Method for providing personalized mental health assesment results

Abstract

A method for providing a user-customized mental health diagnosis result, performed by a computing device to solve the aforementioned problem, is disclosed. The method may include: receiving mental health input information from a user; generating a list of candidate mental illnesses suspected of being the user from the mental health input information using a plurality of artificial intelligence models; determining a target mental illness suffered by the user from the list of candidate mental illnesses using a disease identification module that generates a first user interaction question to check the presence or absence of symptoms for each of the candidate mental illnesses within the list of candidate mental illnesses; generating a verification result for the target mental illness using a validity review module that generates a second user interaction question to check the temporal validity of the target mental illness; and generating the user's mental health diagnosis result based on the verification result for the target mental illness.

Description

Technology Field

[0001] The present disclosure relates to the field of information processing, and more specifically, to a method for generating a mental health diagnosis result from an individual's mental health input information using an artificial intelligence model. Background Technology

[0002] In modern society, the importance of mental health is becoming increasingly prominent. Various mental health issues, such as stress, anxiety, and depression, not only lower an individual's quality of life but also incur enormous social costs. Consequently, there has been a continuous demand for technologies to accurately diagnose and manage mental health conditions at an early stage. Traditional mental health diagnosis has primarily relied on face-to-face consultations with psychiatrists or counselors, supplemented by the use of standardized questionnaires (e.g., the PHQ-9 depression screening tool and the GAD-7 anxiety disorder screening tool).

[0003] To address these issues, active attempts have been made to integrate Information and Communication Technology (ICT) into the field of mental health. The convergence of ICT and mental health technologies has primarily taken the form of providing digital mental health management programs, such as meditation and cognitive behavioral therapy (CBT)-based training, through mobile applications.

[0004] As such, conventional mental health diagnosis is primarily conducted through face-to-face consultations with psychiatrists or counseling specialists. However, this approach may be difficult for many people to access easily due to constraints on time and cost. Furthermore, the expert's subjective judgment may be involved in the diagnostic process, and there may be limitations in continuously monitoring changes in the patient's mental state due to the short duration of the consultations. Prior art literature

[0005] Republic of Korea Registered Patent No. 10-2783562 The problem to be solved

[0006] The present disclosure is devised in response to the aforementioned background technology and aims to efficiently provide mental health diagnosis results from individual mental health input information using an artificial intelligence model. means of solving the problem

[0007] A method for providing a user-customized mental health diagnosis result, performed by a computing device to solve the aforementioned problem, is disclosed. The method may include: receiving mental health input information from a user; generating a list of candidate mental illnesses suspected of being the user from the mental health input information using a plurality of artificial intelligence models; determining a target mental illness suffered by the user from the list of candidate mental illnesses using a disease identification module that generates a first user interaction question to check the presence or absence of symptoms for each of the candidate mental illnesses within the list of candidate mental illnesses; generating a verification result for the target mental illness using a validity review module that generates a second user interaction question to check the temporal validity of the target mental illness; and generating the user's mental health diagnosis result based on the verification result for the target mental illness.

[0008] In one embodiment, the plurality of artificial intelligence models may include: a first artificial intelligence model that generates information on symptoms related to the user's mental health using self-report data included in the mental health input information; a second artificial intelligence model that generates a classification result of the user's emotional state using the self-report data; and a third artificial intelligence model that generates language pattern keywords of the user using text data as input.

[0009] In one embodiment, the first artificial intelligence model may be trained by a first training dataset in which sentences and symptoms are paired, comprising first synthetic data generated by inputting first seed data, in which at least one symptom type corresponding to a pre-collected sentence is labeled, into a large language model (LLM).

[0010] In one embodiment, the second artificial intelligence model may be trained by a second training dataset composed of voice and emotion pairs using second seed data comprising voice data with pre-attached emotion labels and a text transcript corresponding to the voice data.

[0011] In one embodiment, the third artificial intelligence model may be trained by a third training dataset consisting of sentence data containing language pattern keywords and pairs of mental illnesses, which is generated by inputting third seed data constructed using previously collected representative language pattern keywords for each mental illness into a large-scale language model (LLM).

[0012] In one embodiment, the mental health input information may include the user's identification information, question-and-answer data obtained from the user, and self-report data obtained from the user; the user's identification information may include age, gender, and region as information for categorizing the user within a population group; the user's identification information may be used to establish a disease baseline representing the user's baseline probability value for each of a plurality of mental disorders; the question-and-answer data may be used to correct the user's baseline probability value as user response data to a query regarding whether the user has experienced symptoms; and the self-report data may include at least one user speech data among voice and text related to the user's state and symptoms, and may be used as input to the plurality of artificial intelligence models.

[0013] In one embodiment, a second mental illness classification result can be generated by applying the second output of a second artificial intelligence model among the plurality of artificial intelligence models to a first mental illness classification result determined using the first output of a first artificial intelligence model among the plurality of artificial intelligence models, and a third mental illness classification result can be generated by applying the third output of a third artificial intelligence model among the plurality of artificial intelligence models to the second mental illness classification result, and a list of mental illness candidates suspected for the user can be generated using the third mental illness classification result.

[0014] In one embodiment, the first mental illness classification result, the second mental illness classification result, and the third mental illness classification result may each represent the probability of occurrence of the user's mental illness, and the first mental illness classification result may be generated by applying a first comparison result obtained by comparing the first output with a first reference table defining the relationship between a predefined symptom and a mental illness to a disease baseline representing the user's basic probability value for each of a plurality of mental illnesses, and the second mental illness classification result may be generated by applying a second comparison result obtained by comparing the second output with a second reference table defining the relationship between a predefined emotion and a mental illness to the first mental illness classification result, and the third mental illness classification result may be generated by applying a third comparison result obtained by comparing the third output with a third reference table defining the relationship between a predefined language pattern keyword and a mental illness to the second mental illness classification result.

[0015] In one embodiment, the first mental illness classification result may be generated by: a step of generating information on symptoms related to the user's mental health using a first artificial intelligence model that takes self-report data included in the mental health input information as input; a step of comparing the symptom information with the first reference table to determine, for each combination of a plurality of mental illnesses and a plurality of symptoms, a first multiplier value for upwardly correcting the probability of occurrence of a specific mental illness when a specific symptom exists and a second multiplier value for downward correcting the probability of occurrence of a specific mental illness when a specific symptom does not exist; and a step of generating the first mental illness classification result in which the basic probability value of the disease baseline is corrected by applying the first multiplier value and the second multiplier value to the disease baseline representing the user's basic probability value for each of the plurality of mental illnesses.

[0016] In one embodiment, the second mental illness classification result may be generated by: a step of generating an emotional state classification result indicating the user's emotion identification information and emotion intensity using a second artificial intelligence model that takes self-report data included in the mental health input information as input; a step of determining a correlation between a specific emotion and a specific disease by comparing the emotional state classification result with the second reference table; a step of calculating a likelihood ratio by comparing the probability that a specific emotion appears in a group of mental illness patients and the probability that it appears in a group of non-mental illness patients based on the emotional state classification result and the correlation; and a step of generating the second mental illness classification result in which the first mental illness classification result is corrected by applying the likelihood ratio to the first mental illness classification result.

[0017] In one embodiment, the third mental illness classification result may be generated by: a step of generating a language pattern keyword of the user using a third artificial intelligence model that takes self-report data included in the mental health input information as input; a step of determining a mapping value corresponding to the generated language pattern keyword for each combination of a plurality of mental illnesses and a plurality of language pattern keywords by comparing the language pattern keyword with the third reference table; and a step of generating the third mental illness classification result in which the second mental illness classification result is corrected by applying the mapping value to the second mental illness classification result.

[0018] In one embodiment, the step of generating the mental illness candidate list may include: setting a disease baseline representing the initial probability value for each mental illness of the user using the user identification information included in the mental health input information; first adjusting the initial probability value for each mental illness in the disease baseline using mental health-related symptom information generated from a first artificial intelligence model that takes the user's self-report data included in the mental health input information as input and question-and-answer data included in the mental health input information; second adjusting the first adjusted probability value for each mental illness using an emotional state classification result generated from a second artificial intelligence model that takes the user's self-report data included in the mental health input information as input; third adjusting the second adjusted probability value for each mental illness using language pattern keywords generated from a third artificial intelligence model that takes the user's self-report data included in the mental health input information as input; and generating the mental illness candidate list suspected for the user based on the third adjusted probability value for each mental illness.

[0019] In one embodiment, the step of generating the mental illness candidate list comprises: determining first candidate mental illnesses that are likely to occur within the user's category using the user's identification information included in the mental health input information; determining second candidate mental illnesses by filtering out some of the first candidate mental illnesses using mental health-related symptom information generated from a first artificial intelligence model that takes the user's self-report data included in the mental health input information as input (the number of second candidate mental illnesses is smaller than the number of first candidate mental illnesses); and determining third candidate mental illnesses by filtering out some of the second candidate mental illnesses using an emotional state classification result generated from a second artificial intelligence model that takes the user's self-report data included in the mental health input information as input (the number of third candidate mental illnesses is smaller than the number of second candidate mental illnesses). and may include the step of determining third candidate mental disorders by filtering out some of the second candidate mental disorders using language pattern keywords generated from a third artificial intelligence model that takes the user's self-report data included in the mental health input information as input (the number of the third candidate mental disorders is smaller than the number of the second candidate mental disorders).

[0020] In one embodiment, the disease identification module can determine at least one target mental disorder among the candidate mental disorders by generating an independent first user interaction question for each of the candidate mental disorders and scoring the user response to the first user interaction question.

[0021] In one embodiment, the disease determination module may generate a first-1 user interaction question among a plurality of user interaction questions corresponding to a first candidate mental disease included in the mental disease candidate list, and determine whether to generate a first-2 user interaction question based on whether the user response to the first-1 user interaction question satisfies the condition corresponding to the first candidate mental disease.

[0022] In one embodiment, the validity review module generates a second user interaction question that queries symptom time conditions, including the time of symptom onset, duration, and frequency for the target mental disease determined by the disease identification module, determines whether the user response to the second user interaction question satisfies the target time conditions pre-assigned to the target mental disease, and can generate a verification result for the target mental disease based on whether the target time conditions are satisfied.

[0023] In one embodiment, the validity review module may set the target mental illness as a reference mental illness if the user response to the second user interaction question does not satisfy the target time condition, generate a third user interaction question to review the user's medication and underlying disease if the user response to the second user interaction question satisfies the target time condition, and determine whether to include the target mental illness as a confirmed mental illness in the verification result based on the user response to the third user interaction question.

[0024] In one embodiment, the validity review module can determine a mental illness grade for the target mental illness based on the time value related to the user's symptom time condition obtained through the second user interaction question when the user's response to the second user interaction question satisfies the target time condition.

[0025] In one embodiment, the step of generating the mental health diagnosis result may include: generating a fourth user interaction item including a question for measuring the stress level for the confirmed mental illness and a question for examining functions related to the user's daily life, when the validation result includes a confirmed mental illness by the validity review module; and determining the stress level and the risk level for the user's daily life based on the response to the fourth user interaction item.

[0026] In one embodiment, the step of generating the mental health diagnosis result may include: generating the mental health diagnosis result including the mental health input information, the mental illness candidate list, the target mental illness and the confirmed mental illness included in the verification result, the mental illness grade for the confirmed mental illness, the reference mental illness among the target mental illnesses that is not classified as a confirmed mental illness, the user's emotional state classification result, and the user's language pattern keyword.

[0027] A computer program stored on a computer-readable storage medium for solving the aforementioned problem is disclosed. When executed by a computing device, the computer program may perform a method of providing a user-customized mental health diagnosis result, and the method may include: receiving mental health input information from a user; generating a list of candidate mental illnesses suspected of being the user from the mental health input information using a plurality of artificial intelligence models; determining a target mental illness suffered by the user from the list of candidate mental illnesses using a disease identification module that generates a first user interaction question to check the presence or absence of symptoms for each of the candidate mental illnesses within the list of candidate mental illnesses; generating a verification result for the target mental illness using a validity review module that generates a second user interaction question to check the temporal validity of the target mental illness; and generating the user's mental health diagnosis result based on the verification result for the target mental illness.

[0028] A computing device is disclosed that provides user-customized mental health diagnosis results to solve the aforementioned problem. The computing device may include a processor, wherein the processor: receives mental health input information from a user; generates a list of candidate mental illnesses suspected of being the user from the mental health input information using a plurality of artificial intelligence models; determines a target mental illness suffered by the user from the candidate mental illness list using a disease identification module that generates a first user interaction question to check the presence or absence of symptoms for each of the candidate mental illnesses within the candidate mental illness list; generates a verification result for the target mental illness using a validity review module that generates a second user interaction question to check the temporal validity of the target mental illness; and generates the user's mental health diagnosis result based on the verification result for the target mental illness. Effects of the invention

[0029] The present disclosure can efficiently provide mental health diagnosis results from an individual's mental health input information using an artificial intelligence model.

[0030] The effects obtainable from the present disclosure are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art to which the present disclosure belongs from the description below. Brief explanation of the drawing

[0031] Various aspects are now described with reference to the drawings, wherein similar reference numbers are used to collectively refer to similar components. In the following embodiments, for illustrative purposes, a number of specific details are presented to provide a comprehensive understanding of one or more aspects. However, it will be apparent that such aspect(s) may be practiced without these specific details. In other examples, known structures and devices are illustrated in block diagram form to facilitate the description of one or more aspects. FIG. 1 is a block diagram showing a computing device that generates a mental health diagnosis result from an individual's mental health input information according to one embodiment of the present disclosure. FIG. 2 is a diagram showing an exemplary structure of an artificial intelligence-based model according to one embodiment of the present disclosure. FIG. 3 is a flowchart illustrating a method for generating a mental health diagnosis result from an individual's mental health input information according to one embodiment of the present disclosure. FIG. 4 is a diagram showing modules that generate mental health diagnosis results from individual mental health input information according to one embodiment of the present disclosure. FIG. 5 is a diagram illustrating a method in which a mental illness candidate list generation module according to one embodiment of the present disclosure generates a mental illness candidate list. FIG. 6 is a diagram illustrating a method in which a disease identification module according to one embodiment of the present disclosure generates a target mental disease. FIG. 7 is a diagram illustrating a method in which a feasibility review module according to one embodiment of the present disclosure generates a confirmed mental illness. FIG. 8 is a diagram illustrating a method in which a function test module according to one embodiment of the present disclosure tests stress and function. FIG. 9 is a diagram illustrating a method in which a diagnostic result generation module according to one embodiment of the present disclosure generates a mental health diagnostic result. FIG. 10 is a graph for illustrating an embodiment of the present disclosure. FIG. 11 is a general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure may be implemented. Specific details for implementing the invention

[0032] Further scopes of the applicability of the present disclosure will become apparent from the following detailed description. However, since various changes and modifications within the spirit and scope of the invention are clearly understood by those skilled in the art, specific embodiments, such as the detailed description and preferred embodiments of the invention, should be understood as being given merely as examples.

[0033] Various embodiments and / or aspects are now disclosed with reference to the drawings. For illustrative purposes, numerous specific details are disclosed in the following description to aid in a general understanding of one or more aspects. However, it will be apparent to those skilled in the art that these aspects may be practiced without such specific details. The following description and the accompanying drawings describe specific exemplary aspects of one or more aspects in detail. However, these aspects are exemplary, and some of the various methods in the principles of the various aspects may be used, and the descriptions are intended to include all such aspects and their equivalents. Specifically, terms such as “exemplary,” “example,” “aspect,” and “example” as used herein may not be interpreted as implying that any described aspect or design is superior or advantageous over other aspects or designs.

[0034] Hereinafter, identical or similar components are assigned the same reference numeral regardless of drawing symbols, and redundant descriptions thereof are omitted. Furthermore, in describing the embodiments disclosed in this specification, detailed descriptions of related prior art are omitted if it is determined that such detailed descriptions may obscure the essence of the embodiments disclosed in this specification. Additionally, the attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification, and the technical concept disclosed in this specification is not limited by the attached drawings.

[0035] Expressions such as "first," "second," etc. are used to describe various components, but these components are not limited by these terms. These terms are used merely to distinguish one component from another. Therefore, the first component mentioned below may be the second component within the technical scope of the present invention.

[0036] Unless otherwise defined, all terms used in this specification (including technical and scientific terms) may be used in a meaning that is commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise.

[0037] Furthermore, the term "or" is intended to mean an implicit "or" rather than an exclusive "or." That is, unless otherwise specified or evident from the context, "X uses A or B" is intended to mean one of the natural implicit substitutions. In other words, if X uses A; if X uses B; or if X uses both A and B, "X uses A or B" may apply to any of these cases. Additionally, the term "and / or" as used herein should be understood to refer to and include all possible combinations of one or more of the enumerated related items.

[0038] Additionally, the terms “comprising” and / or “comprising” should be understood to mean that such features and / or components are present, but not to exclude the presence or addition of one or more other features, components, and / or groups thereof. Furthermore, unless otherwise specified or clearly evident from the context to indicate a singular form, the singular in this specification and claims should generally be interpreted to mean “one or more.”

[0039] The purpose and effects of the present disclosure, and the technical configurations for achieving them, will become clear by referring to the embodiments described in detail below in conjunction with the accompanying drawings. In describing the present disclosure, if it is determined that a detailed description of known functions or configurations might unnecessarily obscure the essence of the present disclosure, such detailed description will be omitted. Furthermore, the terms described below are defined considering their functions in the present disclosure, and these may vary depending on the intentions or practices of the user or operator.

