Artificial intelligence-based integrated management system for the full lifecycle of medical services for foreign patients
The AI-based system integrates medical service stages for foreign patients, addressing information inconsistencies and language barriers, and dynamically adjusts procedures for improved efficiency and accuracy.
Patent Information
- Application Number
- KR1020260081494
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2026-05-06
- Publication Date
- 2026-11-11
- Estimated Expiration
- 2046-05-06
AI Technical Summary
Conventional medical service operations for foreign patients lack integration across stages, leading to inconsistencies in information management, compromised communication accuracy due to language barriers, and inadequate decision-making based on multiple criteria, resulting in reduced quality and satisfaction.
An AI-based integrated management system that generates patient context information, configures medical service provision procedures, and utilizes multiple AI agents to dynamically adjust procedures based on various judgment criteria, while addressing language differences and maintaining information consistency.
Improves the efficiency and accuracy of medical services by providing patient-centered, context-based management, enhancing communication, and enabling dynamic procedural adjustments.
Smart Images

Figure 112026054682311-PAT00001_ABST
Abstract
Description
Technology Field
[0001] Embodiments of the present invention relate to the field of artificial intelligence and medical service management technology. More specifically, the invention relates to an AI-based integrated management system for the entire lifecycle of medical services for foreign patients that comprehensively manages the entire process from the influx of foreign patients to consultation, matching of medical service providers, reservation, medical support, and follow-up care, configures medical service provision procedures based on patient context information, and intelligently operates full-cycle medical services by reflecting decision-making results using multiple AI agents into the medical services. Background Technology
[0002] With the growth of the global medical tourism industry, there is a continuous increase in cases where foreign patients visit specific countries to utilize a variety of medical services, including medical consultations, procedures, treatments, and follow-up care. These medical services for foreign patients go beyond the simple provision of medical care and encompass a full-cycle service structure that links multiple stages, such as patient influx, consultation, hospital selection, appointment scheduling, treatment execution, interpretation support, and follow-up care.
[0003] However, conventional medical service operations for foreign patients have a structure where each stage is performed as a separate system or independent service unit, resulting in a problem where patient inflow channels, consultation systems, hospital matching systems, appointment management systems, and follow-up care systems are not interconnected. Consequently, information regarding the same patient is managed in a decentralized manner across stages, leading to limitations in maintaining information consistency between these stages.
[0004] For example, health status information and requirements collected during the initial consultation phase may not be sufficiently reflected during the hospital matching phase, or the patient's schedule and preferences may not be properly considered during the reservation phase; this can lead to a decline in the quality of medical services and reduced patient satisfaction. Furthermore, there is a problem in the follow-up care phase after treatment where continuous management is difficult due to the lack of linkage between previous treatment history and patient status information.
[0005] Furthermore, language barriers pose a significant challenge in medical services for foreign patients. Conventional systems typically rely on interpreters or simple translation tools, and voice and text data generated during real-time interactions between patients and healthcare providers is often not systematically managed, resulting in merely simple transmission. Consequently, communication accuracy is compromised during the service delivery process, and there is a risk of errors occurring in the transmission of critical medical information.
[0006] Furthermore, decision-making in the process of providing medical services tends to be based mostly on a single criterion or limited information, and there is a lack of intelligent decision-making structures that comprehensively consider multiple conditions and situations. For example, although various judgment criteria exist in processes such as hospital matching, treatment scheduling, and procedure selection, conventional technology does not provide a systematic decision-making system that simultaneously reflects these criteria.
[0007] Furthermore, existing medical service management systems fail to adequately provide a patient-centered, context-based processing structure that comprehensively reflects the patient's condition, requirements, treatment history, and schedule information. Consequently, these systems face limitations in that medical service delivery procedures are statically defined or manually adjusted. In particular, the lack of functionality to dynamically modify or correct delivery procedures in response to changes in the patient's condition or the external environment makes real-time response difficult.
[0008] As such, conventional systems have limitations in that they fail to adequately reflect the full-cycle characteristics of medical services for foreign patients and cannot provide patient-centered integrated management, intelligent decision-making, support for real-time interaction, or dynamic procedural correction functions; therefore, there is a need for the development of a new system capable of performing decision-making and procedural control based on artificial intelligence while comprehensively managing full-cycle medical services. Prior art literature
[0009] Registered Patent Publication No. 10-2074805 (Registration Date: February 4, 2020) Published Patent Publication No. 10-2022-0077653 (Publication Date: June 9, 2022) The problem to be solved
[0010] An embodiment of the present invention provides an artificial intelligence-based integrated management system for the entire lifecycle of medical services for foreign patients, which generates patient context information based on foreign patient information, configures a medical service provision procedure using said patient context information, and can comprehensively manage linkages with medical service providers according to the configured medical service provision procedure.
[0011] In addition, we provide an AI-based integrated management system for the entire lifecycle of medical services for foreign patients that utilizes multiple AI agents to generate decision results reflecting various judgment criteria for medical services, and incorporates these generated decision results into the medical service delivery procedures, thereby improving the accuracy and efficiency of medical service operations.
[0012] In addition, it provides an AI-based integrated management system for the entire medical service lifecycle of foreign patients that resolves language differences based on voice and text data generated during the interaction process between foreign patients and medical service providers, and structures interaction information to reflect it in the medical service delivery process.
[0013] In addition, it provides an AI-based integrated management system for the entire lifecycle of medical services for foreign patients that dynamically adjusts medical service delivery procedures in response to various situational changes occurring during the medical service delivery process and maintains a medical service flow suitable for the patient's condition and requirements. means of solving the problem
[0014] An AI-based integrated management system for the entire lifecycle of medical services for foreign patients according to an embodiment of the present invention comprises: a context generation unit that generates patient context information based on foreign patient information; a service execution management unit that configures a medical service provision procedure based on the patient context information generated by the context generation unit and manages the linkage of medical services between the foreign patient and the medical service provider according to the configured medical service provision procedure; and a multi-agent processing unit that obtains processing results by a plurality of AI agents based on the patient context information generated by the context generation unit and the medical service information managed by the service execution management unit, generates a decision result for the medical service using the obtained processing results, and reflects the generated decision result in the medical service.
[0015] Additionally, the context generation unit may include: an input conversion unit that receives foreign patient information and converts it into a form usable for providing medical services; a feature extraction unit that identifies patient status information and requirement information based on the foreign patient information converted by the input conversion unit and generates feature information for medical service matching using the identified information; and a context configuration unit that configures patient context information reflecting the patient status information and requirement information using the feature information generated by the feature extraction unit.
[0016] Additionally, the service execution management unit may include: a procedure division unit that generates step information by dividing the medical service provision procedure into multiple steps based on patient context information generated by the context generation unit; a sequence determination unit that defines the execution order between steps by reflecting the preceding and succeeding conditions between the step information generated by the procedure division unit; and an execution management unit that manages the progress status of the medical service provision procedure by updating the execution status and progress status of each step according to the execution order defined by the sequence determination unit.
[0017] Additionally, the multi-agent processing unit may include: an agent configuration unit that configures a plurality of artificial intelligence agents having different judgment criteria using patient context information generated by the context generation unit and medical service information managed by the service execution management unit; a multi-decision execution unit that generates decision results for a medical service provision procedure through the plurality of artificial intelligence agents configured by the agent configuration unit; a result analysis unit that identifies decision items where differences occur between results by comparing the plurality of decision results generated by the multi-decision execution unit; a result derivation unit that generates a final decision result by selecting one or more of the plurality of decision results based on the decision items identified by the result analysis unit; and a result reflection unit that provides the final decision result generated by the result derivation unit to the service execution management unit to reflect it in the medical service provision procedure.
[0018] Additionally, the system may further include an interaction support unit that supports interaction between a foreign patient and a medical service provider during the medical service provision process based on medical service information managed by the service performance management unit, wherein the interaction support unit includes: an interaction receiving unit that receives at least one interaction data among voice data and text data generated or exchanged during the interaction process between the foreign patient and the medical service provider; a language processing unit that generates language information converted into different languages based on the interaction data received by the interaction receiving unit; an interaction information generating unit that generates interaction information that can be reflected in the medical service provision procedure using the language information generated by the language processing unit and the interaction data received by the interaction receiving unit; a channel control unit that sets a communication channel corresponding to each of the foreign patient and the medical service provider based on the interaction information generated by the interaction information generating unit and the language information generated by the language processing unit, and controls the transmission of interaction information and language information through the set communication channel; and an output processing unit that provides at least one of the interaction information and language information transmitted by the channel control unit to the foreign patient or the medical service provider.
[0019] Additionally, the system may further include a procedure correction unit that corrects a medical service provision procedure based on a decision result generated by the multi-agent processing unit, wherein the procedure correction unit may include: a correction target identification unit that identifies a step or execution condition requiring correction among the medical service provision procedures managed by the service execution management unit using the decision result generated by the multi-agent processing unit; and a procedure correction execution unit that modifies the medical service provision procedure by reflecting the decision result generated by the multi-agent processing unit on the step or execution condition identified by the correction target identification unit, and updates the medical service provision procedure information managed by the service execution management unit by reflecting the modified medical service provision procedure.
[0020] Additionally, the service execution management unit further includes: a step standard setting unit that sets standard information for a plurality of steps based on the type of medical service or procedure category; a time point generation unit that calculates the execution time of each step using the standard information and standard time point set by the step standard setting unit; and a journey generation unit that generates a medical service provision procedure including a step execution schedule using the time point information generated by the time point generation unit. The system further includes a service matching unit that queries information about a medical service provider based on patient context information generated by the context generation unit and derives a medical service provider corresponding to the patient context information based on the queried information. The service matching unit may include: a query generation unit that generates query information for querying a medical service provider using patient context information generated by the context generation unit; an information query unit that queries information about a medical service provider from an external data repository using the query information generated by the query generation unit; and a matching result generation unit that determines a medical service provider corresponding to the patient context information based on the results queried by the information query unit.
[0021] Additionally, the system further includes a coordinator agent processing unit that verifies the suitability of each step based on patient context information and medical service provision procedure information, and controls the execution of each step of the medical service provision procedure according to the verification result. The coordinator agent processing unit may include: a rule extraction unit that extracts rule information from an unstructured document; a rule base management unit that generates a rule structure based on the rule information extracted by the rule extraction unit and stores and manages the rule structure in a hierarchical structure; a step verification unit that receives as input the patient context information generated by the context generation unit and the planned actions for each step of the medical service provision procedure configured by the service execution management unit, and verifies the suitability of each step using the rule structure managed by the rule base management unit; and an action decision unit that determines whether to allow or restrict the execution of each step by providing a control signal to the service execution management unit based on the verification result of the step verification unit. Effects of the invention
[0022] According to the present invention, by managing full-cycle medical services from the influx of foreign patients to post-treatment care in a single integrated system, information that was previously dispersed at each stage can be consistently integrated and managed, thereby improving the efficiency of medical service operations.
[0023] In addition, by configuring and managing medical service delivery procedures based on patient context information, it becomes possible to provide customized medical services that reflect the patient's condition and requirements, thereby improving patient satisfaction.
[0024] In addition, since decision results that comprehensively reflect various judgment criteria can be generated through a decision structure utilizing multiple AI agents, the accuracy of hospital matching, scheduling, and medical service selection can be improved.
[0025] In addition, by generating language information based on voice and text data generated during the interaction process and reflecting it in the medical service provision procedure, it is possible to improve the accuracy of communication between foreign patients and medical service providers and reduce errors in the medical service provision process.
[0026] In addition, through the function of dynamically adjusting medical service delivery procedures, it is possible to flexibly respond to changes in the patient's condition and the external environment, thereby improving the stability and reliability of medical services. Brief explanation of the drawing
[0027] FIG. 1 is a schematic diagram showing the entire hardware-based operation process of an artificial intelligence-based integrated management system for the entire lifecycle of medical services for foreign patients according to an embodiment of the present invention. FIG. 2 is a block diagram showing the overall configuration of an artificial intelligence-based integrated management system for the entire lifecycle of medical services for foreign patients according to an embodiment of the present invention. FIG. 3 is a block diagram showing the detailed configuration of a context generation unit according to an embodiment of the present invention. FIG. 4 is a block diagram showing the detailed configuration of a service execution management unit according to an embodiment of the present invention. FIG. 5 is a block diagram showing the detailed configuration of a multi-agent processing unit according to an embodiment of the present invention. FIG. 6 is a block diagram showing the detailed configuration of an interaction support unit according to an embodiment of the present invention. FIG. 7 is a block diagram showing the detailed configuration of a procedure correction unit according to an embodiment of the present invention. FIG. 8 is a block diagram showing the detailed configuration of a service matching unit according to an embodiment of the present invention. FIG. 9 is a block diagram showing the detailed configuration of a coordinator agent processing unit according to an embodiment of the present invention. Specific details for implementing the invention
[0028] The terms used in this specification will be briefly explained, and the invention will be described in detail.
[0029] The terms used in this invention have been selected based on currently widely used general terms, taking into account their functions within the invention; 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 arbitrarily selected by the applicant, and in such cases, their meanings will be described in detail in the relevant description of the invention. Therefore, the terms used in this invention should be defined not merely by their names, but based on their meanings and the overall content of the invention.
[0030] 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.
[0031] Embodiments of the present invention are described below with reference to the attached drawings so that those skilled in the art can easily implement them. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.
[0032] FIG. 1 is a schematic diagram showing the entire hardware-based operation process of an artificial intelligence-based integrated management system for the entire lifecycle of medical services for foreign patients according to an embodiment of the present invention.
[0033] Referring to FIG. 1, the artificial intelligence-based integrated management system (1000) for the entire lifecycle of medical services for foreign patients according to the present embodiment may be implemented by including a foreign patient terminal (10), a central processing server (20), a multi-artificial intelligence agent server (30), and a medical service provision server (40). It may integrate and manage the entire process from the influx of foreign patients to consultation, matching of medical service providers, reservation, treatment, and follow-up care based on patient context, and may reflect the decision-making results using multiple artificial intelligence agents in the medical service provision procedure. It may also include a coordinator agent function that controls the execution of each step of the medical service provision procedure based on rule information extracted from unstructured medical documents.
[0034] The above foreign patient terminal (10) can be implemented as an input / output device used directly by the foreign patient. For example, the foreign patient terminal (10) can be implemented as at least one of a smartphone, tablet, laptop, or desktop computer, and can input or check information related to medical service requests, consultation messages, voice calls, appointment requests, and follow-up care through a mobile application or a web-based interface. Additionally, the foreign patient terminal (10) can be used as a means to generate interaction data including voice data and text data and transmit it to a central processing server (20), and can be used as an interface to receive output of language conversion results, medical service provision procedures, appointment schedules, and medical service provision results.
