KR20190070430A - Estimation method and apparatus for name of disease based on big data analysis - Google Patents

Abstract

A method for estimating a disease diagnosis based on a big data analysis comprises the steps of: generating structured data by performing text-mining on unstructured data of a target electronic medical record (EMR) by using a computer apparatus; extracting a keyword from the structured data according to a correlation between a specific disease and the keyword by using the computer apparatus; and classifying a target diagnosis corresponding to the keyword by inputting the keyword to a previously provided naive Bayes classifier by using the computer apparatus.

Description

TECHNICAL FIELD [0001] The present invention relates to a method and an apparatus for estimating a disease diagnosis name based on big data analysis.

The technique described below relates to a technique for deriving a diagnosis name from electronic medical records based on a big data analysis technique.

Many hospitals now use electronic medical records (EMR). At present, EMR is used only to digitize the contents that had been written by hand and to be used for patient management. Basically, the author writes the contents using a computer device or a smart device, and there may be an error in a part of the contents.

Especially, in the case of military medical systems used in military units, accuracy of data is low, and various errors are generated in the process of inputting or updating the diagnosis name, which makes it difficult to accurately manage diagnostic records. In particular, the classification accuracy of the diagnosis of infectious diseases depends on the nature of the disease, which occurs irrespective of the past history of the patient, or in which the symptom is difficult to distinguish, and verification is carried out only by the conventional form data such as the patient's physical examination record, There are limitations that are not easy.

US Patent No. 9,477,756

The main content of EMR is atypical data. Therefore, it is difficult to use EMR for other purposes besides the medical treatment. Further, as described above, the EMR may contain inaccurate information such as a misidentification of the disease.

The technique described below is intended to provide a technique for deriving a diagnosis name from an EMR based on a big data analysis technique.

The Big Data Analysis-based disease diagnosis name estimation method includes the steps of: generating formatting data by text mining of unstructured data of a target electronic medical record (EMR) of a computer device; Extracting a keyword from the keyword, and inputting the keyword to a Naïve Bayes classifier provided in advance, and classifying the target diagnosis name corresponding to the keyword. The Naive Bayes classifier includes information defining a relationship between a keyword and a disease.

The technique described below improves the accuracy of diagnosis by deriving an accurate diagnosis name from EMR. The techniques described below analyze the EMR and provide new medical information by providing a keyword or pattern associated with the disease.

1 is an example of a big data analysis-based disease diagnosis name estimation system.
2 is an example of a flow chart for a Big Data Analysis-based disease diagnosis name estimation method.
3 is an example of a process of generating a big data analysis model in electronic medical records.
FIG. 4 is an example of a process of estimating the disease diagnosis name in the inputted electronic medical record.
FIG. 5 shows an example of a process for formatting unstructured data of electronic medical records.
6 is an example of visualizing disease information estimated from electronic medical records.

The following description is intended to illustrate and describe specific embodiments in the drawings, since various changes may be made and the embodiments may have various embodiments. However, it should be understood that the following description does not limit the specific embodiments, but includes all changes, equivalents, and alternatives falling within the spirit and scope of the following description.

The terms first, second, A, B, etc., may be used to describe various components, but the components are not limited by the terms, but may be used to distinguish one component from another . For example, without departing from the scope of the following description, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component. And / or < / RTI > includes any combination of a plurality of related listed items or any of a plurality of related listed items.

As used herein, the singular " include "should be understood to include a plurality of representations unless the context clearly dictates otherwise, and the terms" comprises & , Parts or combinations thereof, and does not preclude the presence or addition of one or more other features, integers, steps, components, components, or combinations thereof.

Before describing the drawings in detail, it is to be clarified that the division of constituent parts in this specification is merely a division by main functions of each constituent part. That is, two or more constituent parts to be described below may be combined into one constituent part, or one constituent part may be divided into two or more functions according to functions that are more subdivided. In addition, each of the constituent units described below may additionally perform some or all of the functions of other constituent units in addition to the main functions of the constituent units themselves, and that some of the main functions, And may be carried out in a dedicated manner.

