KR101869438B1 - Method and system for predicting prognosis from diagnostic histories using deep learning - Google Patents

Abstract

A method and system for predicting a disease prognosis from a patient ' s diagnostic history using deep learning is disclosed. The disease prognosis prediction method includes generating sequence data by expressing a diagnostic classification code indicating a diagnosis history of a patient in a word sequence form; Generating learning data by word sequence learning using Recurrent Neural Networks (RNN) for the sequence data; And predicting a disease prognosis using the learning data.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to a method and system for predicting a disease prognosis from a patient's diagnosis history using deep learning,

The following discussion is about techniques for predicting disease prognosis.

Diseases such as hypertension (HBP), diabetes, and hyperlipidemia are among the most common diseases worldwide in the middle-aged. Although adult disease itself is not fatal to health, it can lead to serious complications.

In particular, it has been reported that adult diseases are highly associated with high-risk diseases such as heart, brain, and vascular disease, which are the main causes of death. Therefore, predicting the high-risk prognosis of geriatric patients is an important issue in the medical and biotechnological fields.

In order to predict the high-risk prognosis, various methods using age, sex, current status, family history, etc., biomarker gene expression amount, medical images such as CT, and blood analysis are used.

For example, Korean Patent Laid-Open No. 10-2014-0098561 (published on Aug 08, 2014) discloses a method for predicting a user's risk of developing a disease based on the combination of single nucleotide polymorphism (SNP) Technology is disclosed.

The data used in the existing methods for predicting high-risk prognosis are not only diverse, but also require a lot of time, effort, and cost to acquire necessary data because the data processing process such as preprocessing is complicated.

We provide a method and system for predicting high-risk prognosis from the diagnostic history of adult patients using neural networks such as deep learning (Recurrent Neural Networks).

The present invention provides a method and system for classifying and predicting a high risk prognosis of a patient by expressing a patient classification, a diagnosis, and a treatment code as features of a word sequence in order to learn by using RNN.

A computer-implemented disease prognosis prediction method, comprising: generating sequence data by expressing a diagnostic classification code representing a diagnostic history of a patient in a word sequence; Generating learning data by word sequence learning using Recurrent Neural Networks (RNN) for the sequence data; And predicting a disease prognosis using the learning data.

According to an aspect of the present invention, the step of generating the learning data may include calculating the disease occurrence probability according to the real vector by expressing the sequence data as a real number vector.

According to another aspect of the present invention, the step of generating the learning data comprises: expressing a word vector including sequence information through word sequence learning in the RNN with the sequence data as an input of the RNN; And calculating a disease occurrence probability using the word vector including the sequence information.

According to another aspect, the step of generating the learning data comprises the steps of: embedding a layer for a distributed representation of a word sequence; recurrent layers for sequence modeling; And fully-connected layers for the RNN.

According to another aspect, the learning of the RNN is performed in an end-to-end manner by propagating a prediction error to the input feature layer through the full connection layer, the regression layer, and the embedding layer .

According to another aspect of the present invention, the step of generating the sequence data may generate the sequence data using a last last predetermined number of codes among the diagnostic classification codes.

According to another aspect, the step of generating the sequence data may replace a code corresponding to a given disease among the diagnostic classification codes with a label of the disease.

According to another aspect, the step of generating the sequence data may generate sequence data for each disease when a plurality of codes corresponding to a given disease among the diagnostic classification codes are present.

According to another aspect, the method further comprises generating sequence data for at least one additional clinical history of the patient's time of diagnosis and disease duration, treatment history and pathological measurement data, Step may include learning each of the individual neural networks according to the characteristics of the data for the diagnostic history of the patient and the additional clinical history.

A computer program recorded on a computer-readable recording medium for executing a disease prognosis prediction method in combination with a computer system, the disease prognosis prediction method comprising the steps of: generating a diagnostic classification code indicating a diagnosis history of a patient as a word sequence Generating sequence data by expressing the sequence data; Generating learning data by word sequence learning using Recurrent Neural Networks (RNN) for the sequence data; And predicting a disease prognosis using the learning data.

CLAIMS What is claimed is: 1. A computer implemented disease prognosis prediction system, comprising: at least one processor configured to execute a computer-readable instruction, wherein the at least one processor is configured to generate a diagnostic classification code, ), Generating learning data by word sequence learning using RNN (Recurrent Neural Networks) for the sequence data, and predicting a disease prognosis using the learning data Provides a disease prognosis prediction system.

