Health consultation solution based on artificial intelligence trained with clinical experience information of medical specialist

Abstract

The present disclosure relates to a method of operation of a server. The method of operation comprises the steps of: identifying clinical experience information based on the medical history of a specialist for a plurality of patients by medical specialty; acquiring a counseling model trained to perform counseling based on the clinical experience information by medical specialty; and identifying diagnostic information for a user through at least one of the counseling models acquired by medical specialty. The step of identifying diagnostic information for a user includes the steps of: identifying a medical specialty that matches the user's symptoms based on basic counseling content between a classification model for identifying medical specialty and the user's terminal; and acquiring diagnostic information for a user based on advanced counseling content between a counseling model trained on clinical experience information for the medical specialty that matches the user's symptoms and the user's terminal.

Description

Technology Field

[0001] The present disclosure relates to an artificial intelligence-based health consultation solution trained based on clinical experience information of a medical specialist, and more specifically, to a method of operation of a server that generates diagnostic information reflecting a user's genomic data through a consultation model trained based on clinical experience information obtained by integrating linguistic and non-linguistic data generated during the medical specialist's treatment process. Background Technology

[0002] While modern medicine aims for patient-centered, personalized services, existing AI medical systems had a fundamental limitation in that they could not fully learn the tacit knowledge accumulated by medical specialists over decades.

[0003] While medical specialists intuitively sense illness by synthesizing non-verbal signals such as subtle facial expressions and tone of voice, and establish personalized treatment plans by comprehensively considering individual patients' lifestyles, family history, and psychological state, existing AI has limitations in that it relies solely on structured data and explicit medical knowledge, failing to replicate the intuitive judgment and experiential insights of specialists. Prior art literature

[0004] Registered Patent Publication No. 10-2298330 The problem to be solved

[0005] The present disclosure aims to provide a method of operation for a server that generates diagnostic information reflecting a user's genomic data through a counseling model trained based on clinical experience information obtained by integrating linguistic and non-linguistic data generated during the medical treatment process of a specialist.

[0006] The purposes of the present disclosure are not limited to those mentioned above, and other purposes and advantages of the present disclosure not mentioned may be understood from the following description and will be more clearly understood from the embodiments of the present disclosure. Furthermore, it will be readily apparent that the purposes and advantages of the present disclosure can be realized by the means and combinations thereof set forth in the claims. means of solving the problem

[0007] A method of operation of a server according to one embodiment of the present disclosure comprises: a step of identifying clinical experience information based on the medical history of a specialist for a plurality of patients by medical specialty; a step of acquiring a counseling model trained to perform counseling based on the clinical experience information by medical specialty; and a step of identifying diagnostic information for a user through at least one of the counseling models acquired by medical specialty. The step of identifying diagnostic information for a user comprises: a step of identifying a medical specialty that matches the user's symptoms based on basic counseling content between a classification model for identifying medical specialty and the user's terminal; and a step of acquiring diagnostic information for the user based on in-depth counseling content between a counseling model trained on clinical experience information for the medical specialty that matches the user's symptoms and the user's terminal.

[0008] Here, the step of identifying the clinical experience information may include the step of identifying audio data containing the voice of the specialist performing treatment for each of the plurality of patients from the treatment history, the step of identifying the interaction pattern of the specialist based on the audio data, and the step of identifying the decision-making process of the specialist based on the audio data.

[0009] At this time, the step of identifying the interaction pattern of the specialist involves inputting the audio data into a speaker separation model and, based on the output data, dividing it into at least one first unit segment containing the voice of the specialist and at least one second unit segment containing the voice of the patient, extracting speech feature information related to the speech characteristics of the specialist based on the voice of the specialist included in the first unit segment, and identifying the interaction pattern including the extracted speech feature information by medical specialty.

[0010] In addition, the step of identifying the decision-making process of the specialist can be performed by, for each medical specialty, by performing speech recognition on the audio data to obtain medical text corresponding to the audio data, inputting the medical text into an analysis model for identifying the process of treatment, and based on the output data, identifying medical result data from each of the medical texts obtained for each medical specialty, and obtaining the decision-making process including the medical text and the medical result data for each medical specialty.

[0011] Meanwhile, the step of identifying a medical specialty that matches the above user may involve obtaining symptom information from the user's terminal, inputting the symptom information into the classification model, and based on the output data, obtaining the above basic consultation content that includes information about a medical specialty that matches the symptom information.

[0012] At this time, the step of obtaining diagnostic information for the user may involve obtaining the user's genomic data from a medical record server where the user's medical data is stored, inputting the symptom information and the genomic data into the consultation model and, based on the output data, obtaining at least one query to obtain the user's response necessary to identify the user's disease and transmitting it to the user's terminal, and inputting at least one response received from the user's terminal into the consultation model and, based on the output data, obtaining the diagnostic information including at least one target disease finally diagnosed as the user's disease. Effects of the invention

[0013] According to the present disclosure, medical consultation services can be provided even in areas lacking medical infrastructure, thereby bridging regional disparities and shortening diagnosis time.

[0014] In addition, according to the present disclosure, decision support can be provided through multi-faceted analysis in complex cases.

[0015] In addition, according to the present disclosure, it is possible to improve the quality of life through prediction and prevention of disease before it occurs. Brief explanation of the drawing

[0016] FIG. 1 is a flowchart for explaining the operation of a server according to one embodiment of the present disclosure. FIG. 2 is a diagram illustrating the operation of a server acquiring clinical experience information according to one embodiment of the present disclosure. FIG. 3 is a diagram illustrating the operation of a server communicating with a user's terminal according to one embodiment of the present disclosure. FIG. 4 is a diagram illustrating the operation of a server communicating with a medical record server according to one embodiment of the present disclosure. FIG. 5 is a block diagram illustrating the configuration of a server according to one embodiment of the present disclosure. Specific details for implementing the invention

[0017] Before specifically describing the present disclosure, the method of description in the specification and drawings is described.

[0018] First, the terms used in this specification and claims have been selected based on general terms considering their functions in the various embodiments of this disclosure. However, these terms may vary depending on the intent of those skilled in the art, legal or technical interpretations, and the emergence of new technologies. Additionally, some terms have been arbitrarily selected by the applicant. Such terms may be interpreted according to the meanings defined in this specification; in the absence of specific definitions, they may be interpreted based on the overall content of this specification and common technical knowledge in the relevant field.

