KR20200001777A - Method for predicting of depression and device for predicting of depression risk using the same - Google Patents

Abstract

The present invention relates to a depression prediction method and a depression prediction device using the same. The depression prediction method, which is implemented by a processor, comprises the steps of: receiving an instant mood assessment score of a subject from a wearable device worn by the subject; and evaluating depression for the subject by using a depression prediction model configured to predict depression on the basis of the instant mood assessment score of the subject.

Description

Prediction method for depression and device for predicting depression occurrence using the same {METHOD FOR PREDICTING OF DEPRESSION AND DEVICE FOR PREDICTING OF DEPRESSION RISK USING THE SAME}

The present invention relates to a method for predicting the onset of depression and a device for predicting the onset of depression using the same, and more specifically, it is possible to predict the onset of depression based on various subjective or objective data received from a wearable device worn by a subject. It relates to a method and a device.

Depression refers to a disease that causes various cognitive and psychosomatic symptoms, resulting in deterioration of daily function, with a decrease in motivation and depression. People who are depressed usually have symptoms that they can't control their mood or aren't interested in fun activities. The causes of depression may include biochemical factors caused by neurotransmitters, genetic factors associated with family history, and environmental factors caused by the surrounding environment.

On the other hand, as a method for diagnosing depression, there may be a method based on in-depth interviews of psychiatric practitioners, a self-report questionnaire, or an electroencephalogram analysis. More specifically, according to the method for predicting depression based on EEG analysis, the subject who is suspected of depression is subjected to a certain electrical stimulation to the brain, and then the depression can be determined by measuring the amount of the peak occurring between 300 ms and 600 ms. Can be. However, the diagnosis of depression through EEG analysis requires a low signal to noise ratio, complicated configuration of EEG electrodes necessary for accurate EEG measurement, and high CPU performance required for analyzing EEG obtained. There is such a problem. In addition, the diagnosis using such a complicated measuring device and method, there is a limit that can not be easily performed at home without a general practitioner, can put a great burden on the elderly or disabled people with limited mobility.

Therefore, there is a continuous demand for the development of a system that can overcome the above limitations and diagnose depression. In particular, the development of a system for predicting depression, which can predict depression more easily than a conventional method, and provide diagnostic information in connection with a medical system, can be used in preventing and diagnosing depression for those with limited mobility. It can be very important.

The background art of the invention has been created to facilitate understanding of the present invention. It should not be understood that the matters described in the background of the invention exist as prior art.

The wearable device is used in the form of a smart band, a smart watch, a wearable glass, and the like, and detects a biorhythm using a sensor provided in the wearable device, and provides a service using the wearable device.

The wearable device may enable continuous monitoring of the subject over days to months without disturbing the subject's daily life. Thus, the observation data obtained from the wearable device, together with various other clinical diagnostic results can be used for disease diagnosis or treatment. However, the above diagnostic system is limited in the scope of application only to the diseases that can determine whether the disease by a simple change of observation data. In other words, diseases that are difficult to diagnose or predict may be difficult to apply to the above-described system, as long-term treatment such as depression is required and data is easily changed according to external factors.

On the other hand, the inventors of the present invention have noted the instantaneous mood scores, activity levels, and environmental factors, in particular, light exposure, as assessed by the subject in relation to depression, and were able to recognize that these factors are associated with depression.

Furthermore, the inventors of the present invention have recognized that in order to increase the accuracy of predictions for the development of depression, a prediction model learned by data associated with depression received from a subject's wearable device can be used.

As a result, the inventors of the present invention are configured to predict the onset of depression based on data such as an instant mood evaluation score input from a subject's wearable device, activity amount measured in real time, illuminance amount, red light level, green light level and blue light level. It has led to the development of a new depression prediction system based on the prediction model.

Thus, an object that you want, and the invention is resolved, and the patient is receiving a mood moments that can be used as objective data evaluation for depression diagnosis score with respect to the subject from a wearable device worn by the depression predictive models for patient The present invention provides a method for predicting depression onset and a device for predicting depression onset using the same.

An object of the present invention is to further receive at least one selected from the group consisting of activity, illuminance, red light level, green light level and blue light level measured in real time from the wearable device, and exposed to the subject using a depression prediction model The present invention provides a method for predicting depression onset and a device for predicting depression onset using the same, configured to evaluate whether depression develops according to a predetermined environment.

Another problem to be solved by the present invention is to generate a state pattern for the subject based on the instantaneous mood evaluation score, to evaluate the diagnosis of depression based on the state pattern in the depression prediction model, and depression based on the depression prediction model It is to provide an onset prediction device.

Another object of the present invention is to provide a method for predicting depression and a depression onset predictive device configured to predict senile depression based on a depression prediction model by inputting gender and age information for a subject. .

The objects of the present invention are not limited to the above-mentioned objects, and other objects that are not mentioned will be clearly understood by those skilled in the art from the following description.

