Apparatus and method for generating symptom occurrence prediction information

Abstract

The present specification discloses an apparatus and method for generating symptom occurrence prediction information. The apparatus for generating symptom occurrence prediction information, according to the present specification, may include: a transmitting and receiving unit that receives, from a digital device, demographic information, psychological test scales, biometric information, and daily information of a patient and transmits, to the digital device, prediction information and feedback information about whether a panic symptom has occurred; a pattern information generation unit for generating information related to a pattern change of the biometric information (hereinafter, referred to as "pattern information"); a prediction information generation unit for generating the prediction information by using the demographic information, the psychological test scales, the pattern information, and the daily information of the patient; and a feedback generation unit for generating the feedback information according to the prediction information.

Description

Device and method for generating symptom occurrence prediction information

[0001] The present invention relates to a device and method for generating symptom occurrence prediction information.

[0002] This application claims priority to Korean Patent Application No. 10-2024-0018297, filed on February 6, 2024, the entire disclosure of which is incorporated herein by reference.

[0003] The material described in this section merely provides background information on the embodiments described herein and does not necessarily constitute prior art.

[0004] Panic attacks, characterized by sudden, intense fear and anxiety, are a common psychiatric condition experienced by modern people. Patients experience intense anxiety, agitation, and a sense of impending doom, along with a variety of physical symptoms, including a racing heart, shortness of breath, chest pain, dizziness, numbness in the hands and feet, and fever. One of the core symptoms of panic attacks, which can interfere with patients' daily lives, is anticipatory anxiety. Patients experience anxiety about the potential for a panic attack in their daily lives, which can exacerbate the symptoms of panic attacks.

[0005] Medication and psychotherapy are used to treat panic attacks. Medication carries the burden of long-term medication, and psychotherapy must be combined with medication to gradually taper off medication.

[0006] Psychotherapy utilizes cognitive behavioral therapy (CBT), which consists of six stages: assessment or psychological assessment, reconceptualization, skills acquisition, skills consolidation and application training, generalization and maintenance, and post-treatment assessment follow-up. However, CBT requires patients to recall or imagine past panic attacks in the absence of panic attacks, and has limitations such as its inability to objectify symptoms, identify their causes, and observe and regulate physical reactions in real time.

[0007] Therefore, there is a need for a technology that can objectively analyze the symptoms of panic attacks, identify the cause of the symptoms, and predict the occurrence of panic attacks in advance to help treat patients.

[0008] The purpose of this specification is to provide a device and method for generating symptom occurrence prediction information.

[0009] This specification is not limited to the above-mentioned tasks, and other tasks not mentioned will be clearly understood by those skilled in the art from the description below.

[0010] A device for generating symptom occurrence prediction information according to the present specification for solving the above-described problem may include a transceiver unit for receiving demographic information, a psychological test scale, biometric information, and daily life information of a patient from a digital device, and transmitting prediction information and feedback information on whether a panic symptom occurs to the digital device; a pattern information generation unit for generating information related to a pattern change of the biometric information (hereinafter referred to as “pattern information”); a prediction information generation unit for generating the prediction information using the demographic information, psychological test scale, pattern information, and daily life information of the patient; and a feedback generation unit for generating the feedback information according to the prediction information.

[0011] According to one embodiment of the present specification, the biometric information may include at least one of the patient's heart rate, number of steps, and sleep-related information.

[0012] According to one embodiment of the present specification, the pattern information generation unit can generate the pattern information by extracting at least one of the time when the circadian rhythm of the biometric information is at its highest point, the difference between the highest point and the lowest point, and the average value.

[0013] At this time, when the biometric information is related to heart rate, the pattern information generation unit can extract multiple frequency areas from the circadian rhythm of the heart rate and further generate information related to the intensity of each frequency area.

[0014] According to one embodiment of the present specification, the daily information may include at least one of the following: time of occurrence of panic symptoms, mood changes, energy changes, anxiety level, stress level, alcohol intake, caffeine intake, smoking amount, exercise amount, medication dosage, and information related to a woman's menstruation.

[0015] According to one embodiment of the present specification, the prediction information generation unit can generate the prediction information in steps, and the feedback generation unit can generate the feedback information in steps according to the prediction information.

[0016] The symptom occurrence prediction information generation device according to the present specification may further include a learning unit that uses at least one of the demographic information, psychological test scale, pattern information, and daily life information to teach the prediction information generation unit content related to the occurrence of panic symptoms.

[0017] At this time, the daily life information includes information about the time when panic symptoms occurred, and the learning unit can learn information related to the occurrence of panic symptoms by using the demographic information, psychological test scale, pattern information, and daily life information entered during a preset period according to the time when panic symptoms occurred.

[0018] According to one embodiment of the present specification, the learning unit can extract a plurality of elements having a relatively high contribution in predicting the occurrence of panic symptoms of the patient, and set the weights of nodes into which elements other than the plurality of elements are input to preset values.

[0019] The symptom occurrence prediction information generation device according to the present specification may be a component of a symptom occurrence prediction information generation system including a wearable device that measures a patient's bio-signal to generate bio-information or generates information on the time when panic symptoms occur; and a digital device that is linked to the wearable device and receives demographic information, psychological test scales, and daily life information from the patient.

[0020] A method for generating symptom occurrence prediction information according to the present specification for solving the above-described problem may include a patient information receiving step in which a processor receives a patient's demographic information, a psychological test scale, biometric information, and daily life information from a digital device; a pattern information generating step in which the processor generates information related to a change in a pattern of the biometric information (hereinafter referred to as 'pattern information'); a prediction information generating step in which the processor generates prediction information on whether a panic symptom occurs using the patient's demographic information, psychological test scale, pattern information, and daily life information; a feedback information generating step in which the processor generates feedback information according to the prediction information; and a prediction information transmitting step in which the processor transmits the prediction information and feedback information to the digital device.

[0021] According to one embodiment of the present specification, the biometric information may include at least one of the patient's heart rate, number of steps, and sleep-related information.

[0022] According to one embodiment of the present specification, the pattern information generating step may be a step in which the processor generates the pattern information by extracting at least one of the time when the circadian rhythm of the biometric information is at its highest point, the difference between the highest point and the lowest point, and the average value.

[0023] At this time, when the biometric information is related to heart rate, the pattern information generation step may further include a step in which the processor extracts multiple frequency domains from the circadian rhythm of the heart rate and generates content related to the intensity of each frequency domain.

[0024] According to one embodiment of the present specification, the daily information may include at least one of the following: time of occurrence of panic symptoms, mood changes, energy changes, anxiety level, stress level, alcohol intake, caffeine intake, smoking amount, exercise amount, medication dosage, and information related to a woman's menstruation.

[0025] According to one embodiment of the present specification, the prediction information generating step may be a step in which the processor generates the prediction information step by step, and the feedback information generating step may be a step in which the processor generates the feedback information step by step according to the prediction information.

[0026] The method for generating symptom occurrence prediction information according to the present specification may further include a learning step in which the processor learns content related to the occurrence of panic symptoms by using at least one of the demographic information, the psychological test scale, the pattern information, and the daily life information.

[0027] According to one embodiment of the present specification, the daily life information includes information about the time when panic symptoms occurred, and the learning step may be a step in which the processor learns information related to the occurrence of panic symptoms by using the demographic information, psychological test scale, pattern information, and daily life information input during a preset period according to the time when the panic symptoms occurred.

[0028] According to one embodiment of the present specification, the learning step may be a step in which the processor extracts a plurality of elements having a relatively high contribution in predicting the occurrence of panic symptoms in the patient, and sets the weights of nodes into which elements other than the plurality of elements are input to preset values.

[0029] The method for generating symptom occurrence prediction information according to the present specification can be implemented in the form of a computer program written to perform each step of the method for generating symptom occurrence prediction information on a computer and recorded on a computer-readable recording medium.

[0030] Other specific details of the present invention are included in the detailed description and drawings.

[0031] According to one aspect of the present specification, the onset of panic symptoms can be predicted in advance using the patient's biometric information and daily life information.

[0032] According to another aspect of the present disclosure, it is possible to help patients manage their panic symptoms on their own by informing them in advance of the onset of panic symptoms.

[0033] According to another aspect of the present specification, an artificial intelligence model for predicting the occurrence of panic symptoms is trained using a patient's biometric information and daily life information to create a personalized panic symptom prediction model, thereby providing personalized feedback.

[0034] According to another aspect of the present specification, a model for predicting the occurrence of panic symptoms can be made lighter by extracting multiple factors that have a high contribution to the patient's panic symptoms.

[0035] The effects of the present invention are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

[0036] FIG. 1 is a block diagram of a symptom occurrence prediction information generation system according to one embodiment of the present specification.

[0037] Figure 2 illustrates an example of elements of pattern information generated by a pattern information generation unit.

[0038] Figure 3 illustrates an example of patient demographic information and elements of a psychological test scale.

[0039] Figure 4 illustrates an example of elements of a patient's daily life information.

[0040] FIG. 5 is a block diagram of a symptom occurrence prediction information generation device according to another embodiment of the present specification.

[0041] FIG. 6 is an example image showing the prediction performance of a personalized neural network model using an XGBoost classifier according to one embodiment of the present specification.

[0042] Figure 7 illustrates an example of multiple extracted elements.

[0043] Figure 8 is an example image comparing the performance of neural network models trained using different amounts of data.