[0019] In addition, the same reference numbers or symbols described in each drawing attached to this specification represent parts or components that perform substantially the same function. For convenience of explanation and understanding, the same reference numbers or symbols are used to describe different embodiments. That is, even if components having the same reference number are all depicted in multiple drawings, the multiple drawings do not imply a single embodiment.

[0020] Additionally, in this specification and claims, terms including ordinal numbers, such as "first," "second," etc., may be used to distinguish between components. These ordinal numbers are used to distinguish identical or similar components from one another, and the meaning of the terms should not be limited by the use of such ordinal numbers. For example, the order of use or arrangement of components combined with such ordinal numbers should not be restricted by the number. If necessary, each ordinal number may be used interchangeably.

[0021] In this specification, singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, terms such as "comprising" or "consisting of" are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0022] In the embodiments of the present disclosure, terms such as "module," "unit," "part," etc. are used to refer to a component that performs at least one function or operation, and such component may be implemented in hardware or software, or in a combination of hardware and software. Additionally, a plurality of "modules," "units," "parts," etc. may be integrated into at least one module or chip and implemented as at least one processor, except where each needs to be implemented in specific individual hardware.

[0023] Furthermore, in the embodiments of the present disclosure, when a part is described as being connected to another part, this includes not only a direct connection but also an indirect connection through another medium. Additionally, the meaning that a part includes a certain component implies that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0024] In the present disclosure, the expression “at least one of a, b, or c” may refer to “a”, “b”, “c”, “a and b”, “a and c”, “b and c”, “a, b, and c all”, or variations thereof.

[0025] The terms used in this disclosure have been selected to be as widely used and general as possible, taking into account their functions within this disclosure; however, these terms may vary depending on the intent of those skilled in the art, case law, the emergence of new technologies, etc. Additionally, in specific cases, terms have been selected at the applicant's discretion, and in such cases, their meanings will be described in detail in the relevant explanatory sections. Therefore, terms used in this disclosure should be defined not merely by their names, but based on their meanings and the overall content of this disclosure.

[0026] Singular expressions may include plural expressions unless the context clearly indicates otherwise. Terms used herein, including technical or scientific terms, may have the same meaning as generally understood by those skilled in the art as described in this specification. Additionally, terms including ordinal numbers, such as "first" or "second," used in this specification may be used to describe various components, but said components should not be limited by said terms. Such terms are used solely for the purpose of distinguishing one component from another.

[0027] When a part of a specification is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components. Furthermore, terms such as "part" or "module" as used in the specification refer to a unit that processes at least one function or operation, and this may be implemented in hardware or software, or as a combination of hardware and software.

[0028] Embodiments of the present disclosure are described below with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present disclosure may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present disclosure in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0029] FIG. 1 is a flowchart for explaining the operation of a server according to one embodiment of the present disclosure.

[0030] The server (100) can identify and provide diagnostic information based on the content of the consultation performed by mimicking the diagnostic process of a specialist.

[0031] To this end, the server (100) can perform consultation with the user's terminal (200) through a consultation model (20) trained based on clinical experience information related to the process of each specialist in different medical departments (e.g., internal medicine, surgery, pediatrics, psychiatry, etc.) performing medical treatment on a patient.

[0032] Meanwhile, the server (100) may be implemented as a server device or system including at least one computer. For example, the server (100) may be an FTP server (File Transfer Protocol Server), a web server, a database server, or a cloud server, but is not limited thereto. The server (100) may perform one or more functions and operations, and may be implemented as a single device or distributed across multiple devices so that each function and operation is implemented.

[0033] Referring to FIG. 1, the server (100) can identify clinical experience information based on the medical history of specialists by medical specialty (S110).

[0034] Specifically, the server (100) can identify clinical experience information based on the medical history of specialists for multiple patients by medical specialty.

[0035] Here, the medical history may include video data of a specialist performing treatment on each of multiple patients, audio data containing the voices of the patient and the specialist, recorded during the treatment of each of multiple patients, the patient's medical data (e.g., treatment schedule, medical history, diagnostic history, prescription, test results, treatment records, medical records, test data, etc.), the patient's genomic data, and the types of tests performed according to the treatment (e.g., ultrasound, blood test, CT, MRI, X-ray, etc.). In this case, the video data and audio data for the same patient may be matched and recorded in the medical history.

[0036] Here, genomic data may correspond to result data of genomic analysis tests and may include information on genomes such as DNA, RNA, epigenome, transcriptome, proteome, metabolome, and microbiome.

[0037] Here, clinical experience information may include multiple decision-making processes that a specialist goes through to diagnose a patient (e.g., specialist's inquiries based on the patient's symptoms, test instructions, etc.), and interaction patterns for each decision-making process (e.g., specialist's speech rate, gestures, facial expressions based on the patient's response, etc.).

[0038] A more detailed explanation regarding this will be provided later through Fig. 2.

[0039] The server (100) can obtain a counseling model (20) trained to perform counseling based on clinical experience information (S120).

[0040] In one embodiment, the server (100) can train a counseling model (20) for performing counseling by medical specialty based on clinical experience information identified by medical specialty.

[0041] That is, the server (100) may include a consultation model (20) for each medical specialty, and different consultation models (20) may correspond to models trained based on clinical experience information of different medical specialties, and accordingly, the server (100) may select a consultation model (20) according to a medical specialty (matching the user's symptom information) and perform consultation (question and answer) with the user's terminal (200) using the selected consultation model (20).

[0042] The counseling model (20) may be a model based on a language model that performs natural language processing (NLP) that recognizes and interprets natural language, and natural language generation (NLG) that generates, reinforces, and reconstructs words and sentences.

[0043] Here, the language model is a model trained to output the statistically most appropriate output value (e.g., a vector value corresponding to the answer to the input text) based on input values ​​(e.g., text such as natural language, sentences, words, etc.), and may be a large language model, such as, but not limited to, GPT, Phi, BERT, LaMDA, LLaMA, RoBERTa, etc.

[0044] Accordingly, the consultation model (20) can convert input text data into vector values ​​in a latent space through text embedding, identify the context of words and phrases with similar meanings as well as other relationships between words such as parts of speech based on the vector values ​​in the latent space, and output result values ​​for natural language processing, natural language generation, etc. based on the relationships between these vector values.

[0045] In addition, the counseling model (20) may be trained to generate counseling content including a query or target disease based on a Dynamic Knowledge Graph, Temporal Graph Networks, Dynamic Embedding, Causal Inference, etc.

