WO2019225798A1

WO2019225798A1 - Method and device for selecting question in multiple psychological test sheets on basis of machine learning to promptly diagnose anxiety and depression symptoms

Info

Publication number: WO2019225798A1
Application number: PCT/KR2018/007056
Authority: WO
Inventors: 정범석; 채명수; 윤석호
Original assignee: 한국과학기술원
Priority date: 2018-05-23
Filing date: 2018-06-22
Publication date: 2019-11-28

Abstract

Provided are a method and a device for selecting a question in multiple psychological test sheets on the basis of machine learning to promptly diagnose anxiety and depression symptoms. A method for selecting a question in multiple psychological test sheets on the basis of machine learning to promptly diagnose anxiety and depression symptoms according to an embodiment may comprise the steps of: training a machine learning-based interactive diagnosis tool by using mental health research data as a training data set; and updating status information of a subject every time the subject responds to the interactive diagnosis tool.

Description

Method and device based on machine learning for rapid diagnosis of anxiety and depressive symptoms in multiple psychological papers

The following examples relate to methods and apparatuses for classifying psychological risk groups, and more particularly, to machine-based question screening methods for rapid diagnosis of anxiety and depressive symptoms in multiple psychological papers. And to an apparatus.

In psychological counseling, it is important to understand the person's condition. Therefore, several interactive mental disorder diagnostic tools have recently been developed. Existing psychiatric diagnostic tools are time-consuming and difficult to assess for a variety of diseases, because all symptoms of diagnostic criteria, such as the Diagnostic and Statistical Manual of Mental Disorders (DSM), must be identified. For this reason, most interactive diagnostic tools should only be evaluated for a single disease or identify different symptoms. However, if the interactive diagnostic tool asks too many questions, the time required may be too long to reduce the listener's concentration. As a result, efforts should be made to solve existing problems using diagnostic systems used in psychiatric clinical diagnosis and machine learning techniques.

On the other hand, many diseases can be caused by stress, but most of the previous virtual conversational tools developed to provide recent psychological counseling were designed to diagnose a single disorder. Existing tools also had to examine all symptoms according to diagnostic criteria such as DSM.

Embodiments describe a machine learning-based question screening method and apparatus for rapid diagnosis of anxiety and depression in a plurality of psychological papers, and more specifically, a risk group classification algorithm for use in an interactive diagnostic tool of psychiatry. Develop and evaluate its performance.

Embodiments provide interactive diagnostic tools based on machine learning using real clinical data to screen for a variety of mental illnesses that can occur in real stress situations rather than a single disease. To provide a machine learning-based item screening method and apparatus for rapid diagnosis of anxiety and depressive symptoms in a number of psychological papers.

Machine learning-based item screening method for rapid diagnosis of anxiety and depressive symptoms in a plurality of psychological paper according to an embodiment, machine learning based interactive diagnosis using mental health survey data as a training data set Learning an interactive diagnosis tool; And updating status information of the subject every time the user responds to the interactive diagnostic tool.

The method may further include a preprocessing step of applying an Rpart tree algorithm, which is an R package to which a resampling technique is applied, for the design of the interactive diagnostic tool.

The method may further include diagnosing a mental disorder through a specific question using the interactive diagnostic tool.

The mental health survey data includes the Fatty Health Questionnaire-9 (PHQ-9), the Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz social anxiety scale (LSAS). It can be configured as.

The learning of the interactive diagnostic tool includes the Depression Screening Tool (Patient Health Questionnaire-9, PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz Social Anxiety Scale (Liebowitz social). training an decision tree to find subjects at high risk of depression, generalized anxiety disorder, and social anxiety disorder using anxiety scale (LSAS).

And learning the interactive diagnostic tool may further include asking additional questions to determine the presence or absence of a mental illness when the decision tree is identified as a risk group.

Each target value of the Depression Screening Tool (Patient Health Questionnaire-9, PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz social anxiety scale (LSAS) The target value may be defined as the total score of each item exceeding a specific cut-off value.