[0044] FIG. 9 is a flowchart of a method for generating symptom occurrence prediction information according to one embodiment of the present specification.

[0045] Fig. 10 is a flowchart of a method for generating symptom occurrence prediction information according to another embodiment of the present specification.

[0046] FIG. 11 is a block diagram of a symptom occurrence prediction information generation system according to one embodiment of the present specification.

[0047] Fig. 12 is a flowchart of a method for extracting symptom occurrence elements according to one embodiment of the present specification.

[0048] Fig. 13 is a flowchart of a method for generating symptom occurrence prediction information according to one embodiment of the present specification.

[0049] The advantages and features of the invention disclosed in this specification, and the methods for achieving them, will become clearer with reference to the embodiments described in detail below together with the accompanying drawings. However, this specification is not limited to the embodiments disclosed below and may be implemented in various different forms. These embodiments are provided only to ensure that the disclosure of this specification is complete and to fully inform those of ordinary skill in the art (hereinafter referred to as "skilled workers") of the scope of this specification, and the scope of rights of this specification is defined only by the scope of the claims.

[0050] The terminology used herein is for the purpose of describing embodiments and is not intended to limit the scope of the present disclosure. In this specification, singular forms also include plural forms, unless specifically stated otherwise. As used herein, the terms "comprises" and / or "comprising" do not exclude the presence or addition of one or more other components in addition to the components mentioned.

[0051] Throughout the specification, the same reference numerals refer to the same elements, and the term "and / or" includes each and every combination of the elements mentioned. Although terms such as "first," "second," etc. are used to describe various elements, these elements are not limited by these terms. These terms are used only to distinguish one element from another. Therefore, it should be understood that a first element mentioned below may also be a second element within the technical scope of the present invention.

[0052] Unless otherwise defined, all terms (including technical and scientific terms) used herein may be used in their common sense to those skilled in the art to which this specification pertains. Furthermore, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise.

[0053] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

[0054] FIG. 1 is a block diagram of a symptom occurrence prediction information generation system according to one embodiment of the present specification.

[0055] Referring to FIG. 1, a symptom occurrence prediction information generation system (1) according to one embodiment of the present specification may include a symptom occurrence prediction information generation device (10), a wearable device (20), and a digital device (30).

[0056] The above symptom occurrence prediction information generation device (10) may include a transceiver (100), a pattern information generation unit (110), a prediction information generation unit (120), and a feedback generation unit (130).

[0057] The transceiver (100) can receive demographic information, psychological test scales, biometric information, and daily life information of the patient from the digital device (30). The biometric information may include at least one of the patient's heart rate, number of steps, and sleep-related information. The biometric information may be measured using the wearable device (20) worn by the patient. The wearable device (20) may correspond to a smart watch including a display. This is an example and is not limited to a specific device.

[0058] The wearable device (20) may be equipped with an application for measuring biometric information. The wearable device (20) may measure heart rate using photoplethysmography technology. In addition, the wearable device (20) may measure the number of steps using an accelerometer and / or an inertial measurement unit (IMU) sensor. In addition, the wearable device (20) may extract information related to sleep using the measured heart rate and / or number of steps. This is merely an example, and the method for measuring heart rate, number of steps, and information related to sleep is not limited to a specific method.

[0059] The above-mentioned information related to heart rate, number of steps, and sleep may be measured through an application installed on the wearable device (20) and / or digital device (30) and transmitted to the transceiver (100). The above-mentioned information related to heart rate, number of steps, and sleep may be measured in real time and transmitted to the transceiver (100).

[0060] The wearable device (20) can transmit the biometric information to the digital device (30). At this time, the digital device (30) may correspond to a smartphone and / or tablet computer that is linked to the wearable device (20), which is an example and is not limited to a specific device.

[0061] The wearable device (20) and the digital device (30) can be paired and synchronized via Bluetooth. This is an example, and any communication technology capable of synchronizing the wearable device (20) and the digital device (30) can be used.

[0062] The above-mentioned transceiver (100) can be connected to a network configured to transmit and receive data with various communication terminals, such as between wearable devices, between wearable devices and digital devices, between digital devices, between digital devices and server devices, and between server devices. The network can be a wired or wireless communication network and is not limited by a specific communication protocol.

[0063] The above pattern information generation unit (110) can generate information (hereinafter referred to as "pattern information") related to changes in the pattern of the biometric information. The biometric information can have a specific pattern that repeats in a 24-hour cycle in a circadian rhythm. The pattern information generation unit (110) can extract at least one of the time when the circadian rhythm of the biometric information is at its highest point (Acrophase), the difference between the highest point and the lowest point (Amplitude), and / or the average value.

[0064] For example, when the biometric information is heart rate, the time of the peak point may refer to the time when the patient's heart rate is the highest in the circadian rhythm. The difference between the peak point and the lowest point may refer to the difference between the highest heart rate and the lowest heart rate in the circadian rhythm. The mean value may refer to the mean value (Mesor) of the heart rate in the circadian rhythm and / or the mean value (Mean) of the heart rate measured at a preset time interval during the day. The mean value of the heart rate in the circadian rhythm may refer to the median value in the circadian rhythm. The mean value of the heart rate measured at a preset time interval during the day may refer to the arithmetic mean of the heart rates.

[0065] The above pattern information generation unit (110) can generate pattern information from the circadian rhythm of the biometric information using Cosinor analysis. This is an example and is not limited by the above analysis method.

[0066] In addition, when the biometric information is related to heart rate, the pattern information generation unit (110) can extract multiple frequency domains from the circadian rhythm of the heart rate. The multiple frequency domains may correspond to frequency domains of 0.001-0.0005 Hz, 0.0005-0.0001 Hz, 0.0001-0.00005 Hz, and / or 0.00005-0.00001 Hz. The pattern information generation unit (110) can further generate content related to the intensity of each frequency domain of the heart rate. The pattern information generation unit (110) can convert the measured heart rate information into a frequency domain through Fourier transform. Thereafter, the pattern information generation unit (110) can calculate a bandpower value of each frequency domain using a multi-taper method. The method by which the above pattern information generation unit (110) generates content related to the intensity of each frequency range is an example and is not limited by the above method.

[0067] The pattern information generation unit (110) may interpolate missing values ​​using an interpolation method if biometric information is not input during a specific time period. In this case, the pattern information generation unit (110) may interpolate missing values ​​using linear interpolation and / or spline interpolation. This is merely an example and is not limited to a specific method.

[0068] Figure 2 illustrates an example of elements of pattern information generated by a pattern information generation unit.

[0069] Referring to FIG. 2, the pattern information generation unit (110) can generate pattern information from information related to heart rate, number of steps, and sleep. Pattern information related to heart rate may include elements such as 'HR_variance', 'HR_maximum', 'HR_mean', 'HR_hvar_mean', 'Bandpower(0.001-0.0005Hz)', 'Bandpower(0.0005-0.0001Hz)', 'Bandpower(0.0001-0.00005Hz)', 'Bandpower(0.00005-0.00001Hz)', 'HR_acrophase', 'HR_amplitude', 'HR_mesor', 'HR_acrophase_difference', 'HR_acrophase_difference_2d', 'HR_amplitude_difference', 'HR_amplitude_difference_2d', 'HR_mesor_difference' and / or 'HR_mesor_difference_2d'. Pattern information related to step count may include elements such as 'Steps_total', 'Steps_maximum', 'Steps_mean', 'Steps_variance', and / or 'Steps_hvar_mean'. Pattern information related to sleep may include elements such as 'Sleep_onset_time', 'Sleep_out_time', and / or 'Sleep_duration'. A detailed description of each element is provided in Figure 2, so a detailed description is omitted.

[0070] The above prediction information generation unit (120) can generate prediction information on whether panic symptoms occur by using the patient's demographic information, psychological test scale, pattern information, and daily life information.

[0071] Figure 3 illustrates an example of patient demographic information and elements of a psychological test scale.

[0072] Referring to (a) of Figure 3, the patient's demographic information may include factors such as age, gender, marital status, occupation, smoking history, drinking history, and / or suicide attempt history.

[0073] Referring to Figure 3 (b), the patient's psychological test scales include the Seasonal Pattern Assessment Questionnaire (SPAQ) score, the Center for Epidemiological Studies-Depression Scale (CES_D) score, the Generalized Anxiety Disorder-7 (GAD_7) score, the State-Trait Anxiety Inventory (STAI_X2) score, the Korean Occupational Stress Scale-Short Form (KOSSSF) score, the Patient Health Questionnaire_9 (PHQ_9) score, the Social Anxiety Distress Scale (SADS) score, the State-Trait Anxiety Inventory (STAI_X1) score, the Biological Rhythms Interview of Assessment in Neuropsychiatry (BRIAN) score, the Composite Scale of Morningness (CSM) score, Childhood Trauma Questionnaire (CTQ) score, Attention Questionnaire Scale (AQS) score, Albany Panic & Phobia Questionnaire (APPQ) score, Korean Resilience Quotient (KRQ) score, Body Sensation Questionnaire (BSQ) score, Mood Disorder Questionnaire (MDQ) score, and / or Brief-Fear of Negative Evaluation (BFE) scaleThis may include factors such as BFNE scores,

[0074] Patients can enter demographic information through the application of the wearable device (20) and / or digital device (30). Additionally, the patient can conduct a self-administered psychological test through the application. The results of the psychological test can be used as a measure of the psychological test.

[0075] Figure 4 illustrates an example of elements of a patient's daily life information.