[0046] The counseling model (20) can mimic the diagnostic process of a specialist based on clinical experience information, identify at least one candidate disease that can be diagnosed as a user's disease according to the user's symptom information and genomic data input into the counseling model (20), and generate counseling content including at least one query for selecting at least one of the candidate diseases as a target disease.

[0047] Here, the target disease refers to a disease that is finally diagnosed as corresponding to the user's disease as a result of in-depth consultation with the user, and the query included in the consultation content may correspond to a query intended to obtain the user's response necessary to identify whether a candidate disease corresponds to the target disease.

[0048] At this time, the server (100) can provide the consultation content generated by the consultation model (20) to the user's terminal (200), and can obtain a response to a query included in the consultation content from the user's terminal (200).

[0049] Accordingly, the server (100) can input the response obtained from the user's terminal (200) into the consultation model (20) and, based on the output data, identify at least one of the identified candidate diseases as a target disease.

[0050] Specifically, the counseling model (20) can identify the probability that each candidate disease is expected to be a target disease based on the response obtained from the user's terminal (200) regarding the query included in the counseling content, select at least one candidate disease in which the identified probability exceeds a threshold as a target disease, and generate counseling content including information about the selected target disease.

[0051] If there is no candidate disease identified with a probability exceeding a threshold, the counseling model (20) may generate counseling content including at least one additional query based on the response obtained from the user's terminal (200).

[0052] At this time, the counseling model (20) can generate counseling content including at least one additional query to diagnose at least one of the remaining candidate diseases as the user's disease, excluding at least one candidate disease among the candidate diseases whose probability of being the target disease is less than the reference probability (e.g., 30%).

[0053] If consultation content including additional queries is obtained from the consultation model (20), the server (100) provides the consultation content including additional queries to the user's terminal (200) and can obtain a response to the additional queries from the user's terminal (200).

[0054] Accordingly, the server (100) may input a response to an additional query into the consultation model (20) and, based on the output data, select at least one of the candidate diseases as a target disease and generate consultation content including information about the selected target disease, or generate consultation content including at least one additional query.

[0055] That is, the counseling model (20) can generate counseling content including a query to obtain the user's response necessary to select a target disease according to the decision process included in the clinical experience information based on the response obtained from the user's terminal (200).

[0056] The server (100) can obtain diagnostic information about the user based on the consultation content performed by the consultation model (20) based on the user's symptom information (S130).

[0057] In one embodiment, the server (100) can identify diagnostic information about the user based on the consultation content generated by the consultation model (20) based on the user's symptom information.

[0058] At this time, the server (100) may use a classification model (10) trained to identify medical departments according to symptoms in order to identify a medical department that matches the user's symptom information.

[0059] In this case, the server (100) can identify the consultation content obtained from the classification model (10) as basic consultation content through consultation (question and answer) between the classification model (10) and the user's terminal (200), and the consultation content obtained from the consultation model (20) as advanced consultation content through consultation (question and answer) between the consultation model (20) and the user's terminal (200).

[0060] In one embodiment, the server (100) can identify a medical specialty that matches the user's symptoms based on the (basic) consultation content between the classification model (10) and the user's terminal (200), and obtain diagnostic information about the user based on the (advanced) consultation content between the user's terminal (200) and a counseling model (20) trained based on clinical experience information regarding the medical specialty that matches the user's symptoms.

[0061] Specifically, the server (100) can obtain user symptom information from the user's terminal (200), input the user symptom information into the classification model (10), and based on the output data, obtain (basic) consultation content including a medical department corresponding to the user symptom information.

[0062] Here, symptom information may include text about the symptoms (e.g., fever, body temperature 38 degrees, stomach ache, abdominal pain, dizziness, etc.).

[0063] To this end, the classification model (10) may be a model trained to identify medical specialties according to symptoms. For example, the classification model (10) may be a model trained based on training data in which medical specialties are labeled for each of the different symptoms by at least one keyword (e.g., fever, body temperature, vomiting, depression, etc.), and may be a model trained based on learning algorithms such as Logistic Regression, Support Vector Machine (SVM), Random Forest, Gradient Boosting Machine, and Convolutional Neural Network (CNN), but is not limited thereto.

[0064] Additionally, the server (100) can identify at least one symptom type based on symptom information and identify the urgency of medical treatment for the user based on the identified symptom type.

[0065] In one embodiment, the server (100) can identify at least one symptom type for a user based on the output data by inputting symptom information into an analysis model for identifying symptom types.

[0066] Here, types of symptoms may include, but are not limited to, headache, abdominal pain, muscle pain, chest pain, back pain, delirium, amnesia, vomiting, dizziness, cold sweats, fatigue, phlegm, cough, bleeding, etc.

[0067] The analysis model may be a model trained based on Natural Language Processing (NLP) to identify at least one symptom type, the time of onset or duration of the symptom type, etc., from text in which symptoms are described in natural language.

[0068] To this end, the analysis model may correspond to a model trained based on training data consisting of multiple training texts in which symptoms are described in natural language and entity labels such as symptoms, onset time, and duration included in each training text.

[0069] Here, the analysis model may be a model trained based on various learning algorithms such as CRF (Conditional Random Field), BERT (Bidirectional Encoder Representations from Transformers), SVM (Support Vector Machine), RNN (Recurrent Neural Network), LSTM (Long Short-Term Memory), and GRU (Gated Recurrent Unit), but is not limited thereto.

[0070] At this time, the server (100) can identify the urgency of medical treatment for the user based on the value calculated by summing the preset values ​​for each symptom type.

[0071] Here, as the type of symptom that has an adverse effect on the user's health (e.g., difficulty breathing, unconsciousness, life-threatening condition, etc.) if left untreated is set to a larger value, an urgency level with a larger value can be identified as a situation requiring rapid medical treatment for the user. For example, the urgency level may have a value between 0 and 100, but is not limited thereto.

[0072] Meanwhile, the server (100) can identify a medical specialty that matches the user's symptom information based on the basic consultation content, and select a trained consultation model (20) based on clinical experience information regarding the medical specialty that matches the user's symptom information.

[0073] In this case, the server (100) can obtain diagnostic information about the user based on the in-depth consultation content between the selected consultation model (20) and the user's terminal (200).