[0040] However, the present disclosure is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to make the present disclosure complete and to fully inform those skilled in the art of the scope of the disclosure, and the present disclosure is defined only by the scope of the claims. Therefore, such definitions should be based on the content throughout this specification.

[0041] Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Identical or similar components regardless of drawing symbols will be assigned the same reference number, and redundant descriptions thereof will be omitted. The suffix "bu" used for components in the following description is assigned or used interchangeably solely for the ease of drafting the specification and does not have a distinct meaning or role in itself. Furthermore, in describing the embodiments disclosed in this specification, if it is determined that a detailed description of related prior art could obscure the essence of the embodiments disclosed in this specification, such detailed description will be omitted. Additionally, the attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification; the technical concept disclosed in this specification is not limited by the attached drawings, and it should be understood that they include all modifications, equivalents, and substitutions that fall within the spirit and technical scope of the present invention.

[0042] Terms containing ordinal numbers, such as 1, 1-1, 1-2, 2, 2-1, 2-2, etc., may be used to describe various components, but said components are not limited by said terms. These terms are used solely for the purpose of distinguishing one component from another.

[0043] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to that other component, or that there may be other components in between. On the other hand, when it is stated that a component is "directly connected" to another component, it should be understood that there are no other components in between.

[0044] A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0045] FIG. 1 is a block diagram showing a computing device that generates a mental health diagnosis result from an individual's mental health input information according to one embodiment of the present disclosure.

[0046] Referring to FIG. 1, a computing device (100) that generates a mental health diagnosis result from an individual's mental health input information is disclosed.

[0047] In one embodiment, the computing device (100) may include a processor (110) that performs a method for generating a mental health diagnosis result and a storage unit (130) that stores data, etc. for performing a method for generating a mental health diagnosis result.

[0048] In one embodiment, the processor (110) may be composed of one or more cores. The processor (110) may receive and process data to generate a mental health diagnosis result from an individual's mental health input information, and may transmit the processed data to an external device (e.g., a user terminal). Additionally, the processor (110) may read a computer program stored in a storage unit (130) to generate a mental health diagnosis result from an individual's mental health input information according to one embodiment of the present disclosure.

[0049] In one embodiment, the processor (110) may include all types of software or hardware capable of processing the operation and data of the computing device (100). For example, the processor (110) may refer to a data processing device embedded in hardware having a physically structured circuit to perform a function expressed by code or instructions included in a program. Examples of such data processing devices embedded in hardware may include a microprocessor, a central processing unit (CPU), a processor core, a multiprocessor, an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), etc., but the scope of the present invention is not limited thereto.

[0050] Additionally, the processor (110) may be composed of one or more cores and may include any type of core for data processing, data analysis, and / or deep learning, such as a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), a neural processing unit (NPU), and a tensor processing unit (TPU) of the computing device (100). Additionally, the processor (110) may perform computations for the training of a neural network. The processor (110) may perform computations for the training of a neural network, such as processing input data for training in machine learning or deep learning (DL) (e.g., preprocessing such as noise removal or tokenization of input data), feature extraction from input data, error calculation, and updating the weights of the neural network using backpropagation. At least one of the CPU, GPGPU, and TPU of the processor (110) may process the training of the network function. For example, a CPU and a GPGPU can work together to generate inference results such as learning a network function, data classification using a network function, image processing, or mental health diagnosis results. Additionally, in one embodiment of the present disclosure, processors of a plurality of computing devices can be used together to generate inference results such as learning a network function, data classification using a network function, image processing, or mental health diagnosis results.

[0051] In one embodiment, the storage unit (130) may store information for providing user-customized mental health diagnosis results. For example, the storage unit (130) may store a program for providing user-customized mental health diagnosis results, user input, and personal mental health input information obtained from an external device (e.g., user terminal). In one embodiment, the storage unit (130) may store a pre-trained artificial intelligence model (e.g., learned parameters of the model, etc.) used to generate mental health diagnosis results. In one embodiment, the storage unit (130) may store information of any form generated or determined by the processor (110). For example, the storage unit (130) may store information and cache data for providing user-customized mental health diagnosis results. However, the functions and uses of the storage unit (130) are not limited to the examples listed.

[0052] In one embodiment, the storage unit (130) may include memory and / or a permanent storage medium. The memory may include at least one type of storage medium among flash memory type, hard disk type, multimedia card micro type, card type memory (e.g., SD or XD memory, etc.), RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), magnetic memory, magnetic disk, and optical disk, but the scope of the present invention is not limited thereto.

[0053] In one embodiment, the storage unit (130) may be included in another computing device (e.g., another server or other device) separate from the computing device (100). In this case, the computing device (100) may communicate with the other computing device to obtain desired data from the storage unit (130) included in the other computing device. For example, a server (not shown) providing user-customized mental health diagnosis results including the storage unit (130) may exist separately from the computing device (100). In one embodiment, the computing device (100) may obtain data necessary for performing the method according to the embodiments of the present disclosure from the server. Such a user interface may perform interactions with the user, such as outputting information to the user and receiving input from the user.

[0054] In one embodiment, the storage unit (130) of the computing device (100) stores an artificial intelligence model used to generate user-customized mental health diagnosis results, and the processor (110) can train and / or execute (e.g., infer) the artificial intelligence model.

[0055] In one embodiment, the computing device (100) may be used to encompass a server and / or user terminal.

[0056] In one embodiment, the computing device (100) may further include an input unit. The input unit may receive signals from a user that are necessary to provide user-customized mental health diagnosis results. The input unit may include a camera or video input unit for video signal input, a microphone or audio input unit for audio signal input, and a user input unit for receiving information from a user (e.g., a touch key, a mechanical key, etc.).

[0057] In one embodiment, the user input unit may receive information from the user (e.g., voice by user speech, text by user input, etc.). When information (e.g., self-report data, speech data, question-and-answer data for interaction items, etc.) is input through the user input unit, the computing device (100) may perform an operation corresponding to the input information. Such a user input unit may include mechanical input means (or mechanical keys, e.g., buttons, dome switches, jog wheels, jog switches, etc. located on the front, rear, or side of a mobile terminal) and touch input means. As an example, the touch input means may consist of a virtual key, soft key, or visual key displayed on a touchscreen through software processing, or a touch key placed on a part other than the touchscreen. Meanwhile, the virtual key or visual key may have various forms and be displayed on the touchscreen. For example, a virtual key or visual key can consist of a graphic, text, icon, video, or a combination thereof.

[0058] The term “user input” in the present disclosure may refer to any form of user input related to a user request performed within a user interface (or within a webpage). For example, user input may include user input that moves a pointer object. As another example, user input may include user input that selects a specific object on the user interface. For example, user input regarding an object may be made by touching or clicking any object (e.g., a module, a tab, etc.). When user input related to a selection is received, a new object may be displayed on the user interface or output unit in response to the input, or the attributes of the object may be changed and displayed.

[0059] As another example, user input may include user input to move or rotate the position of an arbitrary object within the user interface or output section, and / or user input to enlarge or reduce the size of the object. For instance, selection input such as dragging while clicking an arbitrary object can cause the object to move, rotate, or enlarge or reduce. User input is not limited to the examples mentioned above, and various forms of user actions are possible, such as mouse cursor control, mouse wheel scrolling, keyboard arrow keys, mouse clicks, and touch.

[0060] A computing device (100) according to one embodiment of the present disclosure may include an output unit. The output unit is intended to generate outputs related to sight, hearing, or touch, and may include at least one of a display unit, an audio output unit, a haptic module, and an optical output unit. The display unit may implement a touch screen by forming a layered structure with a touch sensor or by being formed integrally. Such a touch screen functions as a user input unit that provides an input interface between the computing device (100) and a user, and at the same time, may provide an output interface between users. For example, the output unit may output interaction questions and mental health diagnosis results provided to the user.

[0061] In one embodiment, the display unit visually outputs or displays the analysis results of a medical image. For example, the display unit may include at least one of a liquid crystal display (LCD), a thin film transistor-liquid crystal display (TFT LCD), an organic light-emitting diode (OLED), a flexible display, a 3D display, and an e-ink display. The display unit displays (outputs) data processed by the computing device (100). Specifically, the display unit may display (output) user-customized mental health diagnosis results.

[0062] In one embodiment, when the computing device (100) includes an input unit and an output unit, the computing device (100) may correspond to a user terminal.

[0063] The computer device (100) may have more or fewer components than the listed components to provide user-customized mental health diagnosis results.

[0064] FIG. 2 is a diagram showing an exemplary structure of an artificial intelligence-based model according to one embodiment of the present disclosure.

[0065] Throughout this specification, artificial intelligence model, artificial intelligence-based model, computational model, neural network, network function, and neural network may be used interchangeably. A neural network may consist of a set of interconnected computational units, which may generally be referred to as nodes. These nodes may also be referred to as neurons. A neural network is composed of at least one node. The nodes (or neurons) constituting the neural networks may be interconnected by one or more links.

[0066] In a neural network, one or more nodes connected via links can form relative input and output node relationships. The concepts of input and output nodes are relative; any node in an output node relationship with respect to one node may be in an input node relationship with respect to another node, and vice versa. As described above, the input node versus output node relationship can be generated based on links. One or more output nodes may be connected to a single input node via links, and vice versa.

[0067] In a relationship between an input node and an output node connected through a single link, the value of the output node's data can be determined based on the data input to the input node. Here, the link interconnecting the input node and the output node may have a weight. The weight can be variable and can be varied by the user or an algorithm to enable the neural network to perform the desired function. For example, if one or more input nodes are interconnected to a single output node by respective links, the output node's value can be determined based on the values ​​input to the input nodes connected to the output node and the weights set on the links corresponding to each input node.

[0068] As described above, a neural network consists of one or more nodes interconnected through one or more links, forming input-output node relationships within the network. The characteristics of a neural network can be determined by the number of nodes and links within the network, the relationships between the nodes and links, and the weight values ​​assigned to each link. For example, if two neural networks exist with the same number of nodes and links but different weight values ​​for the links, the two neural networks may be recognized as different from each other.

[0069] A neural network can be composed of a set of one or more nodes. A subset of nodes constituting a neural network can form a layer. Some of the nodes constituting a neural network can form a layer based on their distances from an initial input node. For example, a set of nodes with a distance of n from an initial input node can form n layers. The distance from the initial input node can be defined by the minimum number of links that must be traversed to reach that node from the initial input node. However, this definition of a layer is arbitrary for illustrative purposes, and the degree of a layer within a neural network can be defined in a way different from that described above. For example, a layer of nodes may be defined by its distance from a final output node.

[0070] Initial input nodes may refer to one or more nodes within a neural network to which data is directly input without passing through links in their relationships with other nodes. Alternatively, in terms of link-based relationships between nodes within the neural network, they may refer to nodes that do not have other input nodes connected by links. Similarly, final output nodes may refer to one or more nodes within a neural network that do not have output nodes in their relationships with other nodes. Furthermore, hidden nodes may refer to nodes constituting the neural network that are neither initial input nodes nor final output nodes.

[0071] A neural network according to one embodiment of the present disclosure may have the number of nodes in the input layer equal to the number of nodes in the output layer, and may be a neural network in which the number of nodes decreases and then increases again as it progresses from the input layer to the hidden layer. Additionally, a neural network according to another embodiment of the present disclosure may have the number of nodes in the input layer less than the number of nodes in the output layer, and may be a neural network in which the number of nodes decreases as it progresses from the input layer to the hidden layer. Additionally, a neural network according to yet another embodiment of the present disclosure may have the number of nodes in the input layer greater than the number of nodes in the output layer, and may be a neural network in which the number of nodes increases as it progresses from the input layer to the hidden layer. A neural network according to yet another embodiment of the present disclosure may be a neural network in which the above-described neural networks are combined.

[0072] A neural network that can be used in an artificial intelligence-based model of the present disclosure may be trained in at least one of supervised learning, unsupervised learning, semi-supervised learning, transfer learning, active learning, or reinforcement learning. Training of a neural network may be a process of applying knowledge to the neural network to perform a specific operation.

[0073] Neural networks can be trained to minimize the error in their output. The training process involves repeatedly inputting training data into the network, calculating the error between the network's output and the target for the training data, and updating the weights of each node by backpropagating the error from the output layer to the input layer in a direction that reduces the error. In the case of supervised learning, training data is used where the correct answer is labeled for each data point (i.e., labeled training data), whereas in the case of unsupervised learning, the correct answer may not be labeled for each training data point. For instance, in the case of supervised learning for data classification, the training data may consist of data where each training point is labeled with a category. The labeled training data is input into the neural network, and the error can be calculated by comparing the network's output (category) with the labels of the training data. As another example, in the case of unsupervised learning for data classification, the error can be calculated by comparing the input training data with the neural network's output. The calculated error is backpropagated in the neural network (i.e., from the output layer to the input layer), and through backpropagation, the connection weights of each node in each layer of the neural network can be updated. The amount of change in the connection weights of each node being updated can be determined by the learning rate. The neural network's calculation of the input data and the backpropagation of the error can constitute a learning cycle (epoch). The learning rate can be applied differently depending on the number of iterations of the neural network's learning cycle. For example, a high learning rate can be used in the early stages of training to quickly achieve a certain level of performance and increase efficiency, while a low learning rate can be used in the later stages to improve accuracy.

[0074] In the training of neural networks, the training data is generally a subset of the real-world data (i.e., the data intended to be processed by the trained neural network). Consequently, a training cycle may exist where errors decrease on the training data but increase on the real-world data. Overfitting is a phenomenon where the network learns excessively on the training data, leading to increased errors on the real-world data. For example, a neural network trained on yellow cats might fail to recognize cats when seeing anything other than yellow, which can be considered a type of overfitting. Overfitting can act as a cause for increased errors in machine learning algorithms. Various optimization methods can be used to prevent this overfitting. To prevent overfitting, methods such as increasing the training data, regularization, dropout (which disables some nodes in the network during training), and the use of batch normalization layers can be applied.

[0075] A deep neural network (DNN) may refer to a neural network that includes multiple hidden layers in addition to input and output layers. Using a deep neural network allows for the identification of latent structures in data. That is, it is possible to identify the latent structures of photos, text, videos, voice, and music (e.g., what objects are present in a photo, what the content and emotions of the text are, what the content and emotions of the voice are, etc.). Deep neural networks may include convolutional neural networks (CNN), recurrent neural networks (RNN), autoencoders, restricted Boltzmann machines (RBM), deep belief networks (DBN), Q networks, U networks, Siamese networks, Generative Adversarial Networks (GAN), etc. The description of deep neural networks described above is merely illustrative and the present disclosure is not limited thereto.

[0076] In the present disclosure, a report generation model may refer to a program that takes a predefined medical image as input, diagnoses the progression of a lesion such as the disease name, location of the lesion, and size of the lesion, and outputs the corresponding content as a report. For example, the artificial intelligence model may include a Large Language Model (LM) model capable of performing text-based tasks to understand and generate text by learning pre-learned language capabilities. For example, the report generation model may include a plurality of LLM models.

[0077] An artificial intelligence model according to one embodiment of the present disclosure may include a Large Language Model (LLM). In the present disclosure, a Large Language Model may refer to an artificial intelligence model trained using a vast amount of training data to perform Natural Language Processing. The Large Language Model may include a transformer, a transformer encoder family model, and / or a transformer decoder family model. A transformer encoder family model may correspond to an artificial intelligence model that uses a transformer encoder structure. A transformer decoder family model may correspond to an artificial intelligence model that uses a transformer decoder structure.

[0078] In one embodiment, a large-scale language model can process various data formats, such as natural language text, image data, audio data, and video data. The large-scale language model can embed data to convert data of various data formats into a series of computationally operable data. The large-scale language model can process additional data that represents the relative positional or topological relationships between a series of input data. Alternatively, a series of input data may be embedded by additionally reflecting vectors that represent the relative positional or topological relationships between the input data. In one example, the relative positional relationships between a series of input data may include, but are not limited to, word order within a natural language sentence, the relative positional relationships of each segmented image, and the temporal order of segmented audio waveforms. The process of adding information that represents the relative positional or topological relationships between a series of input data may be referred to as positional encoding.

[0079] FIG. 3 is a flowchart illustrating a method for generating a mental health diagnosis result from an individual's mental health input information according to one embodiment of the present disclosure.

[0080] Referring to FIG. 3, a computing device (100) can receive mental health input information from a user (310) to generate a mental health diagnosis result from the individual's mental health input information.

[0081] In one embodiment, mental health input information may refer to initial data collected to analyze a user's mental health status in a multi-layered manner. Mental health input information may include user identification information, question-and-answer data obtained from the user, and self-report data obtained from the user.

[0082] In one embodiment, user identification information may include age, gender, occupation, and region as information for categorizing users within a population group.

[0083] In one embodiment, user identification information may be used to establish a disease baseline representing the user's basic probability value for each of a plurality of mental disorders. For example, a disease baseline may refer to a personalized initial probability value representing the basic probability of occurrence of each mental health disorder for a user within a specific population group, independent of the user's individual symptom reports. The value in the disease baseline may not be a fixed value applied equally to all users, but may be personalized by taking into account demographic characteristics such as the user's age, gender, and region. The disease baseline may serve as a basic probability value or reference point in generating a list of mental disorder candidates, target mental disorders, validation results for target mental disorders, and / or mental health diagnosis results from mental health input information.

[0084] In one embodiment, question-and-answer data may be used to correct the user's baseline probability value as user response data to a query regarding whether the user has experienced symptoms (e.g., standardized questions on an initial medical history questionnaire). For example, the question-and-answer data may include user response data to a screening question asking whether the user has experienced major symptoms (e.g., 'O' / 'X' responses to 'Do you usually experience mood swings?'). The question-and-answer data is utilized to primarily correct the baseline probability value of the disease.

