KR102860323B1 - System and method for providing patient-specific health information based on artificial intelligence - Google Patents

Abstract

A system and method for providing patient-tailored health information based on an artificial intelligence chatbot are provided. According to embodiments of the present invention, by providing a plurality of patient-tailored predicted questions to a patient terminal based on the patient's personal information, medical information, and prescription information, the patient can be provided with predicted questions that are expected to be most curious about in the current situation. Furthermore, according to embodiments of the present invention, by dynamically changing the predicted questions according to the patient's symptoms and condition, the patient can be provided with predicted questions optimized for the patient's symptoms and condition. Furthermore, according to embodiments of the present invention, by dynamically changing the priority of each predicted question according to the patient's number of consultations and treatment stage, the patient can be provided with predicted questions optimized for the patient's number of consultations and treatment stage.

Description

System and Method for Providing Patient-Specific Health Information Based on Artificial Intelligence Chatbot

Embodiments of the present invention relate to a technology for providing patient-tailored health information based on an artificial intelligence chatbot.

Recently, the market for non-face-to-face medical services has expanded, leading to attempts at non-face-to-face treatment in various fields. Typically, in-person treatment involves a sequential process: patient visit to the hospital, registration and waiting, in-person consultation and prescription by medical staff, and then a visit to the pharmacy. However, non-face-to-face treatment involves a more streamlined process: patient reservation, remote consultation and prescription by medical staff, and delivery of medication by a delivery company.

Furthermore, with the advancement of AI technology, AI chatbots are emerging that automatically perform simple appointment and consultation tasks, providing non-face-to-face medical services. However, conventional AI chatbots only provide simple answers to pre-defined questions, limiting their ability to provide personalized health information tailored to individual patients.

Korean Patent Publication No. 10-2020-0049256 (May 8, 2020)

Embodiments of the present invention are intended to provide a high-quality artificial intelligence chatbot service for more efficient non-face-to-face treatment.

According to one embodiment, a data collection unit that collects a patient's personal information, treatment information, and prescription information generated during a non-face-to-face treatment process of the patient; an expected query provision unit that derives a plurality of expected queries and priorities for each expected query based on the collected personal information, treatment information, and prescription information of the patient and a plurality of previously learned first learning data, and provides each expected query to a patient terminal according to the derived priorities; And an information providing unit that receives an inquiry from the patient terminal and provides patient-tailored health information for the inquiry to the patient terminal based on the collected personal information, treatment information, and prescription information of the patient and previously learned second learning data, wherein the first learning data includes first query data, which is the inquiry of another patient collected from a conversation between the other patient and the medical staff during a non-face-to-face treatment process in the past, second query data, which is the inquiry input from the other patient terminal, and the number of appearances of the first query data and the second query data, and the second learning data is learning data of a large language model (LLM) that serves as the basis for generating the customized information, wherein an artificial intelligence chatbot-based patient-tailored health information providing system is provided.

The above-mentioned expected query providing unit calculates a degree of relevance between the collected personal information, medical information, and prescription information of the patient and the plurality of first learning data based on a first keyword included in the collected personal information, medical information, and prescription information of the patient and a second keyword included in the plurality of first learning data, derives a plurality of expected queries for which the calculated degree of relevance is greater than a reference value, and derives a priority for each of the expected queries based on the number of appearances of the first query data or the second query data corresponding to each of the derived plurality of expected queries.

The above-mentioned expected query provision unit determines the symptoms and patient condition of the patient corresponding to the patient terminal, and can assign different weights to the patient's personal information, treatment information, and prescription information when calculating the relevance for deriving the plurality of expected queries according to the symptoms and patient condition of the patient.

The above-mentioned expected query provision unit can determine the number of visits and the treatment stage of a patient corresponding to the patient terminal, and dynamically change the priority for each expected query according to the number of visits and the treatment stage of the patient.

The above-mentioned expected query provision unit can newly derive the plurality of expected queries and priorities for each expected query based on the query content input from the patient terminal.

According to another embodiment, a step of collecting, in a data collection unit, personal information, treatment information, and prescription information of a patient generated during a non-face-to-face treatment process of the patient; a step of deriving, in a predicted query provision unit, a plurality of predicted queries and priorities for each predicted query based on the collected personal information, treatment information, and prescription information of the patient and a plurality of previously learned first learning data; a step of providing, in the predicted query provision unit, each of the predicted queries to a patient terminal according to the derived priorities; a step of receiving, in an information provision unit, query content from the patient terminal; And, in the information providing unit, a step of providing patient-tailored health information for the query content to the patient terminal based on the collected personal information, treatment information, and prescription information of the patient and the previously learned second learning data is provided, wherein the first learning data includes first query data, which is the query content of another patient collected from a conversation between another patient and a medical staff during a non-face-to-face treatment process in the past, second query data, which is the query content input from another patient's terminal, and the number of appearances of the first query data and the second query data, and the second learning data is learning data of a large language model (LLM) that serves as the basis for generating the customized information. A method for providing patient-tailored health information based on an artificial intelligence chatbot is provided.