[0074] Specifically, the server (100) can input the user's symptom information and genomic data into the counseling model (20) and, based on the output data, obtain at least one query to determine whether at least one candidate disease identified based on the user's symptom information corresponds to the target disease and provide it to the user's terminal (200).

[0075] Accordingly, the server (100) may obtain a response to the above query from the user's terminal (200), input the response into the consultation model (20) and obtain information (e.g., disease name, probability, etc.) regarding a candidate disease corresponding to a target disease among the candidate diseases based on the output data, and may also obtain at least one additional query based on the response.

[0076] When an additional query is obtained, the server (100) can provide the additional query to the user's terminal (200) and obtain a response to the additional query from the user's terminal (200).

[0077] Accordingly, the server (100) may input the response (to an additional query) obtained from the user's terminal (200) into the consultation model (20) and obtain information about candidate diseases corresponding to the target disease based on the output data, and may also obtain at least one additional query based on the response.

[0078] At this time, when information regarding a target disease is obtained from the consultation model (20), the server (100) can identify the target disease included in the consultation content between the consultation model (20) and the user's terminal (200), and can identify diagnostic information including the identified target disease.

[0079] Here, the diagnostic information may include the name of the target disease, the identified probability for candidate diseases corresponding to the target disease, etc.

[0080] In addition, when multiple additional queries are obtained, the server (100) can identify the recognition accuracy regarding the degree to which the user recognizes symptoms based on the similarity between the multiple additional queries, and identify whether remote consultation or face-to-face consultation with medical staff is necessary according to the identified recognition accuracy.

[0081] In one embodiment, the server (100) maintains consultation through the consultation model (20) if the recognition accuracy is greater than or equal to a preset value (e.g., 60%), and if the recognition accuracy is less than a preset value, it can identify whether remote consultation or face-to-face consultation is necessary based on the urgency and recognition accuracy (identified according to the content of consultation with the classification model (10)).

[0082] To this end, the server (100) can identify the user's recognition accuracy according to the order in which each additional query with a similarity exceeding a threshold is provided to the user's terminal (200) and the number of additional queries with a similarity exceeding a threshold.

[0083] For example, the server (100) can convert each of the multiple additional queries into a vector value in a latent space through text embeddings and calculate the similarity between the additional queries based on the vector value in the latent space.

[0084] Specifically, the server (100) may identify similarity based on the angle between vectors for different additional queries by utilizing cosine similarity. In this case, the server (100) can calculate a higher similarity as the angle between vectors becomes smaller. Additionally, the server (100) may identify similarity based on the distance between vectors for different additional queries. Here, the server (100) can calculate the distance between vectors based on the difference in vector values ​​by dimension, and can calculate a higher similarity as the distance between vectors becomes closer.

[0085] At this time, the server (100) sets one of the multiple additional queries as a reference query and identifies the similarity between the additional query provided to the user's terminal (200) after the reference query and the reference query, and can identify the similarity with the reference query sequentially according to the order in which the additional queries are provided.

[0086] For example, the server (100) may set the additional query provided to the user's terminal (200) first among a plurality of additional queries as the reference query, and then if no additional query is identified whose similarity to the reference query exceeds a threshold, the additional query provided to the user's terminal (200) in the order following the additional query set as the reference query may be set as the reference query.

[0087] Here, when an additional query is identified in which the similarity with the reference query exceeds a threshold, the server (100) identifies the additional query provided to the user's terminal (200) next to the reference query as the first target query, and identifies the additional query provided to the user's terminal (200) next to the additional query in which the similarity with the reference query exceeds a threshold as the second target query.

[0088] In this case, the server (100) can identify the similarity between the first target query and the second target query, and if the similarity between the first target query and the second target query is below a threshold (identifying that the user has entered a different response to a similar query), the user's recognition accuracy can be lowered.

[0089] If the similarity between the first target query and the second target query exceeds a threshold, the server (100) may identify each of the additional queries provided to the user's terminal (200) in the order following each of the first and second target queries as the third target query and the fourth target query, and perform an operation to identify the similarity between the third target query and the fourth target query, and may repeat this process until the similarity between the identified target queries is below the threshold.

[0090] In this case, the server (100) can reduce the intensity of lowering the user's recognition accuracy as the number of times the operation to identify similarity between target queries is repeated (since there are many cases where the user has entered similar responses to similar queries).

[0091] If no additional query is identified in which the similarity exceeds a threshold, the server (100) can increase the user's recognition accuracy, and the greater the number of additional queries provided to the user's terminal (200), the greater the intensity of increasing the user's recognition accuracy.

[0092] Here, the server (100) can set an initial value of the user's recognition accuracy and identify the recognition accuracy by adjusting it upward or downward from the initial value.

[0093] For example, the server (100) can set an initial value for the user's recognition accuracy based on the user's urgency, and can set the initial value low by determining that the user's judgment ability is degraded as the user's urgency increases.

[0094] Meanwhile, the server (100) can set a standard urgency level that serves as a criterion for identifying whether remote counseling or face-to-face counseling is necessary when the recognition accuracy is less than a preset value, and can identify whether remote counseling or face-to-face counseling is necessary by comparing the standard urgency level with the user's urgency level (identified based on the user's symptom information).

[0095] For example, the server (100) can set the standard urgency lower as the difference between the user's recognition accuracy and the preset value increases.

[0096] Accordingly, if the user's urgency is below the standard urgency level, the server (100) maintains consultation between the consultation model (20) and the user's terminal (200), and if the user's urgency is above the standard urgency level, it identifies that remote consultation or face-to-face consultation is required and can provide a message to the user's terminal (200) recommending remote consultation or face-to-face consultation with medical staff for a more accurate diagnosis.

[0097] Through this, the server (100) can reduce the possibility of misdiagnosis and enhance user safety by comprehensively considering the user's level of self-awareness and urgency regarding the user's symptoms and requesting a timely transition to appropriate counseling (remote counseling or face-to-face counseling) rather than stopping at counseling through a simple counseling model (20).

[0098] FIG. 2 is a diagram illustrating the operation of a server acquiring clinical experience information according to one embodiment of the present disclosure.

[0099] Referring to FIG. 2, the server (100) can obtain video data including at least one image frame of a specialist performing medical treatment for each of the multiple patients, audio data including the voice of the patient and the voice of the specialist obtained as medical treatment is performed for each of the multiple patients, and medical history including medical data for each of the multiple patients.