[0035] The central processing server (20) described above may be implemented as a central computing unit that performs core control and data processing functions of the system (1000). For example, the central processing server (20) may be implemented as a server computer including one or more processors and memory, a cloud-based computing instance, or a distributed processing server cluster. This central processing server (20) may receive foreign patient information from a foreign patient terminal (10), convert the received information into a form usable for providing medical services, and generate feature information by identifying the patient's condition information and requirement information. Additionally, the central processing server (20) may configure patient context information based on the generated feature information and configure a medical service provision procedure using the patient context information. Furthermore, the central processing server (20) may divide the medical service provision procedure into multiple stages, define the execution order between stages, manage the execution status of each stage, and provide information on the planned actions for each stage as information subject to rule-based verification.
[0036] Additionally, the central processing server (20) can receive interaction data between a foreign patient and a medical service provider and can generate language information converted into different languages based on the interaction data. This language information can be combined with the interaction data to generate interaction information that can be reflected in the medical service provision procedure, and the generated interaction information can be transmitted to the foreign patient terminal (10) and the medical service provision server (40) through a communication channel. Additionally, the central processing server (20) can set step standard information based on the type of medical service or procedure category, and can generate a medical service provision procedure including a step execution schedule by calculating the timing of each step execution.
[0037] The above-mentioned multi-artificial intelligence agent server (30) may be implemented as a computational processing device including multiple artificial intelligence agents. For example, the multi-artificial intelligence agent server (30) may be implemented as a GPU-based computational server, an AI model serving server, or a cloud-based artificial intelligence processing infrastructure, and may execute multiple artificial intelligence agents having different judgment criteria or models in parallel. This multi-artificial intelligence agent server (30) may generate decision results for each agent based on patient context information and medical service information transmitted from the central processing server (20), and may identify decision items by analyzing the differences between the generated multiple decision results. Additionally, the multi-artificial intelligence agent server (30) may generate a final decision result by selecting one or more of the multiple decision results based on the identified decision items, transmit the generated decision result to the central processing server (20), and the decision result may be reflected in the medical service provision procedure after undergoing rule conformity verification.
[0038] The medical service provision server (40) described above may be implemented as a system used in a hospital, clinic, or medical institution. For example, the medical service provision server (40) may be implemented as a hospital computer system, an electronic medical record system, an appointment management system, or a medical service platform server. This medical service provision server (40) may receive medical service provision procedures, appointment information, interaction information, and decision results transmitted from the central processing server (20), and may provide medical appointments, consultation responses, procedure execution, and follow-up care services based on the information. Additionally, the medical service provision server (40) may generate voice data and text data generated during the interaction process with the foreign patient terminal (10) and transmit them to the central processing server (20).
[0039] The overall operation process of the artificial intelligence-based integrated management system (1000) for the entire lifecycle of medical services for foreign patients according to the present embodiment is described below.
[0040] First, a foreign patient can input their health status information, requirements information, or consultation request information through a foreign patient terminal (10), and the input information can be transmitted to a central processing server (20). At this time, the central processing server (20) can generate patient context information using the received foreign patient information, and can configure a medical service provision procedure based on the generated patient context information, and each step of the configured procedure can be structured into a unit of action scheduled to be performed step by step so that rule-based verification is possible.
[0041] Afterwards, the central processing server (20) can control the connection between the foreign patient and the medical service provision server (40) according to the configured medical service provision procedure so that consultation, reservation, and medical treatment are performed, and rule conformity verification can be performed before each step is performed.
[0042] Next, the central processing server (20) can transmit patient context information and medical service information to the multiple artificial intelligence agent server (30), and the multiple artificial intelligence agent server (30) can generate decision results for medical services through multiple artificial intelligence agents. The generated decision results can be transmitted to the central processing server (20), and the central processing server (20) can modify or correct the procedure by reflecting the decision results in the medical service provision procedure, and in the reflection process, unsuitable decision results can be excluded through rule-based verification.
[0043] Additionally, voice data or text data generated during the interaction process between the foreign patient terminal (10) and the medical service provider server (40) can be transmitted to a central processing server (20), and the central processing server (20) can generate language information converted into different languages. The generated language information is composed of interaction information and transmitted to both parties to support smooth communication.
[0044] Subsequently, the central processing server (20) can generate a schedule-based journey by calculating the execution time of each step of the medical service provision procedure, and the generated medical service provision procedure can be transmitted to the medical service provision server (40) and applied to actual medical treatment and service execution. In addition, the medical service provision procedure can be continuously corrected according to the decision results transmitted from the multi-artificial intelligence agent server (30), thereby enabling dynamic medical service provision in response to changes in the patient's condition or external conditions, and the consistency and safety of the entire procedure can be maintained through knowledge graph verification based on a rule structure during the correction process.
[0045] In this way, the artificial intelligence-based full-cycle integrated management system (1000) for foreign patient medical services according to the present embodiment can operate as a full-cycle integrated management structure that continuously improves the medical service provision procedure by generating a patient context-based medical service provision procedure through organic linkage between a foreign patient terminal (10), a central processing server (20), a multi-artificial intelligence agent server (30), and a medical service provision server (40), reflecting the decision-making results of multiple artificial intelligence agents, and controlling the execution of each step based on a rule structure and a hierarchical knowledge graph generated from unstructured documents.
[0046] Hereinafter, the detailed configuration of an artificial intelligence-based integrated management system for the entire lifecycle of medical services for foreign patients according to an embodiment of the present invention will be described in more detail with reference to the attached drawings.
[0047] FIG. 2 is a block diagram showing the overall configuration of an artificial intelligence-based integrated management system for the entire lifecycle of medical services for foreign patients according to an embodiment of the present invention, FIG. 3 is a block diagram showing the detailed configuration of a context generation unit according to an embodiment of the present invention, FIG. 4 is a block diagram showing the detailed configuration of a service execution management unit according to an embodiment of the present invention, FIG. 5 is a block diagram showing the detailed configuration of a multi-agent processing unit according to an embodiment of the present invention, FIG. 6 is a block diagram showing the detailed configuration of an interaction support unit according to an embodiment of the present invention, FIG. 7 is a block diagram showing the detailed configuration of a procedure correction unit according to an embodiment of the present invention, FIG. 8 is a block diagram showing the detailed configuration of a service matching unit according to an embodiment of the present invention, and FIG. 9 is a block diagram showing the detailed configuration of a coordinator agent processing unit according to an embodiment of the present invention.
[0048] Referring to FIG. 2, the artificial intelligence-based full-cycle integrated management system for foreign patient medical services (1000) according to the present embodiment may include at least one of a context generation unit (100), a service execution management unit (200), a multi-agent processing unit (300), an interaction support unit (400), a procedure correction unit (500), a service matching unit (600), and a coordinator agent processing unit (700).
[0049] The above context generation unit (100) can generate patient context information based on foreign patient information. To this end, the context generation unit (100) may include at least one of an input conversion unit (110), a feature extraction unit (120), and a context configuration unit (130), as shown in FIG. 3.
[0050] The above input conversion unit (110) can receive foreign patient information and convert it into a form that can be used for providing medical services.
[0051] More specifically, the input conversion unit (110) can receive various forms of foreign patient information transmitted through the foreign patient terminal (10) or an external input channel, and then convert the information into structured data suitable for medical service processing. This foreign patient information may include natural language-based consultation messages, voice input data, multilingual medical questionnaire responses, appointment request information, image data (e.g., symptom photos), stay schedule information, etc., and may be composed of different data formats and languages.
[0052] In addition, the input conversion unit (110) can perform a natural language processing-based normalization process on text data to remove unnecessary expressions and convert it into a standard sentence form that retains the core meaning, and if input in multiple languages, it can consistently convert it into a common processing language (e.g., Korean or English). In addition, in the case of voice data, the same normalization process can be applied after converting it into text data through speech recognition processing (STT), and in this process, recognition accuracy can be improved by utilizing a medical domain-specific term dictionary.
[0053] Additionally, the input conversion unit (110) can perform the function of extracting structured field information from unstructured data, and, for example, can automatically identify items such as the type of procedure, desired schedule, budget range, preferred hospital type, and language requirements within a consultation message and separate them into independent data fields. This process can be performed using an LLM-based information extraction method or a predefined slot filling structure.
[0054] Additionally, the input conversion unit (110) can unify data types to resolve format inconsistencies between input data, for example, by converting date expressions into a standard time format or normalizing country and language information into an ISO code form, and in the case of monetary information, by unifying currency units or converting based on exchange rates. These conversion results can be reconstructed to fit the data schema so that they can be processed according to consistent standards in subsequent steps.
[0055] Additionally, the input conversion unit (110) can perform reliability or completeness verification on the converted data, and if essential items are missing, it can request additional input from the user or generate supplementary information through linkage with existing conversation history or external data. For example, if a patient briefly inputs, such as "I want to consult about nose surgery," it can supplement by estimating additional conditions based on past consultation history or basic profile information.
[0056] Additionally, when a patient signs an electronic consent form, the input conversion unit (110) can generate and store the contents of the consent form and related input data at the time of signing in the form of an immutable snapshot. Here, the snapshot is managed so that it cannot be subsequently modified or deleted to maintain legal validity, and can be managed as historical data including consent form version information, time of signing, and user identification information.
[0057] The above feature extraction unit (120) can identify patient status information and requirement condition information based on foreign patient information converted by the input conversion unit (110), and generate feature information for medical service matching using the identified information.
[0058] More specifically, the feature extraction unit (120) receives foreign patient data normalized and structured by the input conversion unit (110) as input, and analyzes the semantic elements included in the data to distinguish and identify the patient's condition information and requirement information. Here, the condition information may include information directly used for medical judgment, such as the patient's current physical condition, past medical history, type of symptoms, degree of pain, whether medication is being taken, and allergy information, and the requirement information may include condition information related to service selection, such as the type of procedure the patient desires, budget range, duration of stay, visit schedule, preferred region, type of hospital, and language requirements.
[0059] Additionally, the feature extraction unit (120) can separate information by semantic unit from consultation data input in the form of natural language and convert it into structured attribute values, and in this process, natural language processing techniques such as named entity recognition (NER), intent classification, and slot extraction can be utilized. For example, if a patient inputs, "I want to go to Korea next month to get a skin treatment, my budget is about 2 million won, and I want a hospital where English is available," the treatment type can be identified by separating it into multiple attributes such as 'skin treatment', the time of visit as 'next month', the budget as '2 million won', and the language condition as 'English available'.
[0060] Additionally, the feature extraction unit (120) can generate complex feature information by analyzing the relationships between identified information, and can generate derived features based on combinations of attributes as well as simple individual attributes. For example, it can generate a feature that determines the suitability of a short-term procedure by combining the length of stay and the type of procedure, or derive a feature value that indicates a realistic matching possibility by combining the budget range and hospital average procedure cost data.
[0061] Additionally, the feature extraction unit (120) can correct or enhance feature information by integrating multiple input data sources, and can generate a more precise feature vector by comprehensively reflecting consultation history, previous visit records, image data (e.g., photos of skin condition), medical questionnaire responses, etc. For example, if the same patient has a history of considering plastic surgery procedures during a previous consultation, it can generate features that reflect the possibility or priority of complex procedures by combining them with the currently entered request for skin procedures.
[0062] Additionally, the feature extraction unit (120) can perform a quantification or vectorization process on the generated feature information to convert it into a form usable in a machine learning model or search system, and can assign weights to each feature to reflect its importance. For example, features directly related to the patient's health condition can be assigned high weights, while simple preference conditions can be assigned relatively low weights.
[0063] Additionally, the feature extraction unit (120) can perform inference-based supplementary processing for incomplete or ambiguous inputs, and can generate estimated values based on similar patient data or predefined rules when some conditions are unclear. For example, if a patient does not specify a budget, the budget range can be estimated based on the average cost range of the selected procedure type and reflected in the feature information.
[0064] The above context configuration unit (130) can configure patient context information that reflects patient status information and requirement condition information using feature information generated by the feature extraction unit (120).
[0065] More specifically, the context configuration unit (130) can construct patient-centered context information by receiving multiple feature information generated by the feature extraction unit (120) as input and combining the corresponding features into a single integrated data structure that is mutually related. This context information is not merely a set of individual attributes, but can be organized in an organized form that reflects the patient's state information and requirement condition information, as well as the temporal flow, priority, and relationship structure.
[0066] Additionally, the context configuration unit (130) can analyze the semantic relationships between each feature information and convert them into data in a hierarchical or graph structure, thereby explicitly expressing the influence relationship between a specific feature and another feature. For example, when a requirement condition such as "desire for plastic surgery procedure" is combined with status information such as "recovery period needed after surgery," the requirement for the duration of stay according to the procedure can be automatically expanded and reflected in the context.
[0067] Additionally, the context configuration unit (130) can configure a time-based context that includes not only the patient's current status but also their past history and future schedule, and can integrate consultation history, appointment history, medical history, and follow-up care progress status into a single continuous journey information. For example, the context can be updated by readjusting the recommendation priority by combining a list of hospital candidates generated from a previous consultation with additional requirements currently entered.
[0068] Additionally, the context configuration unit (130) can construct a multimodal-based context by integrating various data formats, and can generate a more precise representation of the patient's condition by combining not only text information but also image data (e.g., photos of the affected area), voice-based consultation records, and medical history response data. In this process, the accuracy of the context can be corrected by reflecting reliability or recency information for each data type.
[0069] Additionally, the context configuration unit (130) can configure the context in the form of a standardized data schema or object so that it can be utilized in the process of providing medical services, and can provide an interface that can be commonly utilized in the service matching unit (600), service execution management unit (200), and multi-agent processing unit (300). For example, the patient context can be defined in a JSON-based structure and stored in fields where procedure requirements, schedule conditions, risk factors, language requirements, etc., are clearly distinguished.
[0070] Additionally, the context configuration unit (130) can perform a dynamic update function of context information, and when new input information occurs or the status of medical service progress changes, it can modify or expand some items while maintaining the existing context. For example, if a patient increases their budget during the consultation process, the existing hospital matching results can be re-evaluated and the context can be updated to reflect the new recommendation results.
[0071] Additionally, the context configuration unit (130) can restore the context based on the previous interaction history even after the conversation session has ended or been interrupted, and thus can be configured so that the service can continue while maintaining the previous consultation state even if the patient reconnects after a certain period of time. For example, if a patient disconnects while checking hospital candidates and then reconnects, the previously viewed list of hospitals and consultation content are restored as they are, allowing for the input of additional conditions or the progress of the reservation.