The step of deriving the plurality of expected queries and priorities for each expected query may include calculating the correlation between the collected personal information, medical information, and prescription information of the patient and the plurality of first learning data based on first keywords included in the collected personal information, medical information, and prescription information of the patient and second keywords included in the plurality of first learning data, deriving a plurality of expected queries for which the calculated correlation is greater than or equal to a reference value, and deriving priorities for each of the plurality of expected queries based on the number of appearances of the first query data or the second query data corresponding to each of the derived plurality of expected queries.

The step of deriving the plurality of expected queries and priorities for each expected query may determine the symptoms and patient condition of the patient corresponding to the patient terminal, and when calculating the relevance for deriving the plurality of expected queries according to the symptoms and patient condition of the patient, different weights may be given to the patient's personal information, treatment information, and prescription information.

The step of deriving the above multiple expected queries and priorities for each expected query may determine the number of treatments and the treatment stage of the patient corresponding to the patient terminal, and dynamically change the priority for each expected query according to the number of treatments and the treatment stage of the patient.

The method for providing patient-tailored health information based on the artificial intelligence chatbot may further include, after the step of receiving the query content from the patient terminal, a step of newly deriving the plurality of expected queries and priorities for each expected query based on the query content input from the patient terminal in the expected query providing unit.

According to embodiments of the present invention, by providing a plurality of patient-tailored expected questions to a patient terminal based on the patient's personal information, treatment information, and prescription information, it is possible to provide the patient with a patient-tailored expected question that the patient is most likely to be curious about in the current situation.

In addition, according to embodiments of the present invention, by dynamically changing the expected query according to the patient's symptoms and condition, it is possible to provide the patient with an expected query optimized for the patient's symptoms and condition.

In addition, according to embodiments of the present invention, by dynamically changing the priority of each expected query according to the number of patient visits and the stage of treatment, it is possible to provide the patient with an expected query optimized for the number of patient visits and the stage of treatment.

Figure 1 is a block diagram showing a detailed configuration of a patient-tailored health information provision system according to one embodiment of the present invention.
Figure 2 is a block diagram showing a detailed configuration of a server according to one embodiment of the present invention.
Figure 3 is an example of a screen of a patient terminal according to one embodiment of the present invention.
Figure 4 is an example of a screen of a patient terminal according to one embodiment of the present invention.
FIG. 5 is a diagram for explaining a process of deriving multiple expected queries and priorities for each expected query through an artificial intelligence model according to one embodiment of the present invention.
Figure 6 is a flowchart for explaining a method for providing patient-tailored health information according to one embodiment of the present invention.
FIG. 7 is a block diagram illustrating a computing environment including a computing device suitable for use in exemplary embodiments.

Hereinafter, specific embodiments of the present invention will be described with reference to the drawings. The following detailed description is provided to facilitate a comprehensive understanding of the methods, devices, and/or systems described herein. However, these are merely examples and the present invention is not limited thereto.

In describing embodiments of the present invention, if a detailed description of a known technology related to the present invention is judged to unnecessarily obscure the gist of the present invention, the detailed description will be omitted. In addition, the terms described below are terms defined in consideration of their functions in the present invention, and this may vary depending on the intention or custom of the user or operator. Therefore, the definitions should be made based on the contents throughout this specification. The terminology used in the detailed description is only for the purpose of describing embodiments of the present invention and should not be limited in any way. Unless clearly used otherwise, the singular form includes the plural form. In this description, expressions such as "comprises" or "having" are intended to indicate certain features, numbers, steps, operations, elements, parts or combinations thereof, and should not be construed to exclude the presence or possibility of one or more other features, numbers, steps, operations, elements, parts or combinations thereof other than those described.

FIG. 1 is a block diagram illustrating a detailed configuration of a patient-tailored health information provision system (100) according to one embodiment of the present invention. As illustrated in FIG. 1, the patient-tailored health information provision system (100) according to one embodiment of the present invention includes a hospital terminal (110), a patient terminal (120), and a server (130).