[0100] At this time, the server (100) can identify a target object corresponding to a specialist from each of the multiple image frames constituting the video data.

[0101] Specifically, the server (100) can identify a target object corresponding to a person among the objects included in each of the multiple image frames based on the output data from inputting multiple image frames into an object recognition model.

[0102] For example, the object recognition model may be a model trained to perform preprocessing (e.g., image resizing, normalization, etc.) on an input image input to the object recognition model, then extract features (e.g., edges, textures, contours, etc.) from the preprocessed input image to generate a feature map, identify at least one candidate region (bounding box) containing an object on the input image based on the generated feature map, and identify the type of object (e.g., person, animal, furniture, etc.) contained in each of the identified candidate regions.

[0103] At this time, the server (100) can identify the motion and facial expression of the target object by comparing the target object included in each of the multiple image frames.

[0104] For example, the server (100) can track the movement of the target object by analyzing the degree of change between consecutive image frames and identify changes in facial expression based on the face area of ​​the target object.

[0105] Through this, the server (100) can obtain behavioral feature information regarding the motion and facial expressions of the medical staff from the video data.

[0106] In addition, the server (100) can separate the voices included in the audio data by speaker and identify at least one first unit section containing the voice of a speaker corresponding to a specialist and at least one second unit section containing the voice of a speaker corresponding to a patient.

[0107] Specifically, the server (100) can identify the speech voice of a specialist and the speech voice of a patient based on the output data by inputting audio data into a speaker separation model for distinguishing speech voices included in audio data by speaker.

[0108] Here, the speaker separation model divides audio data into multiple unit segments, obtains at least one vector reflecting speech feature information (e.g., pitch, pronunciation, intonation, etc.) included in each unit segment, calculates similarity between unit data based on the distance between the obtained vectors, identifies unit segments with a calculated similarity greater than or equal to a preset value as containing the voice of the same speaker, and may be a model trained to match the label of each speaker to each of the multiple unit segments.

[0109] Meanwhile, the speaker separation model may be a model trained based on various learning algorithms such as GMM, HMM, GMM-HMM, DNN, and DNN-HMM, but is not limited thereto.

[0110] Accordingly, the server (100) can compare speaker-specific voice feature information identified from different audio data (obtained by performing medical treatment on each of the multiple patients) to identify a speaker repeatedly identified from each of the different audio data as a specialist and identify a speaker who is not a specialist as a patient.

[0111] At this time, the server (100) can extract (speech) feature information related to speech characteristic elements of a specialist (e.g., change in voice tone, speech speed, length of silence interval, etc.) from a first unit section corresponding to a speech section of a specialist in audio data.

[0112] Specifically, the server (100) can extract at least one feature value related to at least one speech characteristic element from a first unit section corresponding to a speech section of a specialist in audio data, and obtain (speech) feature information including at least one extracted feature value.

[0113] For example, the server (100) can extract at least one feature value from audio data based on FFT (Fast Fourier Transform), MFCC (Mel-Frequency Cepstral Coefficient), PLP (Perceptual Linear Prediction), etc.

[0114] Through this, the server (100) can obtain characteristic information of the specialist (speech) from the audio data.

[0115] At this time, the server (100) can identify the interaction pattern of the specialist based on the specialist's behavioral characteristic information and speech characteristic information.

[0116] Here, the interaction pattern is a pattern related to the way a specialist communicates with a patient, and may include interaction aspects such as the conversational flow leading to the specialist's questions and the patient's responses or the patient's questions and the specialist's responses, changes in the tone and speed of speech used by the specialist to convey information to the patient, changes in the specialist's facial expressions or gestures in response to the patient's reaction, and the frequency and length of silence intervals occurring during speech.

[0117] In addition, the server (100) can obtain medical text by performing speech recognition on audio data.

[0118] Specifically, the server (100) can input audio data into a pre-trained speech recognition model to obtain medical text corresponding to the audio data.

[0119] Here, the speech recognition model may be a model that performs the process of converting input speech data into text. The speech recognition model may consist of, but is not limited to, a feature extraction module, an acoustic model (AM), a pronunciation dictionary, a language model (LM), and a decoder.

[0120] Speech recognition models can analyze signals to extract acoustic features (Feature Extraction). Specifically, in the feature extraction step, representative acoustic features such as Mel-Frequency Cepstral Coefficients (MFCC) or spectrograms can be extracted by analyzing the input speech signal.

[0121] The extracted features are converted into pronunciation units through an Acoustic Model (AM), combined with a Language Model (LM) to generate contextually correct sentences, and finally, through a decoding step, a final text output can be generated.

[0122] Acoustic models (AMs) analyze feature vectors and convert them into phonemes, and Hidden Markov Models (HMMs), deep learning-based RNNs, or Transformer models can be used.

[0123] An Acoustic Model (AM) can be a component that performs the role of converting input speech data into pronunciation units (phonemes). The acoustic model learns which phoneme corresponds to a specific sound in a given speech signal and can be implemented using Hidden Markov Models (HMM), Deep Neural Networks (DNN), Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), or Transformer-based models. By training on a large amount of text data corresponding to the speech data, the acoustic model optimizes the relationship between specific acoustic patterns and pronunciations, and can perform noise removal and correction for acoustic variations (speaker intonation, environmental noise, etc.).

[0124] A pronunciation dictionary can perform the operation of mapping the output of an acoustic model to a word-by-word basis by providing phoneme combinations of specific words.

[0125] Language models (LMs) can perform the task of reducing errors by predicting contextually appropriate word sequences. Language models (LMs) can convert phoneme and word combinations predicted by acoustic models (AMs) into contextually correct sentences. Specifically, N-gram methods capable of predicting the probability of word occurrence, deep learning-based LSTM (Long Short-Term Memory), and Transformer (GPT, BERT) models can be applied to language models.

[0126] The decoder can perform the operation of combining the outputs of the acoustic model and the language model to ultimately determine the most suitable text.

[0127] Here, the server (100) can divide the medical text obtained through voice recognition into a first text corresponding to the voice included in the first unit section and a second text corresponding to the voice included in the second unit section, and obtain the consultation text divided into the first text and the second text.

[0128] For example, the consultation text may correspond to text in which the specialist's questions or answers and the patient's questions or answers are listed sequentially, identified according to the conversation (consultation) between the specialist and the patient.

[0129] At this time, the server (100) can identify treatment result data based on the output data by inputting consultation text into an analysis model for identifying the treatment process.

