KR20190069047A - Apparatus and method for predicting disease - Google Patents

Abstract

The present invention relates to a method for predicting a disease which comprises the steps of: receiving health data of a target to be predicted; generating a disease prediction model by using electronic health records (EHR) and integrated data set generated based on a mobile personal health records (mPHR) obtained through a mobile apparatus; and predicting a disease possibility for the target to be predicted corresponding to the health data based on the disease prediction model.

Description

[0001] APPARATUS AND METHOD FOR PREDICTING DISEASE [0002]

The present invention relates to an apparatus and method for predicting a disease.

Recently, a large amount of medical data stored in hard copy format has been rapidly digitized and accumulated in the medical industry. In addition, with the rapid development of medical devices and platforms based on Internet of Things (IoT), and with the spread of mobile devices, medical big data collection for individual users has become possible. Many organizations and hospitals can obtain valuable information from the vast collection of digitized electronic health records (EHR). In recent years, a variety of medical studies using EHR have been performed to obtain meaningful information. In other words, there is a growing awareness of the importance of medical big data analysis, and it is therefore necessary to find meaningful information effectively from vast amounts of data using techniques such as data mining.

The EHR used to improve the quality of care and patient health in most medical institutions consists of health information (data) related to the overall health care of the electronically recorded patient. Medical big data such as EHR are difficult to analyze by existing methods due to existence of various data types and high dimensional data. Furthermore, since the EHR is relatively static compared to the mPHR described later, it is difficult to grasp the current health state of the patient only by using the EHR. Nevertheless, the EHR has a great advantage in that it allows the patient to provide a higher level of health care services, since the EHR contains the care and prescription information provided by the healthcare professional.

On the other hand, since mobile personal health records (mPHRs) are collected through mobile medical devices, it is difficult to confirm the accuracy and reliability of data. This is because the mPHR data is not collected by a medical professional or collected using a hospital medical device. These mPHRs are difficult to use for in-depth analysis related to disease. However, in the case of mPHR, there is a great merit in that the state of the patient is measured and updated in a relatively short period of time as compared with the EHR, so that the present state of the patient is more accurately reflected. Conventionally, extensive studies related to EHR and mPHR have been performed in various aspects in order to provide improved medical services considering the limitations of EHR and mPHR.

For example, in a paper entitled "A Hybrid Outlier Detection Method for Healthcare Big Data," Ke Yan, Xiaoming You, Xiaobo Ji, Guangqiang Yin, Fan Yang, 2016 IEEE International Conference on Big Data and Cloud Computing (BDCloud) A new hybrid called pruning-based K-Nearest Neighbor (PB-KNN) is proposed in [2], Networking (SocialCom), Sustainable Computing and Communications (SustainCom) (2016) outlier detection method. This approach overcomes the difficulties of data analysis in the medical field, which involves large amounts of data, diverse data types, and high-dimensional data by integrating density-based, cluster-based methods and KNN algorithms. However, since the proposed method uses only EHR, it is difficult to reflect the current state of users.

Giles Clermont, Journal of Biomedical Informatics, Volume 46, Issue 1 (2013) pp. 47-47. [CrossRef], [Web of Science ®] 55] proposed a new data-based monitoring and alerting framework. The purpose of this paper is to detect medical outlier information using past patient cases stored in EHR. However, since the above-described technique utilizes only the previously recorded EHR, it is difficult to confirm the actual status of the patient. In other words, the above paper also has limitations in the analysis because it is difficult to reflect on the current state of the user.

It is an object of the present invention to solve the problems of the prior art described above and to solve the problem that is difficult to confirm the current state of the patient when using the EHR in providing the medical service and the problem that the accuracy and reliability of the data can not be confirmed when using the mPHR The goal is to provide an integrated data set with integrated EHR and mPHR.

The present invention has been made to solve the above-mentioned problems of the prior art, and it is an object of the present invention to provide an integrated data set generating apparatus and method and a disease predicting apparatus and method for enabling more accurate and reliable health related diagnosis or analysis, The purpose is to provide.

It is to be understood, however, that the technical scope of the embodiments of the present invention is not limited to the above-described technical problems, and other technical problems may exist.

According to a first aspect of the present invention, there is provided a method of generating a data set for providing a healthcare service, the method including generating a first data set including electronic health records (EHR) Performing preprocessing on data contained in each second data set including mobile personal health records (mPHR) obtained via the device; And generating an integrated data set in which the data of the preprocessed first data set and the data of the preprocessed second data set are integrated through matching between the data of the preprocessed first data set and the data of the preprocessed second data set Step < / RTI >

The pre-processing may further include, when it is determined that a missing value or an ideal value exists in the data included in each of the first data set and the second data set, By replacing the attribute corresponding to the missing value or the outliers with the average value of attribute values belonging to the same attribute.

In addition, the generating step may generate the integrated data set through the matching in consideration of whether or not the data attribute of the first data set is the same as the data attribute of the second data set.

If the data attribute of the first data set is not equal to the data attribute of the second data set, the generating of the data attribute of the second data set may be performed on the data of the first data set And if the data attribute of the first data set is determined to be the same as the data attribute of the second data set, the data attribute of the first data set may be overwritten with the data attribute of the second data set have.

The method of generating a data set for providing a healthcare service according to the first aspect of the present invention may further include generating the first data set and the second data set prior to the step of performing the preprocessing have.

In addition, the step of generating the first data set and the second data set may include generating a first data set and a second data set, the data being provided from a University of California-Irvine (UCI) Machine Learning Repository, Data may be integrated to generate the first data set, and the second data set may be generated such that an attribute of data acquired through the mobile device matches an attribute of the first data set.