Also, in performing a method or an operation method, each of the processes constituting the method may take place differently from the stated order unless clearly specified in the context. That is, each process may occur in the same order as described, may be performed substantially concurrently, or may be performed in the opposite order.

The technique described below estimates a disease diagnosis name that is presumed to be a specific medical record based on the information recorded in the electronic medical record (EMR). This allows the technology described below to update the diagnostics name incorrectly listed in the EMR. Further, the technology described below can proactively estimate the current patient's disease through EMR. The techniques described below utilize techniques used in big data analysis.

FIG. 1 is an example of a big data analysis-based disease diagnosis name estimation system 100. FIG. The system 100 includes a client device 110, an analysis server 120, and an EMR DB 130.

The client device 110 sends an instruction to the analysis server 120 to estimate the disease diagnosis name based on the EMR data. In addition, the client device 110 can receive the diagnosis name, disease-related information, and the like according to the analysis result. The client device 100 corresponds to a user device such as a PC, a smart device, and the like.

The EMR DB 130 is a device that stores the EMR. The EMR stores information generated during diagnosis and treatment for a particular patient. Therefore, the EMR DB 130 basically stores very much information for each patient.

Meanwhile, the client device 110 may transmit to the analysis server 120 the information (identifier, IP, etc.) about the specific EMR DB and specific EMR data (data identifier, patient identifier, etc.) stored in the corresponding EMR DB.

The analysis server 120 accesses the specific EMR data from the specific EMR DB 130 based on the command and information received from the client device 110. [ The analysis server 120 estimates the diagnosis diagnosis name based on the received EMR data. The analysis server 120 can estimate the diagnosis name using a model for disease diagnosis name estimation. The analysis model DB 125 has a model for estimating the disease diagnosis name by inputting the EMR. In FIG. 1, the analysis model DB 125 is shown as a separate object, but it may be included in the analysis server 120.

Furthermore, the technique of estimating the diagnostic name based on the EMR may operate not only on a networked system but also on individual computer devices such as a PC. That is, the computer device executes a program for estimating the diagnosis name of the disease, and can input a certain EMR to estimate the diagnosis name. However, for convenience of explanation, the Big Data Analysis based disease diagnosis name estimation system 100 described in FIG. 1 will be described. The analysis server 120 is also a kind of computer device. Therefore, it can be said that the computer device estimates the diagnosis diagnosis name based on the EMR.

FIG. 2 is an example of a flowchart for the Big Data Analysis-based disease diagnosis name estimation method 200. FIG. The computer device first generates an analysis model based on the Big Data (210). Here, the computer device includes the above-described analysis server 120, individual computer devices, and the like. The analysis model corresponds to a model for estimating the disease diagnosis name using EMR data. For example, the Naive Bayes classification model used for text classification can be used.

In a situation where a model for analysis is prepared, the computer device receives EMR data for the diagnosis name estimation. EMR data for diagnosis name estimation is hereinafter referred to as target EMR. The target EMR is basically composed of unstructured data. Therefore, for analysis, the computer device first generates textual data while performing text mining on the target EMR (220).

The computer device extracts the disease related keywords from the formatted EMR data (230). The computer device inputs the extracted keyword into the analysis model to estimate the disease diagnosis name (240). Further, the computer device can process the estimated disease diagnosis name and the diagnosed disease information in a uniform manner (250). For example, the computer device can visualize and output relevant information along with a disease diagnosis name.

3 is an example of a process 300 for creating a big data analysis model in electronic medical records. The EMR DB 130 stores various information related to the treatment / treatment. The EMR DB 130 includes an admission / discharge recording sheet, a medical / nursing recording sheet, a surgical / nursing recording sheet, a emergency recording sheet, and the like. The admission record is the data about the patient during admission. The hospital record sheet records the progress of the hospitalization, the patient's condition, and so on. The discharge report is written about the patient during discharge. The discharge record sheet records information on the patient's condition at the time of discharge and subsequent treatment. A medical / nursing record is information about the patient in the course of treatment. The medical care / nursing record sheet records information on patient condition, measured biometric information (blood pressure, body temperature, pulse, etc.) over time, and prescription drugs. Surgical / nursing records store information generated during the surgical procedure and during the treatment of patients after surgery. The emergency record book stores information about emergency care for the patient. The type of EMR shown in FIG. 3 is one example, and various information about the patient can be stored according to various types of classification.