In order to solve the above problems, there is provided a method for predicting depression onset according to an embodiment of the present invention. A method for predicting whether or not depression is implemented by a processor, the method includes: receiving an instant mood evaluation score for a subject from a wearable device worn by the subject, and whether the depression is on the basis of the instant mood evaluation score for the subject Evaluating whether the subject develops depression using a depression prediction model configured to predict the depression.

According to an aspect of the present invention, in the receiving step, the wearable device may further receive at least one selected from the group consisting of activity amount, illuminance amount, red light level, green light level, and blue light level measured in real time, and the depression prediction model may be an activity amount. And based on at least one selected from the group consisting of illuminance, red light level, green light level, and blue light level, and an instant mood evaluation score, it may be further configured to predict whether to develop depression according to the environment exposed to the subject.

According to another feature of the invention, in the receiving step, the wearable device may further receive at least one selected from the group consisting of red light level, green light level and blue light level measured in real time, the depression prediction model is red light level, green light Based on the at least one selected from the group consisting of the level and the blue light level and the instant mood evaluation score, it may be further configured to predict whether the onset of depression according to the type of light exposed to the subject.

According to another feature of the present invention, the receiving step may receive a plurality of instant mood evaluation scores at predetermined time intervals, and evaluating whether the onset of depression is a subject at a predetermined time interval using a depression prediction model Assess whether depression develops

According to another feature of the present invention, the receiving step further receives sleep quality-related data including at least one selected from the group consisting of wake after sleep onset, sleep time, amount of activity during sleep and freshness during wake up The depression prediction model may be further configured to predict whether depression develops for the subject based on the instantaneous mood assessment score and sleep quality association data.

According to another feature of the invention, the depression prediction model receives an instant mood assessment score for each of the depressive and non-depressive individuals different from the subject, and receives the instantaneous mood scores for the depressive and non-depressive individuals. It may be a model trained to calculate a difference in mood assessment scores and to predict whether depression develops for a subject different from a depressive subject and a non-depressive subject based on the calculated instantaneous mood assessment score.

According to another aspect of the present invention, in the step of evaluating the occurrence of depression, using the pattern generation model configured to generate a state pattern for the subject based on the instant mood evaluation score, to generate a state pattern for the subject and predict depression The model can be used to assess the onset of depression in a subject based on the state pattern.

According to another feature of the present invention, the depression prediction model receives an instant mood evaluation score for each of the sample individuals composed of depressive individuals and non-depressive individuals different from the subject, and uses the pattern generation model to receive the sample. A state pattern is generated based on the instantaneous mood evaluation scores for each individual, and among the plurality of state patterns generated for each of the sample individuals, a pattern having a significant difference of more than a predetermined level for a depressive individual and a non-depressive individual is generated. It may be a model trained to determine the pattern associated with the onset of depression and to predict whether to develop depression for a subject different from the depressed and non-depressive individuals based on the pattern associated with the onset of depression.

According to another aspect of the invention, in the step of determining the pattern associated with the onset of depression, odds ratio, coefficients and marginal effects for each of the plurality of state patterns generated for each sample subject effect may be calculated and a pattern associated with the onset of depression may be determined based on the calculated evaluation value.

According to another feature of the invention, the receiving step may further receive the age and sex for the subject, the disclosed depression may be senile depression.

In order to solve the above problems, there is provided a device for predicting depression onset according to an embodiment of the present invention. The device is a device configured to predict whether an onset of depression implemented by a processor, comprising: a receiver configured to receive an instant mood evaluation score for a subject from a wearable device worn by the subject, a processor connected to communicate with the receiver, the instant mood assessment And a processor configured to evaluate depression onset for the subject using a depression prediction model configured to predict depression on the basis of the score, and a storage configured to store the instant mood evaluation score and the depression prediction model.

According to a feature of the invention, the receiving unit is configured to further receive at least one selected from the group consisting of activity amount, illuminance amount, red light level, green light level and blue light level measured in real time from the wearable device, the processor uses a depression prediction model It may be further configured to evaluate whether the onset of depression according to the environment exposed to the subject.

According to another feature of the invention, the receiving unit is configured to further receive at least one selected from the group consisting of red light level, green light level and blue light level measured in real time from the wearable device, the processor is exposed to the subject using a depression prediction model It may be further configured to evaluate whether the onset of depression according to the type of light.

According to another feature of the invention, the device may further comprise a data preprocessor and the data preprocessor may be configured to correct a score having an error value among the received instant mood evaluation scores.

According to another feature of the present invention, the method may further include a pattern generator configured to generate a state pattern for the subject based on the instant mood evaluation score.

According to another feature of the invention, the pattern generator is further configured to generate a pattern associated with depression onset based on the instantaneous mood assessment score from a sample subject different from the subject, and the processor uses the depression prediction model to determine the pattern associated with depression onset And comparing the state pattern generated for the subject, and probabilistically calculating whether or not the onset of depression for the subject is based on the result of the comparison.

According to another feature of the invention, the receiver is further configured to receive sleep quality-related data including at least one selected from the group consisting of awakening time, sleep time, amount of activity during sleep and freshness during wake up for the subject, processor The may be further configured to assess whether the subject developed depression based on the instantaneous mood evaluation score and sleep quality association data using the depression prediction model.