According to a second aspect of the present invention, there is provided a data set generating apparatus for providing a healthcare service, comprising: a first data set including electronic health records (EHR); and a mobile personal health a preprocessing unit for performing preprocessing on data included in each of the second data sets including the first and second data sets; And generating an integrated data set in which the data of the preprocessed first data set and the data of the preprocessed second data set are integrated through matching between the data of the preprocessed first data set and the data of the preprocessed second data set And may include a matching unit.

According to a third aspect of the present invention, there is provided a data set generation system for providing a healthcare service, comprising: a plurality of medical institution's apparatuses providing electronic health records (EHR) for a plurality of users; A plurality of mobile devices measuring and providing mobile personal health records (mPHR) for each of a plurality of users; And generating a second data set comprising the first data set including the electronic health record and the personal health record acquired through the mobile device to generate data included in each of the first data set and the second data set And the data of the preprocessed first data set and the data of the preprocessed second data set are combined through the preprocessing of the data of the preprocessed first data set and the preprocessed second data set, And a data set generation device for generating an integrated data set.

A computer program according to the fourth aspect of the present invention may be stored on a recording medium for executing the method of generating a dataset according to the first aspect of the present application.

According to a fifth aspect of the present invention, there is provided a disease predicting method comprising: receiving health data of a subject to be predicted; Generating a disease prediction model using an integrated data set generated based on electronic health records (EHR) and mobile personal health records (mPHR) obtained via a mobile device; And predicting a disease potential for a predictive subject corresponding to the health data based on the disease predictive model.

Also, the generating step may generate the disease prediction model considering at least one of the characteristics of the predictive subject and the type of disease to be predicted after the health data is received.

Also, the generating step may include generating a plurality of disease prediction models in which characteristics of a plurality of users are taken into consideration based on data included in the integrated data set, and after the health data is received, And selecting an optimal disease prediction model optimized for the predicted subject among the plurality of disease prediction models considering at least one of the types of disease to be diagnosed based on the optimal disease prediction model, The possibility of disease for the predicted subject can be predicted.

In addition, the disease prediction model may include a plurality of rules related to a plurality of attributes of data included in the integrated data set, the number of combinations of the plurality of rules and the combination order may be determined based on characteristics of the predicted subject, Type, and the like.

In addition, the disease prediction model may be generated by applying a decision tree to the data included in the integrated data set.

The integrated data set may also be configured to perform pre-processing on data contained in each of a first data set comprising the electronic health record and a second data set including a personal health record obtained through the mobile device, Can be generated by matching between the data of the first data set and the data of the preprocessed second data set.

According to a sixth aspect of the present invention, there is provided a disease predicting apparatus comprising: a receiving unit for receiving health data of a subject to be predicted; A generating unit for generating a disease prediction model using an integrated data set generated based on electronic health records (EHR) and mobile personal health records (mPHR) obtained through a mobile device; And a predictor for predicting a disease probability of a predictive subject corresponding to the health data based on the disease predictive model.

A disease prediction system according to a seventh aspect of the present invention comprises: a first terminal device for providing health data of a predicted subject; And means for receiving the health data from the first terminal device and generating an integrated data set based on electronic health records (EHR) and mobile personal health records (mPHR) And a disease predicting device for predicting a disease potential of a predictive subject corresponding to the health data based on the disease predictive model and providing a prediction result.

The computer program according to the eighth aspect of the present invention may be stored in a recording medium for executing the disease predicting method according to the fifth aspect of the present invention.

The above-described task solution is merely exemplary and should not be construed as limiting the present disclosure. In addition to the exemplary embodiments described above, there may be additional embodiments in the drawings and the detailed description of the invention.

According to the above-described task solution of the present invention, by providing an integrated data set in which the EHR and the mPHR are integrated, it is possible to solve the problem that is difficult to confirm the current state of the patient when using the EHR in providing the medical service, It is possible to solve the problem that is difficult to confirm reliability.

According to the above-described task solution of the present invention, it is possible to provide a disease prediction model generated using an integrated data set and an integrated data set to provide a more accurate and reliable medical service (that is, Analysis) can be performed.

However, the effects obtainable here are not limited to the effects as described above, and other effects may exist.

FIG. 1 is a block diagram illustrating a data set generation system for providing a healthcare service according to an embodiment of the present invention. Referring to FIG.
2 is a diagram illustrating attributes of an EHR considered in a data set generating apparatus for providing a healthcare service according to an embodiment of the present invention.
3 is a diagram showing a schematic configuration of a disease prediction system according to an embodiment of the present invention.
FIG. 4 is a diagram illustrating an example of a disease prediction model generated by applying a decision tree to an integrated data set according to an embodiment of the present invention. Referring to FIG.
5 is a diagram showing an example of a disease prediction model generated by applying a decision tree to a conventional EHR.
6 is a diagram illustrating an example of a disease prediction model generated by applying a decision tree to a conventional mPHR.
Figure 7 illustrates the accuracy of an integrated data set based disease prediction model in accordance with one embodiment of the present disclosure versus EHR or mPHR.
8 is a flowchart illustrating a method of generating a data set for providing a healthcare service according to an embodiment of the present invention.
9 is a flowchart illustrating an operation of the disease prediction method according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention. It should be understood, however, that the present invention may be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In the drawings, the same reference numbers are used throughout the specification to refer to the same or like parts.

Throughout this specification, when an element is referred to as being "connected" to another element, it is intended to be understood that it is not only "directly connected" but also "electrically connected" or "indirectly connected" "Is included.

It will be appreciated that throughout the specification it will be understood that when a member is located on another member "top", "top", "under", "bottom" But also the case where there is another member between the two members as well as the case where they are in contact with each other.

Throughout this specification, when an element is referred to as "including " an element, it is understood that the element may include other elements as well, without departing from the other elements unless specifically stated otherwise.

We propose the creation of integrated data sets for healthcare services, including electronic health records (EHR) and mobile personal health records (mPHR) obtained via mobile devices.