[0076] Referring to Figure 4, the patient's daily life information may include factors such as the time of panic symptom onset, level of positive emotion, level of negative emotion, level of positive energy, level of negative energy, level of anxiety, level of irritability, amount of alcohol intake, amount of caffeine intake, amount of smoking, amount of exercise, whether or not suicidal thoughts have occurred within the past month, and amount of psychiatric medication taken within the past month. For women, the daily life information may further include factors such as whether or not they had their period that day and / or when their previous menstrual period occurred.

[0077] The patient can input the daily life information into the application installed on the wearable device (20) and / or the digital device (30). The input daily life information can be digitized according to the algorithm of the application. In addition, the screen of the application may have a virtual button that generates information about the time when the panic symptom occurred when the panic symptom occurred. Alternatively, a widget button that generates information about the time when the panic symptom occurred may be present on the home screen of the wearable device (20) and / or the digital device (30). Alternatively, a physical button that generates information about the time when the panic symptom occurred may be present on a portion of the wearable device (20) and / or the digital device (30). The patient can generate information about the time when the panic symptom occurred by using one of the virtual button, the widget button, and the physical button.

[0078] The elements described in FIGS. 2 and 4 are merely examples and are not limited thereto, and various embodiments may be created. For example, the biometric information may further include elements related to electrocardiogram, blood pressure, and / or respiration rate measured using a wearable device (20). The demographic information may further include elements such as religious affiliation, education level, and / or income level. The daily information may further include information related to eating patterns.

[0079] The above prediction information generation unit (120) may be input with a neural network model for generating the prediction information in advance. The prediction information generation unit (120) may input the patient's demographic information, psychological test scale, pattern information, and daily life information into the neural network model to generate the prediction information. The neural network model may correspond to a neural network model trained to predict the occurrence of panic symptoms in advance by using the demographic information, psychological test scale, pattern information, and daily life information of a plurality of people with panic symptoms. The neural network model may be a neural network model trained by a learning technique included in machine learning and / or deep learning technology, and is not limited by a specific learning technique.

[0080] For example, the neural network model may be a neural network model that uses one or more of a multiclass classification model, such as an XGBoost classifier, a RandomForest classifier, and a GradientBoost classifier. This is an example and is not limited to a specific model.

[0081] The above-mentioned predictive information generation unit (120) can extract, modify, or remove outliers from the demographic information, psychological test scales, pattern information, and daily life information. Thereafter, the predictive information generation unit (120) can normalize the information from which outliers have been removed and input it into the neural network model.

[0082] The above-mentioned predictive information generation unit (120) can generate the predictive information for a predetermined time. For example, the predictive information generation unit (120) can generate predictive information regarding the occurrence of panic symptoms the following day. Alternatively, the predictive information generation unit (120) can generate predictive information regarding the occurrence of panic symptoms in the afternoon and / or evening of the current day. This is merely an example and is not limited to a specific time.

[0083] Additionally, the prediction information generation unit (120) may generate the prediction information for a predetermined time period afterward. For example, the prediction information generation unit (120) may generate prediction information regarding the occurrence of panic symptoms 24 hours from the current time. Alternatively, the prediction information generation unit (120) may generate prediction information regarding the occurrence of panic symptoms 8 hours from the current time. This is merely an example and is not limited to a specific time period.

[0084] The feedback generation unit (130) may generate feedback information according to the prediction information. When the prediction information predicts that panic symptoms will not occur the next day, the feedback generation unit (130) may generate feedback information having the content of a daily greeting message, such as, "It is predicted that panic symptoms will not occur tomorrow. Have a nice day." When the prediction information predicts that panic symptoms will occur the next day, the feedback generation unit (130) may generate feedback information including, for example, "It is predicted that panic symptoms will occur tomorrow. Get enough rest. Take deep breaths when panic symptoms occur. Take your prescribed medication if necessary." This is an example and is not limited to the content of the feedback information.

[0085] In addition, the prediction information generation unit (120) may generate the prediction information in stages. For example, the prediction information generation unit (120) may generate the prediction information into four levels of 'normal', 'concern', 'caution', and 'danger' according to the possibility of occurrence of panic symptoms. The 'normal' level may have a relatively lowest possibility of occurrence of panic symptoms. The 'concern' level may have a relatively higher possibility of occurrence of panic symptoms than the 'normal' level. The 'caution' level may have a relatively higher possibility of occurrence of panic symptoms than the 'concern' level. The 'danger' level may have a relatively highest possibility of occurrence of panic symptoms.

[0086] The above feedback generation unit (130) can generate feedback information step by step according to the prediction information. The feedback information may include content regarding the four levels. When the prediction information is at the 'normal' level, the feedback information may include content such as 'panic symptoms are not expected to occur tomorrow.' When the prediction information is at the 'concern' level, the feedback information may include content regarding the level of panic symptom occurrence, such as 'the possibility of panic symptoms occurring tomorrow is at the 'concern' level.' When the prediction information is at the 'caution' and 'danger' levels, it is the same as the 'concern' level, so repeated explanation is omitted.

[0087] Additionally, when the predicted information is at a 'normal' level, the feedback information may further include everyday greetings, such as 'Have a comfortable day.' When the predicted information is at a 'concern' level, the feedback information may further include behavioral guidelines to prevent the occurrence of panic symptoms, such as 'Please get enough rest.' When the predicted information is at a 'caution' and / or 'danger' level, the feedback information may further include behavioral guidelines to prevent the occurrence of panic symptoms, such as 'Please get enough rest. Take deep breaths when panic symptoms occur. Take prescribed medication if necessary.', and content on how to deal with panic symptoms when they occur.

[0088] The stepwise generation of the above prediction information and feedback information according to the possibility of occurrence of the panic symptoms described above is an example and is not limited to the above four levels and feedback information.

[0089] The above-mentioned transceiver (100) can transmit the prediction information and feedback information to the digital device (30).

[0090] FIG. 5 is a block diagram of a symptom occurrence prediction information generation device according to another embodiment of the present specification.

[0091] Referring to FIG. 5, a symptom occurrence prediction information generation device (10') according to another embodiment of the present specification may include a transceiver (100), a pattern information generation unit (110), a prediction information generation unit (120), a feedback generation unit (130), and a learning unit (140). Since the transceiver (100), the pattern information generation unit (110), the prediction information generation unit (120), and the feedback generation unit (130) have been described above, a repetitive description thereof will be omitted.

[0092] The above learning unit (140) can teach the neural network model stored in the prediction information generation unit (120) information related to the occurrence of panic symptoms in the patient by using at least one of demographic information, psychological test scale, pattern information, and daily life information of a specific patient.

[0093] The neural network model may be a generalized neural network model trained using information on multiple people with panic symptoms. The learning unit (140) may train the generalized neural network model using at least one of demographic information, psychological test scales, pattern information, and daily life information of a specific patient. The learning unit (140) may train the generalized neural network model after normalizing the information. Through this, the generalized neural network model may be adjusted to a neural network model personalized for a specific patient. The personalized model may refer to a model in which the generalized neural network model is adjusted to suit the individual characteristics of the patient. In this case, the method by which the learning unit (140) trains the neural network model may be trained using a learning technique included in machine learning and / or deep learning technology, and is not limited to a specific learning technique.

[0094] The learning unit (140) may receive information regarding the time at which panic symptoms occurred. The learning unit (140) may train the generalized neural network model using demographic information, psychological test scales, pattern information, and daily life information input during a preset period based on the time at which panic symptoms occurred.

[0095] For example, the preset period may correspond to 24 hours. The learning unit (140) may train the generalized neural network model using information entered within 24 hours of the time of panic symptom onset. This is merely an example, and the preset period may vary depending on the intensity, frequency, and / or living environment of a specific patient's panic symptoms, and is not limited to a specific period.

[0096] FIG. 6 is an example image showing the prediction performance of a personalized neural network model using an XGBoost classifier according to one embodiment of the present specification.

[0097] Referring to Fig. 6, the prediction performance of a personalized neural network model using an XGBoost classifier can be analyzed using a Receiver Operating Characteristic curve (ROC curve). The area under the ROC curve (Area Under Curve, AUC) has a value between 1 and 0.5 and can represent a measure of the prediction performance of the personalized neural network model. According to the previous paper, "Muller, Matthew P, et al. 'Can routine laboratory tests discriminate between severe acute respiratory syndrome and other causes of community-acquired pneumonia?' Clinical Infectious Diseases 40.8 (2005) 1079-1086", the prediction performance of a personalized neural network model is 'Excellent' according to the area under the ROC curve. ', 'Good ', 'Fair ' and 'Poor ' can be evaluated as.

[0098] In Fig. 6, the first graph (200) is an ROC curve when panic symptoms do not occur the next day, and the area under the ROC curve can be seen to have a value of 0.907. This area may represent a measure of the accuracy with which the personalized neural network model predicts that panic symptoms will not occur the next day.

[0099] Additionally, the second graph (210) is an ROC curve when panic symptoms occur the next day, and the area under the ROC curve can be seen to have a value of 0.777. The area may represent a measure of the accuracy with which the personalized neural network model predicts that panic symptoms will occur the next day.

[0100] Through this, it can be seen that the personalized neural network model can predict whether panic symptoms will occur the next day.