Machine learning-based question screening device for the rapid diagnosis of anxiety and depressive symptoms in a plurality of psychological test paper according to another embodiment, the input unit for inputting the mental health survey data to the training data set; A learning unit for learning an interactive diagnosis tool based on machine learning using the training data set; And an update unit for updating the subject's status information each time the user responds to the interactive diagnostic tool.

The apparatus may further include a preprocessor configured to apply an Rpart tree algorithm, which is an R package to which a resampling technique is applied, for the design of the interactive diagnostic tool.

The apparatus may further include a diagnosis unit for diagnosing a mental disease through a specific question using the interactive diagnostic tool.

The learning unit uses the Patient Health Questionnaire-9 (PHQ-9), the Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz Social Anxiety Scale (LSAS). Training decision trees to find people at high risk for depression, generalized anxiety, and social anxiety disorders.

When the decision tree is identified as a risk group, the learning unit may ask additional questions to determine whether there is a mental illness.

According to embodiments, screening for various mental disorders that may occur in a real stress situation rather than a single disease by providing an interactive diagnostic tool based on machine learning using actual clinical data. A number of psychological test papers can provide a machine learning-based item screening method and apparatus for rapid diagnosis of anxiety and depression.

In addition, according to embodiments, a machine learning-based item screening method for rapid diagnosis of anxiety and depressive symptoms in a large number of psychological test papers that perform relatively accurate mental disease judgment with less conversation through a risk group classification algorithm for an interactive diagnostic tool. A method and apparatus can be provided.

1 is a view schematically showing the structure of an item sorting apparatus according to an embodiment.

2 is a flowchart illustrating a question screening method according to an exemplary embodiment.

3 is a diagram illustrating a decision tree of a depression risk group according to an embodiment.

4 illustrates a result of a pruning process of a decision tree of a depression risk group according to an exemplary embodiment.

5 illustrates a decision tree of a GAD risk population, according to one embodiment.

6 illustrates a result of a pruning process of a decision tree of a GAD risk group according to an embodiment.

7 is a diagram illustrating a decision tree of an SAD risk group according to an embodiment.

8 is a diagram illustrating a result of pruning a decision tree of a SAD risk group according to an embodiment.

Hereinafter, exemplary embodiments will be described with reference to the accompanying drawings. However, the described embodiments may be modified in many different forms, and the scope of the present invention is not limited to the embodiments described below. In addition, various embodiments are provided to more fully describe the present invention to those skilled in the art. Shape and size of the elements in the drawings may be exaggerated for more clear description.

The following embodiments may provide a machine learning based question screening method and apparatus for rapid diagnosis of anxiety and depressive symptoms in a plurality of psychological papers to classify a risk group of mental illness. To do this, we develop a risk group classification algorithm for use in interactive diagnostic tools in psychiatry and evaluate its performance.

More specifically, the real clinical data may be used to provide an interactive diagnosis tool based on machine learning. Interactive diagnostic tools based on machine learning can be screened for a variety of mental illnesses that can occur in real stress situations rather than a single disease, and can be designed to collect and judge appropriate additional information step by step. have.

Referring to FIG. 1, the structure of an item selection apparatus according to an embodiment is schematically illustrated and is a view for explaining a decision making process of an interactive diagnostic tool. In the following, a machine learning-based item screening device for quickly diagnosing anxiety and depressive symptoms in a number of psychological papers will be referred to simply as an item screening device.

The item sorting apparatus according to an embodiment may include an input unit 110, a learner 120, and an updater 130. According to an embodiment, the item selection device may further include a preprocessor and a diagnosis unit 140.

The input unit 110 inputs basic information for determining whether there is a mental disease as a training data set, and may input collected mental health survey data as a training data set. Mental health survey data can be collected from a subject, such as a number of psychological papers obtained from multiple subjects.

Here, mental health survey data include the Patient Health Questionnaire-9 (PHQ-9), the Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz social anxiety scale (LSAS). It can be composed of).

The learner 120 may train an interactive diagnosis tool based on machine learning using a training data set. For example, machine learning-based interactive diagnostic tools can be designed to perform a screening test for depression, diagnose depression, and finally diagnose depression by collecting additional information about depression. In addition to depression, general anxiety disorders and social anxiety disorders may be designed to perform a screening test, diagnose each mental disorder, and collect additional information step by step to finally diagnose the mental disorder.