Briefly, the EHR can be made up of 920 records that facilitate high-quality care. These records may include many of the characteristics associated with heart disease measured at a medical facility. However, the EHR is relatively static compared to the mPHR, which can be attributed to the fact that the information is updated only when the user (patient) receives medical treatment at the hospital. Therefore, it may be difficult to accurately diagnose the current health condition of a patient when analyzing or diagnosing a user's health condition using only EHR.

Therefore, in order to overcome these limitations, we propose a technique to generate an integrated data set with integrated EHR and mPHR. Here, the mPHR may mean a set (data set) of data periodically measured by the mobile device. When the mPHR is used in the analysis or diagnosis of the health status of the user, accurate analysis / diagnosis of the user's current state Lt; / RTI > A detailed description of how to create an integrated dataset follows.

FIG. 1 is a diagram showing a schematic configuration of a data set generation system (hereinafter referred to as 'present data set generation system') for providing a healthcare service according to an embodiment of the present invention.

1, the data set generation system 100 includes a plurality of medical institution apparatuses 10, a plurality of mobile apparatuses 20, and a data set generation apparatus 30 for providing healthcare services Generating device ').

The devices 10 of a plurality of medical institutions may provide electronic health records (EHR) for a plurality of users. A plurality of medical institution apparatuses 10 can electronically record and provide information (data) related to the health of a plurality of users (patients) as EHRs. That is, the EHR can be provided by the apparatus 10 of a plurality of medical institutions. The health information (data) of the user obtainable from the EHR may include, but is not limited to, information such as age, sex, cholesterol, blood sugar, and blood pressure. The medical institutions corresponding to the plurality of medical institution apparatuses 10 may include hospitals, oriental hospitals, dental hospitals, university hospitals, midwives, clinics, oriental clinics, dental clinics, nursing hospitals, general hospitals, and the like. But is not limited thereto.

A plurality of mobile devices 20 may measure and provide mobile based personal health records (mPHR) for each of a plurality of users. That is, mPHR, a mobile-based personal health record, can be measured and provided through mobile devices owned by each user. The health-related information (data) of the user obtainable from the mPHR may include, but is not limited to, blood pressure, heart rate, blood sugar, and active calorie.

The mobile device 20 is a mobile communication device with guaranteed portability and mobility, for example, a PCS (Personal Communication System), a GSM (Global System for Mobile communication), a PDC (Personal Digital Cellular), a PHS (Personal Handyphone System) , PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication) -2000, CDMA (Code Division Multiple Access) -2000, W-CDMA (WCode Division Multiple Access), Wibro (Wireless Broadband Internet) , Smart pads, tablet PCs, notebooks, wearable devices, and the like, but is not limited thereto.

The data set generating apparatus 30 includes a first data set including an electronic health record (EHR) acquired from a plurality of medical institution apparatuses 10 and a mobile-based personal health record obtained from a plurality of mobile apparatuses 20 (mPHR) to perform preprocessing on the data contained in each of the first data set and the second data set, and preprocesses the data of the preprocessed first data set and the preprocessed second data set The data of the first data set and the data of the preprocessed second data set, which are preprocessed through matching, can be generated.

Here, the EHR obtained by the data set generating apparatus 30 from the apparatus 10 of the plurality of medical institutions and the mPHR obtained from the plurality of mobile apparatuses 20 can be obtained through the network 40. [

The network 40 may be, for example, a 3rd Generation Partnership Project (3GPP) network, a Long Term Evolution (LTE) network, a World Interoperability for Microwave Access (WIMAX) network, (LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), a Bluetooth network, an NFC (Near Field Communication) network, a satellite broadcasting network, an analog broadcasting network, a DMB (Digital Multimedia Broadcasting) But is not limited thereto.

Since the present data set generating apparatus 30 includes the medical care and prescription information provided by the medical professional in the case of the EHR, it can provide the reliable medical information to the user, while it is relatively static compared to the mPHR It is difficult to accurately grasp the user's current health state only by using the EHR. In the case of mPHR, since the health information of the user is measured and updated in a relatively short period of time as compared with the EHR, EHR and mPHR generate an integrated data set, taking into account the fact that the information obtained versus being available is not collected by a medical professional or collected using a hospital medical device, Technology. A more detailed description of the data set generating apparatus 30 is as follows.

Hereinafter, the data set generating apparatus 30 will be described in detail. The data set generating apparatus 30 includes an integrated data generating unit 30 for reflecting a current state of a user, for example, You can create a set. In other words, the present invention describes a technique for generating an integrated data set that enables an analysis focusing on heart diseases to be performed, for example, among various types of diseases, including, but not limited to, analysis of various types of diseases It is also possible to create an integrated dataset that facilitates it.

The data set generating apparatus 30 may include a data set generating unit 31, a preprocessing unit 32, and a matching unit 33.

The data set generation unit 31 generates a first data set including electronic health records (EHR) and a second data set including mobile personal health records (mPHR) Can be generated.

The data set generation unit 31 generates a first data set by integrating data belonging to a plurality of data sets having different locations as data provided by a University of California-Irvine (Machine Learning Repository) can do. The data set generating unit 31 may generate the second data set such that the attribute of the data (i.e., mPHR data) acquired through the mobile device 20 matches the attribute of the first data set.

Specifically, the data set generation unit 31 uses the data of the heart disease data set directory of the UCI machine learning repository so that an analysis focusing on heart disease can be performed as described above, for example, You can create datasets. The Heart Disease Data Set Directory can contain four sets of data related to the diagnosis of heart disease. At this time, each of the four datasets was stored at four locations (ie Cleveland Clinic Foundation, OH, USA, Hungarian Institute of Cardiology, Budapest, Hungary, Veterans Affairs Medical Center, CA, USA, and University Hospital, Zurich, Switzerland) (Generated) by collecting data from any one of them. In addition, each of the four data sets may be composed of fourteen raw attributes in the same instance format. The data set generating unit 31 may generate a first data set by integrating the four data sets collected at different places.