The hospital terminal (110) is a terminal carried by a hospital or medical staff within the hospital, and may be, for example, a desktop computer, laptop, tablet computer, or smartphone. In the present embodiments, the term "hospital" is used in a broad sense to include all medical institutions such as clinics, medical centers, oriental medicine clinics, dental clinics, and general hospitals. In addition, the term "medical staff" is used in a broad sense to include all personnel capable of prescribing patients, such as doctors, oriental medicine doctors, and dentists. The medical staff can treat patients remotely through the hospital terminal (110) and input prescription contents into the hospital terminal (110) based on the treatment results. The hospital terminal (110) can generate an electronic prescription based on the input of the medical staff and transmit the generated electronic prescription to the server (130). The electronic prescription may include the patient's personal information, the patient's treatment information, and prescription information. Here, the prescription information may include information regarding the type of medication to be administered to the patient, the dosage, the number of times the medication should be administered, the number of days of the medication, precautions, etc. The non-face-to-face treatment, electronic prescription creation and transmission process of the hospital terminal (110) described above can be performed through a non-face-to-face treatment application provided by the server (130).

The patient terminal (120) is a terminal possessed by the patient, the patient's guardian, the patient's assistant, or the patient's care facility, and may be, for example, a desktop, laptop, tablet computer, or smartphone. Here, the care facility is used in a broad sense to include all facilities established for the care, nursing, and assistance of patients, such as nursing hospitals, social welfare facilities, senior welfare facilities, and medical welfare facilities. The patient can receive non-face-to-face treatment by conversing with medical staff through the patient terminal (120). As described below, the patient terminal (120) receives a non-face-to-face treatment application from the server (130) and can be connected to the hospital terminal (110) through the non-face-to-face treatment application.

The server (130) is connected to the hospital terminal (110) and the patient terminal (120) through a network, and provides a non-face-to-face treatment application to the hospital terminal (110) and the patient terminal (120), respectively. Here, the network may include, for example, the Internet, a Wi-Fi network, a mobile network such as a 3G network, an LTE network, a wide area network, etc. The hospital terminal (110) and the patient terminal (120) can receive various services related to non-face-to-face treatment through the non-face-to-face treatment application. Accordingly, the hospital terminal (110) and the patient terminal (120) can access the server (130) through the non-face-to-face treatment application, and receive various services related to non-face-to-face treatment from the server (130). The non-face-to-face treatment application may be linked to an electronic medical record system (EMR) to provide various services related to non-face-to-face treatment.

Specifically, the server (130) can collect the patient's personal information, treatment information, and prescription information generated during the non-face-to-face treatment process of the patient, and provide an artificial intelligence chatbot service to the patient terminal (120) based on this. The server (130) can derive a plurality of expected queries and priorities for each expected query based on the collected patient's personal information, treatment information, and prescription information and a plurality of previously learned first learning data, and provide each expected query to the patient terminal (120) according to the derived priorities.

In addition, the server (130) can receive an inquiry from a patient terminal (120) and provide customized health information for the inquiry to the patient terminal (120) based on the collected patient's personal information, treatment information, and prescription information and the previously learned second learning data.

In the present embodiments, the first learning data may include first query data, which is the content of a query from another patient collected from a conversation between another patient and medical staff during a non-face-to-face treatment process of another patient in the past, second query data, which is the content of a query input from another patient's terminal, and the number of times the first query data and the second query data appear.

Additionally, the second learning data may be learning data (medical data) of large language models (LLMs) that serve as the basis for generating customized information.

Below, the detailed configuration of the server (130) will be examined in more detail with reference to FIGS. 2 to 5.

Fig. 2 is a block diagram showing a detailed configuration of a server (130) according to one embodiment of the present invention. As illustrated in Fig. 2, the server (130) according to one embodiment of the present invention includes a data collection unit (210), an expected query provision unit (220), an information provision unit (230), a first database (240), and a second database (250).

The data collection unit (210) collects the patient's personal information, treatment information, and prescription information generated during the non-face-to-face treatment process. In the present embodiments, the patient's personal information may include, for example, the patient's gender, age, height, weight, waist circumference, body mass index, eyesight, hearing, blood pressure, fasting blood sugar, family history or presence of genetic diseases, information on the patient's institution, information on hospitals the patient has visited, and health information about the patient's family. In addition, the treatment information may include the patient's symptoms, preliminary diagnosis data, treatment data entered by medical staff, and conversations between the patient and medical staff. In addition, the prescription information may include the type of medication to be administered to the patient, dosage, frequency of administration, number of days of administration, precautions, and information on the dispensing pharmacy.

The data collection unit (210) can be linked to an electronic medical record system to collect patient personal information, treatment information, and prescription information generated during non-face-to-face treatment. For example, the data collection unit (210) can collect patient personal information whenever the patient terminal (120) registers or logs in to the non-face-to-face treatment application, and can collect patient treatment information and prescription information whenever non-face-to-face treatment is performed via the patient terminal (120).