[0101] In addition, the learning unit (140) can extract a plurality of factors that have a relatively high contribution in predicting the occurrence of a patient's panic symptoms. The learning unit (140) can train the generalized neural network model or the personalized neural network model using the extracted plurality of factors. Through this, the generalized neural network model can be adjusted to a personalized neural network model for a specific patient. Furthermore, the generalized neural network model can be adjusted to a personalized and lightweight neural network model by being trained using the extracted plurality of factors. This may mean that the personalized neural network model uses relatively fewer resources. Furthermore, it may mean that the personalized neural network model has a relatively faster computational speed.

[0102] The learning unit (140) may set the weights of nodes into which elements other than the multiple elements extracted from the generalized neural network model or personalized neural network model are input to preset values. For example, the learning unit (140) may set the weights of the nodes to '0'. This is merely an example and is not limited by the weights.

[0103] Figure 7 illustrates an example of multiple extracted elements.

[0104] Referring to FIG. 7, the learning unit (140) can calculate the Shapley value for each element in order to extract the contribution of each element. The learning unit (140) can extract a plurality of elements that have a relatively high contribution in predicting that panic symptoms will not occur the next day. The learning unit (140) can use this to train the generalized neural network model. This is an example, and is not limited to the method of calculating the Shapley value to extract the contribution, and various methods such as feature importance, permutation importance, and / or partial dependence plot (PDP) can be used.

[0105] Alternatively, the learning unit (140) may extract multiple factors that have a relatively high contribution to predicting the occurrence of panic symptoms the next day. The learning unit (140) may use these factors to train the generalized neural network model.

[0106] Alternatively, the learning unit (140) may train the generalized neural network model using a plurality of factors having a relatively high contribution in predicting that panic symptoms will not occur the next day and a plurality of factors having a relatively high contribution in predicting that panic symptoms will occur the next day.

[0107] Figure 8 is an example image comparing the performance of neural network models trained using different amounts of data.

[0108] Figure 8 is a graph showing the accuracy score, receiver operating area under the curve (ROC-AUC) score, precision score, recall score, and F1 score of a personalized neural network model trained using different amounts of data. The accuracy score may refer to an indicator of whether the personalized neural network model accurately predicted whether panic symptoms would occur the next day. The precision score may refer to the ratio of days when the personalized neural network model predicted that panic symptoms would occur the next day and actually caused panic symptoms. The recall score may refer to the ratio when panic symptoms actually occurred the next day and were predicted by the personalized neural network model to the ratio when panic symptoms actually occurred the next day. The F1 score is the harmonic mean of the precision score and the recall score.

[0109] In Fig. 8, the first bar (220) represents the performance of the first model trained using all of the patient's demographic information, psychological test scale, pattern information, and daily life information. The second bar (230) represents the performance of the second model trained using information excluding the patient's pattern information. The third bar (240) represents the performance of the third model trained using information excluding the patient's psychological scale. The fourth bar (250) represents the performance of the fourth model trained using information excluding the patient's daily life information. The fifth bar (260) represents the performance of the fifth model trained using 10 factors that have a relatively high contribution in predicting the occurrence of panic symptoms the next day.

[0110] Referring to Figure 8, the accuracy score, ROC-AUC score, precision score, and recall score of the fifth model are the third highest. Furthermore, the F1 score of the fifth model is the second highest. Furthermore, the fifth model has similar values ​​to the first model in each score. This suggests that predictive performance is maintained even when training a neural network model using 10 factors that contribute significantly to predicting the occurrence of panic symptoms.

[0111] When generating prediction information on the occurrence of panic symptoms using the personalized and lightweight neural network model, the feedback generation unit (130) may generate feedback information based on a plurality of extracted factors. For example, the plurality of extracted factors may include factors related to the patient's sleep time, smoking amount, alcohol intake, and / or exercise amount. If the patient has insufficient sleep time, the patient may be more likely to develop panic symptoms. Furthermore, the patient may be more likely to develop panic symptoms due to excessive smoking, alcohol intake, and / or exercise amount. At this time, if the prediction information generation unit (120) generates information predicting that the patient will develop panic symptoms the next day, the feedback generation unit (130) may generate personalized feedback information for the patient, such as, "Please get at least 8 hours of sleep. Please refrain from smoking and drinking. Please avoid excessive exercise." This is merely an example and is not limited by the above factors, and various embodiments may arise depending on the characteristics of each patient.

[0112] The above-described transceiver (100), pattern information generation unit (110), prediction information generation unit (120), feedback generation unit (130), and learning unit (140) may include a processor, an application-specific integrated circuit (ASIC), another chipset, a logic circuit, a register, a communication modem, a data processing device, etc. known in the technical field to which the present invention pertains in order to execute calculations and various control logics. In addition, when the above-described control logic is implemented in software, the transceiver (100), pattern information generation unit (110), prediction information generation unit (120), feedback generation unit (130), and learning unit (140) may be implemented as a set of program modules. At this time, the program modules may be stored in the memory device and executed by the processor.

[0113] Hereinafter, a method for generating symptom occurrence prediction information using a symptom occurrence prediction information generating device (10, 10') according to the present specification will be described. However, in describing the method for generating symptom occurrence prediction information according to the present specification, repetitive descriptions of each component are omitted.

[0114] FIG. 9 is a flowchart of a method for generating symptom occurrence prediction information according to one embodiment of the present specification.

[0115] Referring to FIG. 9, in step S10, the processor may receive demographic information, psychological assessment scales, biometric information, and daily life information from the digital device. The biometric information may include at least one of the patient's heart rate, step count, and sleep-related information. The biometric information may be measured using a wearable device.

[0116] In step S11, the processor may generate information related to a change in the pattern of the biometric information (hereinafter referred to as “pattern information”). The processor may generate the pattern information by extracting at least one of the time of the peak, the difference between the peak and the lowest point, and / or the average value from the circadian rhythm of the biometric information. In this case, when the biometric information is related to heart rate, the processor may extract multiple frequency ranges from the circadian rhythm of the heart rate, and further generate information related to the intensity of each frequency range.

[0117] The above daily information may include factors such as the time of day when the panic attack occurred, level of positive emotion, level of negative emotion, level of positive energy, level of negative energy, level of anxiety, level of irritability, amount of alcohol consumed, amount of caffeine consumed, amount of smoking, amount of exercise, whether or not suicidal thoughts occurred in the past month, and amount of psychiatric medication taken in the past month. For women, the above daily information may further include whether or not they had their period that day and / or when their previous period occurred.

[0118] In step S12, the processor can generate predictive information regarding the occurrence of panic symptoms using the demographic information, psychological test scales, pattern information, and daily life information. The processor can generate the predictive information by inputting the information into a pre-input neural network model. The neural network model may be a generalized neural network model trained using information from multiple individuals with panic symptoms.

[0119] The processor may generate the predictive information in stages based on the likelihood of panic symptoms occurring. For example, the processor may generate the predictive information at four levels: "normal," "concern," "caution," and "danger."

[0120] In step S13, the processor may generate feedback information based on the prediction information. At this time, the processor may generate the feedback information in stages based on the level of the prediction information.

[0121] In step S14, the processor can transmit the prediction information and feedback information to the digital device.

[0122] Fig. 10 is a flowchart of a method for generating symptom occurrence prediction information according to another embodiment of the present specification.

[0123] Referring to Fig. 10, steps S20 and S21 are identical to steps S10 and S11, so a repetitive description is omitted.

[0124] In step S20, the processor may train the generalized neural network model on information related to the occurrence of panic symptoms in the patient using at least one of the patient's demographic information, psychological test scales, pattern information, and daily life information. The processor may train the generalized neural network model on information related to the occurrence of symptoms using demographic information, psychological test scales, pattern information, and daily life information input during a preset period based on the time of panic symptom onset. Through this, the generalized neural network model may be adjusted to a personalized neural network model for the patient.

[0125] In addition, the processor can calculate the Shapley value of each element included in the demographic information, psychological test scale, pattern information, and daily life information. Through this, the processor can extract a plurality of elements that have a relatively high contribution in predicting the occurrence of panic symptoms in the patient. The processor can train the generalized neural network model or the personalized neural network model using the extracted plurality of elements and adjust it into a personalized and lightweight neural network model. At this time, the processor can set the weights of the nodes into which elements other than the plurality of elements are input in the generalized neural network model or the personalized neural network model to preset values.

[0126] In step S23, the processor may input the patient's demographic information, psychological test scale, pattern information, and daily life information into the personalized and lightweight neural network model to generate prediction information on whether panic symptoms occur.

[0127] In step S24, the processor may generate feedback information based on the prediction information. At this time, the processor may generate personalized feedback information for the patient using the extracted multiple elements.

[0128] In step S25, the processor can transmit the prediction information and feedback information to the digital device.

[0129] The method for generating symptom occurrence prediction information according to the present specification may be implemented in the form of a computer program written to perform each step on a computer and recorded on a computer-readable recording medium. The aforementioned computer program may include code coded in a computer language, such as C / C++, C#, JAVA, Python, or machine language, that can be read by the processor (CPU) of the computer through the device interface of the computer, so that the computer reads the program and executes the methods implemented as a program. Such code may include functional code related to functions that define the functions necessary to execute the methods, and may include control code related to execution procedures necessary for the processor of the computer to execute the functions according to a predetermined procedure. In addition, such code may further include memory reference-related code regarding which location (address address) in the internal or external memory of the computer should reference additional information or media necessary for the processor of the computer to execute the functions. In addition, if the processor of the computer needs to communicate with any other computer or server located remotely in order to execute the functions, the code may further include communication-related code regarding how to communicate with any other computer or server located remotely using the communication module of the computer, and what information or media to send and receive during communication.