The sample EMR refers to the input data used to construct the analysis model. FIG. 3 illustrates an example in which the pumice server 120 constructs an analysis model. However, a separate computer device other than the analysis server 120 may construct an analysis model in advance. The analysis server 120 receives the sample EMR (310). In FIG. 3, the portion indicated by the right square dotted line corresponds to the process performed by the analysis server 120. The analysis server 120 prepares the unstructured data in the sample EMR to be constant (320). The analysis server 120 converts the unstructured data into the formatted data. The preprocessing process will be described later. Thereafter, the analysis server 120 extracts the keyword related to the disease from the formalized EMR data (330). The analysis server 120 matches the extracted keyword with the disease diagnosis name diagnosed in the corresponding EMR (340). For example, the analysis server 120 may generate a table that matches the true diagnosis name associated with the keyword. The analysis server 120 may store the generated model in the analysis model DB. Then, the analysis model for estimating the disease diagnosis name estimates the disease diagnosis name using the generated table. On the other hand, the analytical model may be created as a different route without using the EMR data. For example, the analysis model may be generated using information well known in the medical field.

4 is an example of a process 400 for estimating a disease diagnosis name in the inputted electronic medical record. In FIG. 4, the portion indicated by the right square dotted line corresponds to the process performed by the analysis server 120. The target EMR means input data for disease diagnosis name estimation. The analysis server 120 receives the target EMR from the EMR DB 130 (410). The analysis server 120 regularly prepares irregular data in the target EMR (420). The analysis server 120 converts the unstructured data into the formatted data. Thereafter, the analysis server 120 extracts the keyword related to the disease from the formalized EMR data (430).

Figure 5 is an example of a process 500 for formatting unstructured data of electronic medical records. FIG. 5 shows an example of preprocessing irregular data to generate formatted data. The analysis server 120 may preprocess the unstructured data using a natural language processing program. For example, if the text of the EMR data is in Korean, it is possible to preprocess irregular data using Korean Natural Language Processing (KoNLP), a natural analysis tool.

The analysis server 120 removes the stopwords from the input text data (510). Thereafter, the analysis server 120 separates the input text data in units of sentences (520). The analysis server 120 extracts the root of the corresponding sentence (word) by sentence unit or converts the sentence (word) into the basic type (530). Finally, the analysis server 120 may assign a predetermined weight to the extracted keywords. For example, the analysis server 120 may assign a word / inverse word frequency combination (TF-IDF) weight to a keyword. TF-IDF (Term Frequency - Inverse Document Frequency) is a weight used in text mining. It is a statistical number that indicates how important a word is in a particular document when it is present. TF (word frequency, term frequency) is a value that indicates how often a particular word appears in a document. However, if the word itself is frequently used within a set of documents, this means that the word appears frequently. This is called DF (document frequency), and the reciprocal of this value is called IDF (inverse document frequency). TF-IDF is the product of TF and IDF.

FIG. 5 shows an example of the text data included in the EMR "Hospital visit due to severe fever, headache and diarrhea during the reward vacation lasting 7 days. &Quot; Quoted words such as illumination are removed from the input text data. In Fig. 5, the deleted abbreviation is indicated by a square box. If the text is separated from the text data in which the phrase is removed, the sentence unit can be classified as "reward vacation / severe / fever / headache / diarrhea / symptom / 7 days / persistence". Subsequently, the sentence (word) is converted into a root or basic form in sentence units. The converted result is "a fever headache diarrhea symptom lasting 7 days" which is a rewarding holiday. Then, the TF-IDF value is given to the keyword by calculating the frequency of occurrence of each keyword and the like.