[0130] Medical outcome data may include the patient's disease determined by the specialist based on the conversation between the specialist and the patient, medications prescribed to the patient, and types of tests instructed to be performed on the patient.

[0131] The analysis model may be a model trained to identify treatment outcome data from data related to medical care (e.g., consultation text, etc.) that proceeds according to the conversation (consultation) between a specialist and a patient.

[0132] For example, the analysis model may be a model trained to identify treatment results from consultation texts based on training data in which multiple training treatment texts for each of various treatment situations and the doctor's treatment results on the training treatment texts are labeled respectively.

[0133] Meanwhile, the analysis model may be a model trained based on various learning algorithms such as SVM (Support Vector Machine), Random Forest, Decision Tree, RNN (Recurrent Neural Network), Transformer, and BERT (Bidirectional Encoder Representations from Transformers), but is not limited thereto.

[0134] Through this, the server (100) can obtain a decision process including consultation text and treatment result data. At this time, each decision process may be matched with the genomic information of each patient.

[0135] Accordingly, the server (100) can obtain clinical experience information including the interaction patterns and decision-making processes of the specialist.

[0136] At this time, the server (100) can train a counseling model (20) based on medical information and clinical experience information including information on treatment types (e.g., radiation therapy, drug therapy, surgical treatment, physical therapy, immunotherapy, dietary therapy, etc.) and treatment drugs for each of a plurality of diseases, and experimental data for each of the various treatment types or various drugs (e.g., changes in the subject's biological data, treatment types applied to each subject, drugs taken by each subject, etc.).

[0137] To this end, the server (100) can compose medical information by obtaining information on treatment types and treatment drugs for each of a plurality of diseases from an external database, and information on experimental data for each of various treatment types or various drugs.

[0138] Through this, the server (100) can train a counseling model (20) based on medical information including explicit medical knowledge and clinical experience information reflecting tacit knowledge acquired by a specialist according to the specialist's clinical experience.

[0139] Meanwhile, when the server (100) provides the deep counseling content (additional inquiry or diagnostic information) obtained through the output of the counseling model (20) to the user's terminal (200), if it is identified that the provision of deep counseling content in audio format has been requested based on user input received through the user's terminal (200), the server can obtain the deep counseling content in audio format with the interaction pattern (described above) reflected through the counseling model (20) and provide it to the user's terminal (200).

[0140] At this time, the counseling model (20) can generate in-depth counseling content in an audio format that reflects interaction patterns based on a text-to-speech (TTS) model.

[0141] For example, a speech synthesis model can convert text into audio data by analyzing the grammatical structure of text, dividing it into sentence units, converting each word into multiple phonemes, generating speech based on the converted phonemes, and converting it into a digital signal.

[0142] Here, the speech synthesis model may correspond to a model trained based on the interaction patterns of a specialist, and as there is a speech synthesis model for each medical specialty consultation model (20), the speech synthesis model for each medical specialty consultation model (20) may be a model trained based on the interaction patterns included in the clinical experience information of each medical specialty, and may be a model trained to convert text-format data (e.g., in-depth consultation content) input into the speech synthesis model into audio-format data and output it by reflecting the changes in speech tone and speech speed used by the specialist to convey information to the patient according to the interaction patterns, and the frequency and length of silent intervals occurring during speech.

[0143] Through this, the counseling model (20) can input the output data (in-depth counseling content) of the counseling model (20) into the speech synthesis model and obtain in-depth counseling content in audio format based on the output data.

[0144] FIG. 3 is a diagram illustrating the operation of a server communicating with a user's terminal (200) according to one embodiment of the present disclosure.

[0145] Referring to FIG. 3, the server (100) can communicate with the user's terminal (200).

[0146] In one embodiment, the server (100) can provide diagnostic information about the user to the user's terminal (200).

[0147] Additionally, the server (100) may identify treatment plan information for the target disease included in the diagnostic information and provide it to the user's terminal (200).

[0148] Here, treatment plan information may include the type of treatment for the target disease (e.g., radiation therapy, drug therapy, surgical treatment, physical therapy, immunotherapy, dietary therapy, etc.), treatment drugs, etc.

[0149] Specifically, as the counseling model (20) is trained based on medical information, the counseling model (20) may generate counseling content including information on the target disease and treatment plan for the target disease.

[0150] Through this, the server (100) can transmit target disease and treatment plan information to the user's terminal (200).

[0151] FIG. 4 is a diagram illustrating the operation of a server communicating with a medical record server according to one embodiment of the present disclosure.

[0152] Referring to FIG. 4, the server (100) can communicate with the medical record server (300).

[0153] The medical record server (300) may correspond to a server that manages the medical system, EMR (Electronic Medical Record) system, or ENR (Electronic Nursing Record) system of a medical institution.

[0154] For example, the medical record server (300) can electronically record and manage medical data including the user's medical schedule, medical history, diagnostic history, prescription, test results, treatment records, medical records, test data, patient registration number, etc., and may be an FTP server (File Transfer Protocol Server), a web server, a database server, or a cloud server, but is not limited thereto.

[0155] In one embodiment, the server (100) can obtain genomic data included in the user's medical data from the medical record server (300).

[0156] Additionally, in one embodiment, the server (100) can transmit at least one of the user's diagnosis information and treatment plan information to the medical record server (300).

[0157] Accordingly, the medical record server (300) can add and record at least one of diagnostic information and treatment plan information to the user's medical data, and as at least one medical staff terminal can view the user's medical data through the medical record server (300), the diagnostic information or treatment plan information recorded in the user's medical data can also be checked.

[0158] FIG. 5 is a block diagram illustrating the configuration of a server according to one embodiment of the present disclosure.

[0159] Referring to FIG. 5, the server (100) may include memory (110), a communication interface (120), and a processor (130).

[0160] The memory (110) stores various programs or data temporarily or non-temporarily and transmits the stored information to the processor upon the call of the processor (130). Additionally, the memory (110) can store various information required for the operation, processing, or control operation of the processor (130) in an electronic format.

[0161] The memory (110) may include, for example, at least one of a main memory and an auxiliary memory. The main memory may be implemented using a semiconductor storage medium such as ROM and / or RAM.