[0085] In one embodiment, self-report data may include at least one user speech data, among voice and text, related to the user's condition and symptoms. For example, self-report data may refer to data in which the user directly describes their current condition and symptoms in a free-form voice or text.

[0086] In one embodiment, self-reporting data can be used as input to a plurality of artificial intelligence models. Self-reporting data can be used as a common input to a plurality of artificial intelligence models described below.

[0087] For example, if a 32-year-old male user (e.g., user identification information) responds with 'O' to symptoms of 'mood swings' and 'fatigue / lethargy' (e.g., question-and-answer data) and says aloud, "Um... I just feel a bit strange lately. When I wake up in the morning, my body feels heavy... and I keep feeling depressed" (e.g., self-report data), such information may be included in the mental health input information described above.

[0088] In one embodiment, the computing device (100) can use a plurality of artificial intelligence models to generate a list of suspected mental illness candidates for the user from mental health input information (320). In one embodiment, the plurality of artificial intelligence models may be included in a mental illness candidate list generation module that generates a list of mental illness candidates.

[0089] The artificial intelligence models in the present disclosure may refer to multiple (e.g., three) different artificial intelligence models that receive and analyze user self-report data and generate key metadata necessary for deriving suspected disease candidate groups. For a specific description of the artificial intelligence models, refer to the description in FIG. 2.

[0090] For example, metadata may refer to structured secondary data extracted or generated by artificial intelligence models and analysis modules using self-reported text or voice, which is raw data input by a user. This metadata converts the user's subjective descriptions into objective and quantitative indicators and can be utilized as key evidence in subsequent stages to calculate the probability of a suspected disease, correct diagnostic results, and / or compose a final report. In this disclosure, multi-layered metadata may be generated at each stage (e.g., each stage from screening diagnosis to functional testing). For example, structured data of symptoms generated from the user's self-report data by a first artificial intelligence model may be included within the scope of the metadata. For example, data representing the probability distribution of multiple (e.g., seven) emotional states by analyzing the user's voice data by a second artificial intelligence model may be included within the scope of the metadata. For example, language pattern keywords extracted from linguistic habits or patterns associated with a specific disease, in addition to direct symptom descriptions from the user's utterance, may be included within the scope of the metadata.

[0091] In one embodiment, a plurality of artificial intelligence models may include a first artificial intelligence model that generates information on a user's mental health-related symptoms using self-report data included in mental health input information. For example, the first artificial intelligence model is a Gemma-27B-based model that can analyze self-report data (e.g., text) to extract clinically standardized information on mental health-related symptoms in a structured form. For example, the first artificial intelligence model can analyze the user's utterance "My head felt foggy, so I just did other things" and output symptom tags in JSON format such as {"label": "Attention problem", "present": true}.

[0092] In one embodiment, symptom information may refer to symptoms related to a mental illness. For example, symptom information may include mood swings, irritability / hypersensitivity, fatigue / lethargy, suicidal thoughts, decreased appetite, attention problems, memory problems, excessive worry, panic attacks, somatic anxiety, alcohol, insomnia, low motivation, guilt, functional decline, and depression.

[0093] In one embodiment, the computing device (100) may obtain sentences related to mental illness in advance from self-report data. For example, the pre-collected sentences may include sentences corresponding to major symptoms of each disease selected by a clinical expert analyzing each utterance. For example, the pre-collected sentences may include actual counseling data stored in a sub.

[0094] In one embodiment, the first artificial intelligence model may be trained by a first training dataset composed of sentences and symptoms pairs, comprising first synthetic data generated by inputting first seed data, in which at least one of the corresponding symptom types is labeled for a pre-collected sentence, into a large-scale language model. For example, the first artificial intelligence model may be trained by a training dataset of sentence-symptom pairs constructed by augmenting seed data labeled symptom types according to DSM-5 criteria through LLM.

[0095] In one embodiment, the first seed data may include data labeled with disease names and / or symptoms (e.g., depression, lethargy, panic attacks, etc.) corresponding to pre-collected sentences.

[0096] In one embodiment, the first seed data may be labeled based on DSM-5 diagnostic logic.

[0097] In one embodiment, the first synthetic data generated by inputting the first seed data into a large-scale language model may include utterances reflecting various expression styles, tones, and situational contexts that the large-scale language model automatically generates by utilizing pre-collected sentences. In this case, the first synthetic data may include diversified sentences of the pre-collected sentences while maintaining symptom-specific label information. In one embodiment, since the first synthetic data includes diversified counseling scenarios, it may include synthetic data that closely approximates an actual counseling environment.

[0098] In one embodiment, the first synthetic dataset may be reviewed by a clinical expert. For example, the clinical expert comprehensively reviews the disease suitability of each sentence, the validity of symptom expression, and the feasibility of actual clinical application, and may remove or modify problematic samples.

[0099] In one embodiment, the first synthetic dataset may be used for training a mental illness candidate list generation module, and in addition to the mental illness candidate list generation module, it may also be used for training in the disease identification module, validity review module, and function test module used in this disclosure. In one embodiment, the disease identification module, validity review module, and function test module trained using the first synthetic dataset may generate various counseling scenarios, etc., to generate questions that closely approximate an actual counseling environment.

[0100] In one embodiment, the first synthetic dataset reflects both actual clinical context and natural language expression, and can be used as a core resource to enhance the performance of a digital mental health system.

[0101] The specific operation methods of the remaining modules will be explained with reference to the drawings below.

[0102] In one embodiment, a plurality of artificial intelligence models may include a second artificial intelligence model that generates a classification result of a user's emotional state using self-report data. For example, the second artificial intelligence model is a model based on Whisper 1.5B, which can analyze self-report data (voice) to classify the user's emotional state into a probability distribution form for multiple categories such as 'joy', 'sadness', and 'anger'. For example, the second artificial intelligence model can analyze the user's voice and output a result in JSON format indicating that a specific emotion is dominant, such as {"label": "anger", "probability": 0.80}.

[0103] In one embodiment, the second AI model may be trained by a second training dataset composed of voice and emotion pairs using second seed data comprising voice data with pre-attached emotion labels and text transcripts corresponding to the voice data. For example, the second AI model may be trained by selecting data associated with mental health assessments from the voice emotion dataset, verifying label consistency, and then fine-tuning the Whisper model.

[0104] In one embodiment, the emotional state classification result may represent the user's emotional state corresponding to the user's voice data and / or text entered by the user. For example, the emotional state classification result may include 'joy', 'loveliness', 'fear', 'anger', 'sadness', 'surprise', and 'none'.

[0105] In one embodiment, the second seed data may be data extracted according to predefined criteria, consisting only of emotion categories highly associated with mental health assessment from a voice call-based emotion dataset provided by a server that pre-stores various speech data. In one embodiment, the emotion categories may be limited to a range that can contribute to mental health judgments based on the advice of clinical experts.

[0106] In one embodiment, the computing device (100) may perform an inspection according to internal criteria to determine whether the emotion label attached to the extracted voice data corresponds to the actual voice content and context. This process may include reviewing whether there is a semantic match between the text transcript and the voice emotion label, heterogeneous speech patterns, or labeling errors, and may include removing or reclassifying data suspected of having errors.

[0107] In one embodiment, the second AI model may be trained with the primary goal of improving the accuracy of speech-based emotion recognition by utilizing refined emotion-labeled speech data. In one embodiment, the second AI model may be trained using a full fine-tuning method on a large-scale multitask model specialized in speech-to-text conversion and speech-based classification (e.g., OpenAI Whisper 1.5B model). In one embodiment, the second AI model may be constructed as a mental health-specialized speech emotion model by maintaining the basic structure of a predefined large-scale multitask model while readjusting it with parameters optimized for emotion classification.

[0108] In one embodiment, a large-scale multitasking model is an automatic speech recognition system that converts speech into text and can be built based on a transformer. The large-scale multitasking model may largely consist of an encoder part that understands audio and a decoder part that generates text.

[0109] In one embodiment, a large-scale multitasking model can perform not only speech recognition but also various tasks simultaneously (e.g., identifying which language is being spoken or distinguishing between sections with and without speech).

[0110] In one embodiment, a plurality of artificial intelligence models may include a third artificial intelligence model that takes text data as input and generates a user's language pattern keyword. For example, the third artificial intelligence model may be a model that extracts unique language pattern keywords associated with a specific disease, rather than direct symptom expressions, from the user's self-report text. For example, the third artificial intelligence model may extract non-symptomatic language habits in JSON format, such as {"label": "filler / hesitation"} from the utterance "Hmm...... I'm not sure what to say..." or {"label": "repetition / continuous emphasis"} from the utterance "I keep feeling depressed."

[0111] In one embodiment, the third artificial intelligence model may be trained by a third training dataset consisting of sentence data containing language pattern keywords and mental illness pairs, generated by inputting third seed data constructed using pre-collected representative language pattern keywords for each mental illness into a large-scale language model (LLM). For example, the third artificial intelligence model may be trained by a dataset of keyword-containing sentence-disease pairs generated by extending representative symptom keywords for each disease using an LLM.

[0112] In one embodiment, language pattern keywords may refer to linguistic features and speaking habits revealed in a user's utterance. For example, language pattern keywords may include filler phrases / hesitations, repetition, sustained emphasis, avoidance / delayed discourse, rumination / regretful narratives, and negative self-evaluations.

[0113] In one embodiment, the third seed data may include representative symptom keywords and similar expressions collected based on authoritative medical websites, diagnostic criteria (DSM-5), counseling cases, etc., for major mental illnesses such as depression, panic disorder, anxiety, and sleep disorders. In one embodiment, the third seed data may be refined into core keywords highly relevant to the diagnosis through review by clinical experts.

[0114] In one embodiment, the third training dataset may be generated based on various contextually similar expressions extended from the third seed data, utterance example data reflecting linguistic diversity, etc., by utilizing a large-scale language model (e.g., Gemma-3-27B).

[0115] In one embodiment, a large-scale language model can generate a large dataset of synthetic sentences containing disease-specific keywords. The generated sentences are automatically saved with disease labels attached, and subsequently, sample verification and training reliability are ensured by clinical experts, and error correction can be performed on the generated sentences.

[0116] In one embodiment, the computing device (100) can finally construct a third training dataset of refined sentence-disease label pairs including disease-specific keywords, expressions, and contexts.

[0117] In one embodiment, the third training dataset can be used for fine-tuning a large-scale language model and for training to classify keywords.

[0118] In one embodiment, a computing device (100) may generate a list of candidate mental illnesses using artificial intelligence models. The list of candidate mental illnesses may include candidate mental illnesses that the user is relatively likely to suffer from (above a certain threshold level) based on the user's mental health input information (e.g., self-report data, etc.). The list of candidate mental illnesses may be displayed sorted in order of likelihood. A specific method for generating the list of candidate mental illnesses is described with reference to the drawings below. For example, the list of candidate mental illnesses may be a product of a screening diagnostic process utilizing artificial intelligence models, and may refer to a list compressed into multiple (e.g., 2 to 5) suspected illnesses that the user is highly likely to suffer from. The list of candidate mental illnesses may be derived based on a final probability value generated by sequentially (or in parallel, depending on the implementation mode) reflecting and adjusting the structured response and the analysis results of multiple (e.g., 3) artificial intelligence models, starting from a disease baseline. The list of candidate mental illnesses may be subject to an in-depth examination to determine a target mental illness.

[0119] In one embodiment, the computing device (100) may generate a first adjustment value by applying the output of a first artificial intelligence model on a disease baseline, change the first adjustment value to a second adjustment value using the output of a second artificial intelligence model, and change the second adjustment value to a third adjustment value using the output of a third artificial intelligence model. Based on the third adjustment value, the computing device (100) may generate a list of mental disease candidates.

[0120] In one embodiment, when the computing device (100) integrates the analysis results of artificial intelligence models and derives probabilities such as depression 98.6%, ADHD 67.3%, and bipolar disorder 70.8% for the user, it may finally apply clinical rules (e.g., including depression when bipolar disorder is suspected) to include 'bipolar disorder' and 'ADHD' in the list of candidate mental disorders.

[0121] In one embodiment, the computing device (100) can determine the target mental illness that the user is suffering from among the mental illness candidate list by using a disease determination module that generates a first user interaction question to check for the presence or absence of symptoms for each of the candidate mental illnesses in the mental illness candidate list (330).

[0122] In the present disclosure, user interaction items may refer to query information configured to be automatically generated by a computing device (100) and to perform interaction with a user. For example, user interaction items may collectively refer to questions generated and presented by the computing device (100) to obtain specific information from the user at each stage of identifying suspected diseases (in-depth examination), verifying (validation review), and / or evaluating (functional examination) after screening diagnosis using artificial intelligence models. In one embodiment, the first user interaction item may include items for identifying the user's symptoms, condition, etc. related to the mental disease, which are generated for each mental disease included in the mental disease candidate list. For example, the first user interaction item is a question used in the disease identification module to determine whether the diagnostic criteria for each disease listed in the mental disease candidate list are met. For example, the first user interaction item may be presented in a responsive structure that first checks the core prerequisites for each disease and determines whether to present subsequent items based on the response. For example, if the candidate disorder is 'major depressive disorder,' to verify the core diagnostic condition, an item such as "Have you felt depressed (sadness, emptiness, hopelessness) for most of the day, almost every day for the past two weeks, or has your interest or pleasure in almost all daily activities decreased significantly?" may be generated as a first user interaction item.

[0123] In one embodiment, the first user interaction question may be automatically generated by the computing device (100) based on a list of mental illness candidates. In one embodiment, the first user interaction question may include a plurality of sub-interaction questions. For example, when the first-1 user interaction question is generated based on the list of mental illness candidates, the computing device (100) may generate the first-2 user interaction question based on the user response to the first-1 user interaction question. For example, if depression is in the list of mental illness candidates, the first-1 user interaction question may be "Have you often missed appointment times or deadlines?", and if the user response to the first-1 user interaction question is "Yes," the computing device (100) generates the first-2 user interaction question "Do you find it difficult to focus on one thing at a time?", and if the user response to this is "No," the computing device (100) may not generate a subsequent question.

[0124] In one embodiment, the disease identification module may generate a first-1 user interaction item among a plurality of user interaction items corresponding to a first candidate mental illness included in a list of mental illness candidates, and determine whether to generate a first-2 user interaction item based on whether the user response to the first-1 user interaction item satisfies the condition corresponding to the first candidate mental illness. For example, the first-2 user interaction item may be an item that includes a follow-up question to the first-1 user interaction item. For instance, if the first-1 user interaction item is "Do you have mood swings?" and the user's response is "Yes," the first-2 user interaction item may adaptively, dynamically, or responsively become "Do you feel depressed?"

[0125] In one embodiment, the disease identification module can determine at least one target mental disorder among the candidate mental disorders by generating an independent first user interaction item for each of the candidate mental disorders and scoring the user response to the first user interaction item. For example, the target mental disorder may refer to a major suspected disorder among the mental disorders included in the list of candidate mental disorders derived from the screening diagnosis that is determined to satisfy the core conditions of the accredited diagnostic criteria through an in-depth examination (e.g., the operation of the disease identification module). For example, the target mental disorder may refer to a central disorder that is subject to subsequent validation, such as a T-Check and / or D-Check. Assume that 'depression' and 'ADHD' were derived as a candidate mental disorder from the screening diagnosis results. Subsequently, when a user responds to the first user interaction item presented by the disease identification module during the in-depth examination (e.g., the core prerequisite item for depression), the diagnostic criteria for depression are met, but the criteria for ADHD are not met. In this case, depression may be determined as the target mental illness to proceed to the feasibility review stage.

[0126] In one embodiment, the computing device (100) can generate a verification result for a target mental illness by using a validity review module that generates a second user interaction question to verify the temporal validity of the target mental illness (340).

[0127] In one embodiment, the validity review module may generate a second user interaction item that queries symptom time conditions, including the time of symptom onset, duration, and frequency for the target mental illness determined by the disease identification module.

[0128] For example, the second user interaction item may be generated by the computing device (100) as an item used for a validity review module (e.g., T-Check (temporal validity review)). The second user interaction item may refer to a question to verify whether temporal conditions, such as the time of onset, duration, and frequency of symptoms of a determined target mental disorder, meet the diagnostic criteria. For example, if the target disorder is determined to be ADHD, an item such as "Were the symptoms of attention deficit present since kindergarten or elementary school (before age 12)?" may be presented as a second user interaction item to confirm the time of onset of the symptoms.

[0129] In one embodiment, when a target mental illness is determined, the computing device (100) may generate a second user interaction question. In one embodiment, the second user interaction question may include a plurality of sub-interaction questions. For example, when a second-1 user interaction question is generated based on the target mental illness, the computing device (100) may generate a second-2 user interaction question adaptively, dynamically, or responsively based on the user response to the second-1 user interaction question. For example, if the target mental illness is ADHD, the second-1 user interaction question may be "Have you felt depressed recently?", and if the user response to this is "Yes," the computing device (100) generates the second-2 user interaction question "Has the feeling of depression lasted for more than 2 weeks?", and if the user response to the second-1 user interaction question is "No," the computing device (100) may not generate a subsequent question.

[0130] In one embodiment, the validity review module can determine whether the user response to the second user interaction item satisfies the target time condition pre-assigned to the target mental illness.

[0131] In one embodiment, the target time condition may refer to a time condition required by diagnostic logic such as DSM (e.g., lasting for more than 2 weeks, occurring within 1 month, etc.).