According to embodiments of the present invention, the diagnosis classification code can be expressed as a feature in the form of a word sequence, and the high-risk prognosis of the patient can be predicted through learning using RNN. Therefore, a simple diagnostic history expressed in code can predict the high-risk prognosis faster and more accurately and provide a quick warning about the high-risk prognosis.

1 is a block diagram for explaining an example of the internal configuration of a computer system according to an embodiment of the present invention.
2 is a diagram illustrating an example of components that a processor of a computer system according to an embodiment of the present invention may include.
3 is a flowchart illustrating an example of a disease prognosis prediction method that can be performed by a computer system according to an embodiment of the present invention.
FIG. 4 shows an example of a process of generating sequence data from a diagnostic classification code in an embodiment of the present invention.
FIG. 5 shows an example of an RNN model structure for predicting disease prognosis in an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Embodiments of the invention relate to techniques for predicting disease prognosis.

Embodiments, including those specifically disclosed herein, achieve disease prognosis prediction and thereby achieve significant advantages in terms of efficiency, accuracy, agility, cost savings, and the like.

1 is a block diagram for explaining an example of the internal configuration of a computer system according to an embodiment of the present invention. For example, a disease prognosis prediction system in accordance with embodiments of the present invention may be implemented through the computer system 100 of FIG. 1, the computer system 100 includes a processor 110, a memory 120, a persistent storage 130, a bus 140, an input / output interface 150 and a network interface 160.

The processor 110 may include or be part of any device capable of processing a sequence of instructions as a component for predicting a high risk prognosis from a patient ' s diagnostic history. The processor 110 may comprise, for example, a processor and / or a digital processor within a computer processor, a mobile device, or other electronic device. The processor 110 may be, for example, a server computing device, a server computer, a series of server computers, a server farm, a cloud computer, a content platform, and the like. The processor 110 may be connected to the memory 120 via a bus 140.

The memory 120 may include volatile memory, permanent, virtual or other memory for storing information used by or output by the computer system 100. Memory 120 may include, for example, random access memory (RAM) and / or dynamic random access memory (DRAM). The memory 120 may be used to store any information, such as the state information of the computer system 100. Memory 120 may also be used to store instructions of computer system 100, including, for example, instructions for predicting a disease prognosis. Computer system 100 may include one or more processors 110 as needed or where appropriate.

The bus 140 may comprise a communication infrastructure that enables interaction between the various components of the computer system 100. The bus 140 may, for example, carry data between components of the computer system 100, for example, between the processor 110 and the memory 120. The bus 140 may comprise a wireless and / or wired communication medium between the components of the computer system 100 and may include parallel, serial, or other topology arrangements.

The persistent storage device 130 may be a component such as a memory or other persistent storage device as used by the computer system 100 to store data for a predetermined extended period of time (e.g., as compared to the memory 120) Lt; / RTI > The persistent storage device 130 may include non-volatile main memory as used by the processor 110 in the computer system 100. The persistent storage device 130 may include, for example, flash memory, hard disk, optical disk, or other computer readable medium.

The input / output interface 150 may include a keyboard, a mouse, voice command inputs, displays, or interfaces to other input or output devices. Inputs for configuration instructions and / or disease prognosis prediction may be received via the input / output interface 150.

The network interface 160 may include one or more interfaces to networks such as a local area network or the Internet. The network interface 160 may include interfaces for wired or wireless connections. Configuration commands and / or input for predicting the disease prognosis may be received via the network interface 160.

Also, in other embodiments, the computer system 100 may include more components than the components of FIG. However, there is no need to clearly illustrate most prior art components. For example, the computer system 100 may be implemented to include at least some of the input / output devices connected to the input / output interface 150 described above, or may include a transceiver, a Global Positioning System (GPS) module, Databases, and the like.

The present invention provides a technique for predicting a high-risk prognosis from the diagnostic history of patients using RNN as a deep-running technique. In the present specification, hypertension, diabetes and hyperlipidemia are defined as representative examples of adult diseases, and cardiovascular diseases and cerebrovascular diseases are defined as typical examples of high-risk diseases, but the present invention is not limited thereto.

FIG. 2 illustrates an example of components that a processor of a computer system according to an embodiment of the present invention may include; and FIG. 3 illustrates an example of a computer system according to an embodiment of the present invention, A flowchart showing an example of a prediction method.