According to another feature of the invention, the receiving unit is further configured to receive a plurality of instantaneous mood evaluation scores at predetermined time intervals, the processor evaluates whether the onset of depression for the subject at a predetermined time interval using a depression prediction model It can be further configured to.

According to another feature of the invention, the receiver is configured to further receive age and gender for the subject, and the disclosed depression may be senile depression.

The present invention provides a method for predicting depression onset based on instantaneous mood evaluation scores received from a wearable device of a subject and various data measured in real time from the wearable device, and a device for predicting depression onset using the same, thereby quickly detecting a risk of depression of a subject. And, information about this can be provided to the examinee and medical personnel. Thus, the present invention can provide a good prognosis for the treatment of depression by advancing the time of treatment for the subject.

In particular, the present invention predicts the occurrence of depression in consideration of the instantaneous mood evaluation score input directly from the subject at predetermined time intervals, and thus has an effect of predicting the occurrence of depression with high accuracy and sensitivity for the user.

Furthermore, the present invention provides information related to the predicted depression diagnosis for the subject based on data measured in real time from the wearable device, thereby preventing problems with the conventional depression diagnosis system, especially the elderly or the disabled with limited mobility. There is an effect that can solve problems such as inconvenience of diagnosis suffered by the subject.

In addition, the present invention uses a depression prediction model configured to predict whether to develop depression based on data received from the wearable device, and furthermore, by using a pattern generation model configured to generate a state pattern, depression occurrence with high accuracy is predicted for each individual subject. There is an effect that can provide a prediction method and a device for predicting depression onset using the same.

The effects according to the present invention are not limited by the contents exemplified above, and more various effects are included in the present specification.

1A is a diagram illustrating a system for predicting depression onset using a device for predicting depression according to an embodiment of the present invention.
1B illustrates the configuration of a device for predicting depression onset according to an embodiment of the present invention.
2 exemplarily illustrates a procedure of a method for predicting depression onset according to an embodiment of the present invention.
3A to 3C illustrate evaluation results of a plurality of prediction models according to input variables for setting a depression prediction model used in various embodiments of the present disclosure.
4A to 4D illustrate evaluation results of a depression prediction model used in various embodiments of the present invention.
4E illustrates evaluation results of a plurality of models for setting a base model of a depression prediction model used in various embodiments of the present disclosure.

Advantages of the invention and methods of achieving them will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but will be implemented in various forms, and only the present embodiments are intended to complete the disclosure of the present invention, and the general knowledge in the art to which the present invention pertains. It is provided to fully convey the scope of the invention to those skilled in the art, and the present invention is defined only by the scope of the claims.

Shapes, sizes, ratios, angles, numbers, and the like disclosed in the drawings for describing the embodiments of the present invention are exemplary, and the present invention is not limited to the illustrated items. In addition, in describing the present invention, if it is determined that the detailed description of the related known technology may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. When 'comprises', 'haves', 'consists of' and the like mentioned in the present specification are used, other parts may be added unless 'only' is used. In case of singular reference, the plural number includes the plural unless specifically stated otherwise.

In interpreting a component, it is interpreted to include an error range even if there is no separate description.

Each of the features of the various embodiments of the present invention may be combined or combined with each other in part or in whole, various technically interlocking and driving as can be understood by those skilled in the art, each of the embodiments may be implemented independently of each other It may be possible to carry out together in an association.

For clarity of interpretation of the present specification, the terms used herein will be defined below.

As used herein, the term "depression" refers to any disease or disorder that causes a variety of cognitive and psychosomatic symptoms, leading to a decrease in motivation and depression, leading to a decrease in daily functioning. Thus, "depression" disclosed herein may be used in the same sense as "depressive disorder". The depression may be senile depression that may appear in old age of about 65 years or older, depending on the age of the subject, but is not limited thereto.

As used herein, the term “subject” may refer to a subject for predicting whether to develop depression. For example, the subject may be an elderly person over 60 years old, especially a single elderly person living alone without a caregiver, but is not limited thereto and may be any subject who suspects depression or wants to diagnose depression.

As used herein, the term “momentary mood evaluation score” may refer to a measure used to momentarily assess the degree of mood. In this case, the instant mood evaluation score may be an EMA (ecological momentary assessment) score that indicates the score of the self-report within a moment or near a certain experience in daily life for a certain period of time. This, EMA score may be associated with a degree of depression. For example, if the subject has depression, the EMA score may be close to zero, otherwise the EMA score may be close to ten. However, the instant mood assessment scores disclosed herein are not limited to EMA. For example, the instant mood assessment score can be a Hamilton depression scale (HDS) score, or a geriatric depression scale (GDS) score, as assessed by the subject.

As used herein, the term "wearable device" may refer to any electronic device attached to or worn on the body. In this case, the wearable device may include an acceleration sensor, a gyroscope, an illuminance sensor, an RGB sensor, a PPG sensor, an event button, a clock, and the like. Accordingly, the wearable device may measure an activity amount of the subject in real time, and measure an illuminance amount, a red light level, a green light level, and a blue light level. In this case, the term "real time" used in the present specification may mean a preset time interval, for example, a 30 second interval.