[0072] The above service execution management unit (200) can configure a medical service provision procedure based on patient context information generated by the context generation unit (100), and can manage the linkage of medical services between a foreign patient and a medical service provider according to the configured medical service provision procedure. To this end, the service execution management unit (200) may include at least one of a procedure division unit (210), a sequence determination unit (220), an execution management unit (230), a step standard setting unit (240), a time point generation unit (250), and a journey generation unit (260), as shown in FIG. 4.
[0073] The above procedure division unit (210) can generate step information by dividing the medical service provision procedure into multiple steps based on the patient context information generated by the context generation unit (100).
[0074] More specifically, the procedure division unit (210) receives patient context information generated by the context generation unit (100) as input, breaks down the flow of medical service usage by a foreign patient into multiple logically distinguishable stages, and generates stage information corresponding to each stage. This stage division is not a simple sequential arrangement, but can be configured to correspond to an actual service execution unit by comprehensively considering the patient's condition information, requirements information, type of procedure, and schedule of stay.
[0075] Additionally, the procedure division unit (210) can divide the procedure into a search and consultation stage, a hospital matching stage, a reservation and document preparation stage, a visit and treatment stage, and a follow-up care stage by reflecting the entire flow of medical tourism, and each stage can be defined as an independently manageable unit. For example, in the initial stage, a consultation stage to identify the patient's requirements and a matching stage to identify hospital candidates can be distinguished, and in subsequent stages, the reservation confirmation, medical questionnaire completion, electronic consent form signing, and visit schedule management can each be subdivided into separate stages.
[0076] Additionally, the procedure division unit (210) can variably set the stage configuration according to the type of procedure or medical service category, and even for the same patient, the number of stages and the configuration method may vary depending on the type of procedure selected. For example, in the case of plastic surgery, it may be composed of multiple detailed stages such as pre-operative preparation, the day of surgery, recovery management, and follow-up observation, and in the case of a health checkup, it may be composed of a relatively simple structure such as examination reservation, examination performance, and result verification stages.
[0077] Additionally, the procedure division unit (210) can generate meta-information for each step that includes the purpose of execution, necessary input information, expected results, and relevant entities (patient, hospital, coordinator, etc.), thereby clearly defining the conditions and roles under which each step must be performed. For example, in the case of the reservation step, the patient's confirmed schedule information and the hospital's acceptance status can be set as input conditions, and the reservation completion status can be defined as result information.
[0078] Additionally, the procedure division unit (210) can perform division considering inter-step dependencies by reflecting time information and condition information included in the patient context, and can form a condition-based step structure so that the next step can proceed only after a specific step is completed. For example, the relationship between steps can be established so that entry into the treatment step is not possible if the electronic consent signature is not completed.
[0079] Additionally, the procedure division unit (210) may generate the step division results in the form of a data structure and provide them for use by the sequence determination unit (220) and the execution management unit (230) thereafter, and each step may be defined in the form of an object including a unique identifier, step type, status value, etc. This step information may be utilized as a basic unit constituting the entire medical service provision procedure.
[0080] The above sequence determination unit (220) can define the execution order between steps by reflecting the preceding and succeeding conditions between the step information generated in the procedure division unit (210).
[0081] More specifically, the sequence determination unit (220) receives each step information generated by the procedure division unit (210) as input, analyzes the execution relationship between steps, and defines the execution order of the entire medical service provision procedure by reflecting preceding and succeeding conditions. This execution order is not a simple fixed order, but can be configured as a logically connected flow by considering the completion status of each step, the satisfaction of input data, the state of the external system, etc.
[0082] Additionally, the sequence determining unit (220) may explicitly set conditions for the start and completion of execution for each step, and may define conditions such as the completion status of a preceding step required to start a specific step, the existence of essential input information, and approval by the medical service provider. For example, conditions may be set so that the reservation step is initiated only after the hospital matching step is completed and the patient has selected a specific hospital, and control may be exercised so that the treatment step does not proceed when the reservation is not confirmed.
[0083] Additionally, the sequence determination unit (220) can configure the dependency relationships between steps in the form of a directed graph, where each step is represented as a node and the preconditions are represented as edges, so that the overall procedure flow can be structurally defined. Through this structure, steps that can be performed in parallel and steps that must be performed sequentially can be distinguished, and some steps can be set to proceed simultaneously. For example, the steps of completing the medical questionnaire and confirming the preliminary guidance can be configured to be performed in parallel.
[0084] Additionally, the sequence determining unit (220) can define a sequence that takes into account time constraints between steps by reflecting time information and schedule conditions included in the patient context, and specific steps may be restricted to be performed at a specific time before or after the scheduled date. For example, the pre-procedure preparation step may be set to be completed a certain period prior to the scheduled date, and the post-procedure care step may be configured to be performed sequentially over a certain period after the procedure.
[0085] Additionally, the sequence determining unit (220) may apply different sequence rules depending on the type of medical service or procedure category, and even if the step configuration is the same, the order of execution may be defined differently depending on the type of service. For example, in the case of a simple skin procedure, the process may proceed directly to the reservation and visit steps after consultation, but if surgery is required, steps such as a preliminary examination, consultation with a specialist, and confirmation of the surgery schedule may be added to form a more complex sequence structure.
[0086] Additionally, the sequence determination unit (220) can generate the result of the sequence definition between steps in the form of an executable workflow so that the execution management unit (230) can track the progress status of the steps, and each step can be configured in a structure that can be automatically activated by triggering the completion event of the previous step.
[0087] The above-mentioned execution management unit (230) can manage the progress status of the medical service provision procedure by updating the execution status and progress status of each step according to the execution order defined in the sequence determination unit (220).
[0088] More specifically, the execution management unit (230) can determine whether each step is executed based on the execution order between steps defined by the sequence determination unit (220), and continuously update the progress status of each step to comprehensively manage the current progress status of the entire medical service provision procedure. This management process can be performed using a state-based control method that includes whether the start condition of each step is met, the processing status during execution, whether it is completed, and whether an exception has occurred.
[0089] Additionally, the execution management unit (230) may set a state transition model for each stage, and each stage may have a state value such as 'waiting', 'activated', 'in progress', 'completed', 'holding', or 'canceled', and each state may change according to a specific event or condition. For example, when a patient creates a consultation request, the consultation stage may transition from the 'waiting' state to the 'activated' state, and when hospital staff accept the consultation, it may change to the 'in progress' state, and when the consultation ends, it may be updated to the 'completed' state.
[0090] Additionally, the performance management unit (230) receives an input of acceptance or rejection of the consultation from the medical service provider regarding the patient's consultation request, and can change the status of the corresponding stage depending on whether it is accepted. If the consultation is accepted, it can automatically activate subsequent stages including the medical history writing stage and the document preparation stage, and if the consultation is rejected, it can control the selection of an alternative medical service provider.
[0091] Additionally, the execution management unit (230) can automate status updates using various event-based triggers and can receive data input from the patient terminal (10), processing results generated from the medical service provision server (40), and internal calculation results from the central processing server (20) in real time and reflect them in the stage status. For example, when an event occurs where an electronic consent form is signed, the document preparation stage is automatically completed and the next stage can be activated.
[0092] Additionally, the execution management unit (230) may restrict the execution of a subsequent step if a preceding step is not completed, taking into account the inter-step dependency relationship, and may keep the corresponding step in a suspended state if certain conditions are not met. For example, if the appointment is not confirmed, the visit step may be controlled so that it is not activated, and if the entry of the patient's essential information is not completed, progress to the step after the consultation step may be restricted.
[0093] Additionally, the execution management unit (230) can manage the progress status of each stage over time, record the start time, end time, and estimated time required for each stage, and calculate the progress of the entire journey based on this. For example, the completion deadline for the previous stage is automatically set based on the procedure appointment date, and if it is not performed within that deadline, it can be displayed as delayed.
[0094] Additionally, the execution management unit (230) can detect exceptional situations occurring during the execution of a step and perform control in response thereto. If an error occurs at a specific step or execution is interrupted, the state of that step can be switched to a 'hold' or 'retry required' state and a notification can be sent to the relevant parties. For example, if a hospital refuses the appointment, that step is switched to a re-selection state, and the patient can be encouraged to select a new hospital.
[0095] Additionally, the execution management unit (230) can provide the progress status of the entire medical service provision procedure to the foreign patient terminal (10) and the medical service provision server (40), and the progress status of each stage can be visualized in the form of a timeline or journey and delivered to the user. For example, the patient can check the current status of consultation, the confirmation status of the appointment, the visit schedule, and the progress status of follow-up care step by step, and the medical service provider can take appropriate action based on the patient's overall progress status.
[0096] The above step standard setting unit (240) can set standard information for multiple steps based on the type of medical service or procedure category.
[0097] More specifically, the stage standard setting unit (240) can analyze information such as the type of procedure, medical specialty, and service purpose included in the patient context to define stage configuration standards suitable for the medical service and set standard information to be applied to each stage. This standard information can be defined in the form of a set of rules including the method of configuring the stages, the number of stages, the purpose of performing each stage, the necessary conditions, and the estimated duration.
[0098] Additionally, the step standard setting unit (240) may select a predefined step template or dynamically generate one according to the type of medical service, and may apply different step structures for different procedure categories. For example, in the case of plastic surgery, a multi-step structure such as pre-operative consultation, detailed examination, surgery preparation, surgery execution, recovery management, and follow-up observation may be set; in the case of dermatology procedures, it may consist of consultation, procedure reservation, procedure execution, and simple aftercare steps; and in the case of health checkups, a simplified structure such as reservation, examination execution, and result verification steps may be set.
[0099] Additionally, the stage standard setting unit (240) can set relative time information that serves as a standard for the time of execution for each stage, and can define a time offset for a stage that must be performed before or after a specific standard time (e.g., scheduled date of procedure or confirmed date of appointment). For example, the surgery preparation stage can be set to be completed 7 days prior to the date of the procedure, and the post-operative care stage can be set to be performed at 1, 7, and 30 days after the procedure.
[0100] Additionally, the stage standard setting unit (240) can define detailed execution items to be included in each stage, and can set checklists, precautions, input data requirements, etc., to be performed at each stage as standard information. For example, the medical history stage may include items such as inputting the patient's health status, checking allergy information, and checking medications being taken, and the follow-up care stage may include items such as checking the recovery status, uploading photos, and checking for pain.
[0101] Additionally, the step standard setting unit (240) can set access rights and performing entities for each step, and can include the roles that each entity, such as a patient, hospital staff, or coordinator, must perform or verify as standard information. For example, the patient can be set as the performing entity for the medical questionnaire completion step, and the hospital staff can be set as the performing entity for the consultation approval step.
[0102] Additionally, the step standard setting unit (240) can set step standards by reflecting regulations or protocols applied during the process of providing medical services, and can strengthen step performance standards by including hospital-specific policies, medical guidelines, and compliance conditions. For example, for specific procedures, it can be set so that a step for verifying pre-examination results is mandatory, or for patients from specific countries, it can be configured so that an additional document verification step is included.
[0103] The above-mentioned time point generation unit (250) can calculate the execution time of each step using the reference information and reference time point set in the step reference setting unit (240).
[0104] More specifically, the time point generation unit (250) receives the step-by-step standard information and the standard time point set in the step-by-step standard setting unit (240) as input, calculates the specific time information at which each step must actually be performed, and can calculate the time point for performing each step. This time point for performing is not a simple schedule value, but can be dynamically determined by reflecting various time factors such as the patient's appointment schedule, procedure plan, and length of stay.
[0105] Additionally, the time point generation unit (250) can calculate the time point of execution by applying a time offset value set for each step centered on the reference time point, and the reference time point can be set as the scheduled date of the procedure, the date of the reservation confirmation, the scheduled date of entry, etc. For example, if the scheduled date of the procedure is confirmed as June 15, the pre-procedure preparation step can be calculated as June 8, which is 7 days prior to the reference time point, the first post-procedure check step can be set as June 16, which is 1 day after the procedure, and the additional check step can be calculated sequentially after 7 days and 30 days.
[0106] Additionally, the time generation unit (250) can correct the time by reflecting actual schedule information included in the patient context, and can realistically adjust the time of step execution by considering the patient's entry date, exit date, period of stay, hospital operating hours, etc. For example, if the patient's entry date is June 10, steps that must be performed before that date can be readjusted to a time that can be performed after entry, and if the hospital's holiday is included, the schedule can be reconfigured to avoid that date.
[0107] Additionally, the time point generation unit (250) can adjust the time points to prevent conflicts between time points and maintain a sequential flow by considering the dependency relationship between the stages, and can automatically recalculate the start time of the subsequent stage based on the completion time of the preceding stage. For example, if the medical questionnaire writing stage is completed later than expected, the time points of the subsequent reservation confirmation stage and visit stage can be adjusted sequentially.
[0108] Additionally, the time point generation unit (250) can set an acceptable time range or window for the timing of step execution, and can define that a specific step does not necessarily have to be performed on a specific date but can be performed flexibly within a certain range. For example, the range for post-procedure checks can be set to be performed between 5 and 7 days after the procedure.
[0109] Additionally, the time point generation unit (250) can generate the calculated execution time point information in the form of standardized schedule data and provide it so that it can be utilized by the journey generation unit (260) thereafter, and each step execution time point can be defined in a structure that includes a start time, an end time, an estimated time required, etc.
[0110] The above journey generation unit (260) can generate a medical service provision procedure including a step-by-step execution schedule using time information generated by the time generation unit (250).
[0111] More specifically, the journey generation unit (260) receives step-by-step execution time information generated by the time point generation unit (250) as input and can configure a medical service provision procedure including the execution schedule of each step into a time-based journey. This journey is not a simple list of schedules, but can be composed of an integrated execution plan that combines the execution order, execution time, execution subject, and execution conditions of each step.
[0112] Additionally, the journey generation unit (260) can reconstruct a timeline structure by combining schedule information including a start time, an end time, and an estimated time required for each stage, thereby allowing the entire medical service flow to be intuitively represented in chronological order. For example, the consultation stage can be set as a schedule that starts on a specific date and is completed within a set time, the procedure stage can be arranged as a single event according to the scheduled date and time, and the aftercare stage can be divided into multiple schedules and configured to be performed repeatedly at scheduled intervals.
[0113] Additionally, the journey generation unit (260) can generate different journey structures depending on the procedure category and service type, and can configure an optimized journey for each category by combining the stage configuration defined in the stage standard setting unit (240) and the time information calculated in the time point generation unit (250). For example, in the case of plastic surgery procedures, a multi-stage journey spanning a long period from the pre-procedure preparation stage to the post-procedure care stage can be generated, while in the case of dermatology procedures, a simplified journey completed within a short period can be configured.
[0114] Additionally, the journey generation unit (260) can configure the journey by including information on the performing entity and access rights for each stage, and the schedules that each entity, such as a patient, hospital staff, and coordinator, must perform or verify can be distinguished and expressed. For example, the medical questionnaire completion stage can be displayed as a patient's performance schedule, and the consultation approval stage can be displayed as a hospital staff's performance schedule.