As shown in FIG. 2, the processor 110 may include a sequence generator 210, a learning processor 220, and a prognostic predictor 230. The components of such a processor 110 may be representations of different functions performed by the processor 110 in accordance with control commands provided by at least one program code. For example, the sequence generator 210 may be used as a functional representation that the processor 110 operates to control the computer system 100 to generate sequence data. The components of the processor 110 and the processor 110 may perform the steps S310 through S340 included in the disease prognosis prediction method of FIG. For example, the components of processor 110 and processor 110 may be implemented to execute instructions in accordance with the at least one program code described above and the code of the operating system that memory 120 contains. Here, at least one program code may correspond to a code of a program implemented to process a disease prognosis prediction method.

The disease prognosis prediction method may not occur in the order shown, and some of the steps may be omitted or an additional process may be further included.

In step S310, the processor 110 may load the program code stored in the program file for the disease prognosis prediction method into the memory 120. [ For example, a program file for a disease prognosis prediction method may be stored in the persistent storage 130 described with reference to FIG. 1, and the processor 110 may retrieve the program files stored in the persistent storage 130 The computer system 110 may be controlled such that the program code is loaded into the memory 120. [ At this time, the sequence generation unit 210, the learning processing unit 220, and the prognostic prediction unit 230 included in the processor 110 and the processor 110 respectively generate the program code of the corresponding part of the program code loaded into the memory 120 And may be different functional representations of the processor 110 for executing subsequent steps (S320 through S340). For the execution of steps S320 through S340, the processor 110 and the components of the processor 110 may process an operation according to a direct control command or control the computer system 100. [

In step S320, the sequence generation unit 210 may generate sequence data for the patient by expressing the diagnostic classification code indicating the patient's diagnosis history as a feature in the form of a word sequence, for each patient.

More specifically, diagnostic data from adult patients can be used to predict the prognosis of cardiovascular and cerebrovascular disease. The diagnostic data may include a code for classifying the patient's disease and symptoms, for example, a diagnostic classification code expressed as ICD-10, the 10th edition of the International Statistical Classification.

A total of 6,667 ICD-10 codes are used as the patient's diagnostic data, and some of these codes may be used or some codes appearing more than a certain number of times in patients, for example, 2,777 codes appearing more than 50 times have. Or by not using diagnostic codes that are not related to predictive illnesses.

Given an ICD-10 code set x and a disease set y, the patient's diagnostic history can be defined as a sequence of ICD-10 codes.

[Equation 1]

d = {x, y} = {(x 1, x 2, x 3, ..., x m), y} st x∈X, y∈Y

&Quot; (2) "

^{d k = {x k, y} } = {(x (mn) -k + 1, ..., x m -n), y}

Where m represents the length of the ICD-10 code sequence, and x ^k represents the last k-length sequence of x. Because the patient sequence length is variable and recent diagnosis has the potential to more accurately characterize the patient's condition, the last k ICD-10 codes (ie, k-length sequence) are used for prognostic prediction. For early prediction of high-risk disease prognosis, models can be learned using codes up to mn (n> 0).

If the diagnostic classification code is an element of a high-risk disease set, the patient can be classified as a high-risk patient.

FIG. 4 shows an example of a process of generating ICD-10 sequence data from a diagnosis history of a patient.

It is assumed that the cardiovascular disease set Y ^H1 {I210, I211, I212} and the cerebrovascular disease set Y ^H2 {I610, I611, I615} are given.

The sequence generation unit 210 can use the patient-specific diagnosis history expressed by the ICD-10 code, and can generate the ICD-10 code sequence indicating the diagnosis history of each patient. At this time, the sequence generation unit 210 may generate at least one data instance according to the disease of each patient for each patient.

The sequence generation unit 210 may include a label indicating a disease that the patient has in the sequence data of each patient. For example, H0 for high-risk disease, H1 for cardiovascular disease, and H2 for cerebrovascular disease.

In the case of the patient 401 having the ICD-10 code set {I10, I100, M870, M870, End} as the diagnosis history, there is no code corresponding to the given disease set (Y ^H1 , Y ^H2 ) 10 sequence data can be generated.

In the case of the patient 402 having the ICD-10 code set {I10, I100, L659, K297, I210} as the diagnosis history, since it contains the code I210 corresponding to the cardiovascular disease among the given disease set (Y ^H1 , Y ^H2 ) It is possible to generate ICD-10 sequence data replacing the code with a label H1 indicating cardiovascular disease.