Furthermore, the wearable device may be further configured to receive the aforementioned instant mood evaluation score directly from the subject at a predetermined time interval. For example, the wearable device may include six sections (0 to 04 hours, 04 hours to 08 hours, 08 hours to 12 hours, 12 hours to 16 hours, 16 hours to 20 hours, and 20 hours from the subject) at 4 hour intervals. To 24 hours) for each instant mood evaluation score can be input. However, the predetermined time interval is not limited to this.

Meanwhile, the wearable device may be configured to further measure and receive data related to sleep quality configured as wake after sleep onset, sleep time, amount of activity during sleep, and freshness during wake up.

Data associated with the subject measured or directly input from the wearable device may be used for predicting depression. For example, the data described above may be used as an input variable in a depression prediction model configured to predict whether a subject develops depression.

As used herein, the term “depression prediction model” may refer to a prediction model configured to predict whether to develop depression based on data received from a wearable device worn by a subject. More specifically, the depression prediction model is based on data of at least one of an instantaneous mood evaluation score, measured activity amount, illuminance amount, red light level, green light level, and blue light level for each of the depressed and non-depressed individuals. As a result, it may be a model based on an algorithm configured to extract feature data associated with the onset of depression, and to learn it to determine whether the onset of depression. For example, the depression prediction model may include data about a subject received from a wearable device based on a randomized decision forest algorithm, a decision tree algorithm, an adaptive boosting algorithm, a machine learning algorithm of a penalized logistic regression algorithm, or a deep learning algorithm. It may be a model configured to predict whether the onset of depression based on. However, the depression prediction model of the present invention is not limited thereto and may be based on more various learning algorithms.

As used herein, the term “pattern generation model” may refer to a model configured to generate a state pattern based on data received from a wearable device worn by a subject. For example, the pattern generation model may be a model based on k-Means of unsupervised learning, or a clustering algorithm of SOM. However, the pattern generation model of the present invention is not limited thereto and may be based on more various clustering algorithms. On the other hand, the state pattern generated by the pattern generation model may show a significant difference depending on whether the onset of depression, it can be used to predict depression. For example, the state pattern generated by the pattern generation model may be input to the above-described depression prediction model, and the depression prediction model may be trained to predict whether to develop depression based on the state pattern.

Hereinafter, referring to FIGS. 1A and 1B, a device for predicting depression and a system for predicting depression using the same will be described in detail. 1A is a diagram illustrating a system for predicting depression onset using a device for predicting depression according to an embodiment of the present invention. 1B illustrates the configuration of a device for predicting depression onset according to an embodiment of the present invention.

Referring to FIG. 1A, a depression occurrence prediction system 1000 according to an embodiment of the present invention includes a depression occurrence prediction device 100, a wearable device 300, a database 400, and a medical staff device 500. It is.

The depression occurrence prediction device 100 in the depression occurrence prediction system 1000 may include data received from the wearable device 300, for example, an instant mood evaluation score, an activity amount, an illuminance amount, a red light level, a green light level, a blue light level, an awakening. It may be configured to probabilistically predict the risk of depression or the risk of the invention for the subject based on time, time of sleep, activity during sleep, or the degree of freshness during wake up. In addition, the depression occurrence prediction device 100 may be configured to further receive an additional database 400 such as gender, age, and medical record data for the subject from the outside, to predict whether the depression occurs for the subject. Meanwhile, the depression occurrence prediction device 100 may be further configured to predict whether depression occurs according to the exposed light type based on the red light level, the green light level, and the blue light level received from the wearable device 300. In various embodiments, the depression occurrence prediction device 100 may be further configured to generate a state pattern for the subject based on data received from the wearable device 300, and predict whether to develop depression based on the state pattern. .

The wearable device 300 may be configured to receive an instant mood evaluation score from a subject or to measure an activity amount, illuminance amount, red light level, green light level, and blue light level for the subject. Furthermore, it may be configured to transmit the values of the measured data or the entered instant mood evaluation score to the depression occurrence prediction device 100.

The medical staff device 500 receives the information related to whether the depression has occurred for the subject predicted by the depression occurrence prediction device 100, the medical staff monitors the subject based on the received information, and takes action based on the illness. Can be configured to provide feedback.

More specifically, referring to FIG. 1B, the depression occurrence prediction device 100 includes a receiver 110, an inputter 120, a display 130, a storage 140, and a processor 150.

In detail, the receiver 110 may receive an instant mood evaluation score, an activity amount measured for the subject, an illuminance amount, a red light level, a green light level, and a blue light level from the wearable device 300 worn by the subject, and may include a gender and an age. And additional database 400, such as medical record data.

According to an embodiment of the present invention, the receiver 110 may, from the wearable device 300, correlate sleep quality including at least one selected from the group consisting of awakening time, sleep time, active amount during sleep, and refreshment during wake up. It may be configured to further receive data.