[0132] In one embodiment, temporal validity (e.g., temporal validity for the target mental disorder) is a verification procedure performed by the validity review module through a T-Check, and may refer to whether the symptoms of the determined target mental disorder satisfy the essential temporal conditions required by an accredited diagnostic system, such as DSM-5 (e.g., duration of symptoms, time of onset, and / or frequency of occurrence). The disorder can be considered clinically significant only when temporal validity is secured. For example, if the target mental disorder is determined to be 'ADHD', the validity review module may generate a question as a second user interaction item, "Were these symptoms of attention deficit present before the age of 12?" to verify temporal validity. If the user responds 'Yes', the 'time of onset' condition essential for diagnosing ADHD is satisfied, and temporal validity can be secured. If the user responds 'No', temporal validity is not satisfied, so the disorder may be excluded from the final confirmed diagnosis and classified as a reference disorder.

[0133] In one embodiment, the feasibility review module can generate a verification result for a target mental illness based on whether the target time condition is satisfied.

[0134] In one embodiment, the validity review module can determine the mental illness grade for the target mental illness based on the time value related to the user's symptom time condition obtained through the second user interaction question when the user response to the second user interaction question satisfies the target time condition.

[0135] In one embodiment, the validity review module may set the target mental illness as a reference mental illness if the user response to the second user interaction item does not satisfy the target time condition.

[0136] In one embodiment, the validity review module may generate a third user interaction question to review the user's medication and underlying disease if the user's response to the second user interaction question satisfies the target time condition.

[0137] In one embodiment, the validity review module may determine whether to include the target mental illness in the verification results as a confirmed mental illness based on the user's response to the third user interaction question. For example, the third user interaction question may be generated by a computing device (100) as a question used for the validity review module (e.g., D-Check (drug / disease validity review)). The third user interaction question may be a question to confirm the possibility that the current symptoms were caused by secondary factors such as the user's medication, underlying physical disease, or specific lifestyle habits. For example, when performing the operation of the validity review module (e.g., D-Check), a question such as "Do you have a currently diagnosed physical disease?" or "Do you have any medication currently being taken?" may be presented as the third user interaction question to confirm the underlying disease.

[0138] In one embodiment, when the target mental illness satisfies the target time condition, the computing device (100) may generate a third user interaction question based on the target mental illness. In one embodiment, the third user interaction question may include a plurality of sub-interaction questions. Based on the user's response to a specific sub-interaction question, the next sub-interaction question may be generated adaptively, responsively, or dynamically. For example, if the user includes information that the depression has persisted for more than two weeks in response to the third user interaction question, the computing device (100) may determine that the user has depression and generate a corresponding third-1 user interaction question. Then, based on the user's response to the third-1 user interaction question, the computing device (100) may generate a third-2 user interaction question. For example, the 3-1 user interaction question may be "Do you drink alcohol three or more times a week?" and if the user response to this is "Yes," the computing device (100) generates the 3-2 user interaction question "What is the type and amount of alcohol you drink at one time?" and if the user response to the 3-1 user interaction question is "No," the computing device (100) may not generate a subsequent question.

[0139] In one embodiment, the validity review module can review whether the current symptoms are caused by secondary factors such as the user's medication, underlying diseases, or environmental factors, based on the user's response to the third user interaction question.

[0140] In one embodiment, the computing device (100) can generate a user's mental health diagnosis result based on a verification result for a target mental illness (350).

[0141] In one embodiment, the mental health diagnosis results may present the diagnosis results of a mental illness (e.g., grade of mental illness, validity), functional risk, need for management, functional decline, prevention recommendations, and future monitoring recommendations.

[0142] In one embodiment, the computing device (100) may generate a fourth user interaction item by a validity review module, which includes a question for measuring the stress level of the confirmed mental illness and a question for testing functions related to the user's daily life, when the verification result includes a confirmed mental illness. In one embodiment, the fourth user interaction item may be generated by a function test module. In one embodiment, the fourth user interaction item may be generated by the computing device (100) as a question used in a function test (F-Check). The fourth user interaction item may include questions for quantitatively evaluating the impact of the confirmed mental illness on the user's actual life functions and the related stress level. To assess the level of functional decline, items such as "Are you experiencing difficulties with job (work, study) functioning due to the symptoms you are currently suffering from?" may be presented through fourth-user interaction items, and to identify stress factors, disease-related items such as "Are you experiencing stress due to recent economic or interpersonal problems?" may be presented as fourth-user interaction items to users confirmed to have depressive disorder.

[0143] In one embodiment, if a confirmed mental illness is determined during the preceding question-and-answer process, the computing device (100) may generate a fourth user interaction question for the confirmed mental illness (e.g., based on the confirmed mental illness). In one embodiment, the fourth user interaction question may include a plurality of sub-interaction questions. For example, if the confirmed mental illness is depression, the computing device (100) may generate a fourth-1 user interaction question corresponding to depression. And, based on the user response to the fourth-1 user interaction question, the computing device (100) may adaptively (e.g., responsively) generate a fourth-2 user interaction question. For example, the 4-1 user interaction question may be "Do you have trouble waking up in the morning?", and if the user response to this is "Yes," the computing device (100) generates the 4-2 user interaction question "Does it affect your daily life or work?", and if the user response to the 4-1 user interaction question is "No," the computing device (100) may not generate a subsequent question.

[0144] In one embodiment, the computing device (100) can determine the stress level and the risk to the user's daily life based on the response to the fourth user interaction question.

[0145] In one embodiment, the mental health diagnosis result may include mental health input information, a list of mental illness candidates, a target mental illness and a confirmed mental illness included in the verification result, a mental illness grade for the confirmed mental illness, a reference mental illness among the target mental illnesses that is not classified as a confirmed mental illness, a user's emotional state classification result, and the user's language pattern keywords.

[0146] FIG. 4 is a diagram showing modules that generate mental health diagnosis results from individual mental health input information according to one embodiment of the present disclosure.

[0147] In one embodiment, the computing device (100) receives mental health input information (40) and can generate a mental illness candidate list (41) from the mental illness candidate list generation module (410) using the mental illness candidate list generation module (410).

[0148] In one embodiment, the mental illness candidate list generation module (410) can collect screening diagnosis and metadata (symptom information, emotional state classification results and language pattern keywords) that are used in subsequent stages, and explore the possibility of illness based on the user's mental health input information (40) as an initial process of mental health diagnosis.

[0149] In one embodiment, the computing device (100) can generate a target mental illness (42) from a disease identification module (420) using a mental illness candidate list (41) generated from a mental illness candidate list generation module (410).

[0150] In one embodiment, the computing device (100) can generate a confirmed mental illness (43) from a validity review module (430) using a target mental illness (42) generated from a disease identification module (420).

[0151] In one embodiment, the computing device (100) may generate a stress level and a risk level (44) from a functional test module (440) utilizing a target mental illness (42) generated from a disease identification module (420) and a confirmed mental illness (43) generated from a validity review module (430). For example, the risk level may refer to an indicator that comprehensively evaluates the likelihood of future mental health problems occurring or worsening, regardless of the presence or absence of the disease, as well as the current severity of the confirmed mental illness. For example, the risk level may be used as an indicator representing the stress level and the correlation and / or level of functional decline, going beyond simply judging based on the presence or number of symptoms. The stress level and / or correlation may be used to predict the likelihood of future symptom exacerbation or recurrence by evaluating stress factors closely associated with a specific mental illness. The level of functional decline may reflect the realistic severity of the diagnosis result by quantitatively evaluating the impact of the mental health problem on the user's actual job, daily life, or interpersonal relationships. This risk level explains the current diagnostic status related to mental illness and can be used as a basis for subsequent judgments to predict future changes in status and determine the need for preemptive management.

[0152] In one embodiment, the computing device (100) can generate a mental health diagnosis result (45) from a diagnosis result generation module (450) that uses the stress level and risk level generated from a function test module (440).

[0153] For example, it is assumed that a user has been provisionally diagnosed with mild depressive disorder through in-depth examination and validity review. It is assumed that, as a result of evaluating the level of stress and functional decline in a functional test (F-Check), it is confirmed that the user has very high job stress and that their ability to perform actual work and concentration have significantly declined due to depressive symptoms. The computing device (100) may determine that, even if the symptoms themselves are mild, the impact on actual life functioning is significant, and thus assess the overall risk level as 'high'. Based on this result, the final report may include the diagnosis of mild depressive disorder, increase the severity level to reflect the functional decline, and generate specific recommendations requiring immediate stress management and improvement of the work environment.

[0154] For example, assume that a user has completed a validity review but has concluded that there is no confirmed illness because they do not fully meet the diagnostic criteria for a specific mental illness. Assume that a functional test (F-Check) was conducted and it was confirmed that the user is experiencing severe stress in interpersonal relationships, which has led to social withdrawal and a significant decline in vitality in daily life. In this case, the computing device (100) may determine that there is a high potential risk of future mental health problems, as high stress levels and a distinct decline in function are observed, even though there is no confirmed illness. The final report generated by the computing device (100) may specify the current stress situation and the state of functional decline, provide guidance on the possibility of future deterioration, and include specific comments on preemptive care, such as preventive management, suggestions for stress relief strategies, and / or additional condition monitoring, regardless of the reason for the lack of a diagnosis.

[0155] Through the modules mentioned in the present disclosure, the process of completing an initial medical questionnaire and stating subjective symptoms during actual hospital treatment can be structured to suit a digital environment and automated through analysis based on an artificial intelligence model.

[0156] Existing mental health assessment systems utilize only standardized fixed items, failing to reflect the individual symptom characteristics of users; do not collect self-reports separately, resulting in results skewed toward quantitative data; and omit 'verbal subjective symptom descriptions' that occur during actual hospital consultations, making more precise disease classification and metadata generation impossible. These issues can be resolved through methods performed by the modules mentioned in this disclosure during the initial hospital consultation process.

[0157] FIG. 5 is a diagram illustrating a method in which a mental illness candidate list generation module according to one embodiment of the present disclosure generates a mental illness candidate list.

[0158] In one embodiment, the computing device (100) can set a disease baseline (51) using mental health input information (50).

[0159] In one embodiment, the disease baseline (51) may be set to reflect the basic probability of occurrence of each disease within a population group, regardless of the user's symptom report. The disease baseline (51) is a personalized initial probability value that takes into account group characteristics such as age, gender, and region, and can serve as a stable reference point when correcting the probability through structured responses, self-reports, emotional state, and keyword analysis. The disease baseline (51) should not be a fixed value, but should be managed as a dynamic reference value that is periodically updated based on the latest epidemiological statistics and actual user data, thereby continuously ensuring the accuracy and clinical reliability of the screening process.

[0160] In one embodiment, the disease baseline (51) may include Table 1 below.

[0161] Baselines by disease (Table of 1-year prevalence probabilities for the 30–35 year old male group) Bipolar melancholia ADHD Generalized anxiety disorder Panic disorder Alcohol sleep disorders base-line 0.028 0.095 0.068 0.031 0.0273 0.071 0.063

[0162] In one embodiment, Table 1 may include disease-specific baselines corresponding to men aged 30 to 35.

[0163] In one embodiment, the computing device (100) can obtain question-and-answer data (52) obtained from the user and the user's self-report data / utterance data (53).

[0164] In one embodiment, the question-and-answer data (52) may be responses to questions that simply identify the user's condition through screening questions at the symptom level rather than questions by mental illness. Through the question-and-answer data (52), the computing device (100) can collect only whether the user has experienced the corresponding symptom, rather than detailed information such as duration, frequency, and intensity.

[0165] In one embodiment, the question-and-answer data (52) may be as shown below (Table 2), for example.

[0166] number symptoms question User response 1 mood swings Do you tend to have mood swings? O 2 Sensitive / Overly sensitive Are you the type to get easily hurt or be overly sensitive over small things? X 3 Fatigue / Lethargy Do you experience fatigue or lethargy? O 4 Suicide Have you ever thought about wanting to die? X 5 Loss of appetite Have you lost your appetite or weight compared to usual? X 6 attention problems Do you find it difficult to maintain focus during work? O 7 memory problems Do you often feel that your memory is poor? X 8 Excessive worry Do you find it difficult to stop or control various worries or anxieties on your own? X 9 melancholy How often do you feel depressed? O

[0167] In one embodiment, Table 2 may include questions and responses corresponding to symptoms of mental illness.

[0168] In one embodiment, when symptom information such as mood swings, fatigue / lethargy, attention problems, and depression is selected through question-and-answer data (52), the selected information can be integrated with symptom information derived by artificial intelligence models (411, 412, 413) based on the user's self-report data (53) and used to readjust the probability value for each mental illness.

[0169] In one embodiment, through the question-and-answer data (52), the present invention presents questions in symptom units, compared to a method of presenting all questions by disease, thereby reducing the number of questions and minimizing user fatigue while ensuring sufficient screening information. Furthermore, the collected symptom data is linked with a pre-established symptom-disease weighting mapping table and utilized to statistically adjust the probability of occurrence of each disease. Consequently, the structured response is designed to enable initial probability correction for multiple diseases with only a simple selection by the user during the screening diagnosis process, thereby providing an efficient and user-friendly diagnostic experience.

[0170] In one embodiment, when symptom information such as mood swings, fatigue, lethargy, attention problems, and depression is determined through a structured response, the information can subsequently be integrated with symptom information derived by an artificial intelligence model (e.g., a first artificial intelligence model) based on the user's self-report data and used to readjust the probability values ​​of each disease.

[0171] In one embodiment, the computing device (100) may use the set disease baseline (51) and question-and-answer data (52) to perform the following steps.

[0172] In one embodiment, the computing device (100) can determine first candidate mental illnesses (510) that are likely to occur within a user's category (e.g., may correspond to a population group considering group characteristics such as age, gender, and region) by using user identification information included in mental health input information. The first candidate mental illnesses (510) can be determined using mental health-related symptom information (41) generated from a first artificial intelligence model (411) that takes user self-report data and / or speech data (53) included in mental health input information as input.

[0173] In one embodiment, self-report data and / or speech data (53) may include information in which the user freely expresses their current condition and symptoms as if having an interview with a doctor at a hospital. In one embodiment, the computing device (100) can use the self-report data and / or speech data (53) to obtain the user's subjective experience, nuances, and hidden symptom clues that are difficult to capture with screening questions alone, thereby increasing the precision and clinical validity of the diagnosis of suspected diseases.

[0174] For example, self-report data and / or speech data (53) can be exemplified as: “Um... I don’t really know what to say, but I just feel a bit strange lately. When I wake up in the morning, my body feels heavy, and even though nothing particularly happened, I keep feeling depressed. I have no energy all day, and I just don’t want to get out of bed. I do force myself to come out, but when I actually try to do something, I can’t get a grip... Yesterday, I sat at my desk, but after reading two or three lines, my mind went blank, so I just did other things. But I can’t even stay still. Even if it’s nothing serious, I get annoyed easily just by someone talking to me. After it's over, I often regret doing it. I can’t concentrate well, my mind is scattered, and I can’t do anything properly... As a result, work keeps piling up, and when that piles up, I get more depressed, and so my irritation increases... It’s a constant cycle. It’s not a big deal, but I just feel like I’m not feeling well lately, and I don’t know why I’m like this... Honestly, I’m a bit exhausted.”

[0175] In one embodiment, the computing device (100) may determine a first candidate mental illness (510) based additionally on a disease baseline (51) and / or question-and-answer data (52).

[0176] For example, if the user category is a group of men aged 30 to 35, the first candidate mental illness (510) may be bipolar disorder, depression, ADHD, generalized anxiety disorder, panic disorder, alcoholism, sleep disorder, etc.

[0177] In one embodiment, the computing device (100) can determine the second candidate mental disorders (520) by filtering out a portion of the first candidate mental disorders (510). The number of the second candidate mental disorders (520) may be smaller than the number of the first candidate mental disorders (510).

[0178] In one embodiment, the second candidate mental illnesses (520) may be generated using an emotional state classification result (42) generated from a second artificial intelligence model (412) that takes as input the user's self-report data and / or speech data (53) included in the mental health input information.

[0179] For example, the first candidate mental illnesses (510) are bipolar disorder, depression, ADHD, generalized anxiety disorder, panic disorder, alcoholism, and sleep disorder, and if the emotional state classification result (42) includes depression, mood swings, and lethargy, the second candidate mental illnesses (520) may be determined to be bipolar disorder, depression, and ADHD corresponding to the emotional state classification result (42) in the first candidate mental illnesses (510).

[0180] In one embodiment, the computing device (100) may determine third candidate mental disorders (530) by filtering out some of the second candidate mental disorders (520). In one embodiment, the number of third candidate mental disorders (530) may be smaller than the number of second candidate mental disorders (520).

[0181] In one embodiment, the computing device (100) can determine third candidate mental disorders (530) by filtering out some of the second candidate mental disorders (520) using language pattern keywords (43) generated from a third artificial intelligence model (413) that takes user self-report data and / or speech data (53) included in mental health input information as input. The number of third candidate mental disorders (530) may be smaller than the number of second candidate mental disorders (520).

[0182] In one embodiment, the second candidate mental disorders (520) are bipolar disorder, depression, and ADHD corresponding to the emotional state classification result (42) in the first candidate mental disorders (510), and if the language pattern keyword (43) is 'hesitation', the third candidate mental disorders (530) may be depression and ADHD.

[0183] In one embodiment, the computing device (100) can determine the third candidate mental illnesses (530) as a mental illness candidate list (54).

[0184] In one embodiment, the computing device (100) can perform an examination on a list of mental illness candidates (54) and determine a mental illness candidate that satisfies specific conditions as a target mental illness (55).