For example, the learning unit 120 may use the depression screening tool (PHQ-9), the panic anxiety assessment (GAD-7), and the Liebowitz Social Anxiety Scale (LSAS) to determine depression, anxiety disorder, and social anxiety disorder. You can train the decision tree to find high-risk subjects. Thereafter, when the decision tree is identified as a risk group, the learning unit 120 may ask additional questions to determine whether there is a mental disease.

At this time, each target value of the depression screening tool (PHQ-9), generalized anxiety disorder assessment (GAD-7), and the Liebowitz Social Anxiety Scale (LSAS) exceeds a certain cut-off value. It can be defined as the total score of each item.

The updater 130 updates the patient's status, and can update the subject's status information each time the user responds to the interactive diagnostic tool.

According to an embodiment, the item selection device may further include a preprocessor and a diagnosis unit 140.

The preprocessor may apply an Rpart tree algorithm, which is an R package to which the resampling technique is applied, for the design of the interactive diagnostic tool.

In addition, the diagnosis unit 140 may diagnose a mental disease through a specific question using an interactive diagnostic tool. That is, the diagnosis unit 140 may finally diagnose the various mental diseases by identifying the optimal question. At this time, the diagnosis unit 140 may finally diagnose the mental disease through appropriate intervention.

For example, mental health survey data obtained from 5858 subjects at a university may be used as a training data set for machine learning. The survey may consist of the Depression Screening Tool (PHQ-9), the Global Anxiety Disability Assessment (GAD-7), and the Liebowitz Social Anxiety Scale (LSAS). In this case, the Rpart tree algorithm, which is an R package to which the resampling technique is applied, may be applied in the preprocessing step to solve the imbalance problem (Non-Patent Document 1).

You can design an interactive diagnostic tool to update the subject's information each time you respond and reuse it when needed. Each target value may be defined as the total score of each item exceeding a specific cutoff value.

Doctors may use the Depression Screening Tool (PHQ-9), the Global Anxiety Disability Assessment (GAD-7), and the Liebowitz Social Anxiety Scale (LSAS) as described above to identify subjects at high risk of depression, general anxiety disorder, and social anxiety disorder You can train the decision tree. Once the decision tree is identified as a risk group, it can be designed to ask additional questions to determine the presence of a disease. Therefore, the best questions can identify a variety of diseases.

The accuracy, recall and all F1T scores of the trained algorithm are as follows (accuracy / recall / all F1T scores).

Depression 93.1% / 87.6% / 90.3%

Generalized Anxiety Disorder (GAD) 97.2% / 92.6% / 94.8%

94.0% / 86.4% / 90.0% for Social Anxiety Disorder (SAD).

Referring to FIG. 2, a machine learning-based item screening method for rapid diagnosis of anxiety and depression symptoms in a plurality of psychological test papers according to an embodiment may include machine learning-based dialogue using mental health survey data as a training data set. Learning 220 the diagnostic type tool, and updating the status information of the subject every time the user responds to the interactive diagnostic tool 230.

The method may further include a preprocessing step 210 of applying an Rpart tree algorithm, which is an R package to which a resampling technique is applied, for the design of an interactive diagnostic tool.

In addition, the method may further include a step 240 of diagnosing a mental disease through a specific question using an interactive diagnostic tool.

Hereinafter, each step of the machine learning-based item selection method for prompt diagnosis of anxiety and depressive symptoms in a plurality of psychological test papers will be described in more detail.

Machine learning-based item screening method for the rapid diagnosis of anxiety and depressive symptoms in a plurality of psychological test paper according to an embodiment for rapid diagnosis of anxiety and depressive symptoms in a plurality of psychological test paper according to an embodiment described in FIG. The machine learning based item selection device can be described in more detail as an example. In the following, a machine learning-based item screening method for prompt diagnosis of anxiety and depressive symptoms in a plurality of psychological test papers will be briefly referred to as a question screening method. Meanwhile, the item sorting apparatus according to the embodiment may include an input unit, a learning unit, and an update unit, and the item sorting apparatus may further include a preprocessor and a diagnostic unit.