The expected query providing unit (220) derives a plurality of expected queries and priorities for each expected query in order to provide an artificial intelligence chatbot service, and provides each expected query to the patient terminal (120) according to the derived priorities. In the present embodiments, the expected query may be provided to the patient terminal (120) before receiving the query content from the patient terminal (120). Instead of providing a predetermined, fixed format of expected query to the patient terminal (120), the expected query providing unit (220) may dynamically change a customized expected query for each patient and provide it to the patient terminal (120). In addition, the expected query providing unit (220) may not only provide different expected queries for each patient, but may also dynamically change the priorities of each expected query and provide them to the patient terminal (120). For example, an expected query with a high priority may be displayed in the upper area of the patient terminal (120), displayed larger than an expected query with a lower priority, or displayed in a different color.

To this end, the expected query provision unit (220) can derive a plurality of expected queries and priorities for each expected query based on the patient's personal information, treatment information, and prescription information collected by the data collection unit (210) and a plurality of previously learned first learning data. As described above, the first learning data may include the first query data, which is the content of a query from another patient collected from a conversation between another patient and a medical staff during a non-face-to-face treatment process in the past, the second query data, which is the content of a query input from another patient's terminal, and the number of times the first query data and the second query data appear.

First, the expected query providing unit (220) calculates the degree of correlation between the patient's personal information, treatment information, and prescription information collected by the data collecting unit (210) and the plurality of first learning data based on the first keyword included in the patient's personal information, treatment information, and prescription information collected by the data collecting unit (210) and the second keyword included in the plurality of first learning data, and can derive a plurality of expected queries whose calculated degree of correlation is greater than or equal to a reference value.

As an example, the expected query providing unit (220) extracts first keywords, such as male/teenager/atopic disease/antihistamine drug A/moisturizer B, from the patient's personal information, medical information, and prescription information, and compares the first keywords with second keywords included in a plurality of first learning data stored in the first database (240). The expected query providing unit (220) can calculate the degree of correlation between the patient's personal information, medical information, and prescription information collected by the data collecting unit (210) and the plurality of first learning data by analyzing the correlation between the first keyword and the second keyword. For example, the expected query providing unit (220) can determine that the higher the correlation between the first keyword and the second keyword, the higher the degree of correlation between the patient's personal information, medical information, and prescription information collected by the data collecting unit (210) and the first learning data including the second keyword. In the above example, the expected query providing unit (220) can select a plurality of query data that correspond to at least some of male/teenager/atopic disease/antihistamine drug A/moisturizer B above a standard value among a plurality of first query data and a plurality of second query data, and derive them as expected queries. In this case, query data that patients in their teens suffering from atopic disease and who have been prescribed antihistamine drug A and moisturizer B are mainly curious about can be automatically derived as expected queries.

As another example, if a first keyword such as female/60s/glaucoma diagnosis/glaucoma symptoms/intraocular pressure medication C is extracted from the patient's personal information, medical information, and prescription information, the expected query providing unit (220) can select multiple query data that correspond to at least some of female/60s/glaucoma diagnosis/glaucoma symptoms/intraocular pressure medication C by a standard value or more and derive them as expected queries. In this case, query data that patients in their 60s with glaucoma symptoms and who have been prescribed intraocular pressure medication C are mainly curious about can be automatically derived as expected queries.

In this way, instead of providing a predetermined, fixed form of expected query to the patient terminal (120), the expected query provision unit (220) can dynamically provide a customized expected query for each patient.

Next, the expected query providing unit (220) can derive priorities for each expected query based on the number of appearances of the first query data or the second query data corresponding to each of the derived plurality of expected queries. As described above, the first learning data can include the number of appearances of the first query data and the second query data. Here, the number of appearances is an indicator of how many times (frequency) the first query data and the second query data have appeared in the past, which can be used to determine which information patients are curious about and how much. Accordingly, the expected query providing unit (220) can obtain the number of appearances of the first query data or the second query data corresponding to each of the derived plurality of expected queries, and can set a higher priority for the expected query with a higher number of appearances. As an example, if there are five derived expected queries, the priorities of the expected queries 1 to 5 can be automatically set in order of the number of appearances among the derived expected queries 1 to 5.

In this way, the expected query provision unit (220) can not only provide different expected queries for each patient, but can also dynamically change the priority of each expected query.

In addition, the expected query providing unit (220) determines the symptoms and patient condition of the patient corresponding to the patient terminal (120) by using at least one of the patient's personal information, medical information, and prescription information collected by the data collecting unit (210), and can assign different weights to the patient's personal information, medical information, and prescription information when calculating the relevance to derive a plurality of expected queries according to the patient's symptoms and patient condition. As an example, the expected query providing unit (220) can determine the symptoms and patient condition of the patient corresponding to the patient terminal (120) as "severe atopic dermatitis itching", "severe allergic urticaria due to nut consumption", "mid-stage glaucoma with decreased vision and congestion", etc. by using at least one of the patient's personal information, medical information, and prescription information collected by the data collecting unit (210). The expected query providing unit (220) can determine the symptoms and patient condition of the patient corresponding to the patient terminal (120) based on the first keywords included in the patient's personal information, treatment information, and prescription information and the keywords in the second learning data stored in the second database (250). As described above, the second learning data can be collected and learned from various disease/illness/medical-related reports, papers, journals, websites, etc. as learning data of large language models (LLMs). Such large language models can be combined with a Retrieval-Augmented Generation (RAG) model related to medical data.