The input unit 120 is not limited to a keyboard, a mouse, a touch screen panel, and the like. The input unit 120 may set the depression occurrence prediction device 100 and instruct an operation of the depression occurrence prediction device 100.

Meanwhile, the display unit 130 may display data received by the receiver 110. In addition, the display unit 130 may display a state pattern for the subject generated by the processor 150, and may display whether the depression is predicted by the processor 150.

The storage unit 140 stores data or an additional database 400 for a subject measured or input from the wearable device 300, received through the receiver 110, and sets a depression onset through the input unit 120. Can be configured to store an indication of the prediction device 100. Further, the storage unit 140 is configured to store the state pattern for the subject generated by the processor 150 to be described later, and to store the result of whether or not the depression for the predicted subject. However, without being limited to the above, the storage 140 may store various information determined by the processor 150 for predicting depression occurrence.

The processor 150 may be a component for providing accurate prediction results of the depression occurrence prediction device 110. In this case, for accurate prediction, the processor 150 may be configured to evaluate whether depression occurs in a subject by using a depression prediction model configured to predict whether depression occurs on the basis of data about the subject received from the receiver 110. Can be.

According to one embodiment of the invention, the processor 150 is based on the instant mood evaluation score, activity amount, illuminance amount, red light level, green light level and blue light level for the subject received from the receiver 110 using the depression prediction model. It may be further configured to evaluate whether the onset of depression according to the environment exposed to the subject.

According to another embodiment of the present invention, the processor 150 is exposed to the subject based on the instant mood evaluation score, the red light level, the green light level and the blue light level for the subject received from the receiver 110 using the depression prediction model. It may be further configured to evaluate whether the depression according to the light type.

Meanwhile, the depression prediction device 100 according to an embodiment of the present invention may further include a pattern generator configured to generate a state pattern for the subject based on the data received through the receiver 110. In this case, the pattern generator may be further configured to generate a pattern associated with depression on the basis of data obtained from a sample subject different from the subject, and the processor 150 may use the depression prediction model to determine a pattern and the subject associated with depression onset. It may be further configured to compare the generated state pattern with respect to the stochastic probability of occurrence of depression.

Accordingly, according to the depression onset prediction system 1000, the depression onset prediction device 100 of the present invention is based on the data received from the wearable device 300 and the additional database 400, the depression onset of the subject Presence and risk can be predicted. Furthermore, as the depression occurrence prediction device 100 provides the medical staff device 500 with information on whether the depression occurrence is predicted, the medical staff can recognize the onset of depression for the subject more quickly. That is, according to the depression onset prediction system 1000 based on the depression onset prediction device 100, secondary accidents according to depression onset and depression on the subjects who are out of medical services, such as elderly people living alone without a caregiver, in particular Can be prevented and treatment monitoring can also be facilitated for subjects with depression.

Hereinafter, a method for predicting depression onset according to an embodiment of the present invention will be described in detail with reference to FIG. 2. 2 exemplarily illustrates a procedure of a method for predicting depression onset according to an embodiment of the present invention.

2, the procedure of predicting depression onset according to an embodiment of the present invention is as follows. First, at least one selected from the group consisting of an instant mood evaluation score, an activity amount, an illuminance amount, a red light level, a green light level, and a blue light level is received from a wearable device worn by a subject (S210). In operation S220, a state pattern is generated based on data received from the wearable device using the pattern generation model. Next, using the depression prediction model, it is evaluated whether or not the onset of depression (S230).

According to an embodiment of the present invention, in the receiving step (S210), at least one selected from the group consisting of an instant mood evaluation score, activity amount, illuminance amount, red light level, green light level, and blue light level is determined a plurality of times at a predetermined time interval. can do.

According to another embodiment of the present invention, in the step of receiving (S210), sleep quality related data including at least one selected from the group consisting of wake-up time, sleep time, sleep activity, and refreshment during wake-up measured from the wearable device I can receive more.

According to another embodiment of the present invention, the receiving step (S210) may further receive the age or gender for the subject.

According to an embodiment of the present invention, in the step of generating a state pattern (S220), a state pattern for a subject is generated using a pattern generation model trained to generate a state pattern based on data obtained from the wearable device. Can be. In this case, the pattern generation model may be a model based on a clustering algorithm. For example, in the step of generating a state pattern (S220), the clustering algorithm may be repeatedly performed to group patterns having high similarity among patterns generated by values of data received from the wearable device.

According to one embodiment of the present invention, in the step of evaluating the occurrence of depression (S230), data received from the wearable device using the depression prediction model, for example, instant mood evaluation score, activity amount, illuminance amount, red light The onset of the subject may be assessed based on at least one selected from the group consisting of levels, green light levels and blue light levels.

According to another embodiment of the present invention, in the step of evaluating the occurrence of depression (S230), the instant mood evaluation score, activity amount, illuminance amount, red light level, green light received from the wearable device worn by the subject using the depression prediction model At least one of the level and the blue light level may be used to assess whether depression is caused by the environment exposed to the subject.