2 is a diagram illustrating attributes of an EHR considered in a data set generating apparatus for providing a healthcare service according to an embodiment of the present invention.

Referring to FIG. 2, 14 attributes of the EHR considered in the data set generating apparatus 30 may be included. The 14 attributes of EHR include age, sex, chest pain type, cp, resting blood pressure, trest bps, serum cholesterol, chol, In addition to fasting blood sugar (FBS), resting electrocardiographic results, restecg, maximum heart rate achieved, thalach, exercise induced angina, exang, The number of major blood vessels pigmented by fluoroscopy (number of slices), the number of major blood vessels stained by fluoroscopy major vessels colored by flourosopy, ca), Mediterranean anemia (thalassemia, thal), presence of heart disease (num).

A value (i.e., an attribute value) for all attributes of the EHR can be represented by a numerical value. For example, sex can be expressed as 1 for male and 0 for female. The type of chest pain (cp) is represented by 1 in typical angina, 2 in atypical angina, 3 in non-anginal pain, and 4 in asymptomatic . Fasting blood glucose (fbs) may be expressed as 1 for greater than 120 mg / dl, or as 0 for not exceeding 120 mg / dl. The resting electrocardiogram results (restecg) are expressed as normal (normal) 0, with ST-T wave abnormality (1), or with left ventricular hypertrophy (2) . Exercise-induced angina (exang) can be expressed as 1 if present and 0 if not. The slope of the peak motion ST segment can be expressed as 1 for upsloping, 2 for flat, and 3 for downsloping. Mediterranean anemia (thal) can be expressed as 3 in normal, 6 in fixed defect, and 7 in diagnosis of heart disease. The presence (num) of heart disease can be expressed as 0 if it does not exist, or 1 if it is present.

According to this, the first data set generated through the data set generating unit 31 may include information on the fourteen attributes shown in FIG.

In addition, the data set generating unit 31 generates an integrated data set in which the attribute of the data (i.e., the attribute of the mPHR) acquired through the mobile device 20 matches the attribute of the first data set (i.e., the attribute of the EHR) So that the second data set can be generated.

At this time, the second data set generated through the data set generating unit 31 may include four attributes related to heart disease, for example. The attribute information of the four attributes included in the second data set, that is, the data acquired from the mobile device in relation to the mPHR, may include, for example, blood pressure, heart rate, blood sugar and active calories. Accordingly, the data included in the first data set, the second data set, and the integrated data set to be described later may be, for example, heart disease-related data.

Each of the first data set and the second data set generated through the data set generating unit 31 may be composed of 920 records, for example. Herein, 920 records may mean health related records for 920 users, and the number of records is only one example for helping understanding of the present invention.

The preprocessing unit 32 may be configured to generate a first data set including an electronic health record (EHR) and a personal data record obtained through a mobile device for the generation of an integrated data set after the first data set and the second data set are generated (mPHR) for the data contained in each of the second data sets. According to this, the first data set and the second data set can be generated before the preprocessing is performed.

Specifically, the preprocessing unit 32 may perform preprocessing on each individual record in the first data set and the second data set. That is, the preprocessing unit 32 may perform preprocessing on the individual records included in the first data set and the second data set. In other words, the preprocessing unit 32 preprocesses each EHR-related individual record included in the first data set and preprocesses each mPHR-related individual record included in the second data set.

When the pre-processing unit 32 determines that a missing value or an ideal value exists in the data (i.e., the record) included in each of the first data set and the second data set in the pre-processing, the pre- Processing can be performed by replacing the attribute corresponding to the missing value or the outliers with the average value of the attribute values belonging to the same attribute within the set. This preprocessing process can be expressed as a cleansing process.

Since the missing values or anomalies in the dataset degrade the accuracy of the analysis, the dataset generation apparatus 30 can acquire missing values or anomalies in the first dataset and the second dataset through the preprocessing unit 32, , The accuracy of the analysis can be improved. That is, by eliminating the missing value and the ideal value included in the first data set and the second data set through the preprocessing (that is, replacing the missing value and the ideal value with the average value of the attribute values), the integrated data set generated by the present data set generating device 30 It is possible to provide a more accurate analysis when performing the analysis based on the analysis result.

For example, if a value for a chest pain type (cp) property (i.e., a chest pain type attribute value) is missing for any one of 920 records in the first data set (i.e., a health related record for any one user) . It is also assumed that all the values for the chest pain type property are present in 919 records except for one of the above records. In this case, the preprocessing unit 32 compares the attribute value of the missing chest pain type attribute with the attribute value of the chest pain type property that is the attribute corresponding to the missing value in the first data set to which the missing value belongs To the average value of the average values. That is, the preprocessing unit 32 may replace the attribute value of the missing chest pain type with the average value of the 919 chest pain type attribute values belonging to the chest pain type attribute in the 919 records in the first data set.

In addition, the preprocessing unit 32 may set the format (integrated format) of the integrated data set before the matching through the matching unit 33 is performed. That is, the preprocessing unit 32 can set a format suitable for integrating the EHR and the mPHR in generating the integrated data set. At this time, the format of the integrated data set may be set considering the format of the first data set including the EHR and the format of the second data set including the mPHR, or may be set by the user.

After the pre-processing is performed for each of the first data set and the second data set, the matching unit 33 compares the data of the preprocessed first data set and the data of the preprocessed second data set, It is possible to generate an integrated data set in which the data of the set and the data of the preprocessed second data set are integrated. That is, the unified data set can be generated by matching between the data (individual records) of the preprocessed first data set and the data (individual records) of the preprocessed second data.

The matching unit 33 may perform matching for each individual record in the first data set and the second data set and may perform matching between the individual record in the first data set and the individual record in the second data set.