Thereafter, the expected query provision unit (220) defines the symptoms and patient status of the patient corresponding to the patient terminal (120) as “disease name or symptom name” and “stage of disease or symptom (mild/severe, early/middle/late, etc.)”, respectively, and when calculating the relevance to derive multiple expected queries according to the defined “disease name or symptom name” and “stage of disease or symptom (mild/severe, early/middle/late, etc.)”, different weights (e.g., 0.1, 0.3, 0.6) can be given to the patient’s personal information, treatment information, and prescription information, respectively.

As an example, if the patient's symptom and patient condition are "severe allergic urticaria due to nut ingestion," the expected query providing unit (220) may define the "disease name or symptom name" as "nut allergic urticaria" and the "stage of disease or symptom" as "severe." In this case, the expected query providing unit (220) may predict that the patient will be more curious about prescription information, such as which medicine to take and how to take it, and precautions for prescribed medication, than about medical information, such as the next appointment schedule and what to bring to the next appointment, for "nut allergic urticaria" + "severe." Therefore, when calculating the relevance, the expected query providing unit (220) may give more weight to prescription information than to personal information and medical information. That is, the expected query providing unit (220) may predict that the need for query data for prescription information will increase in the case of severe allergic urticaria due to nut ingestion, and thus may give a higher weight to prescription information, thereby deriving multiple expected queries. In this case, query data related to prescription for severe allergic urticaria caused by nut consumption can be derived as expected queries.

Each of the weights assigned to the patient's personal information, medical information, and prescription information may be dynamically changed according to the above-described "disease name or symptom name" and "stage of disease or symptom." At this time, the expected query providing unit (220) selects keywords related to the "disease name or symptom name" and "stage of disease or symptom" included in the second learning data, and analyzes sentences containing the selected keywords using a natural language processing (NLP) technique to select important elements among multiple elements (e.g., gender, age, eyesight, hearing, blood pressure, type of medication, dosage, etc.) in the patient's personal information, medical information, and prescription information. Thereafter, the expected query providing unit (220) may assign more weight to the patient's personal information, medical information, or prescription information related to the selected important elements when calculating the relevance.

In addition, the expected query providing unit (220) may determine the number of consultations and the treatment stage of the patient corresponding to the patient terminal (120) by using at least one of the patient's personal information, treatment information, and prescription information collected by the data collecting unit (210), and may dynamically change the priority for each expected query according to the number of consultations and the treatment stage of the patient. As an example, the expected query providing unit (220) may determine the number of consultations and the treatment stage of the patient corresponding to the patient terminal (120) as "1 time" and "first visit", respectively, by using at least one of the patient's personal information, treatment information, and prescription information collected by the data collecting unit (210).

Thereafter, the expected question providing unit (220) can dynamically change the priority of each expected question according to the number of patient consultations and the treatment stage. As an example, if the number of patient consultations and the treatment stage are “1 time” and “first visit,” the expected question providing unit (220) can give a higher priority to the expected question about the next appointment than to the expected question about the precautions for prescription drugs. As another example, if the number of patient consultations and the treatment stage are “2 times” and “re-visit,” the expected question providing unit (220) can give a higher priority to the expected question about the progress of the patient’s health than to the expected question about the next appointment. Since the patient’s curiosity may vary depending on the number of patient consultations and the treatment stage, the expected question providing unit (220) can dynamically change the priority of each expected question according to the number of patient consultations and the treatment stage, thereby providing the patient terminal (120) with the expected question that the patient is expected to be most curious about. At this time, the priority can be automatically set based on the first learning data. That is, whenever the first query data and the second query data included in the first learning data are generated, the number of patient consultations and the treatment stage corresponding to the first query data and the second query data can be learned together, and the priority can be automatically set based on the number of patient consultations and the treatment stage corresponding to the first learning data. For example, the priority of query data frequently generated in the first-examination stage and query data frequently generated in the re-examination stage can be dynamically changed according to the actual number of patient consultations and the treatment stage.