[0115] Additionally, the journey generation unit (260) can provide a checklist including step-by-step execution items and precaution information combined with a schedule, and specific tasks to be performed at each step can be linked in schedule units. For example, in the pre-procedure preparation stage, items such as checking whether fasting is required, checking whether medication has been discontinued, and preparing for a hospital visit can be displayed along with the schedule.
[0116] Additionally, the journey generation unit (260) can dynamically update the generated journey and can modify or reconfigure the existing journey according to changes in patient context, changes in appointment schedule, or the reflection of decision results of the multi-agent processing unit (300). For example, if the patient's procedure schedule is changed, the timing of the entire journey can be recalculated and updated to a new schedule structure.
[0117] Additionally, the journey generation unit (260) can automatically select a predefined step template according to the procedure category and generate a journey by calculating backward the dates of each step based on the time of reservation. For example, in the case of plastic surgery, a multi-step structure can be automatically generated in which the pre-surgery preparation step starts 14 days prior to the surgery date and the post-surgery recovery management step lasts for 30 days or more, and in the case of dermatology procedures, a short journey consisting of consultation, procedure, and simple aftercare can be generated. Additionally, each step includes a multilingual checklist and precautions so that the patient and medical staff can perform the procedure according to the same standards.
[0118] Additionally, the journey generation unit (260) can convert the generated medical service provision procedure into a visualizable data form and provide it to the foreign patient terminal (10) and the medical service provision server (40), and the step-by-step schedule is displayed in the form of a timeline or calendar so that the user can intuitively understand the overall flow of progress.
[0119] Additionally, the journey generation unit (260) can automatically identify a procedure category based on the time when a reservation confirmation event occurs, select a predefined step template corresponding to the identified procedure category, and automatically generate multiple detailed steps for a pre-procedure step, a procedure step, and a post-procedure step based on the selected step template. Additionally, it can automatically generate a performance schedule, a checklist, and precautions for each step to configure journey information combined with step performance items.
[0120] The multi-agent processing unit (300) can obtain processing results by a plurality of artificial intelligence agents based on patient context information generated by the context generation unit (100) and medical service information managed by the service execution management unit (200), generate decision results for medical services using the obtained processing results, and reflect the generated decision results in medical services. To this end, the multi-agent processing unit (300) may include at least one of an agent configuration unit (310), a multi-decision execution unit (320), a result analysis unit (330), a result derivation unit (340), and a result reflection unit (350), as shown in FIG. 5.
[0121] The agent configuration unit (310) can configure multiple artificial intelligence agents having different judgment criteria using patient context information generated by the context generation unit (100) and medical service information managed by the service execution management unit (200).
[0122] More specifically, the agent configuration unit (310) receives patient context information generated by the context generation unit (100) and medical service procedure information managed by the service execution management unit (200) as input, and can dynamically generate or combine multiple artificial intelligence agents having different judgment criteria and roles. These agents are designed to have different purposes and evaluation criteria rather than the repeated use of a single model, so that they can derive judgment results from various perspectives for the same input.
[0123] Additionally, the agent configuration unit (310) can analyze procedure requirements, budget ranges, schedule conditions, risk factors, language requirements, etc. included in the patient context to select or create an agent type suitable for the situation, and can deploy agents with different roles according to the medical service stage. For example, in the consultation stage, a consultation creation agent that interprets the patient's needs and suggests appropriate procedure options can be configured; in the hospital matching stage, a matching agent that evaluates suitability by comparing patient conditions and hospital information can be configured; and in subsequent stages, a verification agent that determines compliance or a decision agent that establishes a post-management plan can be configured.
[0124] Additionally, the agent configuration unit (310) can individually set prompt structures, reference knowledge, and evaluation criteria to reflect different judgment criteria for each agent, and can be configured to generate different output results even for the same input data. For example, one agent may evaluate a hospital based on cost-efficiency as a priority criterion, another agent may be set to make judgments based on medical quality or expertise as a priority criterion, and yet another agent may be configured to make judgments based on patient safety and risk factors.
[0125] Additionally, the agent configuration unit (310) can be configured to distinguish between a generating agent and a verifying agent, and some agents may be designed to perform the role of generating proposals or decisions regarding medical services, while other agents may be designed to perform the role of reviewing legal accuracy, practical applicability, risk factors, etc., regarding the generated results. Through this structure, decision-making based on multiple verifications rather than a single judgment can be made possible.
[0126] Additionally, the agent configuration unit (310) may apply a prompt assembly structure to generate different responses by reflecting the characteristics of each hospital, even when using the same artificial intelligence model. For example, if the medical departments, pricing policies, language support ranges, and internal consultation protocols of a specific hospital are stored in a database in advance, the agent configuration unit (310) can dynamically generate execution prompts for each individual agent by combining this hospital-specific knowledge data with patient context information and current progress stage information. Through this, even for the same patient request, a consultation response based on Hospital A and a consultation response based on Hospital B can be generated differently, and the structure can be configured such that the accuracy of the response varies depending on the level of information input by the hospital.
[0127] Additionally, the agent configuration unit (310) can form a cooperation and critique structure among multiple agents and can be configured to interact in such a way that another agent critiques or makes suggestions for modification regarding results generated by one agent. For example, regarding a specific hospital matching result, a legal standard verification agent may review whether there is a violation of regulations, and a practical perspective agent may evaluate the feasibility of actual operation and present different opinions.
[0128] Additionally, the agent configuration unit (310) can link external knowledge or data sources necessary for agent execution, and can be configured to utilize hospital databases, medical guidelines, regulatory documents, past case data, etc., in the decision-making process of each agent. For example, a hospital matching agent can make a decision by referring to hospital-specific procedure availability items, price information, and review data, and a compliance agent can make a decision based on relevant laws and guideline data.
[0129] Additionally, the agent configuration unit (310) can variably adjust the number of agents, configuration method, and interaction structure according to the situation, and can be configured with a small number of agents when simple decision-making is required, or with a large number of agents when complex judgment is required to derive more precise results. For example, in the case of simple consultation responses, it can be processed with 1 to 2 generating agents, and in the case of complex judgment such as hospital selection, it can be expanded to a structure including a large number of generating and verifying agents.
[0130] The above-mentioned multiple decision-making unit (320) can generate decision results for medical service provision procedures through a plurality of artificial intelligence agents configured by the agent configuration unit (310).
[0131] More specifically, the multiple decision execution unit (320) can execute multiple artificial intelligence agents configured by the agent configuration unit (310) in parallel or sequentially, so that each agent generates independent decision results according to different judgment criteria while using the same patient context information and medical service information as input. This process can be performed in a structure that generates multiple candidate results simultaneously rather than deriving a single result, so that comparison and verification can be made later.
[0132] Additionally, the multi-decision execution unit (320) can perform agent-specific decisions regarding specific stages or the entire flow of the medical service provision procedure, and can generate results for various decision items such as generating consultation content, recommending hospitals, adjusting appointment schedules, selecting procedure options, and establishing post-treatment plans. For example, for the same patient, one agent may recommend Hospital A based on cost efficiency, while another agent may recommend Hospital B based on medical staff expertise, thereby generating different results.
[0133] Additionally, the multi-decision execution unit (320) may provide the same input data to each agent, but generate results by reflecting the prompts, reference knowledge, and evaluation criteria set for each agent, thereby allowing results from various perspectives to be derived even under the same conditions. This method ensures diversity of results by allowing information included in the patient context (e.g., budget, purpose of procedure, risk factors, etc.) to be interpreted differently according to each agent's judgment criteria.
[0134] Additionally, the multi-decision execution unit (320) may configure a multi-stage execution structure including interactions between agents, and may perform an iterative decision-making process in which some agents generate initial results and other agents perform critique, modification, and supplementation on the results. For example, when a generating agent generates a hospital recommendation result, a legal verification agent reviews whether the recommendation complies with regulations, and a practical verification agent evaluates the actual reservation feasibility or operational realism to derive a modified result.
[0135] Additionally, the multiple decision execution unit (320) can improve the reliability of the decision results through a criticism structure among multiple agents and can identify risk factors or ambiguities based on judgment discrepancies between different agents. For example, if one agent recommends a specific hospital but another agent points out the hospital's schedule discrepancies or potential violation of regulations, this discrepancy information can be identified as a review item in the result analysis unit (330) and reflected in the final decision-making process. Furthermore, the multiple decision execution unit (320) can manage results generated from multiple agents as independent result objects, and each result may include the basis for judgment, data used, reliability information, etc. Through this, the result analysis unit (330) can subsequently compare differences between results and provide foundational data for deriving the final decision.
[0136] Additionally, the multiple decision execution unit (320) can adjust the timing of decision execution by reflecting the progress status of the medical service provision procedure, perform decisions necessary only at specific stages, or perform re-decisions if changes occur during the procedure. For example, if a patient's schedule changes or new requirements are added, the existing decision result can be invalidated and a result reflecting the new conditions can be regenerated.
[0137] The above result analysis unit (330) can identify decision items where differences occur between results by comparing multiple decision results generated by the multiple decision execution unit (320).
[0138] More specifically, the result analysis unit (330) can identify decision items where differences occur between results by collecting multiple decision results generated by the multiple decision execution unit (320), and then sorting and comparing the judgment items, recommendation content, selection basis, priority, constraints, etc. included in each result by item. This identification process can be performed not only to the extent of simply confirming whether the final recommendation results are different, but also in a manner that distinguishes in detail where differences occurred in judgment criteria, where data interpretations differed, and where condition evaluations conflicted.
[0139] Additionally, the result analysis unit (330) can perform comparisons after normalizing multiple decision results into a standard structure that allows for common comparison, and can analyze each result by breaking it down into multiple comparison items such as recommendation target, suitability score, risk assessment, schedule suitability, regulatory compliance, and level of post-management required. For example, even if one agent determines a specific hospital to be the most suitable institution, another agent may assign a lower priority to the same hospital due to schedule conflicts or cost overruns, and this difference can be identified as a difference in detailed judgment within the hospital selection items.
[0140] Additionally, the result analysis unit (330) can classify types of discrepancies between results and manage them by distinguishing them into differences in ranking for the same candidate, differences in the recommended target itself, differences in the basis for judgment, differences in the interpretation of regulations, differences in risk assessment, etc. Through this, it is possible to distinguish between simple result discrepancies and actual important conflicts in judgment. For example, a case where only the rankings of Hospital A and Hospital B are swapped can be classified as a difference in priority, and a case where one agent judges that the reservation can proceed while another agent judges that it cannot proceed due to the lack of required documents can be classified as a core conflict item regarding whether the procedure can proceed.
[0141] Additionally, the result analysis unit (330) can trace the cause of the difference by comparing the judgment basis and reference data included in each decision result, and can analyze which agent prioritized budget conditions and which agent prioritized medical risks or hospital expertise even with the same patient context. Through this cause analysis, subsequent steps can utilize not only the results but also the reasons for the occurrence of the result differences.
[0142] Additionally, the result analysis unit (330) can also analyze conflicts between the generated results and the verification results when a criticism structure between agents is applied. For example, if a consultation generating agent proposes a specific procedure schedule, but a compliance judgment agent determines that the schedule does not meet the specific document preparation deadline, the result analysis unit (330) can identify this as a conflict in the schedule suitability item. For example, if a post-care plan generating agent recommends a remote progress check within 3 days of returning home, but another verification agent determines that it is difficult to respond within 3 days due to the patient's stay schedule and time difference conditions, the result analysis unit (330) can identify this as a conflict in the post-care timing determination item.
[0143] Additionally, the result analysis unit (330) may assign an importance level or a level of priority review required for items where differences between results have occurred, and may prioritize identifying difference items that directly affect patient safety, regulatory compliance, appointment availability, and continuity of care. For example, differences in the expression of simple recommendation phrases may be classified as low importance, but items such as the possibility of surgery, suitability of hospital selection, and fulfillment of required documents may be identified as difference items of high importance.
[0144] Additionally, the result analysis unit (330) can generate the identified difference items into structured comparison result data and provide it to the result derivation unit (340), and can generate a comparison record for each difference item that includes the comparison target agent, conflict details, basis for judgment, priority, and scope of influence. This comparison result data can subsequently be used as reference information for selecting or adjusting the final decision result.
[0145] The result derivation unit (340) can generate a final decision result by selecting one or more of the multiple decision results based on the decision items identified in the result analysis unit (330).
[0146] More specifically, the result derivation unit (340) receives difference information and comparison result data for each decision item identified by the result analysis unit (330) as input, comprehensively evaluates multiple decision results, and can generate a final decision result by combining or selecting one or more results according to selection rules. This process can be performed not simply by selecting a single agent result as is, but by aggregating the optimal judgment for each item.
[0147] Additionally, the result derivation unit (340) can set selection criteria by reflecting importance and priority for each decision item, and can select results by applying stricter criteria to key elements such as patient safety, compliance with regulations, treatment feasibility, and schedule suitability. For example, even if a specific hospital recommendation result is suitable in terms of cost, if there is a problem with the judgment of compliance with regulations, that result is excluded, and an alternative that satisfies the regulatory conditions can be selected.
[0148] In addition, the result derivation unit (340) can generate a new decision result by combining multiple agent results, in addition to selecting a single agent result, and can select the most suitable judgment for each item to form a single integrated result. For example, the hospital selection item can be combined by adopting the result of an agent that applied expertise criteria, and the schedule adjustment item can be combined by adopting the result of an agent that evaluated realistic operational feasibility.
[0149] Additionally, the result derivation unit (340) may apply rule-based or learning-based decision logic to resolve conflicts between results, and may derive an optimal result by utilizing a predefined policy or a past data-based decision pattern. For example, if there is a selection result that has repeatedly shown high satisfaction under the same conditions, that pattern may be reflected preferentially.
[0150] Additionally, the result derivation unit (340) can produce a final result by a weighted summation method that reflects the reliability information or evaluation score included in each decision result, and can set different weights by considering the judgment reliability, data suitability, past performance, etc. of each agent. For example, the result of a compliance judgment agent may be assigned a weight with high reliability, and the result of a simple recommendation generation agent may be assigned a relatively low weight.
[0151] Additionally, the result derivation unit (340) can generate explanatory information including the basis for selection and the judgment process along with the final decision result, and thereby provide the reason why the result was derived so that the user or the medical service provider can understand it. For example, an explanation may be provided stating that cost-effectiveness, expertise, schedule feasibility, and regulatory compliance were comprehensively considered as reasons why a specific hospital was selected.