First, in step 210, the preprocessor may preprocess the Rpart tree algorithm, which is an R package to which the resampling technique is applied, for the design of the interactive diagnostic tool.

In operation 220, the learner may train the machine learning based interactive diagnostic tool by using the mental health survey data as the training data set. In this case, the learning unit may receive the mental health survey data from the input unit as a training data set. Here, mental health survey data may be comprised of a depression screening tool (PHQ-9), a general anxiety disorder assessment (GAD-7), and the Liebowitz Social Anxiety Scale (LSAS).

The Department uses the Depression Screening Tool (PHQ-9), the Global Anxiety Disability Assessment (GAD-7), and the Liebowitz Social Anxiety Scale (LSAS) to find physicians who are at high risk for depression, generalized anxiety and social anxiety disorders. You can train the decision tree. The learning unit can then ask further questions to determine the presence of mental illness when the decision tree is identified as a risk group.

Each target value of the Depression Screening Tool (PHQ-9), Global Anxiety Disability Assessment (GAD-7), and Liebowitz Social Anxiety Scale (LSAS) can be defined as the total score of each item exceeding a specific cutoff value.

In operation 230, the updater may update the subject's status information each time the user responds to the interactive diagnostic tool.

In operation 240, the diagnosis unit may diagnose a mental disease through a specific question using an interactive diagnostic tool.

Therefore, by using real clinical data to provide an interactive diagnostic tool based on machine learning, screening for various mental disorders that can occur in a real stress situation rather than a single disease can be performed. . In addition, the risk group classification algorithm for the interactive diagnostic tool enables relatively accurate mental disease determination with less conversation.

3 is a diagram illustrating a decision tree of a depression risk group according to an embodiment. 4 illustrates a result of a pruning process of a decision tree of a depression risk group according to an embodiment.

Referring to FIGS. 3 and 4, which represent the results of the decision tree and pruning process of the depression risk group, the decision tree can train the depression risk group through only two to five questions.

5 illustrates a decision tree of a GAD risk population, according to one embodiment. 6 illustrates a result of pruning a decision tree of a GAD risk group according to an embodiment.

Referring to Figures 5 and 6, which represent the results of the decision tree and pruning process of a generalized anxiety disorder (GAD) risk group, the decision tree is a generalized anxiety disorder (GAD) risk group through only two or three questions. Can train.

7 is a diagram illustrating a decision tree of an SAD risk group according to an embodiment. 8 illustrates a result of pruning a decision tree of a SAD risk group according to an embodiment.

Referring to Figures 7 and 8, which represent the results of the decision tree and pruning process of the Social Anxiety Disorder (SAD) risk group, the decision tree has only two to four questions for the Social Anxiety Disorder (SAD) risk group. Can be trained.

Decision trees using clinical data showed high sensitivity and specificity through several questions about the risk groups of three mental disorders: depression, generalized anxiety disorder (GAD), and social anxiety disorder (SAD).

In the above, a risk group classification algorithm for use in an interactive diagnostic tool of psychiatry was developed and its performance was evaluated. According to embodiments, the risk group classification algorithm for the interactive diagnostic tool has relatively good results in depression, generalized anxiety disorder (GAD) and social anxiety disorder (SAD). The developed risk group discrimination algorithm is expected to assist in making relatively accurate judgments with less dialogue with interactive diagnostic tools.

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 devices and components described in the embodiments include, for example, processors, controllers, arithmetic logic units (ALUs), digital signal processors, microcomputers, field programmable arrays (FPAs), It may be implemented using one or more general purpose or special purpose computers, such as a programmable logic unit (PLU), 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 the execution of the software. For the convenience of understanding, a processing device may be described as one being used, but a person skilled in the art will appreciate that the processing device includes a plurality of processing elements and / or a plurality of types of processing elements. It can be seen that it may include. For example, the processing device may include a plurality of processors or one processor and one controller. In addition, other processing configurations are possible, such as parallel processors.