According to embodiments of the present invention, by providing a plurality of patient-tailored predicted questions to the patient terminal (120) based on the patient's personal information, treatment information, and prescription information, it is possible to provide the patient with patient-tailored predicted questions that the patient is most likely to be curious about in the current situation. In addition, according to embodiments of the present invention, it is possible to provide the patient with predicted questions optimized for the patient's symptoms and condition by dynamically changing the predicted questions according to the patient's symptoms and condition. In addition, according to embodiments of the present invention, it is possible to provide the patient with predicted questions optimized for the patient's number of consultations and treatment stages by dynamically changing the priority of each predicted question according to the patient's number of consultations and treatment stages.

The information providing unit (230) receives a query from the patient terminal (120), and provides patient-tailored health information for the query to the patient terminal (120) based on the patient's personal information, treatment information, and prescription information collected by the data collecting unit (210) and the previously learned second learning data. The information providing unit (230) analyzes the query through natural language processing to derive main keywords, and selects second learning data corresponding to the main keywords to generate patient-tailored health information for the query. At this time, the information providing unit (230) may consider the patient's personal information, treatment information, and prescription information in the process of generating the patient-tailored health information. That is, even for the same query, if at least some of the patient's personal information, treatment information, and prescription information change, the patient-tailored health information provided to the patient terminal (120) may also dynamically change. The information provider (230) can determine second learning data that serves as the basis for patient-tailored health information by first selecting second learning data related to the query content and secondly selecting second learning data that corresponds to the patient's personal information, treatment information, and prescription information from the first-selected second learning data. Thereafter, the information provider (230) can generate patient-tailored health information based on the second learning data determined in this manner.

As an example, if the query is “Can I take prescription drugs and red ginseng together?”, the information provider (230) may first select second learning data related to the patient’s prescription drugs and red ginseng, and then secondly select second learning data corresponding to the patient’s personal information, medical information, and prescription information from the first-selected second learning data. When taking prescription drugs and red ginseng together, side effects may occur, but may vary depending on each patient’s symptoms, past disease history, etc. Therefore, the information provider (230) may provide optimal patient-tailored health information suitable for the patient to the patient terminal (120) by taking into consideration the patient’s personal information, medical information, and prescription information.

In addition, the above-described expected query providing unit (220) can newly derive the plurality of expected queries and the priorities for each expected query based on the query content input from the patient terminal (120). To this end, the expected query providing unit (220) can calculate the correlation between the query content input from the patient terminal (120) and each expected query derived from the expected query providing unit (220). The expected query providing unit (220) can calculate the correlation between the query content input from the patient terminal (120) and each expected query derived from the expected query providing unit (220), for example, through correlation analysis.

If there is an expected query having a relevance lower than a threshold value with respect to the query content input from the patient terminal (120), the expected query providing unit (220) can dynamically change the expected query. The expected query providing unit (220) can preferentially select an expected query related to the query content input from the patient terminal (120) in the process of deriving a plurality of expected queries through the above-described method. That is, the expected query providing unit (220) can select an expected query based on the relevance described above, and can derive an expected query related to the query content input from the patient terminal (120) by assigning a set weight to the expected query related to the query content input from the patient terminal (120). This real-time dynamic change and derivation of the expected query can be repeatedly performed whenever a new query content is input from the patient terminal (120). When multiple different query contents are sequentially input by the patient terminal (120), patient-tailored health information matching the query contents may be provided to the patient terminal (120) by the information provision unit (230), and at the same time, a new expected query related to the query contents may be provided to the patient terminal (120) by the expected query provision unit (220). Accordingly, the patient may receive an answer to the query contents he or she is curious about, and may click on an additional expected query related to the query contents to receive related information, thereby receiving a high-quality artificial intelligence chatbot service.

The first database (240) is a repository where the first learning data is stored. As described above, the first learning data may include first query data, which is a query from another patient collected from conversations between another patient and medical staff during a non-face-to-face treatment process in the past, second query data, which is a query input from another patient's terminal, and the number of times the first query data and the second query data appear. In addition, not only the first learning data but also metadata related to the first learning data may be stored in the first database (240). Here, the metadata may include personal information, treatment information, prescription information, etc. of another patient corresponding to the first learning data.

The second database (250) is a repository where second learning data is stored. As described above, the second learning data may be training data for a large-scale language model (LLM), which serves as the basis for generating customized information. The second learning data may be collected and trained from, for example, various disease/illness/medical-related reports, papers, journals, websites, etc.

Figures 3 and 4 are examples of screens of a patient terminal (120) according to one embodiment of the present invention. Here, A represents an expected query provided by the expected query provision unit (220), B represents the content of a query input from a patient, and C represents patient-tailored health information provided by the information provision unit (230).

Referring to FIG. 3, the expected query provision unit (220) derives a plurality of expected queries (A) and priorities for each expected query (A) based on the collected patient's personal information, treatment information, and prescription information and a plurality of previously learned first learning data, and can provide each expected query (A) to the patient terminal (120) according to the derived priorities. At this time, an expected query with a high priority may be displayed, for example, in the upper area of the patient terminal (120), displayed larger than an expected query with a lower priority, or displayed in a different color.