According to another embodiment of the present invention, in the step of evaluating the occurrence of depression (S230), the instant mood evaluation score, red light level, green light level and blue light level received from the wearable device worn by the subject using the depression prediction model The onset of depression according to the type of light exposed to the subject may be evaluated based on at least one of the above.

In another embodiment of the present invention, in evaluating whether depression develops (S230), the depression prediction model includes a state pattern for a subject generated in the aforementioned pattern generation step (S220) and a state pattern associated with a predetermined depression. On the basis of this, the development of depression can be assessed. More specifically, after receiving data for each of the sample individuals consisting of different depressive and non-depressive individuals different from the subject, a pattern generation model is used to generate a state pattern for each of the sample individuals, and the generated plurality of state patterns Patterns with significant differences over a predetermined level for heavy depressed individuals and non-depressive individuals may be determined as patterns associated with depression. In another embodiment of the present invention, an evaluation value of at least one of odds ratios, coefficients, and marginal effects for each of a plurality of state patterns generated for each sample individual is calculated, and based on this, a pattern associated with the onset of depression is determined. Can be.

As a result, according to the method of predicting depression onset according to an embodiment of the present invention, whether or not depression onset in the subject can be predicted. Therefore, the medical staff can monitor a plurality of subjects efficiently, and it becomes easier to take measures according to the risk of developing depression for the subjects.

Example 1 Setting Up a Depression Prediction Model

Hereinafter, a process of setting a depression prediction model used in various embodiments of the present invention will be described with reference to FIGS. 3A to 3C. 3A to 3C illustrate evaluation results of a plurality of prediction models according to input variables for setting a depression prediction model used in various embodiments of the present disclosure. More specifically, model 1 is a model using age, gender, activity amount as an input variable, model 2 is a model using age, gender, activity amount, illuminance amount, red light level, green light level and blue light level as input variables, model 3 Is a model that uses age, gender, activity level, illuminance, red light level, green light level, blue light level, and EMA score as input variables. At this time, for the evaluation of each model, data for 19 depression groups and 28 non-depression groups determined as depression based on GDS and HDS assessments are used. Furthermore, as an evaluation criterion, accuracy is the ratio of accurately predicting the depression group to the non-depression group, accuracy is the ratio of the actual depression group to the predicted depression group, and predicted depression is the actual depression group. Recall or sensitivity, which is the ratio for the suspect group, specificity, which is the ratio of the predicted non-depressive group, to the actual non-depressive group, F-score, which is the harmonic mean of precision, and recall, and ROC curves. An AUC value representing the area was set.

Referring to FIG. 3A, Model 3, which is a model using age, gender, activity, illuminance, red light level, green light level, blue light level, and EMA score as input variables, includes accuracy, precision, recall, specificity, F-score, and All assessments of AUC values show higher levels than other models. More specifically, referring to FIG. 3B, Model 1, Model 2, and Model 3 appear to have a high level of prediction for the actual 28 non-depressive groups. However, when referring to the depression group, the model 1 shows a low prediction rate for the depression group consisting of 19 people and the model 2 predicts 9 people as the depression group. In contrast, model 3 predicts 13 people as the depression group, which appears to have a higher predictive rate for depression group prediction than other models. Furthermore, referring to FIG. 3C, the AUC values for the ROC curves for each model are shown. In this case, the closer the AUC value is to 1, the higher the diagnostic ability of the depression model. The AUC value of model 1, which is a model using age, gender, and activity as an input variable, is 0.69, and the AUC value of model 2, which further uses illuminance, red light level, green light level, and blue light as input variables, is 0.76. As the AUC value of Model 3 configured to further use the EMA score as an input variable is 0.84, which is higher than the rest of the model, Model 3 may be a model having high predictive ability in diagnosing depression prediction.

According to the results of Example 1, the depression prediction model used in various embodiments of the present invention, Model 3 using age, gender, activity amount, illuminance, red light level, green light level, blue light level, EMA score as input variables Can be set. However, the depression prediction model of the present invention is not limited thereto, and may be configured to predict the risk of depression using various combinations of the data disclosed herein as input variables.

Example 2: Evaluation of Depression Prediction Model

Hereinafter, an evaluation result of a depression prediction model used in various embodiments of the present invention will be described with reference to FIGS. 4A to 4E. 4A to 4D illustrate evaluation results of a depression prediction model used in various embodiments of the present invention. 4E illustrates evaluation results of a plurality of models for setting a base model of a depression prediction model used in various embodiments of the present disclosure. At this time, the depression prediction model is based on the model 3, which shows a high level in the prediction of depression in Example 1 described above, using the illumination intensity, red light level, green light level, blue light level and EMA scores excluding sex and age as input variables. A predictive model was used. Furthermore, accuracy, precision, recall, specificity, AUC value and F-score were set as evaluation criteria.

More specifically, in this evaluation, as shown in Table 1 below, the learning data used for learning the depression prediction model and the test data used for the evaluation thereof are divided and used.