A patient 403 with the ICD-10 code set {I10, I212, E835, M480, I610} as a diagnostic history corresponds to a cardiovascular disease code I212 of a given disease set (Y ^H1 , Y ^H2 ) Code I610. ≪ / RTI > Two data instances can be created if one patient has two high-risk diseases. In this case, a first code sequence up to code I212 corresponding to cardiovascular disease, a second code sequence up to code I610 corresponding to cerebrovascular disease are generated, and in the first code sequence code I212 is replaced with label H1 indicating cardiovascular disease And in the second code sequence code I610 can be replaced with the label H2 representing the cerebrovascular disease.

Referring back to FIG. 3, in step S330, the learning processing unit 220 can generate learning data through word sequence learning using the RNN with respect to the sequence data generated in step S320. In detail, the learning processing unit 220 can express the sequence data generated by the diagnostic classification code using the RNN model as a real number vector and calculate the disease occurrence probability according to the real number vector. At this time, the learning processing unit 220 can express the word vector including the sequence information through the word sequence learning in the RNN with the sequence data generated from the ICD-10 code indicating the diagnosis history of the patient as the input of the RNN, and the sequence information The probability of occurrence of each high-risk disease can be calculated using the included word vector.

Each ICD-10 code generated by the sequence data can be given as an input of the learned RNN sequentially. When the sequential input is completed, a new output real vector value is generated in the RNN and the probability of occurrence of each disease is calculated Can be output. In other words, the learning processing unit 220 proceeds the sequence learning on the patient's diagnosis history represented by the ICD-10 code using the RNN so that the word vector can express the word sequence information, and then uses the sequence learning result of the RNN The probability of high-risk disease occurrence of the patient can be calculated.

In step S340, the prognosis prediction unit 230 can predict the disease prognosis of the patient based on the disease occurrence probability output through the RNN model with respect to the patient's diagnosis history. In other words, the prognosis prediction unit 230 can predict the disease prognosis of the patient using the learning data generated in step S330. For example, the prognosis prediction unit 230 can determine the prognosis grade for each patient by using the probability of occurrence of each high-risk disease. For example, each patient's disease can be defined as a grade label according to the probability of occurrence, and patient A can be defined as [adult disease: 0, high risk: 1] or [cardiovascular disease: 1, cerebrovascular disease: 0] You can assign a separate label to a potential disease. As another example, the prognosis prediction unit 230 can classify the patient into at least one disease group using the probability of occurrence of each high-risk disease. As another example, the prognosis prediction unit 230 may provide an alert for the disease with the highest probability according to the occurrence probability of each high-risk disease.

Predicting the prognosis from the sequence data representing the diagnostic history is similar to the task of classifying text from a word sequence.

FIG. 5 shows an example of an RNN model structure for predicting disease prognosis in an embodiment of the present invention.

5, an RNN (hereinafter referred to as PP-RNN) for prognosis prediction according to the present invention includes an embedding layer 501 for distributed representation, Multiple recurrent layers 502, and fully-connected layers 503 for prognostic classification.

Unlike other neural networks, RNN is a model that explicitly learns patterns of sequence data considering time factors.

The embedding layer 501 converts the ICD-10 code into a real number vector. The embedding layer 501 can express a word by expressing it as a multi-dimensional real vector and expressing the semantic / structure similarity between words as a distance between two vectors. These distributed expressions characterize the semantics of the code and accommodate it so that the code can be computed by computational operations as in neural language models.

The multiple regression layer 502 may model the sequence pattern of the ICD-10 code. The multi-regression layer 502 corresponds to a neural network layer considered in terms of temporal aspects, in which a sequence can effectively learn a pattern from data given as input.

The full connection layer 503 represents the sequence of ICD-10 codes as a real number vector, which can then be used to calculate the likelihood of disease outbreak through a soft max function.

&Quot; (3) "

&Quot; (4) "

는 k-길이 시퀀스와 질병에 대해 PP-RNN에 의해 계산된 스코어 함수이고,

는 x^k로부터 예측된 질병을 의미한다.here,

Is a scoring function computed by PP-RNN for k-length sequences and disease,

Means the disease predicted from x ^k .

The learning of the PP-RNN is performed by propagating the prediction error to the input feature layer 500 through the full connection layer 503, the multiple regression layer 502, and the embedding layer 501, to-end method.

In the present invention, by using the RNN suitable for the sequence learning, the diagnostic classification code can be easily expressed as a vector value based on the word embedding method.

The PP-RNN model of the present invention can provide a disease prognosis prediction model of E2E type that enables classification of high-risk diseases from embedded word vectors as well as word embedding expressing code sequences as word vectors.