In this state, the patient terminal (120) can receive a query (B) from the patient. In this case, the information provision unit (230) can provide patient-tailored health information (C) for the query to the patient terminal (120) based on the collected patient's personal information, treatment information, and prescription information, as well as the previously learned second learning data.

In addition, referring to FIG. 4, the patient terminal (120) can receive a new query (B) from the patient. In this case, the information provision unit (230) can provide patient-tailored health information (C) for the query to the patient terminal (120) based on the collected patient's personal information, treatment information, and prescription information and the previously learned second learning data. In addition, the expected query provision unit (220) can newly derive a plurality of expected queries (A) and a priority for each expected query (A) based on the query input from the patient terminal (120).

FIG. 5 is a diagram for explaining a process of deriving multiple expected queries and priorities for each expected query using an artificial intelligence model according to one embodiment of the present invention.

Referring to FIG. 5, the artificial intelligence model can be configured to input a patient's personal information, treatment information, and prescription information and output multiple expected queries and priorities for each expected query.

Specifically, the artificial intelligence model may be configured to calculate the degree of relevance between the collected patient's personal information, medical information, and prescription information and the plurality of first learning data based on the first keyword included in the collected patient's personal information, medical information, and prescription information and the second keyword included in the plurality of first learning data, derive a plurality of expected queries whose calculated degree of relevance is greater than a reference value, and derive a priority for each expected query based on the number of appearances of the first query data or the second query data corresponding to each of the derived plurality of expected queries.

At this time, the artificial intelligence model may be configured to determine the symptoms and patient condition of the patient corresponding to the patient terminal (120), and to assign different weights to the patient's personal information, treatment information, and prescription information when calculating the relevance to derive multiple expected queries based on the patient's symptoms and patient condition. In addition, the artificial intelligence model may be configured to determine the number of consultations and treatment stages of the patient corresponding to the patient terminal (120), and to dynamically change the priority for each expected query based on the number of consultations and treatment stages of the patient.

Figure 6 is a flowchart illustrating a method for providing patient-tailored health information according to one embodiment of the present invention. While the illustrated flowchart divides the method into multiple steps, at least some of the steps may be performed in a different order, combined with other steps and performed together, omitted, divided into substeps, or performed with one or more additional steps not illustrated.

In step S102, the data collection unit (210) collects the patient's personal information, treatment information, and prescription information generated during the non-face-to-face treatment process.

In step S104, the expected query provision unit (220) derives a plurality of expected queries and priorities for each expected query based on the collected patient's personal information, treatment information, and prescription information and a plurality of previously learned first learning data.

In step S106, the expected query provision unit (220) provides each expected query to the patient terminal (120) according to the derived priority.

At step S108, the information provider (230) receives inquiry content from the patient terminal (120).

At step S110, the information provision unit (230) provides patient-tailored health information for the query to the patient terminal (120) based on the collected patient's personal information, treatment information, and prescription information and the previously learned second learning data.

Figure 7 is a block diagram illustrating a computing environment including a computing device suitable for use in exemplary embodiments. In the illustrated embodiment, each component may have different functions and capabilities other than those described below, and may include additional components other than those described below.

The illustrated computing environment (10) includes a computing device (12). In one embodiment, the computing device (12) may be a patient-tailored health information provision system (100) or one or more components included in the patient-tailored health information provision system (100).

A computing device (12) includes at least one processor (14), a computer-readable storage medium (16), and a communication bus (18). The processor (14) may cause the computing device (12) to operate according to the exemplary embodiments mentioned above. For example, the processor (14) may execute one or more programs stored in the computer-readable storage medium (16). The one or more programs may include one or more computer-executable instructions, which, when executed by the processor (14), may be configured to cause the computing device (12) to perform operations according to the exemplary embodiments.

A computer-readable storage medium (16) is configured to store computer-executable instructions or program code, program data, and/or other suitable forms of information. A program (20) stored in the computer-readable storage medium (16) includes a set of instructions executable by the processor (14). In one embodiment, the computer-readable storage medium (16) may be a memory (volatile memory such as random access memory, non-volatile memory, or a suitable combination thereof), one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, any other form of storage medium that can be accessed by the computing device (12) and store desired information, or a suitable combination thereof.

A communication bus (18) interconnects various other components of the computing device (12), including the processor (14) and computer-readable storage media (16).