The matching unit 33 may generate an integrated data set through matching based on whether the data attribute of the first data set and the data attribute of the second data set are the same or not.

In particular, if the data attribute of the first data set and the data attribute of the second data set are not identical, the matching unit 33 may combine the data attribute of the second data set with the data attribute of the first data set . The matching unit 33 may overwrite the data attribute of the first data set with the data attribute of the second data set if it is determined that the data attribute of the first data set is identical to the data attribute of the second data set (Otherwise, it can be updated). More specifically, if it is determined that the data attribute of the first data set is the same as the data attribute of the second data set, the data attribute value of the first data set may be overwritten with the data attribute value of the second data set.

For example, the data attributes of the second data set include blood pressure, heart rate, blood glucose and activity calories, wherein the attributes for blood pressure, heart rate, and blood sugar, excluding activity calories, (Trbs), the heart rate (thalach), and the blood sugar (fbs), which are data attributes in the first data set. Accordingly, when the attribute matching between two data sets is performed, The data attributes of the first data set may be overwritten with the data attributes of the second data set if the matching is performed with respect to the heart rate and blood sugar. The matching unit 33 may combine the data attribute of the second data set with the data attribute of the first data set.

At this time, the matching unit 33 combines the attributes in consideration of the format (integrated format) of the integrated data set set by the preprocessing unit 32 when performing the matching between the data attribute of the first data set and the data attribute of the second data set Or overwrite the property. In other words, the matching unit 33 can overwrite or combine the individual attributes (characteristics) in the EHR and mPHR with the unified format of the integrated data set (i.e., the integrated format of the integrated data set) through matching. In a specific example, when it is determined that the data attribute of the first data set is identical to the data attribute of the second data set, the matching unit 33 may compare the attribute value of the first data with the attribute value of the second data set Can be replaced with the attribute values of

When the data attribute of the first data set is equal to the data attribute of the second data set, the data set generator 30 performs a matching process, and sets the data attribute of the first data set to the data attribute of the second data set And may combine the data attributes of the second data set with the data attributes of the first data set if the data attributes of the first data set and the data attributes of the second data set are not the same.

Through this process, the present data set generating apparatus 30 keeps together the advantages of the EHR (i.e., providing reliable medical information) and the advantages of the mPHR (i.e., accurately grasping the current state of the user) An integrated dataset (ie, an integrated dataset with integrated EHR and mPHR) can be generated that allows complementary information access from the mPHR. That is, the present invention can provide the integrated data set so that more accurate and reliable health related diagnosis or analysis can be performed while reflecting the current state of the user.

The integrated data set generated by the matching unit 33 may be configured to include, for example, 15 attributes and 920 records related to heart disease. Here, the fifteen attributes can mean that the 14 items of the EHR and the items of the active calorie, which are the attributes of the mPHR, are considered together.

In addition, the integrated data set can be updated at a predetermined cycle from the second data set so that an accurate analysis (determination) of the current health status of a particular user (patient) can be made. In other words, the data in the integrated data set can be updated at a predetermined cycle from the mPHR, a mobile-based personal health record corresponding to the second data set, so that real-time analysis and accurate analysis of the current health status of a particular user can be made . The predetermined period may be set by a user, and may be set in units of time, day, and the like.

In addition, the greater the number of records in the consolidated data set, the better the accuracy in diagnosing the user's health condition based on the integrated data set.

The present data set generating apparatus 30 can generate and provide an integrated data set in which the EHR and the mPHR are integrated (merged), thereby enabling the integrated data set to diagnose a high-precision disease. Here, the EHR may include data related to the overall health status of the patient, including the past medical history of the user (patient), thereby identifying past medical history, vital signs, medications, radiation records, etc. of the user have. The mPHR is data recorded by a user's personal terminal (mobile device), and can provide information related to the user's real-time health status over time. Accordingly, the user (patient) So that diagnosis of the current health state of the user can be effectively performed.

In consideration of this, the data set generating device 30 can enable the integrated data set to perform an accurate health diagnosis service through the EHR and the mPHR, so that it can play a complementary role in diagnosing the disease of the user (patient). In other words, we provide accurate health checkup information for the user by the integrated data set, and make it possible to access complementary information from EHR and mPHR when diagnosing / predicting / analyzing the user's disease. For example, the integrated data set of the present disclosure allows accurate diagnosis / prediction / analysis of the presence or absence of heart disease.

Hereinafter, a description will be made of a technique for predicting a disease possibility of a user by using the integrated data set generated by the present data set generating apparatus 30 based on the above-described contents.

FIG. 3 is a diagram illustrating a schematic configuration of a disease prediction system 200 according to an embodiment of the present invention. For reference, disease prediction in the present application may be understood in a broad sense such as predicting the possibility of a disease or detecting or measuring the presence of a disease.

Referring to FIG. 3, the disease prediction system 200 according to an embodiment of the present invention may include a first terminal device 50 and a disease prediction device 70.

The first terminal device 50 can provide the health data of the predicted person. That is, the first terminal device 50 may measure the health data of the predicted subject, which is a target for predicting the disease potential, and provide the measured health data to the disease predictor 70. Here, the health data is data related to the health state of the predicted person, which may mean mPHR as data measured through the first terminal device 50. [

Also, the first terminal device 50 may refer to any one of the plurality of mobile devices 20 described with reference to FIG. Therefore, even if omitted in the following description, the description of the mobile device 20 can be applied to the description of the first terminal device 50 as well.

Data transmission / reception between the first terminal device 50 and the disease predicting device 70 may be performed via the network 60. [ Here, the network 60 may mean the same network as the network 40 described above. Therefore, the contents described for the network 40 can be equally applied to the description of the network 60 even if omitted below.