[0185] In one embodiment, the computing device (100) can generate a second mental illness classification result in which the first mental illness classification result is adjusted by applying the second output of the second artificial intelligence model (412) among the plurality of artificial intelligence models to the first mental illness classification result determined using the first output of the first artificial intelligence model (411) among the plurality of artificial intelligence models (520).

[0186] In one embodiment, the computing device (100) can set a disease baseline (51) representing an initial probability value for each mental disease of the user using user identification information included in the mental health input information (50).

[0187] In one embodiment, the computing device (100) can first adjust the initial probability value for each mental disease at the disease baseline (51) by using mental health-related symptom information (41) generated from a first artificial intelligence model (411) that takes user self-report data (53) included in the mental health input information (50) as input and question-and-answer data included in the mental health input information (50).

[0188] In one embodiment, the initial probability value for each mental illness may mean the likelihood of a user being diagnosed with predefined mental illnesses. In one embodiment, the initial probability value for each mental illness being adjusted first may mean the likelihood of a user being diagnosed with predefined mental illnesses determined based on question-and-answer data included in mental health-related symptom information (41) and mental health input information (50) (510). For example, the initial probability value for each mental illness may include bipolar disorder-0.028, depression-0.095, etc.

[0189] In one embodiment, the computing device (100) can secondarily adjust the firstly adjusted probability values ​​for each mental illness using an emotional state classification result (42) generated from a second artificial intelligence model (412) that takes user self-report data included in mental health input information as input (520). For example, if the emotional state classification result (42) includes sadness, the initial probability values ​​for each mental illness can be secondarily adjusted by a predefined rule such as bipolarity-0.033, depression-0.099, etc.

[0190] In one embodiment, the computing device (100) can adjust the second-adjusted probability value for each mental illness to a third-adjusted value (530) using a language pattern keyword (43) generated from a third artificial intelligence model (413) that takes user self-report data (53) included in mental health input information (50) as input. For example, if the language pattern keyword (43) includes 'expression of daze,' the initial probability value for each mental illness can be adjusted to a third-adjusted value by a predefined rule such as bipolar disorder-0.035, depression-0.1, etc.

[0191] In one embodiment, the computing device (100) can generate a third mental illness classification result in which the second mental illness classification result is adjusted as the third output of the third artificial intelligence model (413) among the plurality of artificial intelligence models is applied to the second mental illness classification result (530).

[0192] In one embodiment, the computing device (100) can generate a list of suspected mental illness candidates (54) for the user using the third mental illness classification result.

[0193] In one embodiment, the computing device (100) can generate a list of suspected mental illness candidates (54) for a user by serially adding the output of each of a plurality of artificial intelligence models or information processed based thereon based on the disease guidelines.

[0194] In one embodiment, the first mental illness classification result, the second mental illness classification result, and the third mental illness classification result may each represent the probability of occurrence of a user's mental illness. For example, regarding the probability of occurrence for depression, the first mental illness classification result may be 50%, and the second mental illness classification result may be increased to 60% as the second output is applied to the first mental illness classification result. In one embodiment, the third mental illness classification result may be decreased to 40% as the third output is applied to the second mental illness classification result. Whether subsequent mental illness classification results increase or decrease as outputs are applied to mental illness classification results is not limited to the examples listed.

[0195] In one embodiment, the first artificial intelligence model (411) may utilize Gemma-27B to extract only clinically standardized symptom tags from the user's self-report utterance. In one embodiment, the first artificial intelligence model (411) may structure and return the basis phrases for the presence or absence of symptoms, confidence, negativity, and uncertainty, and may not perform disease estimation or the application of odds ratios. The symptom information thus generated may be integrated with the symptom information collected through the question-and-answer data (52) so as not to overlap.

[0196] In one embodiment, the first artificial intelligence model (411) takes user's self-report voice input data (utterance content converted into text) as input and can extract symptom information from the utterance content. For example, the extracted data may be as follows.

[0197] {

[0198] "model": "symptom-classifier-gemma-27b",

[0199] "version": "2025-08-25",

[0200] "input_id": "utterance_001",

[0201] "language": "ko",

[0202] "symptoms": [

[0203] {"id": 1, "label": "Mood Swings", "present": true},

[0204] {"id": 2, "label": "Sensitive / Oversensitive", "present": true},

[0205] {"id": 3, "label": "Depressed", "present": true},

[0206] {"id": 4, "label": "Fatigue / Lethargy", "present": true},

[0207] {"id": 5, "label": "Suicidal thoughts", "present": false},

[0208] {"id": 6, "label": "Low appetite", "present": false},

[0209] {"id": 7, "label": "Attention Problem", "present": true},

[0210] {"id": 8, "label": "Memory Problems", "present": false},

[0211] {"id": 9, "label": "Excessive worry", "present": false},

[0212] {"id": 10, "label": "Panic Attack", "present": false},

[0213] {"id": 11, "label": "Skin Anxiety", "present": false},

[0214] {"id": 12, "label": "Alcohol", "present": false},

[0215] {"id": 13, "label": "Insomnia", "present": false},

[0216] {"id": 14, "label": "Decreased motivation", "present": true},

[0217] {"id": 15, "label": "self-blame", "present": true},

[0218] {"id": 16, "label": "Degradation", "present": true}

[0219] ]

[0220] }

[0221] In one embodiment, the question-and-answer data (52) may be "emotional fluctuations, fatigue, lethargy, attention problems, depression," and the symptom information collected through the first artificial intelligence model (411) may be "emotional fluctuations, sensitivity / oversensitivity, depression, fatigue / lethargy, attention problems, lack of motivation, self-blame, functional decline."

[0222] In one embodiment, the computing device (100) can adjust the probability value of each disease by removing duplicate values ​​from the collected symptom information and utilizing predefined disease-specific major symptom-odds ratio (OR) data.

[0223] In one embodiment, the first mental illness classification result may be generated by applying the first comparison result (e.g., the difference between the OR value according to the first reference table and the first output) obtained by comparing the first output with the first reference table (Table 3) which defines the relationship between predefined symptoms and mental illnesses, to the disease baseline (51) representing the user's basic probability value for each of the plurality of mental illnesses.

[0224] The reference tables in the present disclosure may refer to predefined data mapping tables for converting and adjusting the qualitative analysis results of artificial intelligence models into quantitative probability values.

[0225] For example, the first reference table (e.g., disease-symptom odds ratio data) is a table that defines, using the Odds Ratio (OR), how much the probability of each disease changes when a specific symptom is present (+) or not present (-). For example, if the symptom 'depression' is present, the odds ratio for depression is applied as 4.0, which significantly increases the base probability, but the odds ratio for ADHD is 1.1, which may have a negligible effect.

[0226] For example, the second reference table (e.g., disease-emotion association data) is a table that defines whether a specific emotional state has a statistically positive (+), negative (-), or neutral (0) relationship with each disease. The above relationship can be used to calculate the Likelihood Ratio (LR). For example, since the emotion of 'anger' has a positive (+) association with bipolar disorder, if the probability of the emotion of 'anger' is high, the final probability of bipolar disorder may be adjusted upward.

[0227] For example, the third reference table (e.g., the disease-keyword mapping table) is a table that defines how much non-symptomatic language pattern keywords are associated with each disease using numerical mapping values ​​(weights). For example, when the keyword 'avoidance / delayed discourse' is extracted, a weight of 2.10 is applied to the probability value of ADHD, which significantly increases the probability, but a weight of 1.00 is applied to panic disorder, which may result in no change.

[0228] In one embodiment, the first mental illness classification result can generate user's mental health-related symptom information (41) using a first artificial intelligence model (411) that takes self-report data included in mental health input information (50) as input.

[0229] In one embodiment, the computing device (100) can compare symptom information (41) with a first reference table to determine, for each combination of a plurality of mental illnesses and a plurality of symptoms, a first multiplier value (ORd,s(+)) for up-correcting the probability of occurrence of a specific mental illness when a specific symptom is present, and a second multiplier value (ORd,s(-)) for down-correcting the probability of occurrence of a specific mental illness when a specific symptom is not present.

[0230] By applying the first multiplication value and the second multiplication value to the disease baseline (51) representing the user's basic probability value for each of the multiple mental diseases, a first mental disease classification result can be generated in which the basic probability value of the disease baseline (51) is corrected.

[0231] Example) Major symptoms by disease - odds ratio (OR) data number symptoms bipolar disorder melancholia ADHD Generalized anxiety disorder Panic disorder Alcohol sleep disorders 1 mood swings 3.0 / 0.7 1.2 / 1.0 1.3 / 1.0 1.1 / 1.0 1.1 / 1.0 1.0 / 1.0 1.1 / 1.0 2 Sensitive / Overly sensitive 1.5 / 1.0 1.3 / 1.0 1.2 / 1.0 1.8 / 0.9 1.3 / 1.0 1.1 / 1.0 1.2 / 1.0 3 melancholy 1.6 / 0.9 4.0 / 0.6 1.1 / 1.0 1.3 / 1.0 1.1 / 1.0 1.2 / 1.0 1.4 / 1.0 4 Fatigue / Lethargy 1.4 / 1.0 3.2 / 0.7 1.4 / 1.0 1.2 / 1.0 1.1 / 1.0 1.0 / 1.0 1.5 / 1.0 5 Suicide accident 2.5 / 0.8 4.5 / 0.6 1.0 / 1.0 1.2 / 1.0 1.3 / 1.0 1.6 / 0.9 1.3 / 1.0 6 Loss of appetite 1.1 / 1.0 2.2 / 0.8 1.0 / 1.0 1.1 / 1.0 1.0 / 1.0 1.0 / 1.0 1.3 / 1.0 7 Attention problems 1.3 / 1.0 1.4 / 1.0 4.5 / 0.6 1.2 / 1.0 1.0 / 1.0 1.1 / 1.0 1.2 / 1.0 8 memory problems 1.2 / 1.0 1.6 / 0.9 1.5 / 1.0 1.2 / 1.0 1.0 / 1.0 1.2 / 1.0 1.3 / 1.0 9 Excessive worry 1.1 / 1.0 1.5 / 1.0 1.1 / 1.0 4.2 / 0.6 1.6 / 0.9 1.0 / 1.0 1.2 / 1.0 10 panic attack 1.0 / 1.0 1.1 / 1.0 1.0 / 1.0 1.7 / 0.9 5.0 / 0.6 1.0 / 1.0 1.1 / 1.0 11 Somatic anxiety 1.1 / 1.0 1.2 / 1.0 1.0 / 1.0 2.0 / 0.8 2.8 / 0.8 1.0 / 1.0 1.2 / 1.0 12 alcohol 1.2 / 1.0 1.3 / 1.0 1.2 / 1.0 1.1 / 1.0 1.0 / 1.0 5.5 / 0.6 1.4 / 1.0 13 Insomnia 1.1 / 1.0 2.0 / 0.8 1.2 / 1.0 1.5 / 1.0 1.3 / 1.0 1.4 / 1.0 5.0 / 0.6 14 Decreased motivation 1.4 / 1.0 3.2 / 0.7 1.4 / 1.0 1.2 / 1.0 1.1 / 1.0 1.0 / 1.0 1.5 / 1.0 15 remorse 1.4 / 1.0 3.2 / 0.7 1.4 / 1.0 1.2 / 1.0 1.1 / 1.0 1.0 / 1.0 1.5 / 1.0 16 functional decline 1.4 / 1.0 3.2 / 0.7 1.4 / 1.0 1.2 / 1.0 1.1 / 1.0 1.0 / 1.0 1.5 / 1.0

[0232] Table 3 may include odds ratios indicating the degree of association between mental illness and symptoms.

[0233] In one embodiment, the odds ratio (OR) may be an indicator of how much the likelihood of a disease occurring increases or decreases when a specific symptom is present. For example, the odds ratio may quantitatively reflect the relative impact of the symptom on the disease.

[0234] The OR table (Table 3) presents correction values ​​for when symptoms are present (+) and when they are absent (-) depending on the combination of each disease (d) and symptom (s).

[0235] ORd,s={ORd,s(+), symptoms present ORd,s(-), symptoms absent

[0236] The selected symptoms are "mood swings, fatigue / lethargy, attention problems, depression, lack of motivation, irritability / oversensitivity, guilt, and functional decline," so symptoms related to these correspond to ORd,s(+) in bold, and the remaining symptoms correspond to ORd,s(-) in bold.

[0237] ORd,s(+) : Applied when symptoms are present (e.g., fatigue / lethargy, depression increases baseline by 3.2 times)

[0238] ORd,s(-) : Applied when symptoms are absent (e.g., memory problems, depression lowers baseline by 0.9 times)

[0239] In one embodiment, the disease baseline (51) probability is increased or decreased depending on the symptoms.

[0240] The summation process combines and normalizes the ORs adjusted for the reference probability of each disease.

[0241] P(d)=σ(logit(Bd)* s *OR d,s) d:disease, s:symptom, P:probability

[0242] Bd :base line prob(probability) of disorder d.

[0243] σ(x) = x / 1 + x σ: x: A correction function to stabilize and correct extreme value distortions of x. In this example, the logistic monotonic saturating function is used, x: input variable

[0244] logit(p)=p / 1-p logit:logistic+unit(log-odds), p:probability

[0245] In one embodiment, in the case of Table 4 below, through question-and-answer data (52) and the first artificial intelligence model (411), suspected diseases with high prevalence in the order of depression (98.6%), ADHD (67.3%), and sleep disorders (43.2%) can be derived.

[0246] Bipolar melancholia ADHD Generalized anxiety disorder Panic disorder Alcohol sleep disorders Probability of change 0.418 0.986 0.673 0.271 0.089 0.117 0.432

[0247] In one embodiment, Table 4 may include the probability of change by mental illness.

[0248] In one embodiment, the second artificial intelligence model (412) processes the user's self-reported voice data using the Whisper 1.5B model and, through this, classifies the utterance content into one or more of seven emotional states ('joy', 'loveliness', 'fear', 'anger', 'sadness', 'surprise', 'nothing') in the form of a probability distribution to generate emotional state metadata in JSON format.

[0249] In one embodiment, the second artificial intelligence model (412) takes voice data spoken by a user in natural language as input and can process it as ['joy', 'loveliness', 'fear', 'anger', 'sadness', 'surprise', 'none']. For example, the result data processed by the second artificial intelligence model (412) may be as follows.

[0250] {

[0251] "model": "emotion-classifier-whisper-1.5b",

[0252] "version": "2025-08-25",

[0253] "input_id": "utterance_001",

[0254] "language": "ko",

[0255] "emotions": [

[0256] {"label": "Joy", "probability": 0.01},

[0257] {"label": "Loveliness", "probability": 0.01},

[0258] {"label": "Fear", "probability": 0.05},

[0259] {"label": "Angry", "probability": 0.80},

[0260] {"label": "Sadness", "probability": 0.01},

[0261] {"label": "surprise", "probability": 0.01},

[0262] {"label": "None", "probability": 0.1}

[0263] ],

[0264] "dominant_emotion": "Hwanam"

[0265] }

[0266] In one embodiment, the generated emotional state metadata is used to correct the diagnostic classification result by referencing predefined disease-emotion association data. Specific diseases have statistically associated emotional patterns, and if these match the emotional classification result, the probability of that disease is reinforced. Conversely, if an emotion with low correlation appears strongly, the possibility of other diseases is additionally considered.

[0267] For example, since Major Depressive Disorder (MDD) has a high association with 'sadness', if the dominant emotion of the first output (e.g., Major Depressive Disorder) by the first artificial intelligence model (411) and the second output (e.g., Sadness) by the second artificial intelligence model (412) matches as 'sadness', the confidence of the diagnosis can be adjusted upward.

[0268] On the other hand, if the first output (major depressive disorder) by the first artificial intelligence model (411) and the second output (e.g., anger) by the second artificial intelligence model (412) are predominantly detected at the same time, this suggests the possibility that mania or hypomania may be present in addition to depressive symptoms, and thus can be identified as bipolar disorder.

[0269] In one embodiment, the second mental illness classification result may be generated by applying the second comparison result obtained by comparing the second output with a second reference table (Table 5) that defines the relationship between a predefined emotion (e.g., 'joy', 'loveliness', 'fear', 'anger', 'sadness', 'surprise', 'none') and a mental illness to the first mental illness classification result.

[0270] Diagnosis / Emotion delight Loveliness awe aggro sorrow surprise None (lack of emotion) Depression (MDD) - - + 0 + 0 + Bipolar (BD) + + 0 + 0 0 - Generalized Anxiety (GAD) - - + 0 + + 0 ADHD 0 0 0 0 0 0 0

[0271] In one embodiment, Table 5 may include the correlation of emotions (positive / negative) corresponding to the diagnosed mental illness.

[0272] In one embodiment, metadata obtained through the second artificial intelligence model (412) may be reflected by referring to the second reference table. In one embodiment, the computing device (100) may first calculate prior odds based on existing data, score the emotion classification results and map them to the Likelihood Ratio (LR), apply this to finally obtain posterior odds, and then convert them into posterior probability.

[0273] In one embodiment, the second reference table and the emotion may have the following association.

[0274] For example, if the user's emotional state is analyzed as anger with an 80% probability, the probability of bipolar disorder (BD), which is highly associated with that emotion, can be reinforced by referring to the disease-emotion association mapping table. On the other hand, other diseases are not corrected as their mapping value is set to neutral (0). In one embodiment, the previously determined probability of bipolar disorder can be adjusted to a final probability by referring to the emotion classification result score data table (Table 6).