[0152] Additionally, the result derivation unit (340) may output the generated final decision result in the form of structured data so that it can be utilized by the result reflection unit (350) or the service execution management unit (200), and the result may be directly reflected in the modification, supplementation, or reorganization of the medical service provision procedure.
[0153] The result reflection unit (350) can provide the final decision result generated in the result derivation unit (340) to the service execution management unit (200) to reflect it in the medical service provision procedure.
[0154] More specifically, the result reflection unit (350) receives the final decision result generated by the result derivation unit (340) as input, converts the result into a form applicable to the medical service provision procedure managed by the service execution management unit (200), and can reflect the execution conditions, execution order, execution timing, or execution content of each step. This reflection process can be performed not by simply replacing the existing procedure, but by selectively modifying or supplementing only the necessary parts while maintaining the current state of progress.
[0155] Additionally, the result reflection unit (350) can analyze item-specific information included in the decision-making result to identify which stage of the procedure is affected and update the attribute value of the corresponding stage. For example, if the hospital matching result is changed, the data of the relevant stage can be updated so that the hospital selection information is reflected in the reservation stage and subsequent treatment stage, and if the schedule adjustment result is included, the timing of the execution of each stage can be reset.
[0156] Additionally, the result reflection unit (350) can perform procedure reconfiguration based on decision results in conjunction with the step standard setting unit (240), the time point generation unit (250), and the journey generation unit (260), and can recalculate the entire procedure structure when a specific step is added, removed, or changed in order. For example, if a decision result is obtained indicating that additional inspection is required, the inspection step is inserted into the existing procedure, and accordingly, the order and time of subsequent steps can be adjusted together.
[0157] Additionally, the result reflection unit (350) can record the change history and manage the difference from the previous state when a procedure change is required, thereby ensuring traceability of the procedure change. For example, if the procedure is modified due to a change in the patient's requirements or a change in the medical staff's judgment, the step configuration and schedule information before and after the change can be stored so that they can be compared.
[0158] Additionally, the result reflection unit (350) can maintain or adjust the stage status by considering the impact on the currently ongoing stage when reflecting the decision result, and can maintain stages that have already been completed and apply changes only to stages that have not yet been performed. For example, if the hospital selection is changed after the consultation stage is completed, the consultation stage is maintained as is, and only the subsequent reservation and treatment stages can be modified.
[0159] Additionally, the result reflection unit (350) can transmit the changed medical service provision procedure to the foreign patient terminal (10) and the medical service provision server (40) based on the reflected result, and can control the changed schedule, step configuration, execution conditions, etc. to be immediately reflected in the user interface. For example, the patient can check the changed appointment schedule and newly added preparation step information, and the hospital can perform follow-up actions according to the modified treatment plan.
[0160] The interaction support unit (400) can support interaction between a foreign patient and a medical service provider during the medical service provision process based on medical service information managed by the service performance management unit (200). To this end, the interaction support unit (400) may include at least one of an interaction receiving unit (410), a language processing unit (420), an interaction information generating unit (430), a channel control unit (440), and an output processing unit (450), as shown in FIG. 6.
[0161] The above interaction receiving unit (410) can receive at least one interaction data among voice data and text data generated or exchanged during the interaction process between a foreign patient and a medical service provider.
[0162] More specifically, the interaction receiving unit (410) can receive various forms of interaction data generated during the communication process between the foreign patient terminal (10) and the medical service providing server (40) or the medical staff terminal in real-time or asynchronously. This interaction data may include not only text data such as consultation messages, questions and answers, and appointment-related communications, but also voice-based data such as voice calls, voice messages, and voice streams during video calls.
[0163] Additionally, the interaction receiver (410) can simultaneously receive data transmitted through multiple communication channels and can distinguish between the patient-side channel and the medical service provider-side channel to process each independently. For example, when voice data transmitted from a foreign patient terminal (10) and voice data transmitted from a medical staff terminal are input through different channels, the data from each channel can be received individually and managed in an identifiable form.
[0164] Additionally, the interaction receiver (410) can continuously receive voice data transmitted via real-time streaming and can sequentially receive voice frames divided into partial units to reconstruct them into a continuous speech in a subsequent processing step. For example, when a conversation between a doctor and a patient is transmitted in real time during medical treatment, voice data divided into short time units can be continuously received and accumulated.
[0165] Additionally, the interaction receiver (410) may generate or assign metadata such as the sender, the time of reception, channel information, and a session identifier to the received data, thereby enabling tracking of the circumstances under which each interaction data occurred. For example, even if the consultation data is for the same patient, it can be stored by distinguishing whether the data occurred during the consultation stage or the treatment stage.
[0166] Additionally, the interaction receiver (410) can support various data formats and protocols and can receive data from different input paths, such as web-based messaging, mobile application input, 'WebSocket'-based real-time communication, and voice call interfaces. For example, chat messages can be received via 'REST API' or 'WebSocket', and voice data can be received via a real-time streaming protocol.
[0167] Additionally, the interaction receiver (410) can perform a preprocessing function to correct the quality of the received data, and in the case of voice data, it can perform noise removal or signal normalization processing, and in the case of text data, it can perform encoding error correction or incomplete sentence correction processing. For example, it can improve voice recognition accuracy by removing background noise that occurs during a call, or convert the input so that truncated sentences can be supplemented to enable semantic interpretation.
[0168] The above language processing unit (420) can generate language information converted into different languages based on interaction data received from the interaction receiving unit (410).
[0169] More specifically, the language processing unit (420) receives voice data or text data received from the interaction receiving unit (410) as input, identifies the language of the data, and can generate language information converted into another language in a form suitable for the medical service provision situation. This conversion is not a general translation, but can be performed based on semantic units by reflecting specialized terminology and context used in medical consultation and treatment environments.
[0170] Additionally, the language processing unit (420) may be configured to support bidirectional real-time interpretation between the patient and the medical staff, and may process voice data received from different communication channels and convert it into the other party's language. This structure may be implemented by operating the patient-side channel and the medical staff-side channel separately, and by independently performing speech recognition (STT) and translation processing on the voice data received from each channel.
[0171] Additionally, the language processing unit (420) can apply medical department-specific term hints to the translation process to reflect medical domain-specific terminology, thereby reducing mistranslations that may occur in general translation and improving the accuracy of conveying medical meaning. For example, in a dermatology practice situation, terms such as 'laser resurfacing', 'pigmentation', and 'dermal layer' can be processed to be translated into their accurate medical meanings, and in a plastic surgery situation, expressions such as 'general anesthesia', 'incision line', and 'post-operative swelling' can be appropriately converted.
[0172] Additionally, the language processing unit (420) can minimize conversational delay by performing translation based on real-time streaming, and can generate conversion results sequentially by dividing input voice data into utterance units of a certain length. This process can maintain a natural conversational flow by providing partial translation results before the entire sentence is completed.
[0173] Additionally, the language processing unit (420) can operate in the following specific manner in an actual medical tourism service environment. For example, when a foreign patient speaks "I want a skin treatment for acne scars, and my budget is around 2,000 dollars" through the foreign patient terminal (10), the voice data is converted into text in real time and then converted into Korean, such as "I want to receive acne scar treatment, and my budget is around 2,000 dollars," and displayed on the medical staff terminal. At this time, "acne scars" is accurately interpreted as "acne scars" rather than simply "acne," and "skin treatment" can be converted into an expression suitable for the context of a dermatological procedure.
[0174] Conversely, if medical staff say, "You should avoid sun exposure for one week after the procedure," the audio can be converted into text and then translated into the patient's language, such as "You should avoid sun exposure for about one week after the procedure," to be provided to the patient's terminal in real time. In this case, expressions such as "UV exposure" and "post-procedure care" can be converted so that they are naturally conveyed as medical precautions, rather than being a general literal translation.
[0175] Additionally, the language processing unit (420) can maintain a correspondence between the generated translation result and the original text to store the entire conversation content in the form of a transcript, and the original text and the translation are recorded together for each utterance unit so that they can be used for managing consultation history or medical records thereafter. For example, all conversations that took place during the consultation phase are stored together with the original text and translation in chronological order, so that hospital staff can check the content later or refer to it during the follow-up care phase.
[0176] Additionally, the language processing unit (420) may compare multiple translation candidates or perform context-based correction processing to maintain the quality of the translation results, and if there is a possibility of errors in conveying medical meaning, it may review and modify it into a more suitable expression. For example, if multiple translation results are generated for the same sentence, the final expression may be selected based on medical term accuracy and contextual suitability to generate language information.
[0177] The above interaction information generation unit (430) can generate interaction information that can be reflected in the medical service provision procedure using the interaction data received from the interaction receiving unit (410) and the language information generated from the language processing unit (420).
[0178] More specifically, the interaction information generation unit (430) can generate structured interaction information that can be directly utilized in the medical service provision procedure by combining the translation result generated by the language processing unit (420) with the original interaction data received by the interaction receiving unit (410). This interaction information is not a simple conversation record, but can be structured in the form of data organized into semantic units so that it can be reflected in decision-making and procedure progress at each stage of the medical service.
[0179] Additionally, the interaction information generation unit (430) can generate interaction information in a structure that includes metadata such as the sender, the time of reception, stage information, and the purpose of conversation, while maintaining a correspondence between the original data and the translation data and sorting the data by utterance unit. Through this, it is possible to clearly distinguish whether each interaction occurred during the consultation stage, the reservation stage, or the treatment stage.
[0180] Additionally, the interaction information generation unit (430) can extract and structure key information that affects the medical service provision procedure from the conversation content, thereby converting it into actionable information rather than simple text. For example, if a patient inputs a utterance requesting to select a specific procedure or change the schedule during the consultation process, the content can be converted into event information such as 'change of procedure selection' or 'request to modify appointment schedule' and reflected in the procedure.
[0181] Additionally, the interaction information generation unit (430) can generate a dual record structure that maintains the original text and the translation together to prevent semantic distortion occurring in a multilingual conversation environment, thereby allowing medical staff or coordinators to verify the original text and the translation simultaneously. For example, the patient's English speech and the corresponding Korean translation result can be stored together in the same record, so that verification based on the original text is possible when necessary.
[0182] Additionally, the interaction information generation unit (430) can generate interaction information in a specific form linked to the actual medical service flow. For example, if a patient says "I would like to change my appointment to next week," the utterance is not stored as a simple translation result, but can be converted into structured event information such as "request to change appointment schedule," "desired time of change: next week," and transmitted to the service performance management unit (200). For example, if a medical staff member says "Please upload a photo to check the progress 3 days after the procedure" during the treatment process, the content can be converted into a form such as "additional post-care step task," "task type: photo upload," "time of execution: 3 days after the procedure," and reflected in the journey information.
[0183] Additionally, the interaction information generation unit (430) can utilize the generated interaction information in conjunction with the medical service provision procedure to change the state of a specific stage or add a new task, thereby configuring the conversation-based input to be directly reflected in the service flow. For example, if a patient inputs a utterance selecting a specific hospital during a consultation, the interaction information can be converted into a hospital selection completion event and reflected to activate the subsequent reservation stage.
[0184] Additionally, the interaction information generation unit (430) can store the generated interaction information so that it can be recorded and tracked, and provide it for use in subsequent consultation history management, medical record management, and follow-up management stages. For example, major interactions that occurred during the entire treatment process are structured and stored in chronological order, so that medical staff can easily identify changes in the patient's condition and request history.
[0185] The above channel control unit (440) can set up a communication channel corresponding to each foreign patient and medical service provider based on the interaction information generated by the interaction information generation unit (430) and the language information generated by the language processing unit (420), and control the transmission of the interaction information and language information through the set communication channel.
[0186] More specifically, the channel control unit (440) receives structured interaction information generated by the interaction information generation unit (430) and translated language information generated by the language processing unit (420) as input, sets up communication channels corresponding to the foreign patient terminal (10) and the medical service provision server (40) or medical staff terminal, respectively, and controls the transmission of data in a form suitable for each entity. These communication channels may be configured as a multi-channel structure separated by user, role, and session, rather than a single channel.
[0187] Additionally, the channel control unit (440) can logically separate the patient-side channel and the medical staff-side channel, and each channel can be configured to have an independent data flow. For example, the foreign patient terminal (10) can be configured to have a channel for transmitting language information converted into the patient's native language, and the medical staff-side channel can be configured to have information converted into the language used by the medical staff transmitted. Through this, even for the same interaction, data can be provided in a language and format suitable for each recipient.
[0188] Additionally, the channel control unit (440) can configure the patient channel and the medical staff channel as physically separated WebSocket connections to control voice and text data to be transmitted independently without mutual interference. For example, in the patient-side channel, the patient's voice data can be transmitted to a server for voice recognition and translation processing, and in the medical staff-side channel, the medical staff's voice data can be processed separately and transmitted to the other party. Each channel can be maintained as a different stream to enable real-time communication without data collision even in situations of simultaneous speech.
[0189] Additionally, the channel control unit (440) can set up a 'WebSocket'-based bidirectional communication channel to support real-time communication, and each channel can be controlled to maintain a continuous connection state so that interaction information and language information are transmitted without delay. For example, when a patient and medical staff speak simultaneously during treatment, the voice recognition result and translation result are transmitted to the other party in real time through each channel, so that a natural flow of conversation can be maintained.
[0190] Additionally, the channel control unit (440) can be configured to manage interactions for the same patient as a single session by assigning session identifiers and user identification information to each channel, and can manage channels by separating them according to each service stage, such as the consultation stage, the reservation stage, and the treatment stage. For example, channels created during the consultation stage and channels created during the treatment stage can be separated so that their respective interaction data can be managed independently.
[0191] Additionally, the channel control unit (440) can distinguish the transmission method according to the type of data being transmitted, and can control text data to be transmitted in message units and voice data to be transmitted in a streaming manner. For example, translated text can be output immediately in the form of a message, and voice interpretation results can be transmitted continuously via real-time streaming.
[0192] Additionally, the channel control unit (440) can operate in the following specific manner in an actual medical tourism service environment. For example, when content spoken in English by a foreign patient through the foreign patient terminal (10) is converted into Korean by the language processing unit (420), the Korean translation information can be transmitted to the medical service provision server (40) through the medical staff side channel, and at the same time, the original English or corrected English expression can be maintained and displayed on the patient side channel. For example, when a medical staff member speaks in Korean, “The procedure will take about 1 hour,” the speech is translated into English and transmitted as “The procedure will take about one hour” through the patient side channel, and the original Korean can be maintained on the medical staff side channel.
[0193] Additionally, the channel control unit (440) can transmit specific notifications or status change messages to appropriate channels based on event information included in the interaction information, and can transmit not only simple conversations but also system messages related to the progress of procedures. For example, if a reservation is confirmed, the information is transmitted to the patient terminal in the form of a notification message, and at the same time, a reservation confirmation event is transmitted to the medical staff channel so that subsequent preparations can be made.