[0162] ROM may include, for example, conventional ROM, EPROM, EEPROM and / or MASK-ROM. RAM may include, for example, DRAM and / or SRAM. The secondary storage device may be implemented using at least one storage medium capable of storing data permanently or semi-permanently, such as a flash memory device, SD (Secure Digital) card, solid state drive (SSD), hard disk drive (HDD), optical recording media such as a magnetic drum, compact disc (CD), DVD or laser disc, magnetic tape, magneto-optical disc and / or floppy disk.

[0163] In one embodiment, the memory (110) may include a classification model (10) and a consultation model (20).

[0164] The communication interface (120) may include a wireless communication interface, a wired communication interface, or an input interface. The wireless communication interface may communicate with various external devices using wireless communication technology or mobile communication technology. Such wireless communication technologies may include, for example, Bluetooth, Bluetooth Low Energy, CAN communication, Wi-Fi, Wi-Fi Direct, ultrawide band (UWB), Zigbee, infrared data association (IrDA), or near field communication (NFC), and mobile communication technologies may include 3GPP, Wi-Max, LTE (Long Term Evolution), 5G, etc.

[0165] A wireless communication interface can be implemented using an antenna, a communication chip, a substrate, etc., capable of transmitting electromagnetic waves to the outside or receiving electromagnetic waves transmitted from the outside.

[0166] A wired communication interface can communicate with various devices based on a wired communication network. Here, the wired communication network can be implemented using physical cables, such as, for example, pair cables, coaxial cables, fiber optic cables, or Ethernet cables.

[0167] Depending on the embodiment, either the wireless communication interface or the wired communication interface may be omitted. Accordingly, the server (100) may include only the wireless communication interface or only the wired communication interface. In addition, the server (100) may be equipped with an integrated communication interface (120) that supports both wireless connection via the wireless communication interface and wired connection via the wired communication interface.

[0168] The server (100) is not limited to having one communication interface (120) that performs a communication connection in one manner, but may include multiple communication interfaces (120) that perform communication connections in multiple manners.

[0169] The processor (130) controls the overall operation of the electronic device. Specifically, the processor (130) is connected to the configuration of the electronic device including memory as described above, and can control the overall operation of the electronic device by executing at least one instruction stored in the memory as described above. In particular, the processor (130) can be implemented as a single processor as well as as a plurality of processors.

[0170] The processor (130) may be implemented in various ways. For example, one or more processors (130) may include one or more of a CPU (Central Processing Unit), GPU (Graphics Processing Unit), APU (Accelerated Processing Unit), MIC (Many Integrated Core), DSP (Digital Signal Processor), NPU (Neural Processing Unit), hardware accelerator, or machine learning accelerator. One or more processors (130) may control one or any combination of other components of an electronic device and may perform operations or data processing related to communication. One or more processors (130) may execute one or more programs or instructions stored in memory. For example, one or more processors (130) may perform a method according to one embodiment of the present disclosure by executing one or more instructions stored in memory.

[0171] In embodiments of the present disclosure, the processor (130) may mean a system-on-chip (SoC) in which one or more processors (130) and other electronic components are integrated, a single-core processor, a multi-core processor, or a core included in a single-core processor or a multi-core processor, wherein the core may be implemented as a CPU, GPU, APU, MIC, DSP, NPU, hardware accelerator or machine learning accelerator, etc., but the embodiments of the present disclosure are not limited thereto.

[0172] In the case where the method according to one embodiment of the present disclosure includes a plurality of operations, the plurality of operations may be performed by a single processor (130) or by a plurality of processors (130). For example, when a first operation, a second operation, and a third operation are performed by the method according to one embodiment, the first operation, the second operation, and the third operation may all be performed by a first processor, or the first operation and the second operation may be performed by a first processor (e.g., a general-purpose processor) and the third operation may be performed by a second processor (e.g., an artificial intelligence dedicated processor).

[0173] One or more processors (130) may be implemented as a single-core processor including one core, or as one or more multicore processors including multiple cores (e.g., homogeneous multicore or heterogeneous multicore). When one or more processors (130) are implemented as multicore processors, each of the multiple cores included in the multicore processor may include internal processor memory such as on-chip memory, and a common cache shared by multiple cores may be included in the multicore processor (130). Additionally, each of the multiple cores included in the multicore processor (130) (or some of the multiple cores) may independently read and execute program instructions for implementing a method according to one embodiment of the present disclosure, or all (or some) of the multiple cores may be linked together to read and execute program instructions for implementing a method according to one embodiment of the present disclosure.

[0174] When a method according to one embodiment of the present disclosure includes a plurality of operations, the plurality of operations may be performed by one of the plurality of cores included in a multi-core processor, or may be performed by a plurality of cores. For example, when a first operation, a second operation, and a third operation are performed by a method according to one embodiment, the first operation, the second operation, and the third operation may all be performed by a first core included in a multi-core processor, or the first operation and the second operation may be performed by a first core included in a multi-core processor and the third operation may be performed by a second core included in a multi-core processor.

[0175] In one embodiment, the processor (130) can control a data collection module (131), a model training module (132), a chatbot module (133), etc. Each of these modules may correspond to a functional unit module implemented in software and / or hardware.

[0176] The data collection module (131) is a module for acquiring data.

[0177] In one embodiment, the data collection module (131) can acquire the medical history of a specialist for multiple patients. For example, the data collection module (131) may acquire at least one image captured through at least one camera installed in the specialist's examination room, and may acquire audio data containing the specialist's voice through at least one microphone installed in the specialist's examination room.

[0178] In addition, the data collection module (131) may also obtain the specialist's medical history for multiple patients from the specialist's terminal.

[0179] The model learning module (132) can train the counseling model (20).

[0180] In one embodiment, the model learning module (132) can extract behavioral characteristic information of a specialist from video data, extract speech characteristic information of a specialist from audio data, and identify clinical experience information based thereon.

[0181] At this time, the model learning module (132) can train the counseling model (20) based on clinical experience information and medical information.

[0182] For example, the model learning module (132) can train the counseling model (20) to generate counseling content based on a node encoder that converts entities such as genes, diseases, symptoms, and treatments into numerical representations (e.g., vectors) and an edge encoder that converts relationships between entities into numerical representations, based on clinical experience information and medical information.

[0183] The chatbot module (133) is a module for performing consultation (question and answer) with the user.

[0184] In one embodiment, the chatbot module (133) can obtain user symptom information based on user input received through the user's terminal (200) and input the symptom information into the classification model (10).

[0185] Additionally, the chatbot module (133) can transmit a query or target disease to the user's terminal (200) based on the consultation content generated by the consultation model (20).