[0275] In one embodiment, the second mental illness classification result may include an emotional state classification result (42) representing the user's emotion identification information (e.g., joy) and emotion intensity (e.g., 0.01) generated using a second artificial intelligence model (412) that takes self-report data included in the mental health input information as input.

[0276] In one embodiment, the computing device (100) can determine the correlation between a specific emotion and a specific disease by comparing the emotional state classification result (42) with a second reference table.

[0277] In one embodiment, the computing device (100) can calculate a likelihood ratio based on the correlation with the emotional state classification result (42), comparing the probability that a specific emotion appears in a group of patients with mental illness and the probability that it appears in a group of non-patients with mental illness.

[0278] In one embodiment, the computing device (100) can generate a second mental illness classification result in which the first mental illness classification result is corrected by applying a likelihood ratio to the first mental illness classification result.

[0279] In one embodiment, the likelihood ratio can be adjusted according to emotional intensity as follows.

[0280] Disease-Emotion Association Data+ : Likelihood Ratio (LR) increases as emotional intensity increases *Probability increases, Disease-Emotion Association Data- : Likelihood Ratio (LR) decreases as emotional intensity increases *Probability decreases, Disease-Emotion Association Data0 : Neutral, No adjustment"

[0281] In one embodiment, the Likelihood Ratio (LR) can be calculated by comparing the probability P(d) that an emotion occurs in the patient group with the probability P(~d) that it occurs in the non-patient group.

[0282] LR(s)= σ(Pd / P(~d))

[0283] LR: Likelihood ratio, P(): Probability of emotion occurring, d: Patient group, ~d: Non-patient group, σ(x): Correction function to stabilize and prevent extreme value distortion of x. In this example, the logistic monotonic saturating function can be used.

[0284] LR > 1: That emotion is observed more frequently in the corresponding disorder group → Signal of increased probability

[0285] LR < 1: Emotional patterns contrary to the disorder → Signal of reduced probability

[0286] In one embodiment, the 'Hwanam' likelihood ratio (LR) calculation data may correspond to Figure 10 and Table 6 below.

[0287] Level of anger P(s|d) P(s|~d) LR(s) 0 0.000001 0.000468 0.7209 5 0.000003 0.000974 0.7612 10 0.000013 0.001876 0.8141 15 0.000048 0.003347 0.8828 20 0.000163 0.005527 0.9709 25 0.000481 0.008448 1.0823 30 0.001252 0.011955 1.2202 35 0.002869 0.015662 1.387 40 0.005786 0.018994 1.5828 45 0.010271 0.021325 1.8048 50 0.016051 0.022163 2.0469 55 0.02208 0.021325 2.3 60 0.026735 0.018994 2.5531 65 0.028496 0.015662 2.7952 70 0.026735 0.011955 3.0172 75 0.02208 0.008448 3.213 80 0.016051 0.005527 3.3798 85 0.010271 0.003347 3.5177 90 0.005786 0.001876 3.6291 95 0.002869 0.000974 3.7172 100 0.001252 0.000468 3.7859

[0288] In one embodiment, Table 6 may include the degree of emotional state (0 to 100) for the emotion of anger, the probability P(d) of occurrence in the patient group and the probability P(~d) of occurrence in the non-patient group, and the likelihood ratio LR(s).

[0289] In one embodiment, the user's self-report data and / or speech data (53) may include not only simple symptom expressions but also linguistic features or speaking habits that frequently appear in a specific patient group.

[0290] In one embodiment, these linguistic patterns, even if not directly linked to symptoms, reflect the speaker's psychological state and speech manner and may be observed relatively more frequently in specific disease groups. In one embodiment, the third artificial intelligence model (413) can detect these non-symptomatic linguistic patterns and map them to predefined disease-linguistic pattern association data to further readjust the disease-specific probabilities.

[0291] In one embodiment, the third artificial intelligence model (413) can take user self-report text as input and process it as follows: "preprocessing: morphological analysis, stop word removal, standardization / keyword extraction: identification and structuring of non-symptomatic related language patterns." For example, the result data processed by the third artificial intelligence model (413) may be as follows.

[0292] {

[0293] "model": "keyword-pattern-extractor",

[0294] "version": "2025-08-25",

[0295] "input_id": "utt_001",

[0296] "language": "ko",

[0297] "keywords": [

[0298] {"label": "Filling / Hesitation"},

[0299] {"label": "Repeat,Continuous Emphasize"},

[0300] {"label": "Avoidance / Delayed Discourse"},

[0301] {"label": "Narrative / Regret"},

[0302] {"label": "Negative self-evaluation"}

[0303] ]

[0304] }

[0305] For example, non-symptomatic language patterns may be as follows.

[0306] Negative self-evaluation: “I’m just not good at it,” “I’m the problem,” “I’m useless”

[0307] Rumination / Regret Narrative: “I shouldn’t have done that,” “I keep thinking about it,” “I keep mulling it over”

[0308] Hedge / Ambiguity: “I think”, “Probably”, “Not really”, “I’m not entirely sure”

[0309] Filler / Hesitation: “Um...”, “Uh...”, “Uh... how should I put it”

[0310] Overgeneralization: “I always fail”, “Nobody [does] me..”, “Absolutely not”

[0311] Emphasis on repetition and continuity (persistence): “continuously,” “all day long,” “always,” “every day”

[0312] Avoidance / Delay Discourse: “I was distracted,” “I put it off,” “I’ll do it later”

[0313] Expressions of loss of control: “I can’t control myself,” “I’m at a loss”

[0314] Warning of Uncertainty and Worry: “Just in case..,” “Worried for no reason,” “A bad feeling”

[0315] Self-blame / guilt framing: “It’s my fault,” “I’m sorry, it’s my mistake”

[0316] Devaluation / Devaluation of meaning: “It’s no big deal,” “It’s all useless”

[0317] Expressions of cognitive cloudiness / blankness: “My head feels blank,” “My mind went blank”

[0318] Oversensitive / Aggressive Tone Clues: “Gets annoyed by even trivial things,” “Gets angry easily”

[0319] Clues for relationship avoidance / withdrawal: “Not answering calls”, “I don’t want to meet people”

[0320] Mention of sleep and rhythm (non-symptomatic context): “Mornings are especially difficult,” “Rhythm is disrupted” (at the level of habits / feelings, not symptom descriptions)

[0321] In one embodiment, non-symptomatic language patterns are based on linguistic features and speaking habits revealed in the user's utterances; while distinct from direct symptom reporting, they can serve as important clues reflecting the patient's psychological state and cognitive characteristics. For example, patients with major depressive disorder may frequently use expressions emphasizing repetition and duration, such as "continuously," "all day long," and "always"; patients with anxiety disorders may use expressions of ambiguity and hedging; and patients with ADHD may frequently use expressions of avoidance and procrastination. These non-symptomatic language patterns can reflect internal states and disease-specific language habits that are difficult to capture through traditional symptom reporting alone.

[0322] In one embodiment, the third artificial intelligence model (413) is combined with the first artificial intelligence model (411) to correct the probability of occurrence of mental illness, thereby increasing the precision and validity of the clinical diagnosis. In one embodiment, if there is a duplicate with keywords already classified based on symptoms through the question-and-answer data (52) and the first artificial intelligence model (411), it can be excluded to prevent the same information from being excessively reflected.

[0323] In one embodiment, the symptom information is mood swings, fatigue / lethargy, attention problems, depression, low motivation, sensitivity / oversensitivity, self-blame, and functional decline, and the language patterns extracted through the third artificial intelligence model (413) are filler phrases / hesitations, repetition / continuous emphasis, avoidance / delayed discourse, rumination / regret descriptions, and negative self-evaluations, and only filler phrases / hesitations are identified as independent patterns that do not overlap with the symptom information and can be used as final correction factors.

[0324] In one embodiment, the third mental illness classification result can be generated by applying the third comparison result obtained by comparing the third output with the third reference table (Table 7) which defines the relationship between a predefined language pattern keyword and a mental illness to the second mental illness classification result.

[0325] In one embodiment, the computing device (100) can generate a user's language pattern keyword (43) using a third artificial intelligence model (413) that takes as input self-report data and / or speech data (53) included in mental health input information.

[0326] In one embodiment, the computing device (100) can compare the language pattern keyword (43) with the third reference table to determine a mapping value corresponding to the generated language pattern keyword (43) for each combination of the plurality of mental disorders and the plurality of language pattern keywords (43).

[0327] In one embodiment, the computing device (100) can generate a third mental illness classification result in which the second mental illness classification result is corrected by applying a mapping value to the second mental illness classification result.

[0328] Keywords Bipolar melancholia ADHD Generalized anxiety disorder Panic disorder Alcohol sleep disorders Filling phrase / hesitation 1.05 1.08 1.12 1.40 1.10 1.00 1.00 Repetition, Continuous Emphasis 1.12 1.15 1.05 1.03 1.00 0.98 1.02 Avoidance / Delayed Discourse 1.05 1.08 2.10 1.05 1.00 1.05 1.00 Rumination / Regret Narration 1.05 1.90 1.00 0.90 1.00 1.00 1.10 Negative self-evaluation 1.10 2.20 1.02 1.10 1.05 1.0 1.02

[0329] In one embodiment, Table 7 may include data that scores the relationship between language pattern keywords and mental illness.

[0330] In one embodiment, mental disorders can be finally classified as follows by reflecting the filling / hesitation data obtained through the third artificial intelligence model.

[0331] Bipolar melancholia ADHD Generalized anxiety disorder Panic disorder Alcohol sleep disorders 0.718 0.987 0.698 0.342 0.097 0.117 0.432

[0332] In one embodiment, Table 8 may include the probability of a specific user developing a mental illness obtained by reflecting the language pattern keywords of Table 7.

[0333] The items showing a probability of 50% or higher in the classification of suspected disorders were bipolar disorder, depression, and ADHD; however, in cases where bipolar disorder is clinically suspected, depression is not classified separately and may be included in the differential diagnosis test for bipolar disorder. Therefore, the final validity review and functional testing are conducted only for bipolar disorder and ADHD.

[0334] FIG. 6 is a diagram illustrating a method in which a disease identification module according to one embodiment of the present disclosure generates a target mental disease.

[0335] Referring to FIG. 6, the disease identification module (420) can automatically construct responsive questions based on core diagnostic conditions of accredited standards such as DSM-5 for 1 to 3 suspected diseases derived during the screening diagnosis stage, perform only valid tests, and automatically omit unnecessary questions. Existing mental health assessment systems provide accredited diagnostic scales such as DSM-5, PHQ-9, and GAD-7 in a fixed question structure, and force users to respond to all questions regardless of whether they have actual symptoms. As a result, questions that are not directly related to the diagnosis are included, leading to problems such as increased user fatigue, time inefficiency, and reduced diagnostic reliability. Accordingly, the present invention implements a responsive question system based on core diagnostic conditions for each disease and applies a condition-based responsive digital scale structure that exposes subsequent questions only when conditions with clinical diagnostic value are met, thereby minimizing user fatigue regarding question solving, improving time inefficiency, and maximizing diagnostic efficiency.

[0336] In one embodiment, the disease identification module (420) can select (610) a question for each disease included in the mental disease candidate list and provide (620) it to the user.

[0337] In one embodiment, the disease determination module (420) generates and provides a first-1 user interaction question among a plurality of user interaction questions corresponding to a first candidate mental disease included in a list of mental disease candidates, and can determine whether to generate a first-2 user interaction question based on whether the user response to the first-1 user interaction question satisfies the condition corresponding to the first candidate mental disease.

[0338] In one embodiment, the disease determination module (420) may generate and provide a first-2 user interaction question (630) if the user response to the first-1 user interaction question satisfies the condition corresponding to the first candidate mental disease. Alternatively, if the user response to the first-1 user interaction question does not satisfy the condition corresponding to the first candidate mental disease, the disease determination module (420) may not generate a first-2 user interaction question and may generate a second-1 user interaction question corresponding to the second candidate mental disease (640).

[0339] In one embodiment, the disease identification module (420) first presents a question (e.g., "Has a depressed mood lasted for 2 weeks or more?") to check for the presence of core symptoms of the disease (e.g., if the disease is major depressive disorder, at least one of 'depressed mood' or 'decreased interest' must last for 2 weeks or more to be considered diagnostically valid) at the start of the test, and if the condition is not met, subsequent questions may be automatically omitted. The user does not proceed further with responses regarding the disease.

[0340] In one embodiment, the disease determination module (420) may provide subsequent questions in stages only when the corresponding condition is met.

[0341] In one embodiment, when items for multiple diseases are provided, the disease determination module (420) designs the items in a complete branching structure so that even if an item for a specific disease is omitted due to non-fulfillment of conditions, the evaluation flow of other diseases is not affected, thereby maintaining overall diagnostic validity without the risk of missing diagnoses.

[0342] In one embodiment, the disease determination module (420) can be configured in the same way that a doctor first determines the presence or absence of key symptoms during diagnosis and determines the diagnosis questions accordingly, so that the flow of the entire set of questions matches the actual clinical treatment flow.

[0343] In one embodiment, the disease identification module (420) can secure the possibility of future application and expansion to a digital therapeutic device (SaMD).

[0344] In one embodiment, the disease identification module (420) can calculate (650) the accumulated scores for each disease by obtaining responses to user interaction questions.

[0345] For example, if the first candidate mental disorder is ADHD, the 1-1 user interaction item may be "Is it difficult to focus on one thing?" and if the user's response to this is "Yes," the 1-2 user interaction item may be "Do multiple thoughts come to mind at the same time?"

[0346] For example, if the first candidate mental disorder is ADHD and the second candidate mental disorder is depression, the first-1 user interaction item may be "Is it difficult to focus on one thing?" and if the user's response to this is "No," the second-1 user interaction item may be "Have you thought about feeling depressed recently?"

[0347] For example, the score can increase in proportion to the number of questions answered by the user.

[0348] In one embodiment, the disease determination module (420) can generate a user interaction question (610) for the subsequent disease if it is determined that there is a subsequent disease based on the disease-specific score for the first candidate mental disease, and can determine the first candidate mental disease as the target mental disease (660) if it is determined that there is no subsequent disease.

[0349] In one embodiment, when a target mental illness is determined, the computing device (100) executes a validity review module (430), and when a target mental illness is not determined, it may execute a function test module (440).

[0350] FIG. 7 is a diagram illustrating a method in which a feasibility review module according to one embodiment of the present disclosure generates a confirmed mental illness.

[0351] Referring to FIG. 7, the validity review module (430) can perform additional review to ensure the clinical validity of the results regarding the suspected disease and its severity derived from the deep examination results.

[0352] In one embodiment, the validity review module (430) can analyze temporal and underlying factors such as the duration of symptoms, the time of occurrence, and triggering factors.

[0353] In one embodiment, the validity review module (430) can eliminate the possibility of diagnostic errors caused by non-psychopathological factors such as drug use or physical illness, and enhance the reliability of the final diagnosis result.

[0354] Most existing mental health self-assessment systems are designed based on standardized, fixed questions. This can lead to problems where diagnostic criteria are met for symptoms that are not actually present or are non-psychopathological, if users include past symptoms in their responses or check for temporary symptoms that occurred while taking medication. Consequently, diagnostic accuracy is reduced, and patients may be unnecessarily classified into high-risk groups.

[0355] Accordingly, the present invention secures the validity of the diagnosis by adding two review procedures included in the validity review module (430).

[0356] In one embodiment, the feasibility review module (430) can review (710) whether the target time condition is met for the target mental illness.

[0357] In one embodiment, the validity review module (430) may generate a second user interaction question that queries symptom time conditions, including the time of symptom onset, duration, and frequency for the target mental illness determined by the disease identification module (420).

[0358] In one embodiment, the validity review module (430) determines whether the user response to the second user interaction question satisfies the target time condition pre-assigned to the target mental illness, and based on whether the target time condition is satisfied, can generate a verification result for the target mental illness.

[0359] For example, if the target mental illness is depression, the user may respond with "symptom onset 2 years ago, duration and frequency about once every 2 months for 1 week."

[0360] In one embodiment, the validity review module (430) may set the target mental illness as a reference mental illness (730) if the user response to the second user interaction question does not satisfy the target time condition. In one embodiment, the reference mental illness may refer to a mental illness or a similar mental illness that has the potential to develop or has developed in the past, rather than a confirmed mental illness.

[0361] In one embodiment, the validity review module (430) can generate a third user interaction question to review the user's medication and underlying disease if the user's response to the second user interaction question satisfies the target time condition.

[0362] In one embodiment, the validity review module (430) can determine whether to include the target mental illness as a confirmed mental illness in the verification results based on the user response to the third user interaction question.

[0363] In one embodiment, the validity review module (430) may determine a mental illness grade for a target mental illness based on a time value related to the user's symptom time condition obtained through the second user interaction question when the user response to the second user interaction question satisfies the target time condition. For example, the mental illness grade may be determined based on a score corresponding to the target mental illness and may be determined according to the range in which the score is included. For example, if the score is 41 to 50, the mental illness grade may be Grade 2; if it is less than 41, the mental illness grade may be Grade 1; and if it is 50 or more, the mental illness grade may be Grade 3.

[0364] In one embodiment, the validity review module (430) may add a garage usage guide to the diagnosis result when the user response to the second user interaction question does not meet the target time value (e.g., period, frequency, timing validity, etc.).

[0365] In one embodiment, the validity review module (430) may add a notice requiring additional verification of the diagnosis result if the user response to the third user interaction question satisfies a predefined condition (e.g., whether the user is taking more than a certain amount of medication, whether there is an underlying disease, etc.).

[0366] In one embodiment, when the operation by the validity review module (430) is completed and a confirmed mental illness is generated, the computing device (100) can measure the stress level for the confirmed mental illness and examine functions related to the user's daily life (800).