More specifically, of the 47 sample subjects, data for 19 non-depressive groups and 11 depression groups were set as training data, and data for 9 non-depressive groups and 8 depression groups were evaluated. Was set as test data. In this evaluation, data segmentation into training data and test data was performed by a random sampling method, and the average prediction accuracy level was calculated after performing 200 predictions by repeating the random sampling process about 200 times. At this time, the training data was used for coefficient estimation and the test data was directly used for predictive evaluation.

Referring to (a) and (b) of FIG. 4A, the cumulative and average result values according to the depression prediction model in which 200 predictions are performed as described above are illustrated.

More specifically, evaluation results based on cumulative result values are shown in Table 2 below.

For the predictive model of depression, the accuracy of predicting depression is 0.7050 ((1546 + 851) / (1546 + 254 + 749 + 851)), and the precision is 0.7701 ((851) / (245 + 851)) and sensitivity The recall means 0.5319 ((851) / (749 + 851)), the specificity is 0.8589 ((1546) / (1546 + 254)), and the F-score and AUC are 0.6292 and 0.7540, respectively. As such, the depression prediction model used in various embodiments of the present invention appears to have a high level for the above evaluation criteria. In particular, a depression prediction model with a high precision of 0.7701 and a high AUC value of 0.7540 may be a suitable model for predicting the onset of depression.

Referring to (a), (b), (c), (d), (e) and (f) of FIG. 4B, each of the evaluation criteria for accuracy, precision, recall, specificity, AUC value, and F-score Results are shown showing the frequency of the estimated result value for each of the 200 iterative predictions calculated for. More specifically, the accuracy in 200 predictions appears to have a constant value in the range of about 0.7 to 0.8, and the AUC value also appears to have a constant value in the range of about 0.7 to 0.8.

Referring to (a), (b), (c), (d), (e) and (f) of FIG. 4C, each of the evaluation criteria for accuracy, precision, recall, specificity, AUC value, and F-score The average level for the cumulative value of the result by 200 iterations, calculated for the plot, is shown. More specifically, the average level calculated for each of the evaluation criteria of accuracy, precision, recall, specificity, AUC value, and F-score is shown at a level similar to that of Table 2 above.

Referring to (a), (b), (c), (d), (e) and (f) of FIG. 4D, the number of prediction iterations is 1, 5, 10, 50, 100 and 150 times. AUC values are shown as increases. More specifically, as the number of predictions by the prediction model approaches 200, the AUC value appears to approach 0.70, and the accuracy of the prediction appears to be formed at the level of 70%.

Referring to FIG. 4E, an evaluation result according to algorithm types of Logit Model, Decision Tree, Boosting, and Random Forest for the depression prediction model is shown. More specifically, when using the Logit Model, the AUC value is 0.7540, the accuracy is 0.7050, the precision is 0.7701, and the specificity is 0.8589. That is, the depression prediction model of the present invention can evaluate the onset of depression at a high prediction rate based on the Logit Model. However, the prediction model used in various embodiments of the present invention without being limited to the Logit Model can predict whether the onset of depression is based on more various algorithms.

As a result of Example 2, the method and device for predicting depression onset according to an embodiment of the present invention using a prediction model for predicting the occurrence of depression with high accuracy and precision, may provide information on the diagnosis of depression of the subject It can be effective. The present invention can prevent the onset of depression and secondary accidents following the onset of depression, particularly for subjects who are distant from medical services, such as elderly people living alone without a caregiver, and it will also be easy to monitor treatment for those with depression. Can be.

In particular, the present invention may predict whether the onset of depression according to the exposed environment and the type of exposed light based on the illuminance, activity, red light level, green light level and blue light level measured from the wearable device worn by the test subject.

Meanwhile, the method for predicting depression and the device using the same according to various embodiments of the present disclosure may predict whether the depression occurs using a pattern generation model. In particular, the present invention can predict the occurrence of depression on the basis of the state pattern associated with depression, and even if there is missing data due to malfunction or non-wear of the wearable device, it is possible to accurately predict the occurrence of depression with high accuracy and accuracy. .

Although the embodiments of the present invention have been described in more detail with reference to the accompanying drawings, the present invention is not necessarily limited to these embodiments, and various modifications can be made without departing from the spirit of the present invention. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention but to describe the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not restrictive. The protection scope of the present invention should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of the present invention.

