who delivered at term and those who delivered prematurely, the magnitude of the difference in data distribution between preterm and term products and / or time series of data distribution The data collection device according to any one of claims 1 to 5, further comprising a premature birth prediction model generation function for generating a preterm birth prediction model using a magnitude of a difference in a global change.   The premature birth prediction model generation function determines a data item for discriminating a pregnant woman who gave birth by the term birth and a pregnant woman who gave birth by the premature birth, and the data distribution difference between preterm delivery and term delivery is determined by the determined data item. The data collection apparatus according to claim 6, wherein the premature birth prediction model is generated using a magnitude or / and a magnitude of a difference in time-series change in data distribution. Obtaining data on pregnant women from a plurality of data source devices;
Performing a preterm birth prediction of the pregnant woman from the acquired data based on a preterm birth prediction model;
Notifying the result of the preterm birth prediction;
Have
The step of executing the preterm birth prediction of the pregnant woman includes obtaining the data at a plurality of time points in a data obtaining period from any time before and after pregnancy of the pregnant woman to any time before and after giving birth. Collection method.

Images