The analysis server 120 generates the formatting data through a text mining process as shown in FIG. Further, although not shown in FIG. 4, the analysis server 120 may select a keyword related to a specific disease without using all the words generated by the regularization data. The analysis server 120 can select a keyword from the stereotyped data using the correlation between the specific disease and the keyword. The correlation may be a function defining the correlation between the specific disease and the specific keyword based on a frequency in which a specific keyword exists for a specific disease in EMR and a pattern in which a specific keyword appears for a specific disease. For example, the correlation may be a pattern in which the keyword appears in each process such as the frequency of occurrence of a specific word (keyword) in an EMR related to a specific disease and / or diagnosis / treatment / hospitalization (for example, initial diagnosis, The concentration is high during a certain period of time).

Returning to the description of FIG. The analysis server 120 estimates the disease diagnosis name based on the extracted keyword (440). The analysis server 120 can classify the diagnosis names based on the keywords appearing in the currently input EMR by using the table described in FIG. For example, the analysis server 120 may estimate the diagnosis name corresponding to the currently input EMR using the frequency of the keyword and the table (matching the keyword and disease diagnosis name). The analysis server 120 can derive the most probable diagnosis name based on the keyword by using the Naive Bayes classification (classifier) technique. The lower part of FIG. 4 shows an example of the diagnosis name estimated by the keyword extracted from the input EMR according to the Naïve Bayes classification. According to Fig. 4, the disease most related to the input keyword is the cold (91%). Therefore, the analysis server 120 derives the disease diagnosis name according to the current EMR as a cold.

The Naive Bayes classifier is a Supervised Learning classifier used for spam filtering and keyword searches. The basic principle of the Naive Bayes classifier is a stochastic classification scheme that applies Bayesian theorem to conditional probability and classifies the input vector assuming the independence of the probability that each element constituting the document or data appears. The conditional probability is the probability that event B occurs when event A occurs. The conditional probability is equal to the probability that A and B will occur simultaneously by the probability that A will occur. The operation of the Naïve Bayes classification is well known in the field, so a detailed description is omitted. In summary, the analysis server 120 uses a table matching the keyword and disease diagnosis name in advance, processes the EMR as formalized data, and estimates a disease diagnosis name having high relevance to the corresponding EMR based on the derived keyword.

Meanwhile, the analysis server 120 can generate new information by processing the diagnosis name or information related to the disease to a certain extent. For example, the analysis server 120 may generate data that visualizes information related to the disease. The analysis server 120 can provide the generated visualization data to the client device. The client device 110 can output the visualization data to the screen. 6 is an example of visualizing disease information estimated from electronic medical records. 6 shows an example of a kind of word network. In the field of text mining, the network type is frequently used to visually express the result of text mining. As a result of text mining, it is possible to express a large number of frequently appearing words. The analysis server 120 can display a word related to the disease among the keywords extracted from the EMR close to or larger than the diagnostic diagnosis name.

Furthermore, the analysis server 120 can update the record with the estimated diagnosis name if the diagnosis names estimated through the disease diagnosis name and the analysis model recorded in the current EMR are different from each other. In this case, the analysis server 120 delivers the estimated diagnosis name to the EMR DB 130, and the EMR DB 130 can record a new disease diagnosis name in the target EMR.

Based on the above-mentioned Big data-based disease diagnosis name estimation method, it is possible to diagnose a disease for a specific patient based on the currently recorded EMR. Also, knowing that there is an error in the diagnosis name recorded in the current EMR, you can update the diagnosis name.

In addition, the Big Data-based disease diagnosis name estimation method as described above can be implemented as a program (or an application) including an executable algorithm that can be executed in a computer. The program may be stored and provided in a non-transitory computer readable medium.

A non-transitory readable medium is a medium that stores data for a short period of time, such as a register, cache, memory, etc., but semi-permanently stores data and is readable by the apparatus. In particular, the various applications or programs described above may be stored on non-volatile readable media such as CD, DVD, hard disk, Blu-ray disk, USB, memory card, ROM,

The present embodiment and drawings attached hereto are only a part of the technical idea included in the above-described technology, and it is easy for a person skilled in the art to easily understand the technical idea included in the description of the above- It will be appreciated that variations that may be deduced and specific embodiments are included within the scope of the foregoing description.