The PP-RNN may include a plurality of RNNs for two or more heterogeneous data features and may be provided to the regression layer 502 of the RNN to prevent the vanishing gradient problem caused by the long sequence, a gated recurrent unit, and a long short-term memory (LSTM).

In the present invention, various methods can be applied to merge a plurality of RNNs. As shown in FIG. 5, the output vectors of the returning layer 502 can be connected to merge them.

In addition to using only one RNN model for the sequence of diagnostic classification codes, it is also possible for the present invention to use multiple RNNs for a sequence of additional clinical histories. Diagnostic categorization codes associated with diagnostic history can be used as additional features such as time of diagnosis, disease duration, treatment history and pathological measurement data, and sequences for each feature can be learned using individual neural networks This can be combined to classify the disease prognosis. For example, in addition to the diagnostic classification code according to the diagnostic history, the type of patient can be utilized as an additional clinical feature, separated into three values: outpatient, inpatient, and emergency. For each additional feature, individual neural networks such as FFNN (feedforward neural network) or CNN (convolutional neural network) as well as RNN can be learned depending on the characteristics of the data.

The use of additional clinical features in conjunction with diagnostic classification codes associated with diagnostic histories in the disease prognosis prediction system and method according to the present invention can provide more accurate predictions of disease prognosis.

As described above, according to embodiments of the present invention, the diagnosis classification code can be expressed as a feature in the form of a word sequence, and high-risk prognosis can be predicted through learning using RNN. Therefore, a simple diagnostic history expressed in code can predict the high-risk prognosis faster and more accurately and provide a quick warning about the high-risk prognosis.

The apparatus described above may be implemented as a hardware component, a software component, and / or a combination of hardware components and software components. For example, the apparatus and components described in the embodiments may be implemented as a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit, a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For ease of understanding, the processing apparatus may be described as being used singly, but those skilled in the art will recognize that the processing apparatus may have a plurality of processing elements and / As shown in FIG. For example, the processing unit may comprise a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as a parallel processor.

The software may include a computer program, code, instructions, or a combination of one or more of the foregoing, and may be configured to configure the processing device to operate as desired or to process it collectively or collectively Device can be commanded. The software and / or data may be in the form of any type of machine, component, physical device, virtual equipment, computer storage media, or device , Or may be embodyed temporarily. The software may be distributed over a networked computer system and stored or executed in a distributed manner. The software and data may be stored on one or more computer readable recording media.

The method according to an embodiment may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer-readable medium. The medium may be one that continues to store computer executable programs, or temporarily store them for execution or download. Further, the medium may be a variety of recording means or storage means in the form of a combination of a single hardware or a plurality of hardware, but is not limited to a medium directly connected to any computer system, but may be dispersed on a network. Examples of the medium include a magnetic medium such as a hard disk, a floppy disk and a magnetic tape, an optical recording medium such as CD-ROM and DVD, a magneto-optical medium such as a floptical disk, And program instructions including ROM, RAM, flash memory, and the like. As another example of the medium, a recording medium or a storage medium that is managed by a site or a server that supplies or distributes an application store or various other software is also enumerated.

Program instructions to be recorded on the medium may be those specially designed and constructed for the embodiments or may be available to those skilled in the art of computer software. 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.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. For example, it is to be understood that the techniques described may be performed in a different order than the described methods, and / or that components of the described systems, structures, devices, circuits, Lt; / RTI > or equivalents, even if it is replaced or replaced.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (17)