[0130] The above storage medium refers to a medium that stores data semi-permanently and can be read by a device, rather than a medium that stores data for a short period of time, such as a register, cache, or memory. Specifically, examples of the storage medium include, but are not limited to, ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage device. That is, the program can be stored in various recording media on various servers that the computer can access or in various recording media on the user's computer. In addition, the medium can be distributed across network-connected computer systems, so that computer-readable code can be stored in a distributed manner.

[0131] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings.

[0132] FIG. 11 is a block diagram of a symptom occurrence prediction information generation system according to one embodiment of the present specification.

[0133] Referring to FIG. 11, a symptom occurrence prediction information generation system (4) according to one embodiment of the present specification may include a symptom occurrence element extraction device (40), a symptom occurrence prediction information generation device (50), and a wearable device (60).

[0134] The above symptom occurrence prediction information generating device (50) may correspond to a digital device such as a smartphone and / or a tablet computer, which is an example and is not limited to a specific device.

[0135] The above symptom occurrence element extraction device (40) can receive patient information from the symptom occurrence prediction information generation device (50). The above symptom occurrence element extraction device (40) can correspond to a server that learns content related to the occurrence of panic symptoms using the patient's information.

[0136] The above symptom occurrence element extraction device (40) may include a transceiver (400), a storage unit (410), a learning unit (420), and an element extraction unit (430).

[0137] The above-mentioned transceiver (400) can receive the patient's biometric information, daily life information, demographic information, psychological test scale, and information on the time when panic symptoms occurred from the symptom occurrence prediction information generating device (50).

[0138] The above biometric information may be information related to a pattern change in a circadian rhythm (hereinafter referred to as “pattern information”). The biometric information may have a specific pattern that repeats in a 24-hour cycle as a circadian rhythm. The pattern information may include content related to at least one or more of the patient’s biometric information related to heart rate, number of steps, and sleep. In this case, the pattern information may include content related to at least one or more of the time when the circadian rhythm of the biometric information is at its highest point (acrophase), the difference between the highest point and the lowest point (amplitude), and / or the average value.

[0139] For example, when the biometric information is heart rate, the time of the peak point may refer to the time when the patient's heart rate is the highest in the circadian rhythm. The difference between the peak point and the lowest point may refer to the difference between the highest heart rate and the lowest heart rate in the circadian rhythm. The mean value may refer to the mean value (Mesor) of the heart rate in the circadian rhythm and / or the mean value (Mean) of the heart rate measured at a preset time interval during the day. The mean value of the heart rate in the circadian rhythm may refer to the median value in the circadian rhythm. The mean value of the heart rate measured at a preset time interval during the day may refer to the arithmetic mean of the heart rates.

[0140] Additionally, when the biometric information is related to heart rate, the pattern information may further include information related to the intensity of multiple frequency domains extracted from the circadian rhythm of the heart rate. The multiple frequency domains may correspond to frequency domains of 0.001-0.0005 Hz, 0.0005-0.0001 Hz, 0.0001-0.00005 Hz, and / or 0.00005-0.00001 Hz.

[0141] Referring to the above Figure 2, the pattern information related to the heart rate may include elements such as 'HR_variance', 'HR_maximum', 'HR_mean', 'HR_hvar_mean', 'Bandpower(0.001-0.0005Hz)', 'Bandpower(0.0005-0.0001Hz)', 'Bandpower(0.0001-0.00005Hz)', 'Bandpower(0.00005-0.00001Hz)', 'HR_acrophase', 'HR_amplitude', 'HR_mesor', 'HR_acrophase_difference', 'HR_acrophase_difference_2d', 'HR_amplitude_difference', 'HR_amplitude_difference_2d', 'HR_mesor_difference' and / or 'HR_mesor_difference_2d'. Pattern information related to step count may include elements such as 'Steps_total', 'Steps_maximum', 'Steps_mean', 'Steps_variance', and / or 'Steps_hvar_mean'. Pattern information related to sleep may include elements such as 'Sleep_onset_time', 'Sleep_out_time', and / or 'Sleep_duration'. A detailed description of each element is omitted as it is described in Fig. 2.

[0142] Referring to (a) of the above drawing 3, the patient's demographic information may include elements such as age, gender, marital status, occupation, smoking history, drinking history, and / or suicide attempt history.

[0143] In addition, referring to (b) of the above Figure 3, the patient's psychological test scale includes the Seasonal Pattern Assessment Questionnaire (SPAQ) score, the Center for Epidemiological Studies-Depression Scale (CES_D) score, the Generalized Anxiety Disorder-7 (GAD_7) score, the State-Trait Anxiety Inventory (STAI_X2) score, the Korean Occupational Stress Scale-Short Form (KOSSSF) score, the Patient Health Questionnaire_9 (PHQ_9) score, the Social Anxiety Distress Scale (SADS) score, the State-Trait Anxiety Inventory (STAI_X1) score, the Biological Rhythms Interview of Assessment in Neuropsychiatry (BRIAN) score, and the Composite Scale of Morningness (Composite Scale of Morningness, CSM) score, Childhood Trauma Questionnaire (CTQ) score, Attention Questionnaire Scale (AQS) score, Albany Panic & Phobia Questionnaire (APPQ) score, Korean Resilience Quotient (KRQ) score, Body Sensation Questionnaire (BSQ) score, Mood Disorder Questionnaire (MODQ) scoreThis may include factors such as the MDQ score and / or the Brief-Fear of Negative Evaluation (BFNE) score.

[0144] Additionally, referring to the above Figure 4, the patient's daily life information may include elements such as the time of panic symptom occurrence, level of positive emotion, level of negative emotion, level of positive energy, level of negative energy, level of anxiety, level of irritability, amount of alcohol intake, amount of caffeine intake, amount of smoking, amount of exercise, whether or not suicidal thoughts have occurred within the past month, and dosage of psychiatric medication within the past month. For women, the above daily life information may further include elements such as whether or not they had their period that day and / or when their previous menstrual period occurred.

[0145] The above biometric information can be measured using the wearable device (60) worn by the patient. The wearable device (60) may correspond to a smart watch including a display. This is an example and is not limited to a specific device.

[0146] An application for measuring biometric information may be installed in the wearable device (60). The wearable device (60) may measure heart rate using photoplethysmography technology. In addition, the wearable device (60) may measure the number of steps using an accelerometer and / or an inertial measurement unit (IMU) sensor. In addition, the wearable device (60) may extract information related to sleep using the measured heart rate and / or number of steps. This corresponds to one example, and is not limited to a specific method for measuring heart rate, number of steps, and / or information related to sleep.

[0147] The wearable device (60) can transmit the biometric information to the symptom occurrence prediction information generation device (50). The wearable device (60) and the symptom occurrence prediction information generation device (50) can be paired and synchronized via Bluetooth. This is merely an example, and any communication technology capable of synchronizing the wearable device (60) and the symptom occurrence prediction information generation device (50) can be used.

[0148] The information related to the heart rate, number of steps, and / or sleep may be measured through an application installed in the wearable device (60) and / or the symptom occurrence prediction information generation device (50) and transmitted to the transceiver (400). The information related to the heart rate, number of steps, and / or sleep may be measured in real time and transmitted to the transceiver (400).

[0149] The above-mentioned transceiver (400) can be connected to a network configured to transmit and receive data with various communication terminals, such as between wearable devices, between wearable devices and digital devices, between digital devices, between digital devices and server devices, and between server devices. The network can be a wired or wireless communication network and is not limited by a specific communication protocol.

[0150] The patient can input demographic and daily life information using an application installed on the wearable device (60) and / or the symptom occurrence prediction information generation device (50). Additionally, the patient can conduct a self-psychological test using the application. The results of the self-psychological test can be used as a measure of the psychological test. The demographic and daily life information can be quantified according to the application's algorithm.

[0151] In addition, the screen of the application may have a virtual button that generates information about the time when the panic symptom occurred when the panic symptom occurred. Alternatively, a widget button that generates information about the time when the panic symptom occurred may be present on the home screen of the wearable device (60) and / or the symptom occurrence prediction information generation device (50). Alternatively, a physical button that generates information about the time when the panic symptom occurred may be present on a portion of the wearable device (60) and / or the symptom occurrence prediction information generation device (50). The patient may generate information about the time when the panic symptom occurred by using one of the virtual button, the widget button, and the physical button.

[0152] The elements described in FIGS. 2 and 4 are merely examples and are not limited thereto, and various embodiments may be developed. For example, the biometric information may further include elements related to electrocardiogram, blood pressure, and / or respiration rate measured using a wearable device (60). The demographic information may further include elements such as religious affiliation, education level, and / or income level. The daily information may further include information related to eating patterns.

[0153] The storage unit (410) may store a neural network model that predicts whether a patient will develop panic symptoms by using demographic information, psychological test scales, daily life information, and the patient's biometric information as input values. The neural network model may correspond to a generalized neural network model trained to predict the onset of panic symptoms in advance by using demographic information, psychological test scales, pattern information, and daily life information of multiple people with panic symptoms. The neural network model may be a neural network model trained by a learning technique included in machine learning and / or deep learning technology, and is not limited by a specific learning technique.