[0194] Additionally, the channel control unit (440) can reset the channel or retransmit data when a network state change or a connection error occurs, and can perform control functions to maintain communication stability. For example, if a temporary disconnection occurs during real-time communication, the channel can be restored based on previous session information to control the continuation of interaction.
[0195] The output processing unit (450) can provide at least one of the interaction information and language information transmitted by the channel control unit (440) to a foreign patient or a medical service provider.
[0196] More specifically, the output processing unit (450) receives interaction information and language information transmitted through the channel control unit (440) as input, processes them into a form suitable for the foreign patient terminal (10) or the medical service providing server (40), and provides them in the form of visual or auditory output. This output may be provided not as simple data transmission, but as information reconstructed to suit the user interface environment.
[0197] Additionally, the output processing unit (450) can display text-based interaction information on a user screen in the form of a message, and can output it in the form of a chat interface arranged in chronological order according to the conversation flow. For example, on a patient terminal, the consultation content is displayed in a messenger format, and each message can be provided separately in the form of the original text or a translation, and can be configured so that the original text and the translation are displayed together as needed.
[0198] Additionally, the output processing unit (450) can provide voice-based interaction information in the form of voice output, and can convert the translation result generated by the language processing unit (420) into a text-to-speech (TTS) method and deliver it to the user. Through this, the user can listen to the other party's speech in real time without reading the text.
[0199] Additionally, the output processing unit (450) can reflect structured event information included in the interaction information in the user interface, and can output information such as reservation status, stage progress, and request processing results in addition to simple conversations in a separate notification or status display area. For example, if a patient requests a change to their reservation, the current processing status may be displayed along with a message indicating that the request has been received.
[0200] Additionally, the output processing unit (450) can operate in the following specific manner in an actual medical service environment. For example, when a foreign patient conducts a consultation in English, the Korean speech of the medical staff can be translated into English and displayed on the chat screen of the patient's terminal as a sentence such as "You should visit the clinic two days before the procedure," and at the same time, the sentence can be played back as voice so that the patient can understand it immediately. For example, if the patient inputs "Can I change my appointment to Friday?", the message is displayed on the medical staff's screen as the Korean translation "Can I change the appointment to Friday?", and the response from the medical staff can be output back to the patient's terminal in English.
[0201] Additionally, the output processing unit (450) can visually provide information linked to the step-by-step medical service provision procedure, and can output the currently ongoing step, the next step to be performed, the scheduled schedule, etc., in the form of a timeline or calendar. For example, on the patient terminal, a step flow such as "consultation completed," "appointment confirmed," and "visit scheduled" is visually displayed, and a schedule and items to be performed corresponding to each step can be provided together.
[0202] Additionally, the output processing unit (450) can differentiate the output content according to the user type, and can be configured to provide understanding-oriented information to foreign patients and detailed information necessary for performing tasks to medical service providers. For example, a brief guide message and translated results are provided to patients, while structured information regarding patient requests, schedule change conditions, and additional verification requirements may be displayed together to medical staff.
[0203] The above procedure correction unit (500) can correct the medical service provision procedure based on the decision result generated by the multi-agent processing unit (300). To this end, the procedure correction unit (500) may include at least one of a correction target identification unit (510) and a procedure correction execution unit (520), as shown in FIG. 7.
[0204] The correction target identification unit (510) can identify a step or execution condition requiring correction among the medical service provision procedures managed by the service execution management unit (200) using the decision result generated by the multi-agent processing unit (300).
[0205] More specifically, the correction target identification unit (510) can analyze the final decision result generated by the multi-agent processing unit (300) together with the current medical service provision procedure information managed by the service execution management unit (200) to detect points where differences or discrepancies occur between the existing procedure and the decision result, and to identify steps or execution conditions that require correction. This identification process can be performed by comprehensively considering various factors such as step structure, execution timing, feasibility of execution, and compliance with regulations, rather than simply comparing values.
[0206] Additionally, the correction target identification unit (510) can track the range of affected stages based on the change items included in the decision result and can map which stage of the procedure a specific decision affects on a stage-by-stage basis. For example, if the hospital matching result is changed, the change may affect not only the reservation stage but also the subsequent treatment stage and follow-up care stage, so all related stages can be identified as candidates for correction.
[0207] Additionally, the correction target identification unit (510) can determine whether there is a timing conflict or schedule discrepancy by reflecting the results of schedule-related decision-making, and accordingly identify the stage requiring correction. For example, if the existing itinerary information is not updated even though the patient's visit schedule has changed, the timing of the procedure preparation stage, the visit stage, the follow-up care stage, etc., can all be identified as stages subject to readjustment.
[0208] Additionally, the correction target identification unit (510) can identify steps or conditions within the procedure that may be in violation by reflecting the results of the compliance judgment, and if a specific step does not meet medical regulations or hospital policies, that step may be designated as a correction target. For example, if a pre-examination step that is essential for performing a specific procedure is omitted, that step may be identified as an additional correction target.
[0209] Additionally, the correction target identification unit (510) can limit the applicable range of correction by considering the state of step execution, and can selectively identify areas requiring correction by distinguishing between steps that have already been completed and steps that have not yet been performed. For example, after a consultation step has already been completed, that step is maintained, and only subsequent steps, such as the reservation and treatment steps, can be identified as correction targets.
[0210] Additionally, the correction target identification unit (510) can operate in the actual medical service flow in the following specific manner. For example, if the multi-agent processing unit (300) generates a decision result stating that "additional pre-operative examination is required in the current state" considering the patient's condition, and the existing procedure does not include the corresponding examination step, the correction target identification unit (510) can identify an item requiring correction, such as "missing examination step," and also identify the location where the examination step should be inserted. For example, if a patient requests a schedule change after completing a reservation following a consultation, the difference between the new schedule reflecting the request and the existing schedule can be analyzed, and all schedule-related steps after the reservation step can be identified as correction targets.
[0211] Additionally, the correction target identification unit (510) can generate the identified correction target steps and execution conditions in the form of structured data and provide them to the procedure correction execution unit (520), and for each correction target, information such as the reason for correction, the scope of influence, and priority can be included and utilized in the subsequent correction processing process.
[0212] The above procedure correction execution unit (520) can modify the medical service provision procedure by reflecting the decision result generated by the multi-agent processing unit (300) regarding the step or execution condition identified by the correction target identification unit (510), and update the modified medical service provision procedure by reflecting it in the medical service provision procedure information managed by the service execution management unit (200).
[0213] More specifically, the procedure correction execution unit (520) receives as input the step or execution condition identified by the correction target identification unit (510) and the decision result generated by the multi-agent processing unit (300), and can generate an updated procedure by modifying, adding, or reconfiguring necessary parts based on the existing medical service provision procedure. This correction is not performed by initializing the entire procedure, but can be selectively performed only on sections requiring change while maintaining the current progress state.
[0214] Additionally, the procedure correction execution unit (520) can perform a step-by-step modification function and can change the attribute value of a specific step, insert a new step, or remove an unnecessary step. For example, if a specific hospital is changed based on the judgment result of the multi-agent processing unit (300), the hospital information of the existing reservation step can be replaced with the new hospital information, and accordingly, the linkage information of the subsequent treatment step and post-treatment step can be modified together.
[0215] Additionally, the procedure correction execution unit (520) can readjust the order and dependency relationships between steps, and can reorganize the order so that the entire procedure flow is logically maintained when a new step is added or an existing step is changed. For example, if a decision result indicating that additional testing is required is reflected, the order can be adjusted so that the testing step is inserted after the consultation step and before the reservation step.
[0216] Additionally, the procedure correction execution unit (520) can correct schedule information in conjunction with the time point generation unit (250) and the journey generation unit (260), and can be configured so that the execution time of each step is recalculated according to the change in the step. For example, if the procedure schedule is changed, the previously set pre-preparation schedule and post-management schedule can both be recalculated to match the new reference time.
[0217] Additionally, the procedure correction execution unit (520) can change the execution criteria of the procedure by modifying the step execution conditions and can control whether a specific step is performed based on conditions. For example, if it is determined that a specific procedure is impossible due to a change in the patient's health condition, the corresponding procedure step may be deactivated or changed to an alternative procedure step.
[0218] Additionally, the procedure correction execution unit (520) can operate in the actual medical service flow in the following specific ways. For example, if the multi-agent processing unit (300) analyzes the patient's condition after the patient has booked a skin procedure and derives a decision that "additional skin examination is needed," an examination step that did not exist in the existing procedure is inserted into the post-consultation step, and the timing of the booking step can be automatically adjusted to after the completion of the examination. For example, if the patient requests a schedule change after completing the booking and the visit date is postponed by one week, the procedure correction execution unit (520) can collectively readjust the timing of the entire journey, including not only the existing procedure schedule but also the pre-procedure preparation step and the post-treatment care schedule.
[0219] Additionally, the procedure correction execution unit (520) can transmit the corrected medical service provision procedure to the service execution management unit (200) to update existing procedure information, and can be configured so that the changed procedure is immediately reflected and applied to subsequent step execution and status management. This update process can be carried out while maintaining the difference from the previous procedure, so that the change history can be tracked.
[0220] The service matching unit (600) can query information about a medical service provider based on patient context information generated by the context generation unit (100), and derive a medical service provider corresponding to the patient context information based on the queried information. To this end, the service matching unit (600) may include at least one of a query generation unit (610), an information query unit (620), and a matching result generation unit (630), as shown in FIG. 8.
[0221] The above query generation unit (610) can generate query information for querying a medical service provider using patient context information generated by the above context generation unit (100).
[0222] More specifically, the query generation unit (610) can receive patient context information generated by the context generation unit (100) as input and generate search query information for searching for a medical service provider. This query information may not be a simple list of keywords, but may be composed of structured search conditions that reflect the patient's condition information and requirements, and may be designed to efficiently query a suitable hospital or medical institution from an external data repository.
[0223] Additionally, the query generation unit (610) can identify key attributes included in the patient context and convert them into query parameters, and can configure information such as procedure type, budget range, preferred region, visit schedule, length of stay, language requirements, and hospital type as independent conditions. These conditions can be divided into mandatory conditions and optional conditions and reflected in the query. For example, the condition that a hospital is capable of performing a specific procedure can be set as a mandatory condition, and a specific region or price range can be set as an optional condition.
[0224] Additionally, the query generation unit (610) can convert natural language-based input into a structured query form and can analyze the patient's consultation content or request sentence to reconstruct it into a semantic-based query. For example, if a patient inputs "I want to find a dermatology clinic in Seoul that speaks English, and my budget is around 2 million won," the query generation unit (610) can convert this into a structured query condition such as "Procedure type: Dermatology, Location: Seoul, Language support: English, Budget: 2 million won or less."
[0225] Additionally, the query generation unit (610) can perform synonym expansion or condition supplementation to improve query performance, and can expand the query to include additional keywords or categories related to specific conditions. For example, if a condition such as 'dermatology procedure' is entered, the query can be expanded to include related detailed procedure categories such as laser procedure, acne treatment, pigmentation treatment, etc.
[0226] Additionally, the query generation unit (610) can perform inference-based condition generation to compensate for the incompleteness of the patient context, and can construct query conditions within a reasonable range even when some information is not specified. For example, if a patient has specified the type of procedure but has not entered a budget, a basic budget condition can be set based on the average cost range of the procedure.
[0227] Additionally, the query generation unit (610) can operate in the following manner in an actual medical tourism platform environment. For example, if the patient context includes information such as "desiring plastic surgery nose surgery, budget of $5,000, planned visit date in early August, preferring a hospital where English is spoken," the query generation unit (610) can generate a complex condition query based on this, such as "Procedure availability: Nose surgery, Budget range: $5,000 or less, Available visit date: Booking available in early August, Language support: English, Location: Korea." For example, if a patient requests acne scar treatment during a skin procedure consultation and the duration of stay is limited to 5 days, the query generation unit (610) can generate a query including conditions such as "Procedure type: Acne scar treatment, Recovery period: Short-term, Duration of stay: Suitable for 5 days or less, Booking availability: Capable of accommodating short-term schedules" so as to prioritize searching for hospitals where short-term procedures are available and quick recovery is possible.
[0228] Additionally, the query generation unit (610) can convert the generated query information into a standardized API request form or a database query form and provide it to the information retrieval unit (620), thereby enabling the search for medical service providers in a consistent manner from an external data repository.
[0229] Additionally, the query generation unit (610) may operate not as a simple static query generation method, but as a dynamic query generation structure that generates queries in real-time during the conversation flow and calls for external data retrieval. For example, if a patient enters additional conditions during the consultation process (e.g., "It is okay if the price is a little higher, and I want to look for a place with good reviews"), the query generation unit (610) may generate a modified query that reflects the new conditions while maintaining the existing query, and the information retrieval unit (620) may be configured to immediately re-retriev based on the query. This process is implemented as an 'LLM'-based 'tool call' structure, allowing for the repeated generation and retrieval of multiple queries within a single conversation session, and enabling the gradual refinement of optimal hospital candidates while maintaining the context of the conversation with the patient.
[0230] The above information inquiry unit (620) can query information about a medical service provider from an external data repository using the query information generated by the query generation unit (610).
[0231] More specifically, the information retrieval unit (620) receives structured query information generated by the query generation unit (610) as input, searches for data related to medical service providers stored in an external data repository, and can retrieve candidate information that meets the conditions. This external data repository may be composed of various types of data sources, including a hospital information database, data on available procedures, price information, medical staff expertise information, available appointment schedules, language support status, etc.
[0232] Additionally, the information inquiry unit (620) can perform filtering and searching based on multiple conditions included in the query information, and can perform the inquiry by first applying essential conditions to reduce the set of candidates and then applying selection conditions to refine the results. For example, the candidate list can be limited by first applying the condition that a hospital is capable of performing a specific procedure, and then a final candidate list can be generated by additionally reflecting conditions such as budget range, location, and language support.
[0233] Additionally, the information lookup unit (620) can perform not only a simple keyword matching method but also a semantic-based search or a vector similarity-based search, and can derive more suitable results based on the semantic similarity between the patient's requirements and the hospital characteristics. For example, if a patient presents a requirement that they "prefer natural results," the condition can be matched with data related to the procedure style or medical staff tendencies rather than simple keywords, and a suitable hospital can be looked up.
[0234] Additionally, the information lookup unit (620) can generate lookup results by integrating multiple data sources, and can generate results containing richer information by linking not only internal databases but also external APIs, affiliated hospital systems, review data, evaluation data, etc. For example, patient review ratings, the number of successful procedures, and medical staff experience information can be retrieved together with the availability of the hospital's procedure and included in the result.