A method for predicting a disease prognosis performed in a computer system,
The computer system comprising at least one processor configured to execute computer-readable instructions contained in a memory,
Wherein the at least one processor comprises:
A sequence generation unit, a learning processing unit, and a prognosis prediction unit as elements for performing the disease prognosis prediction method,
The disease prognosis prediction method comprises:
Generating sequence data by expressing a diagnostic classification code indicating a diagnosis history of a patient in a word sequence form in the sequence generation unit;
Generating learning data by word sequence learning using RNN (Recurrent Neural Networks) on the sequence data in the learning processing unit; And
Wherein the prognosis prediction unit predicts a disease prognosis using the learning data
Lt; / RTI >
Wherein the step of generating the sequence data comprises:
Generating at least one sequence data corresponding to a given disease from the diagnosis classification code,
In the disease prognosis prediction method,
The probability of disease occurrence of the patient is calculated for each given disease using the sequence learning result of the RNN with respect to the sequence data
Wherein said disease prognosis prediction method comprises the steps of: The method according to claim 1,
In the disease prognosis prediction method,
The sequence data is represented by a real number vector, and a disease occurrence probability according to the real number vector is calculated
Wherein said disease prognosis prediction method comprises the steps of: The method according to claim 1,
In the disease prognosis prediction method,
The sequence data is input to the RNN, word vectors including sequence information are represented through word sequence learning in the RNN,
The disease occurrence probability is calculated using the word vector including the sequence information
Wherein said disease prognosis prediction method comprises the steps of: The method according to claim 1,
Wherein the step of generating the learning data comprises:
An RNN consisting of an embedding layer for distributed representation of word sequences, recurrent layers for sequence modeling, and fully-connected layers for disease prognosis classification Using
Wherein said disease prognosis prediction method comprises the steps of: 5. The method of claim 4,
Learning of the RNN is performed in an E2E (end-to-end) manner by propagating a prediction error to the input feature layer through the full connection layer, the regression layer, and the embedding layer
Wherein said disease prognosis prediction method comprises the steps of: The method according to claim 1,
Wherein the step of generating the sequence data comprises:
And generating the sequence data using the latest last predetermined number of codes among the diagnostic classification codes
Wherein said disease prognosis prediction method comprises the steps of: The method according to claim 1,
Wherein the step of generating the sequence data comprises:
Replacing the code corresponding to the given disease among the diagnostic classification codes with the label of the disease
Wherein said disease prognosis prediction method comprises the steps of: The method according to claim 1,
Wherein the step of generating the sequence data comprises:
If there are a plurality of codes corresponding to the given disease among the diagnostic classification codes, generating sequence data for each disease
Wherein said disease prognosis prediction method comprises the steps of: The method according to claim 1,
The disease prognosis prediction method comprises:
Generating sequence data on at least one additional clinical history of the time and disease duration of the patient, the treatment history and the pathological measurement data in the sequence generator,
Further comprising:
Wherein the step of generating the learning data comprises:
Adding each individual neural network to the diagnostic history of the patient and the additional clinical history according to the characteristics of the corresponding data
Wherein said disease prognosis prediction method comprises the steps of: A computer-readable recording medium having recorded thereon a program for causing a computer to execute a disease prognosis prediction method according to any one of claims 1 to 9. A computer-implemented disease prognosis prediction system comprising:
And at least one processor configured to execute computer-readable instructions contained in the memory,
Wherein the at least one processor comprises:
A sequence generator for generating sequence data by expressing a diagnostic classification code indicating a diagnosis history of a patient in a word sequence;
A learning processing unit for generating learning data through word sequence learning using Recurrent Neural Networks (RNN) for the sequence data; And
A prognostic prediction unit for predicting a disease prognosis using the learning data;
Lt; / RTI >
Wherein the sequence generator comprises:
Generating at least one sequence data corresponding to a given disease from the diagnosis classification code,
In the at least one processor,
The probability of disease occurrence of the patient is calculated for each given disease using the sequence learning result of the RNN with respect to the sequence data
Wherein said disease prognosis prediction system comprises: 12. The method of claim 11,
In the at least one processor,
The sequence data is input to the RNN, word vectors including sequence information are represented through word sequence learning in the RNN,
The disease occurrence probability is calculated using the word vector including the sequence information
Wherein said disease prognosis prediction system comprises: 12. The method of claim 11,
The learning processing unit,
An RNN consisting of an embedding layer for distributed representation of word sequences, recurrent layers for sequence modeling, and fully-connected layers for disease prognosis classification Using
Wherein said disease prognosis prediction system comprises: 14. The method of claim 13,
Learning of the RNN is performed in an E2E (end-to-end) manner by propagating a prediction error to the input feature layer through the full connection layer, the regression layer, and the embedding layer
Wherein said disease prognosis prediction system comprises: 12. The method of claim 11,
Wherein the sequence generator comprises:
And generating the sequence data using the latest last predetermined number of codes among the diagnostic classification codes
Wherein said disease prognosis prediction system comprises: 12. The method of claim 11,
Wherein the sequence generator comprises:
If there are a plurality of codes corresponding to the given disease among the diagnostic classification codes, generating sequence data for each disease
Wherein said disease prognosis prediction system comprises: 12. The method of claim 11,
Wherein the sequence generator comprises:
Generating sequence data for an additional clinical history of at least one of the time and the disease duration of the patient, the treatment history and the pathological measurement data of the patient,
Learning by adding each individual neural network according to the characteristic of the data on the diagnosis history of the patient and the additional clinical history
Wherein said disease prognosis prediction system comprises:

Images