[0154] For example, the neural network model may be a neural network model that uses one or more of a multiclass classification model, such as an XGBoost classifier, a RandomForest classifier, and a GradientBoost classifier. This is an example and is not limited to a specific model.

[0155] The learning unit (420) can train the generalized neural network model using the patient's biometric information, daily life information, demographic information, psychological test scale, and information about the time when panic symptoms occurred. The learning unit (420) can check whether there are any outliers in the information. If there are any outliers, the learning unit (420) can correct or remove the outliers. Thereafter, the learning unit (420) can normalize the information. The learning unit (420) can train the generalized neural network model using the normalized information. Through this, the generalized neural network model can be adjusted to a neural network model personalized for the patient. The personalized model may refer to a model in which the generalized neural network model is adjusted to suit the individual characteristics of the patient. In this case, the method by which the learning unit (420) trains the neural network model can be trained using a learning technique included in machine learning and / or deep learning technology, and is not limited to a specific learning technique.

[0156] The above learning unit (420) can train the generalized neural network model using the demographic information, psychological test scale, pattern information, and daily life information input during a preset period based on the time when panic symptoms occurred.

[0157] For example, the preset period may correspond to 24 hours. The learning unit (420) may train the generalized neural network model using information entered within 24 hours of the time of panic symptom onset. This is merely an example, and the preset period may vary depending on the intensity, frequency, and / or living environment of the patient's panic symptoms, and is not limited to a specific period.

[0158] Referring to the above Fig. 6, the prediction performance of a personalized neural network model using an XGBoost classifier can be analyzed using a Receiver Operating Characteristic curve (ROC curve). The area under the ROC curve (Area Under Curve, AUC) has a value between 1 and 0.5 and can represent a measure of the prediction performance of the personalized neural network model. According to the previous paper, "Muller, Matthew P, et al. 'Can routine laboratory tests discriminate between severe acute respiratory syndrome and other causes of community-acquired pneumonia?' Clinical Infectious Diseases 40.8 (2005) 1079-1086", the prediction performance of a personalized neural network model is 'Excellent' according to the area under the ROC curve. ', 'Good ', 'Fair ' and 'Poor ' can be evaluated as.

[0159] In Fig. 6, the first graph (200) is an ROC curve when panic symptoms do not occur the next day, and the area under the ROC curve can be seen to have a value of 0.907. This area may represent a measure of the accuracy with which the personalized neural network model predicts that panic symptoms will not occur the next day.

[0160] Additionally, the second graph (210) is an ROC curve when panic symptoms occur the next day, and the area under the ROC curve can be seen to have a value of 0.777. The area may represent a measure of the accuracy with which the personalized neural network model predicts that panic symptoms will occur the next day.

[0161] Through this, it can be seen that the personalized neural network model can predict whether panic symptoms will occur the next day.

[0162] The above-mentioned element extraction unit (430) can calculate the contribution of each element included in the patient's biometric information, daily life information, demographic information, and psychological test scale to predicting symptom occurrence. Thereafter, the element extraction unit (430) can extract multiple elements in order of relatively high contribution.

[0163] Referring to the above-described FIG. 7, the element extraction unit (430) may calculate Shapley values ​​for each element in order to extract the contribution of each element. This is merely an example, and the method of extracting contributions by calculating the Shapley values ​​is not limited, and various methods such as feature importance, permutation importance, and / or partial dependence plot (PDP) may be used.

[0164] The above element extraction unit (430) can extract multiple elements that have a relatively high contribution in predicting that panic symptoms will not occur the next day.

[0165] Alternatively, the element extraction unit (430) may extract multiple elements that have a relatively high contribution in predicting the occurrence of panic symptoms the next day.

[0166] Alternatively, the element extraction unit (430) may extract both a plurality of elements having a relatively high contribution in predicting that panic symptoms will not occur the next day and a plurality of elements having a relatively high contribution in predicting that panic symptoms will occur the next day.

[0167] The above-mentioned transceiver (400) can transmit information about the plurality of elements to the symptom occurrence prediction information generation device (50).

[0168] The above symptom occurrence prediction information generation device (50) may include a transceiver (500), a pattern information generation unit (510), a storage unit (520), a neural network model adjustment unit (530), a prediction information generation unit (540), and a feedback generation unit (550).

[0169] The transceiver (500) of the symptom occurrence prediction information generating device (50) can transmit the pattern information, daily life information, demographic information, psychological test scale, and information about the time when panic symptoms occurred to the symptom occurrence element extraction device (40). Thereafter, the transceiver (500) of the symptom occurrence prediction information generating device (50) can receive information about the plurality of elements from the symptom occurrence element extraction device (40).

[0170] The pattern information generation unit (510) may generate the pattern information from the circadian rhythm of the biometric information. The pattern information generation unit (510) may generate the pattern information by extracting at least one of the time of the peak, the difference between the peak and the lowest point, and the average value from the circadian rhythm of the biometric information related to the patient's heart rate, number of steps, and / or sleep. The pattern information generation unit (510) may generate the pattern information using Cosinor analysis. This is an example and is not limited by the analysis method.

[0171] In addition, when the biometric information is related to heart rate, the pattern information generation unit (510) can extract multiple frequency domains from the circadian rhythm of the heart rate. The pattern information generation unit (510) can further generate content related to the intensity of the multiple frequency domains. The pattern information generation unit (510) can convert the measured heart rate information into a frequency domain through Fourier transform. Thereafter, the pattern information generation unit (510) can calculate the bandpower value of each frequency domain using a multi-taper method. The method by which the pattern information generation unit (510) generates content related to the intensity of each frequency domain is an example and is not limited by the method.

[0172] The pattern information generation unit (510) may interpolate missing values ​​using an interpolation method if biometric information is not input during a specific time period. In this case, the pattern information generation unit (110) may interpolate missing values ​​using linear interpolation and / or spline interpolation. This is merely an example and is not limited to a specific method.

[0173] The storage unit (520) of the above symptom occurrence prediction information generating device (50) can store a neural network model that predicts whether a patient will develop panic symptoms using the pattern information, daily life information, demographic information, and psychological test scale as input values. The neural network model stored in the storage unit (520) of the above symptom occurrence prediction information generating device (50) may be the same neural network model as the generalized neural network model stored in the storage unit (520) of the above symptom occurrence element extraction device (40).

[0174] The neural network model adjustment unit (530) can adjust the generalized neural network model using information on multiple elements received from the symptom occurrence element extraction device (40). Through this, the generalized neural network model can be adjusted to a personalized and lightweight neural network model for the patient. This may mean that the personalized neural network model uses relatively fewer resources. Furthermore, it may mean that the personalized neural network model has a relatively faster computational speed.

[0175] The neural network model adjustment unit (530) may set the weights of nodes into which elements other than the plurality of elements are input in the generalized neural network model to preset values. For example, the neural network model adjustment unit (530) may set the weights of the nodes to '0'. This is merely an example and is not limited by the weights.

[0176] The above-mentioned predictive information generation unit (540) can input the patient's pattern information and daily life information into the personalized and lightweight neural network model to generate predictive information on the occurrence of panic symptoms. The pattern information and daily life information can be input into the personalized and lightweight neural network model in real time.

[0177] The demographic information and psychological test scales may be input into the personalized and lightweight neural network model in advance. The demographic information and psychological test scales may have a relatively slower update rate than the pattern information and daily life information. For example, the patient's occupational status in the demographic information may not change for a year. Additionally, the psychological test scales may be updated when the patient takes a psychological test on a monthly basis. The predictive information generation unit (540) may update the demographic information and / or psychological test scales input into the personalized and lightweight neural network model when there is a change in the demographic information and / or psychological test scales.

[0178] The above prediction information generation unit (540) can normalize the above information and input it into the personalized and lightweight neural network model. At this time, the transmission / reception unit (500) of the symptom occurrence prediction information generation device (50) can transmit the changed content to the symptom occurrence element extraction device (40). The symptom occurrence element extraction device (40) can train the stored neural network model using the changed content.

[0179] The above-described prediction information generation unit (540) can generate the prediction information for a predetermined time. For example, the prediction information generation unit (540) can generate prediction information regarding the occurrence of panic symptoms the following day. Alternatively, the prediction information generation unit (540) can generate prediction information regarding the occurrence of panic symptoms in the afternoon and / or evening of the current day. This is merely an example and is not limited to a specific time.

[0180] Additionally, the prediction information generation unit (540) may generate the prediction information for a predetermined time period afterward. For example, the prediction information generation unit (540) may generate prediction information regarding the occurrence of panic symptoms 24 hours from the current time. Alternatively, the prediction information generation unit (540) may generate prediction information regarding the occurrence of panic symptoms 8 hours from the current time. This is merely an example and is not limited to a specific time period.

[0181] The above Figure 8 is a graph showing the accuracy score, ROC-AUC score, precision score, recall score, and F1 score of a personalized neural network model trained using different amounts of data. The accuracy score may refer to an indicator of whether the personalized neural network model accurately predicted whether panic symptoms would occur the next day. The precision score may refer to the ratio of days when the personalized neural network model predicted that panic symptoms would occur the next day and actually caused panic symptoms. The recall score may refer to the ratio when panic symptoms actually occurred the next day and were predicted by the personalized neural network model to the ratio when panic symptoms actually occurred the next day. The F1 score is the harmonic mean of the precision score and the recall score.