[0367] FIG. 8 is a diagram illustrating a method in which a function test module according to one embodiment of the present disclosure tests stress and function.

[0368] Referring to FIG. 8, the functional test module (440) can quantitatively evaluate the clinical state of the confirmed mental illness, the impact of the confirmed mental illness on an individual's actual life, and analyze the need for future risk prediction and management.

[0369] In one embodiment, the function test module (440) can be designed to generate more precise and realistic diagnostic results by including the correlation of reduced living function and stress, which were overlooked by existing self-diagnosis systems.

[0370] In one embodiment, the function test module (440) may generate a fourth user interaction item (810) based on the major stress factors of the confirmed mental illness when a confirmed mental illness exists. In one embodiment, the function test module (440) may additionally receive input of a score corresponding to the confirmed mental illness, an emotional state classification result, and / or language pattern keywords when a confirmed mental illness exists.

[0371] In one embodiment, the function test module (440) can generate a fourth user interaction item (810) based on general stress factors when no confirmed mental illness exists.

[0372] In one embodiment, the function test module (440) can obtain a response (820) to a fourth user interaction question.

[0373] In one embodiment, the function test module (440) may generate result data (830) based on the response to the fourth user interaction question. For example, the result data may represent stress factors and / or intensity.

[0374] For example, if the confirmed mental illness is depressive disorder, stress factors may include work, finances, relationships, anxiety about the future, self-esteem, etc.

[0375] For example, if the confirmed mental disorder is panic disorder, stressors may include anticipatory anxiety, loss of control, enclosed spaces, crowded places, and maintaining tension.

[0376] In one embodiment, the function test module (440) may generate comprehensive analysis information (840), such as the degree to which stress and function are affected, based on the response to the fourth user interaction question. For example, the analysis information may include comments on predicting future risks and management measures, and / or, if a disease is present, by comparing and correcting the severity of the mental illness with the decline in function.

[0377] In one embodiment, the function test module (440) can automatically generate (850) items for evaluating the user's job function, daily life function, interpersonal function, physical function, etc. In one embodiment, the function-related items may include items necessary for analyzing the impact of mental illness.

[0378] In one embodiment, the function test module (440) can obtain response data (860) for function questions.

[0379] In one embodiment, the function test module (440) can generate (870) information related to functional decline reactions that specifically evaluate how mental illness actually affects an individual's daily life, job, social life, and physical function using response data to the acquired function items.

[0380] In one embodiment, information related to functional decline responses is configured to reflect the realistic severity of the diagnosis of mental illness and to clearly determine the need for future management and the possibility of maintaining daily life.

[0381] In one embodiment, information related to functional decline responses may be combined with stress assessment results to include an overall risk level associated with actual functional decline. For example, even if the stage of a specific disease (e.g., risk, severity) is determined to be low after undergoing in-depth examination and validity review stages, if functional test results confirm that the specific disease is having a significant impact on daily life, the severity of the specific disease may be configured to be assessed by adjusting it upward.

[0382] Through this, the computing device (100) can provide more precise and realistic diagnostic results that reflect the impact on actual life functions, going beyond a simple numerical evaluation of symptoms.

[0383] Through the present invention, even if a disease is not confirmed as a result of in-depth examination and validity review, if high stress levels and functional decline are indicated in the results of functional tests, it is possible to predict the possibility of future mental health problems in advance and provide specific comments to the user, such as preventive management, additional monitoring, and suggestions for stress relief strategies.

[0384] Through the present invention, the effect of providing a structure capable of proactive care based on real life, going beyond a simple diagnosis of the 'presence or absence' of mental illness, can also be achieved.

[0385] FIG. 9 is a diagram illustrating a method in which a diagnostic result generation module according to one embodiment of the present disclosure generates a mental health diagnostic result.

[0386] Referring to FIG. 9, the computing device (100) can collect data and generate a list of mental illness candidates (411), and if there are mental illness candidates, generate a first user interaction question (421).

[0387] In one embodiment, the computing device (100) can determine whether a mental illness candidate satisfies the disease-specific conditions based on a response corresponding to a first user interaction question.

[0388] In one embodiment, if a mental illness candidate satisfies the conditions for each disease, the computing device (100) can set the mental illness candidate as a target mental illness (55).

[0389] In one embodiment, if a candidate mental illness does not satisfy the conditions for each illness, the computing device (100) may set the candidate mental illness as a reference mental illness (65).

[0390] In one embodiment, if there are no mental illness candidates, the collected data and the list of mental illness candidates can be combined with the reference mental illness (65) and the target mental illness (55) (450).

[0391] In one embodiment, the computing device (100) may generate (431) a second user interaction question to determine whether the target mental illness (55) satisfies a pre-assigned target time condition. In one embodiment, metadata such as response data and results obtained through the second user interaction question may be collected (450).

[0392] In one embodiment, the computing device (100) may generate a third user interaction question (432) for reviewing the user's medication and underlying diseases. In one embodiment, metadata such as response data and results obtained through the third user interaction question may be collected (450).

[0393] In one embodiment, the computing device (100) can verify the presence or absence of a suspected mental illness through question-and-answer data and an artificial intelligence model during the process of generating a list of mental illness candidates (411), and can explore additional symptoms related to the mental illness during the subsequent process of generating a first user interaction question (421). The information collected through this process leads to a condition-based classification procedure, and the application of the second user interaction question varies depending on the characteristics of the disease; it may be performed for a specific disease but omitted for another disease.

[0394] In one embodiment, the computing device (100) can comprehensively analyze multilayer data generated at each stage from screening diagnosis to functional test to comprehensively present the diagnosis results of mental illness, functional risk, need for management, and recommendations for future monitoring (450).

[0395] In one embodiment, the computing device (100) can generate a mental health diagnosis result (45) (450).

[0396] In one embodiment, the mental health diagnosis result (45) may include a stress level determined based on the response to a fourth user interaction question, which includes a question for measuring the stress level for the confirmed mental illness and a question for examining functions related to the user's daily life, when the confirmed mental illness is included in the verification result by the validity review module, and a risk level to the user's daily life.

[0397] In one embodiment, the mental health diagnosis result (45) may include mental health input information, a list of mental illness candidates, a target mental illness (55) and a confirmed mental illness included in the verification result, a mental illness grade for the confirmed mental illness, a reference mental illness (65) that is not classified as a confirmed mental illness among the target mental illnesses (55), a user's emotional state classification result, and a user's language pattern keyword.

[0398] For example, the mental health diagnosis result (45) may include mental health input information and, accordingly, a list of mental illness candidates (e.g., depression, bipolar disorder, ADHD, etc.), a target mental illness (e.g., ADHD, bipolar disorder), a confirmed mental illness (e.g., ADHD), a reference mental illness (65, e.g., bipolar disorder), a user's emotional state classification result (e.g., anger), and a user's language pattern keyword (e.g., avoidance).

[0399] Existing mental health assessment systems have limitations in that they mechanically determine the presence of a disorder based solely on individual scale scores, merely list fragmentary information without linking assessment results, and fail to consider complex information such as emotion, temporality, functional level, and precipitating factors. Consequently, existing systems may result in errors where actual diagnostic reliability is low, appropriate response recommendations are omitted even when functional decline is evident despite mild symptoms, or conversely, excessive judgments are made regarding transient symptoms.

[0400] The present invention can implement a system that comprehensively analyzes multilayer metadata accumulated throughout the entire evaluation stage (from screening diagnosis using artificial intelligence models to functional testing), rather than simple symptom-based results, and produces precise context-based comprehensive results including the grade of mental illness, validity, functional decline, and prevention recommendations based on the results.

[0401] In one embodiment, the third user interaction item and the fourth user interaction item are not direct judgment elements of the diagnosis, but the third user interaction item provides auxiliary information for establishing a more accurate diagnosis and a safe treatment plan, and the fourth user interaction item can provide additional information that contributes to improving the quality of life by reflecting the user's overall function and life context.

[0402] In one embodiment, the computing device (100) can comprehensively present the diagnosis results of a mental illness, functional risk, the need for management, and recommendations for future monitoring, and implements a differentiated judgment system that delivers 'context-based precise interpretation results' to the user rather than simple symptom-based results.

[0403] Existing mental health assessment systems have limitations in that they mechanically determine the presence of a disorder based solely on individual scale scores, merely list fragmentary information without linking assessment results, and fail to consider complex information such as emotion, temporality, functional level, and precipitating factors. Consequently, errors occur where appropriate response recommendations are omitted in cases of low actual diagnostic reliability or significant functional decline even with mild symptoms, while conversely, excessive judgments are made regarding transient symptoms.

[0404] In one embodiment, the artificial intelligence models of the present invention can quantify a user's emotional state, recurring issues, tone, etc. through emotion classification and keyword analysis. The metadata obtained by the artificial intelligence model can subsequently be used as auxiliary weights and explanations when determining the severity of each disease.

[0405] The present invention implements a system that comprehensively analyzes multilayer metadata accumulated during the entire evaluation phase and, based on the results, produces a precise context-based comprehensive result that includes the grade of mental illness, validity, functional decline, and prevention recommendations.

[0406] In one embodiment, the mental health diagnosis result (45) is obtained by confirming whether there is a suspected disease through question-and-answer data and an artificial intelligence model in the screening diagnosis stage, and the information collected through the process of exploring additional symptoms related to the disease in the subsequent deep diagnosis stage leads to a condition-based classification procedure, and in the step of confirming temporal validity, the applicability varies depending on the characteristics of the disease, so it is performed for specific diseases but may be omitted for other diseases.

[0407] In one embodiment, the step of reviewing medications and underlying diseases and the step of measuring stress levels and examining functions related to daily life are not direct diagnostic criteria, but the step of reviewing medications and underlying diseases provides auxiliary information for establishing a more accurate diagnosis and a safe treatment plan, and the step of examining functions related to daily life can provide additional information that contributes to improving the quality of life by reflecting the user's overall functions and life context.

[0408] In one embodiment, the computing device (100) can examine the trigger conditions of a disease that are predefined for each disease and automatically classify and exclude diseases that do not meet the minimum diagnostic requirements as reference mental diseases (450). This process may be a step of verifying whether the disease satisfies the minimum diagnostic criteria required clinically by a validity review module or by adding to a previously calculated probability value.

[0409] In one embodiment, depending on the mental disorder, there are cases where a temporal validity test is necessary and cases where it is not, and the temporal validity test can verify the temporal conditions among the major diagnostic requirements of a specific disorder. For example, in the case of depression, a temporal validity test is necessary to confirm the time of onset and duration, while in the case of generalized anxiety disorder, a temporal validity test may not be necessary.

[0410] In one embodiment, if a temporal validity test for a mental illness is required, the computing device (100) may present additional questions asking about temporal elements such as the time of symptom onset, duration, and frequency for major symptoms of each illness, and may review the temporal validity of the illness based on the user's response. If the temporal conditions are not met as a result of the temporal validity test, the illness may be exposed in the result data as a 'reference mental illness' and may be set not to be classified as a 'confirmed mental illness'.

[0411] In one embodiment, the temporal validity test may include items to determine the temporal relationship in which the symptoms occurred for each mental disorder. For example, in the case of ADHD, the items may include the question, "Were the symptoms of reduced attention present since kindergarten or elementary school (before age 12)?" and if not (i.e., if the reduced attention was due to temporary mood symptoms), the diagnosis of ADHD may be excluded.

[0412] In one embodiment, when both depression and bipolar disorder can be diagnosed according to the definition of the diagnostic system, the diagnosis is made only as "bipolar disorder" (because bipolar disorder itself encompasses all symptoms of depression), so the final diagnosis of the case can be made as "bipolar disorder."

[0413] In one embodiment, information obtained from the first to fourth user interaction questions, metadata derived from said information, and the final diagnosis result, etc., may be aggregated as core data provided to the patient. In this case, the mental health diagnosis result may not merely present the presence of a major disease, but may provide more precise and comprehensive information by reflecting auxiliary evaluation tools such as medication and underlying symptom tests, as well as stress and functional tests.

[0414] In one embodiment, testing for medications and underlying symptoms may be performed to identify the interrelationships between medications, physical conditions, lifestyle factors (e.g., caffeine, alcohol, etc.) and the diagnosed disease. For example, the user may select the relevant item themselves through a question such as, "Do you have any diagnosed physical conditions?" and information regarding the intake of medications being taken or caffeine may also be provided in the same manner.

[0415] In one embodiment, while this information does not directly influence the diagnosis of the disease, it can be utilized as supplementary data for a more accurate diagnosis and the establishment of a safe drug treatment plan. For example, if a user diagnosed with bipolar disorder consumes large amounts of caffeine or alcohol or takes blood pressure medication, the possibility that these factors acted secondarily on the depressive episode may be considered. Additionally, some medications used to treat depressive episodes may cause counterproductive effects or side effects when taken with blood pressure medication. Therefore, the results of a mental health diagnosis can improve the precision of the diagnosis and the safety of treatment, and enhance the user's quality of life, by guiding the prescription of medications that do not interact with the diagnosis.

[0416] In one embodiment, the mental health diagnosis result may be a user interface displayed through an output unit of a computing device (100) (e.g., a display device including a screen such as a monitor).

[0417] In one embodiment, mental health diagnosis results may display a drug icon next to specific symptoms to indicate the possibility that the symptoms were caused by medications being taken or other external factors (e.g., caffeine, alcohol, etc.). This allows the user to distinguish whether the symptoms they are experiencing are caused by a disease or by drug interactions or lifestyle factors, and enables clinicians to establish a safer and more precise treatment plan by identifying the cause of the symptoms.

[0418] In one embodiment, the stress and functionality test is an area for understanding the user's overall function, and can evaluate major stressors experienced in daily life (work, interpersonal relationships, job stress, domestic discord, etc.), the impact received from them, and the stress relief methods utilized by the user. For example, the stress and functionality test may ask, through a fourth user interaction item, in which areas the user experiences stress, whether they suffer from daily difficulties such as decreased concentration or poor physical condition as a result, and whether they utilize healthy relief methods such as exercise or hobbies.

[0419] In one embodiment, stress and functional tests, like medication and underlying symptom tests, do not directly affect the diagnosis of a disease but can be used as additional information to improve the user's quality of life. Through this, the mental health diagnosis results may include identifying causes that exacerbate symptoms and supplementing the treatment plan, in addition to describing the current state of the final mental illness diagnosis (e.g., bipolar disorder). For example, if a user responds that they are experiencing significant stress due to domestic discord or workplace conflict, the mental health diagnosis results may include setting assertiveness training or social skills training as the direction of future treatment. Furthermore, through the stress and functional tests, if inefficient and destructive coping mechanisms such as alcohol dependence or binge eating are identified, the mental health diagnosis results may include motivating the user to switch to positive methods, such as exercise or a balanced diet.

[0420] In one embodiment, mental health diagnostic results, through stress and functional tests, can identify the user's current stress situation and potential risk factors even if no specific disease is diagnosed, and provide guidance on future predictions based on this. Through this, the user can gain a more comprehensive understanding of their mental health status and receive assistance in disease prevention and quality of life improvement.

[0421] In one embodiment, the stress and functional test can generate a fourth user interaction item to evaluate job function, daily living function, interpersonal function, physical function, etc.

[0422] In one embodiment, the fourth user interaction item may be configured to specifically evaluate how the discovered mental illness actually affects an individual's daily life, work, social life, and physical function, thereby reflecting the realistic severity of the diagnosis results and clearly determining the need for future management and the possibility of maintaining life.

[0423] In one embodiment, the computing device (100) can evaluate the overall risk associated with actual functional decline in combination with the results of a stress assessment. For example, even if the computing device (100) determines that the stage of the disease is low after undergoing an in-depth examination and a feasibility review, if the results of a functional test confirm that the disease is having a significant impact on daily life, the severity of the disease may be evaluated by adjusting it upward.

[0424] Through this, the present invention can go beyond a simple numerical evaluation of symptoms and provide more precise and realistic diagnostic results that reflect the impact on actual daily functioning.

[0425] Stress is closely related to mental illness and acts as a major factor in inducing recurrence. Accordingly, the present invention can utilize stress as an essential indicator to predict not only the current state but also the possibility of future deterioration, and can be configured to identify whether major stress factors related to the currently detected disease exist. The present invention does not simply ask about general stress levels, but selectively exposes stress items related to major diseases derived from the results of the preceding validity review, and can analyze the association between the disease and stress through the corresponding items. Through this, the computing device (100) can be configured with an adaptive item structure to predict not only an individual's current state but also the possibility of future symptom deterioration.

[0426] In one embodiment, the computing device (100) can quantify whether there is a decline in actual daily life, job, or social function, and if there is a significant decline in function, the severity of the disease can be increased.

[0427] In one embodiment, if the stress item has a high correlation with the disease, the mental health diagnosis result may include the possibility of future recurrence or the need for management.

[0428] Unlike existing systems that mechanically present the presence or absence of a disease, the final mental health diagnosis result of the present invention can enhance user understanding and acceptance by providing a 'context-based explanation' of the diagnosis result through a multi-layered analysis of the user's symptoms, emotional state, functional decline, stress level, validity information, etc.

[0429] The present invention can provide advance notice regarding the possibility of future deterioration and the need for preventive management in cases where functional decline is evident but the disease has not been clearly confirmed; conversely, if temporary symptoms or symptoms caused by external factors (medication, environment, etc.) are identified, the possibility of misdiagnosis can be minimized by including a comment on whether the disease has been confirmed.