[0186] Specifically, the chatbot module (133) inputs the user's symptom information into the consultation model (20), and if the output consultation content includes a query, it provides the query to the user's terminal (200), obtains a response to the query from the user's terminal (200), and inputs the obtained response into the consultation model (20).

[0187] Additionally, the chatbot module (133) can input the user's symptom information into the consultation model (20), and if the output consultation content includes the target disease, it can provide it to the user's terminal (200) of the target disease.

[0188] Meanwhile, the various embodiments described above may be implemented by combining two or more embodiments, provided that they do not conflict or contradict each other.

[0189] Specifically, each of the multiple operations, steps, and configurations for implementing one embodiment may be embodied in another embodiment, or the operations and steps of another embodiment may be followed by the last operation or step of one embodiment, but are not limited thereto.

[0190] Meanwhile, computer instructions or computer programs for performing processing operations in the various embodiments of the present disclosure described above may be stored in a non-transitory computer-readable medium. When such computer instructions or computer programs stored in the non-transitory computer-readable medium are executed by a processor of a specific device, the specific device described above performs processing operations according to the various embodiments described above.

[0191] A non-transient computer-readable medium refers to a medium that stores data semi-permanently and can be read by a device, unlike media that store data for a short period of time such as registers, caches, and memory. Specific examples of non-transient computer-readable media include CDs, DVDs, hard disks, Blu-ray discs, USBs, memory cards, and ROMs.

[0192] According to one embodiment, the method according to the various embodiments disclosed herein may be provided by being included in a computer program product. The computer program product may be traded between a seller and a buyer as a product. The computer program product may be distributed in the form of a device-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or distributed online (e.g., download or upload) through an application store (e.g., Play Store™) or directly between two user devices (e.g., smartphones). In the case of online distribution, at least a portion of the computer program product (e.g., downloadable app) may be temporarily stored or temporarily created on a device-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0193] Although preferred embodiments of the present disclosure have been illustrated and described above, the present disclosure is not limited to the specific embodiments described above. It is understood that various modifications can be made by those skilled in the art without departing from the essence of the present disclosure as claimed in the claims, and such modifications should not be understood individually from the technical spirit or perspective of the present disclosure. Explanation of the symbols

[0195] 100: Server 110: Memory 120: Communication interface 130: Processor 131: Data Collection Module 132: Model Training Module 133: Chatbot Module 200: User's terminal 300: Medical record server 10: Classification Model 20: Counseling Model

Claims

Claim 1 A method of operation of a server comprises: a step of identifying clinical experience information based on the medical history of a specialist for multiple patients by medical specialty; a step of acquiring a counseling model trained to perform counseling based on the clinical experience information by medical specialty; and a step of identifying diagnostic information for a user through at least one of the counseling models acquired by medical specialty; wherein the step of identifying diagnostic information for a user comprises: a step of identifying a medical specialty that matches the user's symptoms based on basic counseling content between a classification model for identifying medical specialty and the user's terminal; and a step of acquiring diagnostic information for a user based on advanced counseling content between a counseling model trained on clinical experience information for a medical specialty that matches the user's symptoms and the user's terminal; and wherein the method of operation of the server comprises: a step of acquiring symptom information from the user's terminal; and a step of identifying at least one symptom type based on the symptom information. The method of operation of the server includes the step of identifying the urgency of medical treatment for the user based on the identified symptom type; and the method of operation of the server includes the step of obtaining at least one additional query based on the output data obtained by inputting at least one response received from the user's terminal into the consultation model; and the step of providing the additional query to the user's terminal and obtaining a response to the additional query from the user's terminal; and the method of operation of the server includes, when multiple additional queries are obtained, the step of identifying the recognition accuracy regarding the degree to which the user recognizes the symptoms based on the similarity between the multiple additional queries; and the step of identifying whether remote consultation or face-to-face consultation with medical staff is necessary based on the identified recognition accuracy.A method of operation of a server comprising: a step of identifying whether remote consultation or face-to-face consultation with medical staff is necessary, wherein if the identified recognition accuracy is greater than or equal to a preset value, the consultation through the consultation model is maintained; if the identified recognition accuracy is less than the preset value, a reference urgency is set as a criterion for identifying whether remote consultation or face-to-face consultation is necessary; if the identified urgency is less than the reference urgency, the consultation through the consultation model is maintained; and if the identified urgency is greater than or equal to the reference urgency, a message recommending remote consultation or face-to-face consultation with medical staff is provided to the user's terminal. Claim 2 A method of operation of a server according to claim 1, wherein the step of identifying clinical experience information comprises: identifying audio data containing the voice of the specialist performing treatment for each of the plurality of patients from the treatment history; identifying the interaction pattern of the specialist based on the audio data; and identifying the decision-making process of the specialist based on the audio data. Claim 3 In paragraph 2, the step of identifying the interaction pattern of the specialist comprises inputting the audio data into a speaker separation model and, based on the output data, dividing the audio data into at least one first unit segment containing the voice of the specialist and at least one second unit segment containing the voice of the patient, extracting speech feature information related to the speech characteristics of the specialist based on the voice of the specialist included in the first unit segment, and identifying the interaction pattern including the extracted speech feature information by medical specialty. Claim 4 In paragraph 2, the step of identifying the decision-making process of the specialist comprises: performing speech recognition on the audio data for each medical specialty to obtain medical text corresponding to the audio data; inputting the medical text into an analysis model for identifying the process of treatment and, based on the output data, identifying medical result data from each of the medical texts obtained for each medical specialty; and obtaining the decision-making process including the medical text and the medical result data for each medical specialty. Claim 5 A method of operation of a server according to claim 1, wherein the step of identifying a medical specialty matched with the user comprises obtaining symptom information from the user's terminal, inputting the symptom information into the classification model, and based on the output data, obtaining the basic consultation content including information about a medical specialty matched with the symptom information. Claim 6 In claim 5, the step of obtaining diagnostic information for the user comprises obtaining the user's genomic data from a medical record server in which the user's medical data is stored, inputting the symptom information and the genomic data into the consultation model and, based on the output data, obtaining at least one query to obtain the user's response necessary to identify the user's disease and transmitting it to the user's terminal, and inputting at least one response received from the user's terminal into the consultation model and, based on the output data, obtaining the diagnostic information including at least one target disease finally diagnosed as the user's disease.

Images