[0182] In the above Fig. 8, the first bar (220) represents the performance of the first model learned using all of the patient's demographic information, psychological test scale, pattern information, and daily life information. The second bar (230) represents the performance of the second model learned using information excluding the patient's pattern information. The third bar (240) represents the performance of the third model learned using information excluding the patient's psychological scale. The fourth bar (250) represents the performance of the fourth model learned using information excluding the patient's daily life information. The fifth bar (260) represents the performance of the fifth model learned using 10 factors that have a relatively high contribution in predicting the occurrence of panic symptoms on the next day.

[0183] Referring to Figure 8, the accuracy score, ROC-AUC score, precision score, and recall score of the fifth model are the third highest. Furthermore, the F1 score of the fifth model is the second highest. Furthermore, the fifth model has similar values ​​to the first model in each score. This suggests that predictive performance is maintained even when training a neural network model using 10 factors that contribute significantly to predicting the occurrence of panic symptoms.

[0184] The feedback generation unit (550) may generate feedback information according to the prediction information. When the prediction information predicts that panic symptoms will not occur the next day, the feedback generation unit (550) may generate feedback information having the content of a daily greeting message, such as, "It is predicted that panic symptoms will not occur tomorrow. Have a nice day." When the prediction information predicts that panic symptoms will occur the next day, the feedback generation unit (550) may generate feedback information including, for example, "It is predicted that panic symptoms will occur tomorrow. Get enough rest. Take deep breaths when panic symptoms occur. Take your prescribed medication if necessary." This is an example and is not limited to the content of the feedback information.

[0185] In addition, the prediction information generation unit (540) may generate the prediction information in stages. For example, the prediction information generation unit (540) may generate the prediction information into four levels of 'normal', 'concern', 'caution', and 'danger' according to the possibility of occurrence of panic symptoms. The 'normal' level may have a relatively lowest possibility of occurrence of panic symptoms. The 'concern' level may have a relatively higher possibility of occurrence of panic symptoms than the 'normal' level. The 'caution' level may have a relatively higher possibility of occurrence of panic symptoms than the 'concern' level. The 'danger' level may have a relatively highest possibility of occurrence of panic symptoms.

[0186] The above feedback generation unit (550) can generate feedback information step by step according to the prediction information. The feedback information may include content regarding the four levels. When the prediction information is at the 'normal' level, the feedback information may include content such as 'panic symptoms are not expected to occur tomorrow.' When the prediction information is at the 'concern' level, the feedback information may include content regarding the level of panic symptom occurrence, such as 'the possibility of panic symptoms occurring tomorrow is at the 'concern' level.' When the prediction information is at the 'caution' and 'danger' levels, it is the same as the 'concern' level, so repeated explanation is omitted.

[0187] Additionally, when the predicted information is at a 'normal' level, the feedback information may further include everyday greetings, such as 'Have a comfortable day.' When the predicted information is at a 'concern' level, the feedback information may further include behavioral guidelines to prevent the occurrence of panic symptoms, such as 'Please get enough rest.' When the predicted information is at a 'caution' and / or 'danger' level, the feedback information may further include behavioral guidelines to prevent the occurrence of panic symptoms, such as 'Please get enough rest. Take deep breaths when panic symptoms occur. Take prescribed medication if necessary.', and content on how to deal with panic symptoms when they occur.

[0188] The stepwise generation of the above prediction information and feedback information according to the possibility of occurrence of the panic symptoms described above is an example and is not limited to the above four levels and feedback information.

[0189] In addition, the feedback generation unit (550) may generate feedback information according to the plurality of factors. For example, the extracted plurality of factors may include factors related to the patient's sleep time, smoking amount, alcohol intake, and / or exercise amount. If the patient has insufficient sleep time, the patient may be more likely to experience panic symptoms. In addition, the patient may be more likely to experience panic symptoms due to excessive smoking, alcohol intake, and / or exercise amount. In this case, if the prediction information generation unit (540) generates information predicting that the patient will experience panic symptoms the next day, the feedback generation unit (550) may generate personalized feedback information for the patient, such as, "Please get more than 8 hours of sleep. Please refrain from smoking and drinking. Please avoid excessive exercise." This is merely an example and is not limited by the above factors, and various embodiments may be created depending on the characteristics of each patient.

[0190] The transceiver (500) of the symptom occurrence prediction information generating device (50) may transmit the prediction information to the symptom occurrence element extraction device (40). The learning unit (420) may compare the prediction information with information on the time when panic symptoms occurred. If the prediction information and information on the time when panic symptoms occurred do not match, the learning unit (420) may retrain the neural network model. For example, the prediction information may include content predicting that panic symptoms will occur in the patient the next day. However, the patient may not actually experience panic symptoms the next day. Therefore, information on the time when panic symptoms occurred may not be transmitted to the symptom occurrence element extraction device (40). In this case, the learning unit (420) may retrain the neural network model. Accordingly, the element extraction unit (430) may re-extract a plurality of elements according to the adjusted neural network model. This is an example and is not limited by the above situation.

[0191] The transceiver (400), learning unit (420), element extraction unit (430) of the symptom occurrence element extraction device (40), the transceiver (500), pattern information generation unit (510), neural network model adjustment unit (530), prediction information generation unit (540), and feedback generation unit (550) of the symptom occurrence prediction information generation device (50) may include a processor, ASIC (application-specific integrated circuit), other chipset, logic circuit, register, communication modem, data processing device, etc. known in the technical field to which the present invention belongs in order to execute calculation and various control logic. In addition, when the above-described control logic is implemented in software, the transceiver unit (400), learning unit (420), element extraction unit (430) of the symptom occurrence element extraction device (40), the transceiver unit (500), pattern information generation unit (510), neural network model adjustment unit (530), prediction information generation unit (540), and feedback generation unit (550) of the symptom occurrence prediction information generation device (50) may be implemented as a set of program modules. At this time, the program modules may be stored in the memory device and executed by the processor.

[0192] Below, a method for extracting symptom occurrence elements using a symptom occurrence element extraction device (40) according to this specification will be described. However, in explaining the method for extracting symptom occurrence elements according to this specification, repetitive descriptions of each component will be omitted.

[0193] Fig. 12 is a flowchart of a method for extracting symptom occurrence elements according to one embodiment of the present specification.

[0194] Referring to FIG. 12, in step S30, the processor may receive information regarding the patient's biometric information, daily life information, demographic information, psychological test scale, and the time of panic symptom onset from a digital device. The digital device may correspond to a symptom onset prediction information generation device (50) according to the present specification.

[0195] The above biometric information may be information related to pattern changes in the circadian rhythm (hereinafter referred to as "pattern information"). The pattern information may include information related to at least one of the patient's biometric information, including heart rate, number of steps, and sleep. In this case, the pattern information may include information related to at least one of the time of the highest point (acrophase), the difference between the highest point and the lowest point (amplitude), and / or the average value in the circadian rhythm of the biometric information.

[0196] Additionally, when the biometric information is related to heart rate, the pattern information may further include information related to the intensity of multiple frequency domains extracted from the circadian rhythm of the heart rate.

[0197] In step S31, the processor may train a neural network model to predict the onset of panic symptoms using the patient's biometric information, daily life information, demographic information, psychological test scales, and information regarding the time of panic symptom onset. This may correspond to a generalized neural network model trained to predict the onset of panic symptoms in advance using demographic information, psychological test scales, pattern information, and daily life information of multiple individuals with panic symptoms.

[0198] The processor can identify outliers in the information. If outliers are present, the processor can correct or remove them. Subsequently, the processor can normalize the information. The processor can use the normalized information to train the generalized neural network model. Through this, the generalized neural network model can be tailored to the patient's individual needs.

[0199] Additionally, the processor can train the generalized neural network model using the demographic information, psychological test scale, pattern information, and daily life information input during a preset period based on the time when panic symptoms occurred.

[0200] In step S32, the processor may calculate the contribution of each element included in the patient's biometric information, daily life information, demographic information, and psychological test scale to predicting symptom occurrence. Subsequently, the processor may extract multiple elements in descending order of contribution. To extract the contribution of each element, the processor may calculate a Shapley value for each element.

[0201] In step S33, the processor can transmit information about the plurality of elements to the digital device.

[0202] Hereinafter, a method for generating symptom occurrence prediction information using a symptom occurrence prediction information generating device (50) according to the present specification will be described. However, in describing the method for generating symptom occurrence prediction information according to the present specification, repetitive descriptions of each component will be omitted.

[0203] Fig. 13 is a flowchart of a method for generating symptom occurrence prediction information according to one embodiment of the present specification.

[0204] Referring to FIG. 13, in step S40, the processor may generate information related to a pattern change in the biometric information from the circadian rhythm of the patient's biometric information (hereinafter referred to as "pattern information"). The processor may generate the pattern information by extracting at least one of the time of the peak, the difference between the peak and the lowest point, and the average value from the circadian rhythm of the biometric information related to the patient's heart rate, number of steps, and / or sleep. In addition, when the biometric information is related to heart rate, the processor may extract a plurality of frequency domains from the circadian rhythm of the heart rate. The processor may further generate information related to the intensity of the plurality of frequency domains. The biometric information may be measured using a wearable device that is linked to a symptom occurrence prediction information generating device (50).

[0205] In step S41, the processor may transmit information on the patient's pattern information, daily life information, demographic information, psychological test scale, and time of panic symptom occurrence to the server. The server may correspond to the symptom occurrence element extraction device (40) according to the present specification. The daily life information, demographic information, and psychological test scale may be input by the patient through an application installed in the symptom occurrence prediction information generation device (50). Information on the time of panic symptom occurrence may be generated through a virtual button and / or a physical button of the wearable device and / or the symptom occurrence prediction information generation device (50).