The disease predicting apparatus 70 may include a receiving unit 71, a generating unit 72, and a predicting unit 73. [

The receiving unit 71 can receive the health data of the person to be predicted from the first terminal device 50. [

The generating unit 72 may be configured to generate an electronic health record based on electronic health records (EHR) for a plurality of users and mobile personal health records (mPHR) obtained via a mobile device for a plurality of users A disease prediction model can be generated using the generated integrated data set.

Here, the integrated data set performs pre-processing on the data contained in each of the first data set including the electronic health record and the second data set including the mobile-based personal health record, and the data of the preprocessed first data set And the data of the preprocessed second data set. Such an integrated data set may be generated by the data set generating apparatus 30 described above. Since the process of generating the integrated data set has been described in detail above, the detailed description will be omitted.

For example, after the health data is received through the receiving unit 71, the generating unit 72 may generate the disease prediction model considering at least one of the characteristics (state and type) of the subject to be predicted and the type of disease to be predicted have. Then, the predicting unit 73 can predict a disease possibility of the predicted subject corresponding to the health data received by the receiving unit 71 based on the generated disease prediction model.

Here, the characteristics (state, type) of the predicted subject may refer to the current health state of the predicted subject based on the health data. For example, the characteristics of a predictor to be considered when generating a disease prediction model include state information about the current user's blood pressure, heart rate, blood sugar and activity calorie level relative to the mPHR of the predicted subject . Also, the type of disease to be predicted may mean information about what kind of disease possibility is to be predicted for the predicted subject. Examples of the disease type to be predicted include, for example, heart disease, skin disease, cerebrovascular disease, and the like. The type of disease to be predicted can be set by the user.

In another example, the generating unit 72 generates a plurality of disease prediction models in which characteristics of a plurality of users are taken into account based on the data included in the integrated data set, and after the health data is received through the receiving unit 71, It is possible to select an optimal disease prediction model optimized for a prediction subject among a plurality of disease prediction models considering at least one of the characteristics (state) of the subject and the type of disease to be predicted. Thereafter, the predicting unit 73 can predict the possibility of disease to the predicted person corresponding to the health data received by the receiving unit 71 based on the optimal disease prediction model selected by the generating unit 72. [

That is, for example, the disease prediction model used for predicting the disease probability for the predicted subject considers at least one of the characteristics of the predicted subject based on the received health data and the type of disease to be predicted after the health data is received ≪ / RTI > In another example, the disease predicting device 70 may generate a plurality of disease prediction models based on the integrated data set before the health data is received. Thereafter, the disease prediction model used for predicting the disease probability for the predicted subject is determined by taking into account at least one of the characteristics (state) of the predicted subject and the type of disease to be predicted, Can be selected by selecting a disease prediction model.

In addition, the disease prediction model generated through the generation unit 72 may include a plurality of rules related to a plurality of attributes of data included in the integrated data set. That is, the disease prediction model may include a plurality of rules associated with the fifteen attributes included in the integrated data set, for example. Here, the description of the fifteen attributes included in the integrated data set has been described in detail above, and will not be described below.

Further, in generating the disease prediction model, the number of combinations of the plurality of rules of the disease prediction model to be generated and the combination order can be determined in consideration of at least one of the characteristics of the subject to be predicted and the type of disease to be predicted. In addition, the number and combination order of the plurality of rules of the disease prediction model to be generated can be determined in consideration of a plurality of attributes included in the integrated data set in consideration of attribute value criteria defined in the medical industry. In addition, the number of combinations of the plurality of rules of the disease prediction model to be generated and the combination order can be automatically determined.

The disease prediction model can also be generated by applying a decision tree to the data contained in the integrated data set. The generation unit 72 may use a decision tree that provides the IF-THEN rule as an example of generating the disease prediction model, but is not limited thereto.

The predicting unit 73 can predict (or measure or detect the presence or absence of a disease) the disease probability for the predicted subject corresponding to the health data based on the disease prediction model. As a specific example, the predicting unit 73 can predict the possibility of disease to the predicted person corresponding to the health data, based on the disease prediction model generated upon reception of the health data. Or the predicting unit 73 can predict a disease possibility for a predictive subject corresponding to health data based on an optimal disease prediction model selected from a plurality of disease predictive models.

The disease prediction apparatus 70 according to an embodiment of the present invention generates a disease prediction model using an integrated data set in which the EHR and the mPHR are integrated and calculates a disease state related disease probability related to the user based on the generated disease prediction model Prediction accuracy can be improved as compared to the case where only EHR is used or only mPHR is used.

Hereinafter, experimental results for demonstrating the superiority of the disease-based disease prediction based on the integrated data set proposed in the present invention in comparison with the conventional method using only the EHR or using only the mPHR will be described.

Experiments were conducted using R version 3.4.1 in one experimental example of the present invention for demonstrating the superiority of the present invention.

Also, in one experimental example of the present invention, a disease prediction model related to heart disease is generated using a decision tree that provides an IF-THEN rule for data classification for performance comparison of integrated data sets. In addition, in the example of the present invention, the 'num' attribute in the integrated data set is used to determine whether a user (for example, a user recorded in the predictor or integrated data set) has a heart disease (in other words, Whether there is a possibility of existence) or no (there is no possibility, there is no possibility of existence).

Hereinafter, an experimental example of a comparison of performance (accuracy) between a disease prediction model generated based on the integrated data set of the present invention including EHR and mPHR and a conventional EHR or mPHR based disease prediction model will be described in detail .

FIG. 4 is a diagram illustrating an example of a disease prediction model generated by applying a decision tree to an integrated data set according to an embodiment of the present invention. Referring to FIG.

Referring to FIG. 4, the disease prediction model generated according to one experimental example of the present invention may have 15 rules as an example. According to this rule, the predicted person (or the user recorded in the integrated data set) can be classified as 'yes' or 'no' for the possibility of disease (in other words, whether or not the disease exists). Hereinafter, in one experimental example of the present invention, prediction of the possibility of a heart disease can be made, for example, in relation to predictability of a disease, and the possibility of existence of various diseases can be predicted.