[0235] Additionally, the information inquiry unit (620) can perform sorting and ranking processing on the inquiry results, and can sort the results based on priority by comprehensively considering the suitability with the query conditions, whether the price conditions are met, the availability of the schedule, the level of language support, etc. For example, among hospitals capable of performing the same procedure, a hospital that is closest to the patient's budget range and has a matching available schedule may be placed at the top.
[0236] Additionally, the information lookup unit (620) can operate in the following manner in an actual medical tourism platform environment. For example, if the query generation unit (610) generates a condition such as "dermatology acne scar treatment, budget of 2 million won or less, English support, Seoul area," the information lookup unit (620) can look up a list of hospitals satisfying the condition from the database and return a result including the procedure price, available appointment schedule, availability of English consultation, and location information for each hospital. For example, if a patient enters a condition such as "rhinoplasty, available to visit in early August, budget of 5,000 dollars," the information lookup unit (620) can prioritize looking up hospitals that meet the budget range among hospitals available for booking during that period, and simultaneously compose the result by including data on medical staff expertise and procedure experience.
[0237] Additionally, the information inquiry unit (620) can convert the inquiry results into a standardized data structure and provide them to the matching result generation unit (630), and can provide them in the form of an object containing a unique identifier, available procedure items, price information, schedule information, evaluation information, etc. for each medical service provider so that they can be utilized in a subsequent processing step.
[0238] The above matching result generation unit (630) can determine the medical service provider corresponding to the patient context information based on the result retrieved from the information retrieval unit (620).
[0239] More specifically, the matching result generation unit (630) receives multiple candidate medical service provider information retrieved from the information retrieval unit (620) as input, evaluates the suitability with the patient context information, and can ultimately determine the medical service provider to be provided to the patient. This decision process can be performed not by simply returning a retrieval result, but by precisely comparing and analyzing the patient's condition information and requirements information with the characteristics of each candidate hospital to derive an optimal matching result.
[0240] Additionally, the matching result generation unit (630) can perform a multidimensional evaluation for each candidate medical service provider and can calculate a suitability score by comprehensively reflecting multiple evaluation criteria such as procedure suitability, budget suitability, schedule suitability, location conditions, language support availability, medical staff expertise, and patient review evaluation. This suitability score can be calculated by reflecting weights for each criterion, and the weights can be dynamically adjusted according to the patient context.
[0241] Additionally, the matching result generation unit (630) can filter candidates based on whether the patient requirements are met, exclude medical service providers that do not satisfy the essential conditions, and perform a detailed evaluation only on the remaining candidates. For example, hospitals where certain procedures are impossible or language support is not provided may be excluded in the initial stage.
[0242] Additionally, the matching result generation unit (630) may select one of the multiple candidates as a single option, sort the multiple candidates based on priority and provide them in a list form, or configure the recommendation results so that the patient can make a selection. For example, the hospital with the highest suitability may be presented as the first priority, and hospitals with some differences in conditions may be provided together as the second and third priorities.
[0243] Additionally, the matching result generation unit (630) can generate detailed matching grounds between the patient context and hospital characteristics, and can provide the reason for selection for each recommendation result in an explanatory form. For example, it can present that conditions such as "meeting the budget range, having medical staff specializing in the procedure, English consultation available, and matching available appointment schedules" are met as reasons why a specific hospital was recommended.
[0244] Additionally, the matching result generation unit (630) can operate in the following manner in an actual medical tourism platform environment. For example, if the patient context includes conditions such as "acne scar treatment, budget of 2 million won or less, English consultation available, 5-day stay," among the searched hospitals, a hospital that provides the relevant procedure, offers a procedure option that allows for short-term recovery, and offers English consultation can be selected as a priority, and at the same time, a hospital where treatment can be completed within the stay period can be determined as the top result. For example, if a patient presents conditions such as "rhinoplasty, budget of 5,000 dollars, visit in early August," among hospitals where a surgery schedule can be secured during that period and procedures can be performed within the budget range, a hospital with specialized medical staff can be selected as the final matching result.
[0245] Additionally, the matching result generation unit (630) can generate the determined medical service provider information in the form of structured data and provide it to the service performance management unit (200) or the multi-agent processing unit (300), and the result can be directly utilized in subsequent reservation, consultation, and treatment stages.
[0246] The above-mentioned coordinator agent processing unit (700) can verify the suitability of each step based on patient context information and medical service provision procedure information, and control the execution of each step of the medical service provision procedure according to the verification result. To this end, the coordinator agent processing unit may include at least one of a rule extraction unit (710), a rule base management unit (720), a step verification unit (730), and an action decision unit (740), as shown in FIG. 9.
[0247] The rule extraction unit (710) can extract rule information from unstructured documents. More specifically, the rule extraction unit (710) receives data of various formats as input, such as natural language-based descriptions, procedural descriptions, medical guidelines, coordinator manuals, certification textbooks, and internal hospital protocol documents contained in unstructured documents, and can analyze the normative meaning inherent in the documents to structure them into rule information. Here, unstructured documents may include documents created without a defined data schema, such as PDFs, image-based documents, text files, scanned documents, and web pages, and to process these various input formats, optical character recognition (OCR), natural language processing (NLP), semantic analysis algorithms, etc., can be combined and applied.
[0248] Additionally, the rule extraction unit (710) can analyze the meaning of sentences or paragraphs within a document to identify situation, condition, action, and constraint elements, which are key elements constituting the rule, and can separate each element to convert it into a structured rule unit. For example, if a guideline sentence related to a specific medical procedure contains the content "a specific procedure should be restricted if the patient has a specific disease," the sentence can be broken down into situation information corresponding to the patient's condition, condition information determining whether to perform the procedure, action information that is performed or restricted, and constraint information signifying prohibition or restriction, and converted into a rule structure.
[0249] Additionally, the rule extraction unit (710) can combine rule elements distributed across multiple sentences or paragraphs, as well as rules at the single sentence level, to reconstruct them into a single integrated rule, and can connect mutually complementary rule elements by analyzing the association between contexts. For example, if the application conditions of a specific procedure are described in one paragraph and the contraindications of the procedure are described in another paragraph, these pieces of information can be combined to create a more complete rule structure.
[0250] Additionally, the rule extraction unit (710) can perform synonym processing, expression normalization, and entity standardization processes by considering the diversity of expressions within the document, and can integrate various expressions having the same meaning into a consistent rule element. For example, expressions such as "no surgery," "no procedure," and "not applicable" can be recognized as having the same constraint and integrated into a single rule expression.
[0251] Additionally, the rule extraction unit (710) can analyze the logical relationships between rule elements to generate complex rules that include logical operation structures such as AND, OR, and NOT between conditions, thereby enabling more precise rule judgment. For example, a rule can be generated that allows a specific action only when a specific age condition and a specific disease condition are simultaneously satisfied.
[0252] Additionally, the rule extraction unit (710) can perform reliability evaluation or priority setting on the extracted rule information, and can determine the priority of application of the rules by considering the source of the document, the time of creation, medical importance, etc. For example, rules extracted from the latest medical guidelines can be set to have a higher priority than rules extracted from past documents.
[0253] Additionally, the rule extraction unit (710) can convert the extracted rule information into a structured data form that can be utilized in subsequent processing steps, for example, by converting it into JSON, a graph data structure, or an object-based representation and transmitting it to the rule base management unit (720). This structured rule information can be used in subsequent steps to establish relationships between rules, construct a hierarchical structure, and verify the suitability of the steps. For example, if a PDF document for medical coordinator training contains a rule stating "If the patient does not submit specific test results before the procedure, the appointment cannot be confirmed," the sentence can be analyzed to structure it into a rule that includes a condition of "failure to submit test results before the procedure" and a behavioral restriction of "cannot confirm the appointment."
[0254] The above rule base management unit (720) can generate a rule structure based on rule information extracted from the rule extraction unit (710) and store and manage the rule structure in a hierarchical structure.
[0255] More specifically, the rule base management unit (720) receives rule information transmitted from the rule extraction unit (710) as input and can generate a rule structure by reorganizing the rule information into structured rule units including situation, condition, action, and constraint elements. At this time, each rule can be expressed in the form of a single condition-based rule or a composite rule combining multiple conditions, and can be defined in a form in which the permission or restriction of a specific action is determined depending on whether the condition is satisfied.
[0256] Additionally, the rule base management unit (720) can analyze the relationships between multiple generated rules to form a hierarchical structure between upper rules and lower rules. This hierarchical structure can generally be configured to reflect the relationship between standard rules that are applied globally and rules that are applied only to specific hospitals or specific situations, and the upper rules may include more general and universal constraints, while the lower rules may be defined as detailed rules corresponding to specific conditions.
[0257] Additionally, the rule base management unit (720) can be configured to resolve conflicts between multiple rules in the same situation by setting priorities or scopes of application among the rules. For example, even if there is an internal rule defined at a specific hospital, the hospital rule can be controlled so that it is not applied if it violates a higher-level global standard rule.
[0258] Additionally, the rule base management unit (720) can manage the application priority, application scope, and policy attributes together for each rule, and the policy attributes may include information on the control method that can be performed when a rule is violated, whether recommendation information is provided, and whether automatic execution is possible.
[0259] Additionally, the rule base management unit (720) can convert the rule structure into a graph-based data structure and store it, and each rule can be represented as a node and the relationships between rules can be represented as edges. Through this graph structure, inheritance relationships, dependency relationships, and constraint relationships between rules can be clearly expressed and can be efficiently utilized in the rule search and verification process in subsequent stages.
[0260] Additionally, the rule base management unit (720) can perform dynamic update functions for the rule structure and can reorganize the hierarchical structure or redefine the relationships between some rules when new rule information is added or existing rules are modified or deleted. At this time, the rule change history can be managed as separate version information so as to be configured to track the rule status at a specific point in time.
[0261] Additionally, the rule base management unit (720) may provide a lookup interface to enable external components to utilize the stored rule structure, and the step verification unit (730) may search for and apply relevant rules through the interface when verifying the suitability of performing a specific step. For example, if a global standard rule related to a specific procedure is defined as "restriction on procedures for patients with a specific disease," and a rule is defined at a specific hospital to "relax restrictions on some conditions of the disease," these rules can be hierarchically linked, and a structure can be set up to apply the higher-level rule first.
[0262] The above-mentioned step verification unit (730) receives as input patient context information generated by the context generation unit (100) and the planned actions for each step of the medical service provision procedure configured by the service execution management unit (200), and can verify the suitability of each step's execution using a rule structure managed by the rule base management unit (720).
[0263] More specifically, the step verification unit (730) receives as input patient context information generated by the context generation unit (100) and information on planned actions to be performed at each step of the medical service provision procedure provided by the service performance management unit (200), and can convert the information into data to be verified by normalizing it into a form to which rules can be applied. At this time, the patient context information may include patient status information, requirements, schedule information, past history, etc., and the planned actions to be performed at each step may include information on actions such as consultation, reservation, procedure, and aftercare performed or scheduled at a specific step.
[0264] Additionally, the step verification unit (730) can decompose patient context information and actions scheduled to be performed at each step into attribute units comparable to the condition expression of the rule in order to match the input verification target data with the rule structure stored in the rule base management unit (720). For example, the patient's age, presence of disease, type of procedure, schedule conditions, hospital selection information, etc., can be mapped to fields corresponding to the rule conditions to enable rule evaluation.
[0265] Additionally, the step verification unit (730) can search for rules applicable to the current verification target within the rule structure and determine whether the conditions of the corresponding rule are satisfied to derive the application result for each rule. In this process, multiple rules may be applied simultaneously, and each rule may be evaluated individually to generate a judgment result such as allow, restriction, or conditional allowance.
[0266] Additionally, the step verification unit (730) can perform verification by considering the hierarchical structure between rules and can determine the final judgment result by reflecting the relationship between the upper rule and the lower rule. For example, even if a lower rule allows a specific action, if the upper rule restricts it, the judgment result of the upper rule can be configured to be applied first.
[0267] Additionally, the step verification unit (730) can identify whether there is a conflict between the results of applying multiple rules, and if a conflict occurs, determine the final judgment result using the priority of the rules, the scope of application, or predefined conflict resolution criteria. For example, if some rules determine permission and other rules determine restriction for the same step execution, the system can be configured to reflect the judgment of the rule with higher priority or to apply the restriction condition first.
[0268] Additionally, the step verification unit (730) can generate verification results on a step-by-step basis for the actions scheduled to be performed in each step, and can generate verification result data including whether the action is possible, reasons for restriction, and whether additional conditions are required for each step.
[0269] Additionally, the step verification unit (730) can generate an extended verification result that includes not only whether the action to be performed in each step is simply possible, but also the degree of possibility of rule violation, risk level, whether additional verification is needed, and whether a recommendation is possible.
[0270] Additionally, the step verification unit (730) can record rule information used in the verification process, whether conditions are satisfied, and judgment results in a log format, and can be configured to track the basis for restricting the execution of a specific step or to utilize it for future rule improvement. For example, if the patient context information includes specific disease information and there is a step for which a specific procedure is scheduled for the patient, the step verification unit (730) can query the rule related to the procedure in the rule structure and determine whether the patient's disease condition corresponds to the restriction condition of the rule to verify whether the step can be executed.
[0271] The above action decision unit (740) can determine whether to allow or restrict the execution of each step by providing a control signal to the service execution management unit (200) based on the verification result of the step verification unit (730).
[0272] More specifically, the action decision unit (740) receives the step execution suitability verification result transmitted from the step verification unit (730) as input, and determines whether to allow execution for each step based on the verification result, and can generate a control signal to be transmitted to the service execution management unit (200). At this time, the control signal may be configured to include the permission to execute a specific step, restriction of execution, conditional execution, or a request for reconsideration.
[0273] Additionally, the action decision unit (740) may select or apply at least one of a plurality of operation modes by comprehensively considering the verification results provided by the step verification unit (730), the policy attributes of the rules managed by the rule base management unit (720), and the patient context information. Specifically, the action decision unit (740) may select at least one of a control mode that generates a control signal to allow or restrict the execution of each step, a recommendation mode that generates recommendation information including the possibility of violation related to step execution, recommended actions, or priority information and provides it to the medical service provider, and an autonomous execution mode that automatically performs schedule rescheduling, sending notifications, requesting additional information, adding, deleting, or changing the order of steps according to predefined rules or policies.
[0274] Additionally, the action decision unit (740) may select one or more of the control mode, recommendation mode, and autonomous execution mode based on at least one of the patient's condition information, type of medical service, risk level of each stage, degree of rule violation, or policy criteria. For example, if there is a high probability of rule violation, the control mode may be applied to restrict the execution of the stage; if there is a low probability of violation but additional verification is required, the recommendation mode may be applied; and if predefined correction is possible, the autonomous execution mode may be applied to perform schedule rescheduling or information requests.