[0206] In step S42, the processor may receive information about a plurality of factors having relatively high importance in relation to panic symptom prediction from the server.

[0207] In step S43, the processor may adjust a neural network model that predicts whether the patient will experience panic symptoms using information about the plurality of elements. At this time, the processor may set the weights of nodes in the neural network model, into which elements other than the plurality of elements are input, to preset values. The neural network model may be adjusted to a personalized and lightweight neural network model for the patient.

[0208] In step S44, the processor may input the patient's pattern information and daily life information into the personalized and lightweight neural network model to generate predictive information regarding the occurrence of panic symptoms. The processor may input the pattern information and daily life information into the personalized and lightweight neural network model in real time. The demographic information and / or psychological test scales may be input into the personalized and lightweight neural network model in advance. The processor may generate the predictive information in stages based on the likelihood of panic symptoms occurring.

[0209] In step S45, the processor may generate feedback information based on the prediction information. At this time, the processor may generate feedback information in stages based on the prediction information. Furthermore, the processor may generate personalized feedback information for the patient based on information about the plurality of elements.

[0210] There may be instances where the demographic information and / or psychological test scale information is changed from the previously entered information. In this case, in step S41, the processor may transmit the changed information related to the demographic information and / or psychological test scale information to the server. Subsequently, in step S44, the processor may update the demographic information and / or psychological test scale information input into the personalized and lightweight neural network model with the changed information.

[0211] The processor may transmit the prediction information to the server. If the prediction information and the information regarding the time of panic symptom onset do not match, the server may retrain the stored neural network model.

[0212] The method for extracting symptom occurrence elements and / or the method for generating symptom occurrence prediction information according to the present specification may be implemented in the form of a computer program written to perform each step on a computer and recorded on a computer-readable recording medium. The aforementioned computer program may include code coded in a computer language, such as C / C++, C#, JAVA, Python, or machine language, that can be read by the processor (CPU) of the computer through the device interface of the computer, so that the computer reads the program and executes the methods implemented as a program. Such code may include functional code related to functions that define functions necessary for executing the methods, and may include control code related to execution procedures necessary for the processor of the computer to execute the functions according to a predetermined procedure. In addition, such code may further include memory reference-related code regarding which location (address address) of the internal or external memory of the computer should reference additional information or media necessary for the processor of the computer to execute the functions. In addition, if the processor of the computer needs to communicate with any other computer or server located remotely in order to execute the functions, the code may further include communication-related code regarding how to communicate with any other computer or server located remotely using the communication module of the computer, and what information or media to send and receive during communication.

[0213] The above storage medium refers to a medium that stores data semi-permanently and can be read by a device, rather than a medium that stores data for a short period of time, such as a register, cache, or memory. Specifically, examples of the storage medium include, but are not limited to, ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage device. That is, the program can be stored in various recording media on various servers that the computer can access or in various recording media on the user's computer. In addition, the medium can be distributed across network-connected computer systems, so that computer-readable code can be stored in a distributed manner.

[0214] While the embodiments of this specification have been described with reference to the attached drawings, those skilled in the art will appreciate that the present invention can be implemented in other specific forms without altering the technical spirit or essential features thereof. Therefore, the embodiments described above should be understood to be illustrative in all respects and not restrictive.

[0215] [Explanation of symbols]

[0216] 1: Symptom occurrence prediction information generation system

[0217] 10: Symptom occurrence prediction information generation device

[0218] 20: Wearable devices

[0219] 30: Digital devices

[0220] 100: Transmitter and receiver

[0221] 110: Pattern information generation unit

[0222] 120: Prediction information generation unit

[0223] 130: Feedback generation unit

[0224] 140: Learning Department

Claims

1. A transmitter / receiver unit that receives demographic information, psychological test scales, biometric information, and daily life information of a patient from a digital device, and transmits prediction information and feedback information on the occurrence of panic symptoms to the digital device; A pattern information generation unit that generates information related to changes in the pattern of the above biometric information (hereinafter referred to as “pattern information”); A prediction information generation unit that generates the prediction information using the patient's demographic information, psychological test scale, pattern information, and daily life information; and A symptom occurrence prediction information generation device, comprising a feedback generation unit that generates the feedback information according to the prediction information.

2. In claim 1, The above biometric information is, A symptom occurrence prediction information generating device characterized in that it includes at least one of the patient's heart rate, number of steps, and sleep-related information.

3. In claim 2, The above pattern information generation unit, A symptom occurrence prediction information generating device that generates the pattern information by extracting at least one of the time of the peak, the difference between the peak and the lowest point, and the average value from the circadian rhythm of the above bio-information.

4. In claim 3, When the above biometric information is related to heart rate, The above pattern information generation unit, A device for generating symptom occurrence prediction information that extracts multiple frequency domains from the circadian rhythm of heart rate and further generates content related to the intensity of each frequency domain.

5. In claim 1, The above daily information is, A device for generating symptom occurrence prediction information, characterized in that it includes at least one of the following: time of occurrence of panic symptoms, mood changes, energy changes, anxiety level, stress level, alcohol intake, caffeine intake, smoking amount, exercise amount, medication dosage, and information related to a woman's menstruation.

6. In claim 1, The above prediction information generation unit generates the above prediction information in stages, The above feedback generation unit, A symptom occurrence prediction information generation device that generates the feedback information step by step according to the above prediction information.

7. In claim 1, A symptom occurrence prediction information generation device further comprising a learning unit that teaches the prediction information generation unit content related to the occurrence of panic symptoms by using at least one of the demographic information, psychological test scale, pattern information, and daily life information.

8. In claim 7, The above daily information includes the time when panic symptoms occurred, The above learning department, A device for generating symptom occurrence prediction information that learns information related to the occurrence of panic symptoms by using the demographic information, psychological test scale, pattern information, and daily life information entered during a preset period according to the time at which the above panic symptoms occurred.

9. In claim 8, The above learning department, A device for generating symptom occurrence prediction information, which extracts multiple factors having a relatively high contribution in predicting the occurrence of panic symptoms in the patient, and sets the weights of nodes into which other factors excluding the multiple factors are input to preset values.

10. A wearable device that measures the patient's vital signs to generate biometric information or information about the time when panic symptoms occurred; A digital device that is linked to the wearable device and receives demographic information, psychological test scales, and daily life information from the patient; and A symptom occurrence prediction information generation system, comprising a symptom occurrence prediction information generation device according to any one of claims 1 to 9.

11. A patient information receiving step in which the processor receives the patient's demographic information, psychological test scale, biometric information and daily life information from a digital device; A pattern information generation step in which the processor generates information related to a change in the pattern of the biometric information (hereinafter referred to as “pattern information”); A predictive information generation step in which a processor generates predictive information on whether panic symptoms occur by using the patient's demographic information, psychological test scale, pattern information, and daily life information; a feedback information generation step in which the processor generates the feedback information according to the prediction information; and A method for generating symptom occurrence prediction information, comprising a prediction information transmission step in which a processor transmits the prediction information and feedback information to the digital device.

12. In claim 11, The above biometric information is, A method for generating symptom occurrence prediction information, characterized in that it includes at least one of the patient's heart rate, number of steps, and sleep-related information.

13. In claim 12, The above pattern information generation step is: A method for generating symptom occurrence prediction information, wherein the above processor is a step of generating the pattern information by extracting at least one of the time when the circadian rhythm of the above bio-information is at its peak, the difference between the peak and the lowest point, and the average value.

14. In claim 13, When the above biometric information is related to heart rate, The above pattern information generation step is: A method for generating symptom occurrence prediction information, further comprising a step of the processor extracting a plurality of frequency domains from the circadian rhythm of the heart rate and generating content related to the intensity of each frequency domain.

15. In claim 11, The above daily information is, A method for generating symptom occurrence prediction information, characterized in that it includes at least one of the following: time of occurrence of panic symptoms, mood changes, energy changes, anxiety level, stress level, alcohol intake, caffeine intake, smoking amount, exercise amount, medication dosage, and information related to a woman's menstruation.

16. In claim 11, The above prediction information generation step is a step in which the processor generates the prediction information step by step, A method for generating symptom occurrence prediction information, wherein the above feedback information generation step is a step in which the processor generates the feedback information step by step according to the prediction information.

17. In claim 11, A method for generating symptom occurrence prediction information, further comprising a learning step in which the processor learns content related to the occurrence of panic symptoms by using at least one of the demographic information, the psychological test scale, the pattern information, and the daily life information.

18. In claim 17, The above daily information includes the time when panic symptoms occurred, The above learning steps are: A method for generating symptom occurrence prediction information, wherein the above processor learns information related to the occurrence of panic symptoms by using the demographic information, psychological test scale, pattern information, and daily life information input during a preset period of time according to the time at which the above panic symptoms occurred.

19. In claim 18, The above learning steps are: A method for generating symptom occurrence prediction information, wherein the above processor extracts a plurality of elements having a relatively high contribution in predicting the occurrence of panic symptoms in the patient, and sets the weights of nodes into which other elements excluding the plurality of elements are input to preset values.

20. A computer program written to perform each step of the method for generating symptom occurrence prediction information according to any one of claims 11 to 19 on a computer and recorded on a computer-readable recording medium.

Images