According to this, whether or not the possibility of heart disease for all users recorded in the integrated data set or the predicted object (for example, new predicted object) received through the receiving unit 71 is estimated through a decision tree- .

For example, according to the disease prediction model generated according to one experimental example of the present invention, If the value of 'cp' representing the type of chest pain of the health data of a particular user (which may mean either a new predictor or a user recorded in the integrated data set) in relation to the rule is not 1, 2 or 3 , The 'trest bps' indicating the stable blood pressure of the specific patient can be confirmed with respect to the second rule. For example, if the value of 'trestbps' in the second rule is less than 132, confirmation of the attributes related to the third rule in the health data of the specific patient can be made. That is, as an attribute related to the third rule, for example, confirmation of cholesterol 'chol' can be made. Here, if it is determined in the third rule that the value of the 'chol' attribute of a specific user is greater than 430, the possibility of a specific user's heart disease (i.e., presence of a heart disease) can be predicted as YES. For reference, the 'hr' attribute in FIG. 4 may refer to the heart rate (heart rate) associated with the mPHR.

5 is a diagram showing an example of a disease prediction model generated by applying a decision tree to a conventional EHR.

Referring to FIG. 5, in the EHR-based disease predicting model, 'cp' is considered as the first rule, for example, to predict the disease potential. However, the EHR-based disease prediction model has a problem that the accuracy of the disease prediction model is lower than that of the present invention because the number of rules is smaller than that of the integrated disease-based disease prediction model.

6 is a diagram illustrating an example of a disease prediction model generated by applying a decision tree to a conventional mPHR.

Referring to FIG. 6, there are five rules for the mPHR-based disease prediction model. This can be attributed to the fact that mPHR contains only four attributes related to heart disease (ie, blood pressure, heart rate, blood sugar and activity calories). According to this, the mPHR-based disease prediction model, like the EHR-based disease prediction model, has a problem that the accuracy of the disease prediction model is lower than that of the present invention because the number of rules is smaller than that of the integrated data set based disease prediction model.

Hereinafter, the decision tree of each EHR, mPHR, and integrated data set is analyzed to evaluate the accuracy of the prediction results derived from the decision tree.

Figure 7 illustrates the accuracy of an integrated data set based disease prediction model in accordance with one embodiment of the present disclosure versus EHR or mPHR.

Referring to FIG. 7, the values for n and p can be set as follows to calculate the accuracy of the disease prediction model. n represents the number of data elements predicted by the disease prediction model to be 'No' when the actual value is 'No'. p represents the number of data elements predicted by the disease prediction model to be 'Yes' when the actual value is 'Yes'. The accuracy of the disease prediction model (i. E., The decision tree) can be calculated based on "(n + p) / (total number of records in the data set) ".

As a result of the accuracy calculation, the accuracy of the integrated data set based on the integrated data set according to one embodiment of the present invention was 0.82. In addition, the accuracy of the disease-predicting model based on datasets including EHR was 0.79. In addition, the accuracy of the data-based disease prediction model including mPHR was 0.78.

According to this, the integrated data set based disease prediction model of the present invention has the highest accuracy compared to the conventional EHR or mPHR based disease prediction model. In particular, our integrated data set-based disease prediction model shows approximately 3% and 4% higher accuracy than the existing EHR or mPHR-based disease prediction models. In other words, according to the disease prediction model based on the integrated data set of the present invention, it can be confirmed that the accuracy of prediction of the user's disease is higher than that of EHR or mPHR.

The present invention can provide an integrated data set to enable more effective and accurate delivery of medical services. That is, the present invention can extract meaningful information based on the integrated data set, thereby providing a better quality medical service.

 Hereinafter, the operation flow of the present invention will be briefly described based on the details described above.

8 is a flowchart illustrating a method of generating a data set for providing a healthcare service according to an embodiment of the present invention.

The data set generating method shown in FIG. 8 can be performed by the data set generating apparatus 30 described above. Therefore, even if omitted below, the contents described for the data set generating apparatus 30 can be similarly applied to the description of the data set generating method.

Referring to FIG. 8, a method of generating a data set for providing a healthcare service according to an embodiment of the present invention includes generating a first data set including an electronic health record (EHR) and a mobile- Processing may be performed on the data contained in each of the second data sets including the health record (mPHR).

If it is determined in step S11 that there is a missing value or an ideal value for the data included in each of the first data set and the second data set, the missing value or the ideal value in the data set to which the missing value or the ideal value belongs It is possible to perform preprocessing to replace the attribute corresponding to the outlier with the average value of the attribute values belonging to the same attribute.

Next, in step S12, the data of the first data set, which is pre-processed through the matching between the data of the preprocessed first data set and the data of the preprocessed second data set, and the data of the preprocessed second data set, Can be generated.

In addition, in step S12, the integrated data set can be generated through matching based on whether the data attribute of the first data set and the data attribute of the second data set are the same or not.

If it is determined in step S12 that the data attribute of the first data set is not the same as the data attribute of the second data set, the data attribute of the second data set may be combined with the data attribute of the first data set. If it is determined in step S12 that the data attribute of the first data set is identical to the data attribute of the second data set, the data attribute of the first data set may be overwritten with the data attribute of the second data set.

In addition, the method of generating a data set for providing healthcare service according to an embodiment of the present invention may include a step of generating a first data set and a second data set prior to step S11.

At this time, in the step of generating the first data set and the second data set, data belonging to a plurality of data sets having different places as data provided by the University of California-Irvine (UCI) Machine Learning Repository To generate a first data set. Also, in the step of generating the first data set and the second data set, the second data set may be generated such that the attributes of the data acquired through the mobile device match attributes of the first data set.