[0275] Additionally, if the verification result derived from the step verification unit (730) includes multiple judgment elements, the action decision unit (740) can generate a final decision result by comprehensively considering the elements. For example, if some rules determine that execution is allowed and other rules determine that execution is restricted, the final decision on whether to allow or restrict can be made by reflecting the priority, scope of application, or predefined policy criteria of the rules.
[0276] Additionally, when the action decision unit (740) determines the execution restriction of each step, it may identify the condition or rule information that causes the restriction and include it in the control signal, thereby configuring the service execution management unit (200) to perform subsequent processing actions based on the reason for the restriction. For example, if a specific step is restricted due to a patient's condition, a request for additional information input or the execution of an alternative step to resolve the condition may be induced.
[0277] Additionally, the action decision unit (740) can generate control signals by considering the step-by-step execution status and the flow of the entire medical service provision procedure, and can adjust the activation status or order of subsequent steps by reflecting the impact of a specific step's execution restriction on subsequent steps. For example, if a specific procedure step is restricted, subsequent steps dependent on that procedure may also be automatically deactivated or replaced with other steps.
[0278] Additionally, the action decision unit (740) can be configured to transmit a control signal to the service execution management unit (200) and, at the same time, record the control result or provide it to an external component, and enable history management or monitoring of whether a step is performed. For example, if the performance of a reservation step is restricted based on the context information of a specific patient and the result of rule verification, a control signal restricting the performance of the step can be generated and transmitted to the service execution management unit (200), and the service execution management unit (200) can change the state of the step to a restricted state and perform additional measures.
[0279] Meanwhile, the artificial intelligence-based integrated management system for the entire lifecycle of medical services for foreign patients according to the present invention may be implemented in the form of a computing device comprising one or more processors and a memory that stores program instructions executed by said processors, and such a computing device may be operated in a single server, a distributed processing environment composed of multiple servers, or a cloud-based infrastructure environment.
[0280] The above processor may be configured to include at least one of a central processing unit (CPU), a graphics processing unit (GPU), and a neural network processing unit (NPU), and can perform functions related to processing foreign patient information, configuring medical service procedures, making AI-based decisions, processing interactions, and correcting procedures by executing program instructions stored in memory.
[0281] The above memory may include volatile and non-volatile memory and can store program code, artificial intelligence model parameters, patient context information, medical service provision procedure information, interaction data, and processing logs. Such memory may be implemented in the form of RAM, ROM, flash memory, SSD, or a database, and may be configured in a distributed storage structure as needed to reliably manage large amounts of data. In addition, it is connected to a foreign patient terminal and a medical service provision system through a network interface, and can transmit and receive data using internet-based communication, API calls, real-time bidirectional communication methods, etc.
[0282] The computing device described above can receive input data from a foreign patient through program execution and convert it into structured data, and can generate context information reflecting the patient's condition information and requirements based on the converted data. The generated context information can be utilized to configure medical service provision procedures and can be expanded into procedural data including the execution order, conditions, and timing of each step; furthermore, these procedures can be dynamically updated according to changes in the patient's condition or input information.
[0283] In addition, the computing device includes a computational environment capable of executing multiple artificial intelligence models, and can execute multiple AI agents with different judgment criteria in parallel or sequentially, thereby generating various decision-making outcomes for the same input. The generated results can be derived as final outcomes through a process of mutual comparison and analysis, and these results can be reflected in the medical service provision procedure to modify or supplement key elements such as hospital selection, scheduling, and procedure planning.
[0284] Furthermore, the computing device can process interaction data between foreign patients and medical service providers, and can generate interaction information by performing speech recognition and translation processing on voice data, and natural language processing and semantic analysis on text data. The generated interaction information can be transformed beyond simple conversation records into structured data that can be reflected in medical service procedures, and can be directly incorporated into the service flow in forms such as changes to step execution conditions, schedule adjustments, and the creation of additional tasks.
[0285] Furthermore, the computing device can query information about medical service providers in conjunction with an external data repository, determine a suitable hospital or medical institution based on the patient context, and support the patient's selection by providing the results to the user terminal. In this process, factors such as the availability of the hospital's procedure, pricing information, scheduled appointments, medical staff expertise, and language support can be comprehensively considered.
[0286] In addition, the computing device can modify or correct existing procedures in response to state changes or changes in external conditions occurring during the service provision process, and the modified procedures can be updated by reflecting them in the procedure data stored in memory. Such corrections can be performed by selectively modifying only certain steps rather than reconstructing the entire procedure, thereby maintaining the continuity of the procedure.
[0287] Furthermore, computing devices can be implemented by separating each function into independent modules or microservices, and can be configured to allow specific functions to be expanded or relocated as needed. Through this structure, the system can flexibly scale to respond to various service scales and user demands, and the entire medical service delivery process can be operated stably.
[0288] According to the present embodiment, the entire process from the influx of foreign patients to consultation, selection of medical service providers, reservation, treatment, and follow-up care is managed as a single integrated flow, thereby allowing information that was dispersed at each stage to be consistently connected, and consequently, the operational efficiency of the entire medical service provision process can be improved.
[0289] Furthermore, through a context-based processing structure that reflects the patient's condition information and requirements, necessary decision-making at each stage can be performed more precisely, enabling the provision of personalized medical services. Going beyond simple condition matching, this structure allows the service flow to be dynamically adjusted according to changes in the patient's situation, thereby enhancing the flexibility and responsiveness of the service.
[0290] In addition, by comprehensively utilizing multiple AI-based decision-making results, it is possible to derive results that reflect various judgment factors rather than relying on a single criterion, thereby improving the accuracy and reliability of key decisions such as hospital selection, scheduling, and procedure planning.
[0291] In addition, by resolving language differences based on voice and text data generated during the interaction process and converting and utilizing this into structured information, the accuracy of communication between foreign patients and medical service providers can be improved, and errors in information transmission that may occur during the provision of medical services can be reduced.
[0292] The above description is merely one embodiment for implementing an AI-based integrated management system for the entire lifecycle of medical services for foreign patients according to the present invention. The present invention is not limited to the above embodiment, and the technical spirit of the present invention extends to the scope in which various modifications can be made by anyone with ordinary knowledge in the field to which the invention belongs, without departing from the gist of the invention as claimed in the following patent claims. Explanation of the symbols
[0293] 1000: AI-based Integrated Management System for the Full Lifecycle of Medical Services for Foreign Patients 100: Context creation unit 110: Input conversion unit 120: Feature extraction unit 130: Context configuration unit 200: Service Execution Management Department 210: Procedure Division Department 220: Sequence Determination Unit 230: Execution Management Unit 240: Step standard setting section 250: Time point generation section 260: Journey Generation Unit 300: Multi-Agent Processing Unit 310: Agent Configuration Unit 320: Multi-Decision Execution Unit 330: Result Analysis Unit 340: Result Derivation Unit 350: Result Reflection Section 400: Interaction Support Section 410: Interaction receiver 420: Language processor 430: Interaction information generation unit 440: Channel control unit 450: Output processing unit 500: Procedure correction unit 510: Correction Target Identification Unit 520: Procedure Correction Execution Unit 600: Service Matching Unit 610: Query Generation Unit 620: Information retrieval unit 630: Matching result generation unit 700: Coordinator Agent Processing Unit 710: Rule Extraction Unit 720: Rule-Based Management Department 730: Action Decision Department 740: Phase Verification Unit 10: Foreign Patient Terminal 20: Central Processing Server 30: Multiple AI Agent Servers 40: Medical service provision server
Claims
Claim 1 A system implemented by a computing device having at least one processor and memory for storing programs controlling them, comprising: a context generation unit that generates patient context information based on foreign patient information; a service execution management unit that configures a medical service provision procedure based on the patient context information generated by the context generation unit and manages the linkage of medical services between a foreign patient and a medical service provider according to the configured medical service provision procedure; a multi-agent processing unit that obtains processing results by a plurality of artificial intelligence agents based on the patient context information generated by the context generation unit and medical service information managed by the service execution management unit, generates a decision result for a medical service using the obtained processing results, and reflects the generated decision result in a medical service; and a procedure correction unit that corrects a medical service provision procedure based on the decision result generated by the multi-agent processing unit, wherein the procedure correction unit includes a correction target identification unit that identifies a step or execution condition requiring correction among the medical service provision procedure managed by the service execution management unit using the decision result generated by the multi-agent processing unit. An AI-based integrated management system for the full lifecycle of medical services for foreign patients, characterized by including a procedure correction execution unit that modifies a medical service provision procedure by reflecting a decision result generated by the multi-agent processing unit regarding a step or execution condition identified by the correction target identification unit, and updates the modified medical service provision procedure by reflecting it in the medical service provision procedure information managed by the service execution management unit. Claim 2 The AI-based integrated management system for the entire lifecycle of medical services for foreign patients according to claim 1, wherein the context generation unit comprises: an input conversion unit that receives foreign patient information and converts it into a form usable for providing medical services; a feature extraction unit that identifies patient condition information and requirement information based on the foreign patient information converted by the input conversion unit and generates feature information for medical service matching using the identified information; and a context configuration unit that configures patient context information reflecting the patient condition information and requirement information using the feature information generated by the feature extraction unit. Claim 3 A system implemented by a computing device having at least one processor and memory for storing programs controlling them, comprising: a context generation unit that generates patient context information based on foreign patient information; a service execution management unit that configures a medical service provision procedure based on the patient context information generated by the context generation unit and manages the linkage of medical services between a foreign patient and a medical service provider according to the configured medical service provision procedure; and a multi-agent processing unit that obtains processing results by a plurality of artificial intelligence agents based on the patient context information generated by the context generation unit and medical service information managed by the service execution management unit, generates a decision result for a medical service using the obtained processing results, and reflects the generated decision result in a medical service, wherein the service execution management unit comprises: a procedure division unit that divides the medical service provision procedure into a plurality of stages and generates stage information based on the patient context information generated by the context generation unit; and a sequence determination unit that defines the execution order between stages by reflecting the preceding and succeeding conditions between the stage information generated by the procedure division unit. An artificial intelligence-based integrated management system for the full lifecycle of medical services for foreign patients, characterized by including an execution management unit that manages the progress status of the medical service provision procedure by updating the execution status and progress status of each step according to the execution order defined in the sequence determination unit. Claim 4 A system implemented as a computing device having at least one processor and memory for storing programs controlling them, comprising: a context generation unit that generates patient context information based on foreign patient information; and a service execution management unit that configures a medical service provision procedure based on the patient context information generated by the context generation unit, and manages the linkage of medical services between a foreign patient and a medical service provider according to the configured medical service provision procedure. The AI-based integrated management system for the entire lifecycle of medical services for foreign patients includes a multi-agent processing unit that obtains processing results by a plurality of AI agents based on patient context information generated by the context generation unit and medical service information managed by the service execution management unit, generates decision results for medical services using the obtained processing results, and reflects the generated decision results in medical services. The multi-agent processing unit comprises: an agent configuration unit that configures a plurality of AI agents having different judgment criteria using patient context information generated by the context generation unit and medical service information managed by the service execution management unit; a multi-decision execution unit that generates decision results for medical service provision procedures through a plurality of AI agents configured by the agent configuration unit; a result analysis unit that identifies decision items where differences occur between results by comparing a plurality of decision results generated by the multi-decision execution unit; a result derivation unit that generates a final decision result by selecting at least one of a plurality of decision results based on the decision item identified by the result analysis unit; and a result reflection unit that provides the final decision result generated by the result derivation unit to the service execution management unit and reflects it in the medical service provision procedure. Claim 5 A system implemented by a computing device having at least one processor and memory for storing programs controlling them, comprising: a context generation unit that generates patient context information based on foreign patient information; a service execution management unit that configures a medical service provision procedure based on the patient context information generated by the context generation unit and manages the linkage of medical services between a foreign patient and a medical service provider according to the configured medical service provision procedure; and a multi-agent processing unit that obtains processing results by a plurality of artificial intelligence agents based on the patient context information generated by the context generation unit and medical service information managed by the service execution management unit, generates a decision result for a medical service using the obtained processing results, and reflects the generated decision result in a medical service. The system includes an interaction support unit that supports interaction between a foreign patient and a medical service provider during the medical service provision process based on medical service information managed by the service execution management unit, wherein the interaction support unit comprises: an interaction receiving unit that receives at least one interaction data among voice data and text data generated or exchanged during the interaction process between the foreign patient and the medical service provider; a language processing unit that generates language information converted into different languages based on the interaction data received by the interaction receiving unit; an interaction information generating unit that generates interaction information that can be reflected in the medical service provision procedure using the language information generated by the language processing unit and the interaction data received by the interaction receiving unit; and a channel control unit that sets a communication channel corresponding to each of the foreign patient and the medical service provider based on the interaction information generated by the interaction information generating unit and the language information generated by the language processing unit, and controls the transmission of interaction information and language information through the set communication channel.An artificial intelligence-based integrated management system for the full lifecycle of medical services for foreign patients, characterized by including an output processing unit that provides at least one of interaction information and language information transmitted by the channel control unit to a foreign patient or a medical service provider. Claim 6 delete Claim 7 delete Claim 8 A system implemented by a computing device having at least one processor and memory for storing programs controlling them, comprising: a context generation unit that generates patient context information based on foreign patient information; a service execution management unit that configures a medical service provision procedure based on the patient context information generated by the context generation unit and manages the linkage of medical services between a foreign patient and a medical service provider according to the configured medical service provision procedure; and a multi-agent processing unit that obtains processing results by a plurality of artificial intelligence agents based on the patient context information generated by the context generation unit and medical service information managed by the service execution management unit, generates a decision result for a medical service using the obtained processing results, and reflects the generated decision result in a medical service. An AI-based integrated management system for the full lifecycle of medical services for foreign patients, comprising a coordinator agent processing unit that verifies the suitability of each step based on patient context information and medical service provision procedure information, and controls the execution of each step of the medical service provision procedure according to the verification result; wherein the coordinator agent processing unit comprises: a rule extraction unit that extracts rule information from unstructured documents; a rule base management unit that generates a rule structure based on the rule information extracted by the rule extraction unit and stores and manages the rule structure in a hierarchical structure; a step verification unit that receives as input the patient context information generated by the context generation unit and the planned actions for each step of the medical service provision procedure configured by the service execution management unit, and verifies the suitability of each step using the rule structure managed by the rule base management unit; and an action decision unit that determines whether to allow or restrict the execution of each step by providing a control signal to the service execution management unit based on the verification result of the step verification unit.
Citation Information
Patent Citations
Method for providing hospital referral service for foreign medical treatment and electronic device for performing the same
KR102716383B1
Operation server of medical support system supporting medical service for foreign based on platform and operation method of operation server
KR102903537B1