[0430] Presenting such comprehensive results enables the provision of specific and highly reliable reports that go beyond mere diagnosis to explain 'why these results occurred' and 'what management is required in the future,' thereby ensuring sufficient medical explainability and clinical applicability as a future Digital Therapeutic Device (SaMD).

[0431] FIG. 10 is a graph for illustrating an embodiment of the present disclosure.

[0432] Referring to FIG. 10, the graph (1000) may include a calculated value (Y-axis) corresponding to a mental illness score (X-axis) calculated in the previous process.

[0433] In one embodiment, the graph (1000) may include a likelihood ratio (1100, LR(s)=(PdP~d)) and a probability Pd (1200) of an emotion appearing in a patient group.

[0434] In one embodiment, the likelihood ratio (1100) may show a form in which it increases rapidly when the anger score increases and then decreases. For example, the likelihood ratio (1100) may be 1.7 when the anger score is 0, and 3.8 when the anger score is 100.

[0435] In one embodiment, the monotonically increasing saturation function (1200) may exhibit a form that increases gradually as the anger score increases. For example, the monotonically increasing saturation function (1200) may be 0.1 when the anger score is 0, and 0.5 when the anger score is 100.

[0436] FIG. 11 is a general schematic diagram of an exemplary computing environment of a computing device in which embodiments of the present disclosure may be implemented.

[0437] Although the present invention has generally been described in relation to computer-executable instructions that can be executed on one or more computers, those skilled in the art will be well aware that the present invention may be combined with other program modules and / or implemented as a combination of hardware and software.

[0438] Computers typically include various computer-readable media. Any medium accessible by a computer may be a computer-readable medium, and such computer-readable media include volatile and non-volatile media, transitory and non-transitory media, and removable and non-removable media. By example, but not by limitation, computer-readable media may include computer-readable storage media and computer-readable transmission media.

[0439] Computer-readable storage media include volatile and non-volatile media, transient and non-transient media, and removable and non-removable media implemented by any method or technique for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer-readable storage media may include any other media that can be accessed by a computer and used to store desired information. Computer-readable storage media may store a computer program comprising instructions that cause one or more processors to perform a method for generating and verifying statements according to the present disclosure.

[0440] A computer-readable transmission medium typically implements computer-readable instructions, data structures, program modules, or other data, etc., on a modulated data signal, such as a carrier wave or other transport mechanism, and includes all information transmission media.

[0441] An exemplary environment for implementing various aspects of the present disclosure, including a computer (2002), is shown, wherein the computer (2002) includes a processing unit (2004), a system memory (2006), and a system bus (2008). The system bus (2008) connects system components, including the system memory (2006) (but not limited thereto), to the processing unit (2004). The processing unit (2004) may be any of various commercial processors.

[0442] The system bus (2008) may be any of several types of bus structures that can be additionally interconnected to a local bus using any of the memory bus, peripheral bus, and various commercial bus architectures. The system memory (2006) includes read-only memory (ROM) (2010) and random access memory (RAM) (2012). The basic input / output system (BIOS) is stored in non-volatile memory (2010), and this BIOS includes basic routines that help transfer information between components within the computer (2002) at times such as during startup.

[0443] The computer (2002) also includes an internal hard disk drive (HDD) (2014), a magnetic floppy disk drive (FDD) (2016), and an optical disk drive (2020). These can be connected to the system bus (2008) by their respective interfaces (2024, 2026, 2028).

[0444] These drives and associated computer-readable media provide non-volatile storage of data, data structures, computer-executable instructions, etc. In the case of a computer (2002), the drives and media correspond to storing any data in a suitable digital format.

[0445] A number of program modules, including an operating system (2030), one or more application programs (2032), other program modules (2034), and program data (2036), may be stored and cached in the drive and RAM (2012).

[0446] The user can input commands and information into the computer (2002) through one or more wired / wireless input devices, such as pointing devices like a keyboard (2038) and a mouse (2040). These and other input devices are often connected to the processing unit (2004) through an input device interface (2042) connected to the system bus (2008), but may be connected by any other interface.

[0447] A monitor (2044) or other type of display device is also connected to the system bus (2008) via an interface such as a video adapter (2046). In addition to the monitor (2044), the computer generally includes other peripheral output devices (not shown), such as speakers, a printer, and so on.

[0448] The computer (2002) may operate in a networked environment using a logical connection to one or more remote computers, such as remote computer(s) (2048), via wired and / or wireless communication. The remote computer(s) (2048) may be any ordinary network node and generally include many or all of the components described for the computer (2002), but for brevity, only the memory storage device (2050) is shown.

[0449] Those skilled in the art of the present disclosure will understand that information and signals may be represented using any various different technologies and techniques.

[0450] Those skilled in the art will understand that the various exemplary logic blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented by electronic hardware, various forms of programs or design code (referred to herein as software for convenience), or a combination of all of these.

[0451] The various embodiments presented herein may be implemented as methods, devices, or articles using standard programming and / or engineering techniques. The term "article" includes a computer program, carrier, or medium accessible from any computer-readable storage device.

[0452] Various modifications to the embodiments presented will be apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the scope of the disclosure. Thus, the disclosure is not limited to the embodiments presented herein, but should be interpreted in the broadest possible scope consistent with the principles and novel features presented herein.

[0453] Description of the presented embodiments is provided so that a person skilled in the art may use or practice the present disclosure. Various modifications to these embodiments will be apparent to a person skilled in the art, and the general principles defined herein may be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present disclosure is not limited to the embodiments presented herein, but should be interpreted in the broadest possible scope consistent with the principles and novel features presented herein.

Claims

Claim 1 A method for providing user-customized mental health diagnosis results by generating adaptive user interaction questions performed by a computing device, comprising: receiving mental health input information from a user; generating a list of candidate mental illnesses suspected of being the user from the mental health input information using multiple artificial intelligence models; determining a target mental illness suffered by the user from the list of candidate mental illnesses using a disease determination module that generates a first user interaction question to verify the presence or absence of symptoms for each of the candidate mental illnesses within the list of candidate mental illnesses; generating a verification result for the target mental illness; and generating the mental health diagnosis result of the user based on the verification result for the target mental illness; wherein the step of generating the verification result for the target mental illness comprises: automatically generating a second user interaction question that queries a symptom time condition including at least one of the time of symptom onset, duration, and frequency for the target mental illness using a validity review module to determine the temporal validity of the target mental illness in response to the determination of the target mental illness; A method comprising: a step of setting the target mental illness as a reference mental illness if the user response to the second user interaction question does not satisfy the target time condition pre-assigned to the target mental illness; and, if the user response satisfies the target time condition, automatically generating a third user interaction question to determine the user's medication and underlying disease using the validity review module, and determining whether to include the target mental illness as a confirmed mental illness in the verification result by determining whether the target mental illness was caused by a secondary factor including at least one of medication, underlying physical disease, and lifestyle habits based on the user response to the third user interaction question. Claim 2 A method according to claim 1, wherein the plurality of artificial intelligence models comprises: a first artificial intelligence model that generates information on symptoms related to the user's mental health using self-report data included in the mental health input information; a second artificial intelligence model that generates a classification result of the user's emotional state using the self-report data; and a third artificial intelligence model that generates language pattern keywords of the user using text data as input. Claim 3 In paragraph 2, the first artificial intelligence model is learned by a first learning dataset in which sentences and symptoms are paired, the first learning dataset including first synthetic data generated by inputting first seed data, in which at least one symptom type corresponding to a pre-collected sentence is labeled, into a Large Language Model (LLM). Claim 4 In paragraph 2, the method wherein the second artificial intelligence model is trained by a second learning dataset composed of voice and emotion pairs using second seed data comprising voice data with pre-attached emotion labels and text transcripts corresponding to the voice data. Claim 5 In paragraph 2, the method wherein the third artificial intelligence model is trained by a third learning dataset composed of sentence data containing language pattern keywords and mental illness pairs, generated by inputting third seed data constructed using previously collected representative language pattern keywords for each mental illness into a large-scale language model (LLM). Claim 6 In claim 1, the mental health input information comprises the user's identification information, question-and-answer data obtained from the user, and self-report data obtained from the user; the user's identification information comprises age, gender, and region as information for categorizing the user within a population group; the user's identification information is used to establish a disease baseline representing the user's baseline probability value for each of a plurality of mental disorders; the question-and-answer data is used to correct the user's baseline probability value as user response data to a query regarding whether the user has experienced symptoms; and the self-report data comprises at least one user speech data among voice and text related to the user's state and symptoms; and the self-report data is used as input to the plurality of artificial intelligence models. Claim 7 A method according to claim 1, wherein a first mental illness classification result is determined using a first output of a first artificial intelligence model among the plurality of artificial intelligence models, and a second mental illness classification result is generated by applying a second output of a second artificial intelligence model among the plurality of artificial intelligence models to the first mental illness classification result, and a third mental illness classification result is generated by applying a third output of a third artificial intelligence model among the plurality of artificial intelligence models to the second mental illness classification result, and a list of mental illness candidates suspected for the user is generated using the third mental illness classification result. Claim 8 In claim 7, the first mental illness classification result, the second mental illness classification result, and the third mental illness classification result each represent the probability of occurrence of the user's mental illness; the first mental illness classification result is generated by applying a first comparison result obtained by comparing a first output with a first reference table defining the relationship between a predefined symptom and a mental illness to a disease baseline representing the user's basic probability value for each of the plurality of mental illnesses; the second mental illness classification result is generated by applying a second comparison result obtained by comparing a second output with a second reference table defining the relationship between a predefined emotion and a mental illness to the first mental illness classification result; and the third mental illness classification result is generated by applying a third comparison result obtained by comparing a third output with a third reference table defining the relationship between a predefined language pattern keyword and a mental illness to the second mental illness classification result. Claim 9 A method wherein, in claim 8, the first mental illness classification result is generated by: a step of generating information on symptoms related to the user's mental health using a first artificial intelligence model that takes self-reported data included in the mental health input information as input; a step of comparing the symptom information with the first reference table to determine, for each combination of a plurality of mental illnesses and a plurality of symptoms, a first multiplier value for upwardly correcting the probability of occurrence of a specific mental illness when a specific symptom is present and a second multiplier value for downwardly correcting the probability of occurrence of a specific mental illness when a specific symptom is not present; and a step of generating the first mental illness classification result in which the basic probability value of the disease baseline is corrected by applying the first multiplier value and the second multiplier value to the disease baseline representing the user's basic probability value for each of the plurality of mental illnesses. Claim 10 A method wherein, in claim 8, the second mental illness classification result is generated by: a step of generating an emotional state classification result indicating the user’s emotional identification information and emotional intensity using a second artificial intelligence model that takes self-reported data included in the mental health input information as input; a step of determining a correlation between a specific emotion and a specific disease by comparing the emotional state classification result with the second reference table; a step of calculating a likelihood ratio by comparing the probability of a specific emotion appearing in a mental illness patient group and the probability of it appearing in a mental illness non-patient group based on the emotional state classification result and the correlation; and a step of generating the second mental illness classification result in which the first mental illness classification result is corrected by applying the likelihood ratio to the first mental illness classification result. Claim 11 A method wherein, in claim 8, the third mental illness classification result is generated by: a step of generating a language pattern keyword of the user using a third artificial intelligence model that takes self-report data included in the mental health input information as input; a step of determining a mapping value corresponding to the generated language pattern keyword for each combination of a plurality of mental illnesses and a plurality of language pattern keywords by comparing the language pattern keyword with the third reference table; and a step of generating the third mental illness classification result in which the second mental illness classification result is corrected by applying the mapping value to the second mental illness classification result. Claim 12 In claim 1, the step of generating the mental illness candidate list comprises: a step of setting a disease baseline representing an initial probability value for each mental illness of the user using the user's identification information included in the mental health input information; a step of first adjusting the initial probability value for each mental illness in the disease baseline using mental health-related symptom information generated from a first artificial intelligence model that takes the user's self-report data included in the mental health input information as input and question-and-answer data included in the mental health input information; a step of secondarily adjusting the firstly adjusted probability value for each mental illness using an emotional state classification result generated from a second artificial intelligence model that takes the user's self-report data included in the mental health input information as input; a step of thirdly adjusting the secondly adjusted probability value for each mental illness using language pattern keywords generated from a third artificial intelligence model that takes the user's self-report data included in the mental health input information as input; and a step of generating the mental illness candidate list suspected for the user based on the thirdly adjusted probability value for each mental illness. Claim 13 In claim 1, the step of generating the mental illness candidate list comprises: determining first candidate mental illnesses that are likely to occur within the user's category using the user's identification information included in the mental health input information; determining second candidate mental illnesses by filtering out some of the first candidate mental illnesses using mental health-related symptom information generated from a first artificial intelligence model that takes the user's self-report data included in the mental health input information as input - the number of the second candidate mental illnesses is smaller than the number of the first candidate mental illnesses -; and determining third candidate mental illnesses by filtering out some of the second candidate mental illnesses using emotional state classification results generated from a second artificial intelligence model that takes the user's self-report data included in the mental health input information as input - the number of the third candidate mental illnesses is smaller than the number of the second candidate mental illnesses -; A method comprising the step of determining fourth candidate mental disorders by filtering out some of the third candidate mental disorders using language pattern keywords generated from a third artificial intelligence model that takes the user's self-report data included in the mental health input information as input, wherein the number of the fourth candidate mental disorders is smaller than the number of the third candidate mental disorders. Claim 14 A method according to claim 1, wherein the disease identification module generates an independent first user interaction question for each of the candidate mental disorders and determines at least one target mental disorder among the candidate mental disorders by scoring user responses to the first user interaction question. Claim 15 A method according to claim 1, wherein the disease identification module generates a first-1 user interaction item among a plurality of user interaction items corresponding to a first candidate mental disease included in the mental disease candidate list, and determines whether to generate a first-2 user interaction item based on whether the user response to the first-1 user interaction item satisfies the condition corresponding to the first candidate mental disease. Claim 16 delete Claim 17 delete Claim 18 In claim 11, the validity review module determines a mental illness grade for the target mental illness based on a time value related to the user's symptom time condition obtained through the second user interaction question when the user response to the second user interaction question satisfies the target time condition. Claim 19 A method according to claim 1, wherein the step of generating the mental health diagnosis result comprises: generating a fourth user interaction item including a question for measuring the stress level for the confirmed mental illness and a question for examining functions related to the user's daily life when the confirmation result is included by the validity review module; and determining the stress level and the risk level to the user's daily life based on the response to the fourth user interaction item. Claim 20 A method according to claim 1, wherein the step of generating the mental health diagnosis result comprises: generating the mental health diagnosis result including the mental health input information, the mental illness candidate list, the target mental illness and the confirmed mental illness included in the verification result, the mental illness grade for the confirmed mental illness, the reference mental illness among the target mental illnesses that is not classified as a confirmed mental illness, the user's emotional state classification result, and the user's language pattern keyword. Claim 21 A computer program stored on a computer-readable storage medium, wherein the computer program performs a method of generating adaptive user interaction questions and providing user-customized mental health diagnosis results when executed by a computing device, the method comprising: receiving mental health input information from a user; generating a list of candidates for a mental illness suspected of being the user from the mental health input information using a plurality of artificial intelligence models; determining a target mental illness suffered by the user from the list of candidates for a mental illness using a disease identification module that generates a first user interaction question to verify the presence or absence of symptoms for each of the candidate mental illnesses within the list of candidates for a mental illness; and generating a verification result for the target mental illness. The method comprises the step of generating the user's mental health diagnosis result based on the verification result for the target mental illness; and the step of generating the verification result for the target mental illness comprises: the step of automatically generating a second user interaction question that queries symptom time conditions including at least one of the time of symptom onset, duration, and frequency for the target mental illness using a validity review module to determine the temporal validity of the target mental illness in response to the determination of the target mental illness; and the step of setting the target mental illness as a reference mental illness if the user response to the second user interaction question does not satisfy the target time condition pre-assigned to the target mental illness.A computer program stored on a computer-readable storage medium, comprising: a step of, when the user response satisfies the target time condition, automatically generating a third user interaction question to determine the user's medication and underlying disease using the validity review module, and determining whether to include the target mental disease as a confirmed mental disease in the verification result by determining whether the target mental disease was caused by a secondary factor including at least one of medication, underlying physical disease, and lifestyle habits based on the user response to the third user interaction question; Claim 22 A computing device that generates adaptive user interaction questions to provide user-customized mental health diagnosis results comprises: a processor; wherein the processor receives mental health input information from a user; generates a list of candidates for a mental illness suspected of being the user from the mental health input information using a plurality of artificial intelligence models; determines a target mental illness suffered by the user from the list of candidates by using a disease identification module that generates a first user interaction question to verify the presence or absence of symptoms for each of the candidate mental illnesses within the list of candidates for mental illness; and generates a verification result for the target mental illness by using a validity review module. And, based on the verification result for the target mental illness, generating the mental health diagnosis result of the user and generating the verification result for the target mental illness comprises: automatically generating a second user interaction question that queries symptom time conditions including at least one of the time of symptom onset, duration, and frequency for the target mental illness using a validity review module to determine the temporal validity of the target mental illness in response to the determination of the target mental illness; and if the user response to the second user interaction question does not satisfy the target time condition pre-assigned to the target mental illness, setting the target mental illness as a reference mental illness; A computing device that, when the above user response satisfies the above target time condition, automatically generates a third user interaction question to determine the user’s medication and underlying disease using the above validity review module, and determines whether to include the target mental disease as a confirmed mental disease in the verification result by determining whether the target mental disease was caused by a secondary factor including at least one of medication, underlying physical disease and lifestyle habits based on the user response to the above third user interaction question.

Images