In the above description, steps S11 through S12 may be further divided into further steps or combined into fewer steps, according to an embodiment of the present invention. Also, some of the steps may be omitted as necessary, and the order between the steps may be changed.

9 is a flowchart illustrating an operation of the disease prediction method according to an embodiment of the present invention.

The disease prediction method shown in FIG. 9 can be performed by the disease prediction apparatus 70 described above. Therefore, even if omitted below, the contents described for the disease predicting device 70 can be similarly applied to the description of the disease predicting method.

Referring to FIG. 9, the disease prediction method according to an embodiment of the present invention can receive the health data of the predicted subject in step S21.

Next, in step S22, the disease prediction model can be generated using the integrated data set generated based on the electronic health record (EHR) and the mobile-based personal health record (mPHR) obtained through the mobile device.

In addition, in step S22, after the health data is received, the disease prediction model can be generated in consideration of at least one of the characteristics (state) of the subject to be predicted and the type of disease to be predicted.

In addition, in step S22, a plurality of disease prediction models in which characteristics of a plurality of users are taken into account based on the data included in the integrated data set are generated, and after the health data is received, the characteristics (state) The optimal disease prediction model optimized for the prediction subject can be selected from among a plurality of disease prediction models. Thereafter, in step S23, the possibility of disease to the predicted subject can be predicted based on the optimal disease prediction model.

In addition, the disease prediction model in step S22 may include a plurality of rules associated with a plurality of attributes of data included in the integrated data set. Further, the number of combinations of the plurality of rules and the order of combination may be determined in consideration of at least one of the characteristics of the subject to be predicted and the type of disease to be predicted.

The disease prediction model can also be generated by applying a decision tree to the data contained in the integrated data set.

Further, the unified data set at step S22 performs pre-processing on the data included in each of the second data set including the first data set including the electronic health record and the personal health record obtained through the mobile device, May be generated through matching between the data of the preprocessed first data set and the data of the preprocessed second data set.

Next, in step S23, based on the disease prediction model, the disease probability for the predicted subject corresponding to the health data can be predicted.

In the above description, steps S21 to S23 may be further divided into additional steps or combined into fewer steps, according to embodiments of the present disclosure. Also, some of the steps may be omitted as necessary, and the order between the steps may be changed.

The data set generation method and the disease prediction method according to an exemplary embodiment of the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer readable medium. The computer-readable medium may include program instructions, data files, data structures, and the like, alone or in combination. The program instructions recorded on the medium may be those specially designed and constructed for the present invention or may be available to those skilled in the art of computer software. Examples of computer-readable media include magnetic media such as hard disks, floppy disks and magnetic tape; optical media such as CD-ROMs and DVDs; magnetic media such as floppy disks; Magneto-optical media, and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, flash memory, and the like. Examples of program instructions include machine language code such as those produced by a compiler, as well as high-level language code that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

It will be understood by those of ordinary skill in the art that the foregoing description of the embodiments is for illustrative purposes and that those skilled in the art can easily modify the invention without departing from the spirit or essential characteristics thereof. It is therefore to be understood that the above-described embodiments are illustrative in all aspects and not restrictive. For example, each component described as a single entity may be distributed and implemented, and components described as being distributed may also be implemented in a combined form.

The scope of the present invention is defined by the appended claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included within the scope of the present invention.

100: Dataset generation system
30: Data set generating device
31: Data set generation unit
32:
33:
200: disease prediction system
70: Disease prediction device
71: Receiver
72:
73:

Claims (9)

In a disease predicting method,
Receiving health data of a predicted subject;
Generating a disease prediction model using an integrated data set generated based on electronic health records (EHR) and mobile personal health records (mPHR) obtained via a mobile device; And
Predicting a disease potential for a predictive subject corresponding to the health data based on the disease predictive model,
Lt; / RTI > The method according to claim 1,
Wherein the generating comprises:
Wherein the disease prediction model is generated in consideration of at least one of the characteristics of the predictive subject and the type of disease to be predicted after the health data is received. The method according to claim 1,
Wherein the generating comprises:
Generating a plurality of disease prediction models in which characteristics of a plurality of users are taken into consideration based on data included in the integrated data set and generating at least one of the characteristics of the predicted subject and the type of disease to be predicted after the health data is received And selecting an optimal disease prediction model optimized for the predicted subject among the plurality of disease prediction models,
Wherein the predicting comprises:
And predicting a disease possibility for the predicted subject based on the optimal disease prediction model. The method according to claim 1,
Wherein the disease prediction model includes a plurality of rules associated with a plurality of attributes of data contained in the aggregate data set,
Wherein the number of combinations of the plurality of rules and the order of combination are determined in consideration of at least one of the characteristics of the predictive subject and the type of disease to be predicted. The method according to claim 1,
The disease prediction model includes:
And generating a decision tree for the data contained in the integrated data set. The method according to claim 1,
The integrated data set includes:
Processing the data contained in each of the first data set including the electronic health record and the second data set including the personal health record obtained through the mobile device, Is generated through matching between data in the preprocessed second data set. A disease predicting device comprising:
A receiving unit for receiving health data of a predicted subject;
A generating unit for generating a disease prediction model using an integrated data set generated based on electronic health records (EHR) and mobile personal health records (mPHR) obtained through a mobile device; And
A predictor for predicting a disease probability of a predictive subject corresponding to the health data based on the disease predictive model,
And a disease prediction device. In a disease prediction system,
A first terminal device for providing health data of a predicted subject; And
Receiving the health data from the first terminal device and generating an integrated data set based on electronic health records (EHR) and mobile personal health records (mPHR) obtained via the mobile device A disease prediction device that generates a disease prediction model by using the disease prediction model, predicts a disease probability of a prediction subject corresponding to the health data based on the disease prediction model,
A disease prediction system. A computer-readable recording medium on which a program for executing the method of any one of claims 1 to 6 is recorded.

Images