100: depression onset prediction device
110: receiver
120: input unit
130: display unit
140: storage unit
150: processor
300: measurement device
400: database
500: medical staff terminal
1000: Depression Prediction System

Claims (19)

In the method of predicting depression onset implemented by a processor,
Receiving an instant mood evaluation score for the subject from a wearable device worn by the subject, and
And evaluating depression occurrence for the subject using a depression prediction model configured to predict depression occurrence based on the instantaneous mood evaluation score for the subject. The method of claim 1,
The receiving step,
And receiving at least one selected from the group consisting of activity, illuminance, red light level, green light level and blue light level measured in real time from the wearable device,
The depression prediction model,
Depressive onset, further configured to predict whether to develop depression according to the environment exposed to the subject based on at least one selected from the group consisting of the amount of activity, illuminance, red light level, green light level and blue light level and the instant mood assessment score Forecast method. The method of claim 2,
The receiving step,
Receiving at least one selected from the group consisting of a red light level, a green light level, and a blue light level measured in real time from the wearable device,
The depression prediction model,
And predicting the occurrence of depression according to the type of light exposed to the subject based on at least one selected from the group consisting of the red light level, the green light level, and the blue light level and the instant mood evaluation score. The method of claim 1,
The receiving step,
Receiving the instant mood evaluation scores a plurality of times at predetermined time intervals,
Evaluating whether the depression occurs,
Using the depression prediction model, evaluating whether or not depression occurs in the subject at the predetermined time interval. The method of claim 1,
The receiving step,
Further comprising receiving sleep quality related data comprising at least one selected from the group consisting of wake after sleep onset, sleep time, amount of activity during sleep, and wakefulness during wake up,
The depression prediction model,
And predicting the onset of depression for the subject based on the instantaneous mood evaluation score and the sleep quality related data. The method of claim 1,
The depression prediction model,
Prior to the receiving step,
Receiving an instant mood assessment score for each of the depressive and non-depressive individuals different from the subject;
Calculating a difference between the instantaneous mood evaluation scores for the depressed individual and the non-depressive individual; and
And a model trained by predicting depression onset for the subjects different from the depressed individuals and the non-depressive individuals based on the calculated difference in the instantaneous mood evaluation score. The method of claim 1,
Evaluating whether the depression occurs,
Generating a state pattern for the subject using a pattern generation model configured to generate a state pattern for the subject based on the instant mood evaluation score; and
And evaluating whether the subject develops depression based on the state pattern using the depression prediction model. The method of claim 7, wherein
The depression prediction model,
Prior to said receiving, receiving an instantaneous mood assessment score for each of the sample individuals consisting of depressive individuals and non-depressive individuals different from said subject;
Generating a state pattern based on an instantaneous mood evaluation score for each of the sample individuals using the pattern generation model;
Determining, among the plurality of state patterns generated for each of the sample subjects, a pattern having a significant difference over a predetermined level for the depressive subject and the non-depressive subject as a pattern associated with the onset of depression, and
A method for predicting depression onset based on a pattern associated with the onset of depression, a model trained by predicting depression onset in a subject different from the depressed and non-depressive individuals. The method of claim 8,
Determining the pattern associated with the development of depression,
Calculating an evaluation value of at least one of odds ratio, coefficients, and marginal effects for each of the plurality of state patterns generated for each of the sample entities; and
Determining a pattern associated with depression on the basis of the calculated value. The method of claim 1,
The receiving step,
Further receiving an age and a gender for the subject,
And the depression is senile depression. A device configured to predict whether an outbreak of depression implemented by a processor,
A receiver configured to receive an instant mood evaluation score for the subject from a wearable device worn by the subject;
A processor coupled to communicate with the receiver,
A processor configured to evaluate depression onset for the subject using a depression prediction model configured to predict depression onset based on an instantaneous mood assessment score, and
And a storage configured to store the instant mood evaluation score and the depression prediction model. The method of claim 11,
The receiving unit,
Is configured to further receive at least one selected from the group consisting of activity amount, illuminance, red light level, green light level and blue light level measured in real time from the wearable device,
The processor,
A depression occurrence prediction device, further configured to evaluate whether depression develops according to an environment exposed to the subject using the depression prediction model The method of claim 12,
The receiving unit,
Is configured to further receive at least one selected from the group consisting of a red light level, a green light level and a blue light level measured in real time from the wearable device,
The processor,
And predicting the onset of depression according to the type of light exposed to the subject by using the depression prediction model. The method of claim 11,
Further comprising a data preprocessor,
And the data preprocessor is configured to correct a score having an error value among the received instant mood evaluation scores. The method of claim 11,
And a pattern generator configured to generate a state pattern for the subject based on the instantaneous mood evaluation score. The method of claim 15,
The pattern generation unit,
Further generate a pattern associated with the onset of depression based on an instantaneous mood assessment score from a sample subject different from the subject,
The processor,
Using the depression prediction model, comparing the pattern associated with the onset of the depression with the state pattern generated for the subject, and probably calculating whether or not the onset of depression for the subject is based on a result of the comparison, Depression onset prediction device. The method of claim 11,
The receiving unit,
Is configured to further receive sleep quality related data comprising at least one selected from the group consisting of awakening time, sleep time, sleep activity and wakefulness for the subject,
The processor,
And predict the onset of depression for the subject based on the instant mood evaluation score and the sleep quality related data using the depression prediction model. The method of claim 11,
The receiving unit,
Is configured to receive the instant mood assessment score a plurality of times at a predetermined time interval,
The processor,
And predict the onset of depression for the subject at the predetermined time interval using the depression prediction model. The method of claim 11,
The receiving unit,
Further receive an age and a gender for the subject,
And the depression is senile depression.

Images