100: Big data analysis based disease diagnosis name estimation system
110: Client device
120: Analysis server
125: Model DB
130: EMR DB

Claims (13)

Generating formalized data by text mining of unstructured data of a target electronic medical record (EMR) using a natural language analysis tool KoNLP (Korean Natural Language Processing);
Extracting a keyword from the stereotyped data according to a correlation between a specific disease and a keyword;
Inputting the keyword into a Naive Bayes classifier provided in advance and classifying a target diagnosis name corresponding to the keyword; And
Visualizing the target diagnostic name categorized with respect to the target electronic medical record and outputting it to a screen,
The Naive Bayes classifier includes information defining a relationship between a keyword and a disease,
The correlation is information that defines a correlation between the specific disease and the specific keyword based on a frequency of occurrence of a specific keyword with respect to a specific disease and a pattern in which the specific keyword appears in the electronic medical record,
The step of generating the formatting data may include the steps of removing an idiomatic word from the unstructured data, discriminating a sentence from the unstructured data, converting a word included in the unstructured data into a root-type or a basic type, And assigning a TF-IDF (Term Frequency - Inverse Document Frequency) weighting to the converted word. The computer device generating formatting data by text mining for unstructured data of a target electronic medical record (EMR);
Extracting a keyword from the stereotyped data according to a correlation between a specific disease and a keyword; And
And inputting the keyword to the Naive Bayes classifier provided in advance to classify the target diagnosis name corresponding to the keyword,
Wherein the Naive Bayes classifier comprises information defining a relationship between a keyword and a disease. 3. The method of claim 2,
Wherein the electronic medical record comprises at least one of an admission record, a discharge record, a medical record, an operation record, a nursing record, and an emergency record. 3. The method of claim 2,
The step of generating the formatting data
Removing idle words from the unstructured data;
Identifying sentences in the atypical data;
Converting a word included in the atypical data into a root type or a basic type; And
And assigning a TF-IDF (Term Frequency-Inverse Document Frequency) weight to the root-converted or basic-converted word. 3. The method of claim 2,
Wherein the computer device generates the formatted data using KoNLP (Korean Natural Language Processing), a natural language analysis tool. 3. The method of claim 2,
Wherein the correlation is determined based on a frequency in which a specific keyword exists for a specific disease in the electronic medical record and a pattern in which the specific keyword appears in the specific disease, Based disease diagnosis method. 3. The method of claim 2,
Further comprising the computer device visualizing the target diagnosis name categorized with respect to the target electronic medical record and outputting the visualized name to the screen. 3. The method of claim 2,
Further comprising the computer device modifying the diagnosis name stored in the target electronic medical record to the target diagnosis name. A computer-readable recording medium recording a program for executing the Big Data Analysis-based disease diagnosis name estimation method according to any one of claims 2 to 8 on a computer. A database for storing electronic medical records (EMR);
A client device for transmitting a data analysis command for a target electronic medical record out of electronic medical records stored in the database; And
Receiving the analysis command, receiving the target electronic medical record by contacting the database, formatting the unstructured data included in the received target electronic medical record, analyzing the formatted data according to a correlation between a specific disease and a keyword And an analysis server for inputting the keyword into a previously prepared Naive Bayes classifier and deriving a target diagnosis name corresponding to the keyword. 11. The method of claim 10,
The analysis server removes an idiomatic word from the unstructured data by using natural language analysis software, converts a word included in the irregular data into a root type or a basic type, Frequency - Inverse Document Frequency) weights to generate the formatted data. 11. The method of claim 10,
Wherein the analysis server transmits to the client device data obtained by visualizing the target diagnosis name classified with respect to the target electronic medical record to the client device. 11. The method of claim 10,
And the analysis server transmits the target diagnosis name to the database.

Images