The software may include a computer program, code, instructions, or a combination of one or more of the above, and configure the processing device to operate as desired, or process it independently or collectively. You can command the device. Software and / or data may be any type of machine, component, physical device, virtual equipment, computer storage medium or device in order to be interpreted by or to provide instructions or data to the processing device. It can be embodied in. The software may be distributed over networked computer systems so that they may be stored or executed in a distributed manner. Software and data may be stored on one or more computer readable recording media.

The method according to the embodiment may be embodied in the form of program instructions that can be executed by various computer means and recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the media may be those specially designed and constructed for the purposes of the embodiments, or they may be of the kind well-known and available to those having skill in the computer software arts. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks, 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 not only machine code generated by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like.

Although the embodiments have been described by the limited embodiments and the drawings as described above, various modifications and variations are possible to those skilled in the art from the above description. For example, the described techniques may be performed in a different order than the described method, and / or components of the described systems, structures, devices, circuits, etc. may be combined or combined in a different form than the described method, or other components. Or even if replaced or substituted by equivalents, an appropriate result can be achieved.

Therefore, other implementations, other embodiments, and equivalents to the claims are within the scope of the claims that follow.

Claims

Training an interactive diagnosis tool based on machine learning using mental health survey data as a training data set; And

Updating status information of the subject each time the user responds to the interactive diagnostic tool.

Including, the item selection method.
The method of claim 1,

A preprocessing step of applying an Rpart tree algorithm, which is an R package to which a resampling technique is applied, for the design of the interactive diagnostic tool.

Further comprising, item selection method.
The method of claim 1,

Diagnosing mental illness through a specific question using the interactive diagnostic tool

Further comprising, item selection method.
The method of claim 1,

The mental health survey data includes the Fatty Health Questionnaire-9 (PHQ-9), the Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz social anxiety scale (LSAS). Consisting of

Characterized in that the question screening method.
The method of claim 1,

Learning the interactive diagnostic tool,

Depression, malaise using the Depression Screening Tool (Patient Health Questionnaire-9, PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz social anxiety scale (LSAS). Training decision trees to find people at high risk of anxiety disorders and social anxiety disorders.

Including, the item selection method.
The method of claim 5,

Learning the interactive diagnostic tool,

If the decision tree is identified as a risk group, asking further questions to determine the presence of mental illness

Further comprising, item selection method.
The method of claim 5,

Each target value of the Depression Screening Tool (Patient Health Questionnaire-9, PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz social anxiety scale (LSAS) target value is defined as the total score of each item that exceeds a certain cut-off value

Characterized in that the question screening method.
An input unit for inputting mental health survey data into a training data set;

A learning unit for learning an interactive diagnosis tool based on machine learning using the training data set; And

Update unit for updating the status information of the subject each time the response to the interactive diagnostic tool

Including, the item selection device.
The method of claim 8,

Pre-processing unit for applying the Rpart tree algorithm, which is an R package to which the resampling technique is applied, for the design of the interactive diagnostic tool

Further comprising, item selection device.
The method of claim 8,

Diagnosis unit for diagnosing mental disorders through specific questions using the interactive diagnostic tool

Further comprising, item selection device.
The method of claim 8,

The mental health survey data includes the Fatty Health Questionnaire-9 (PHQ-9), the Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz social anxiety scale (LSAS). Consisting of

Characterized in that the item selection device.
The method of claim 8,

The learning unit,

Depression, malaise using the Depression Screening Tool (Patient Health Questionnaire-9, PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz social anxiety scale (LSAS). Training decision trees to find people at high risk of anxiety disorders and social anxiety disorders.

Characterized in that the item selection device.
The method of claim 12,

The learning unit,

If the decision tree is identified as a risk group, asking further questions to determine the presence of mental illness

Characterized in that the item selection device.
The method of claim 12,

Each target value of the Depression Screening Tool (Patient Health Questionnaire-9, PHQ-9), Generalized Anxiety Disorder-7 (GAD-7), and the Liebowitz social anxiety scale (LSAS) target value is defined as the total score of each item that exceeds a certain cut-off value

Characterized in that the item selection device.