The computing device (12) may also include one or more input/output interfaces (22) that provide interfaces for one or more input/output devices (24) and one or more network communication interfaces (26). The input/output interfaces (22) and the network communication interfaces (26) are connected to the communication bus (18). The input/output devices (24) may be connected to other components of the computing device (12) via the input/output interfaces (22). Exemplary input/output devices (24) may include input devices such as pointing devices (such as a mouse or a trackpad), a keyboard, a touch input device (such as a touchpad or a touchscreen), a voice or sound input device, various types of sensor devices and/or photographing devices, and/or output devices such as display devices, printers, speakers and/or network cards. The exemplary input/output devices (24) may be included within the computing device (12) as a component constituting the computing device (12), or may be connected to the computing device (12) as a separate device distinct from the computing device (12).

While the present invention has been described in detail above through representative examples, those skilled in the art will understand that various modifications to the above-described embodiments are possible without departing from the scope of the present invention. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined not only by the claims set forth below but also by equivalents thereof.

100: Patient-tailored health information provision system
110: Hospital terminal
120: Patient terminal
130: Server
210: Data Collection Department
220: Expected Question Provider
230: Information Department
240: First database
250: Second database

Claims (10)

A data collection unit that collects the patient's personal information, treatment information, and prescription information generated during the non-face-to-face treatment process;
A predicted query providing unit that derives a plurality of predicted queries and priorities for each predicted query based on the collected personal information, treatment information, and prescription information of the patient and a plurality of previously learned first learning data, and provides each predicted query to the patient terminal according to the derived priorities; and
An information providing unit that receives an inquiry from the patient terminal and provides patient-tailored health information for the inquiry to the patient terminal based on the collected personal information, treatment information, and prescription information of the patient and previously learned second learning data.
The above first learning data includes first query data, which is the content of a query from another patient collected from a conversation between another patient and medical staff during a non-face-to-face treatment process in the past, second query data, which is the content of a query input from another patient's terminal, and the number of times the first query data and the second query data appear.
The above second learning data is learning data of large language models (LLMs) that serve as the basis for generating the patient-tailored health information.
The above-mentioned expected query providing unit calculates a degree of correlation between the collected personal information, treatment information and prescription information of the patient and the plurality of first learning data based on a first keyword included in the collected personal information, treatment information and prescription information of the patient and a second keyword included in the plurality of first learning data, and dynamically assigns different weights to the personal information, treatment information and prescription information of the patient when calculating the degree of correlation according to the symptom name and the stage of the symptom after determining the symptom name and the stage of the symptom of the patient corresponding to the patient terminal, derives a plurality of expected queries for which the calculated degree of correlation is greater than a reference value, and derives a priority for each of the expected queries based on the number of appearances of the first query data or the second query data corresponding to each of the plurality of derived expected queries, and dynamically changes the priority for each of the expected queries based on the number of times of treatment and the stage of treatment of the patient corresponding to the patient terminal after determining the number of times of treatment and the stage of treatment of the patient.
A patient-tailored health information provision system based on an artificial intelligence chatbot, wherein at least one of the position, size, and color of each expected query displayed on the patient terminal changes according to the above priority, and the priority is automatically set based on the number of consultations and the treatment stage of the patient corresponding to the first query data and the second query data whenever the first query data and the second query data are generated.
delete delete delete In claim 1,
The above-mentioned expected query provision unit is an artificial intelligence chatbot-based patient-tailored health information provision system that newly derives a plurality of expected queries and priorities for each expected query based on the query content entered from the patient terminal.
In the data collection department, a step of collecting the patient's personal information, treatment information, and prescription information generated during the non-face-to-face treatment process;
In the expected query provision unit, a step of deriving a plurality of expected queries and priorities for each expected query based on the collected personal information, treatment information, and prescription information of the patient and a plurality of previously learned first learning data;
In the above expected query provision unit, a step of providing each expected query to the patient terminal according to the derived priority;
In the information provision department, a step of receiving inquiry content from the patient terminal; and
In the above information provision unit, a step is included in which patient-tailored health information for the above inquiry is provided to the patient terminal based on the collected personal information, treatment information, and prescription information of the patient and the previously learned second learning data.
The above first learning data includes first query data, which is the content of a query from another patient collected from a conversation between another patient and medical staff during a non-face-to-face treatment process in the past, second query data, which is the content of a query input from another patient's terminal, and the number of times the first query data and the second query data appear.
The above second learning data is learning data of large language models (LLMs) that serve as the basis for generating the patient-tailored health information.
The step of deriving the plurality of predicted queries and priorities for each predicted query is: calculating the correlation between the collected personal information, clinical information, and prescription information of the patient and the plurality of first learning data based on the first keyword included in the collected personal information, clinical information, and prescription information of the patient and the second keyword included in the plurality of first learning data, and dynamically assigning different weights to the patient's personal information, clinical information, and prescription information when calculating the correlation according to the symptom name and the symptom stage after determining the symptom name and the symptom stage of the patient corresponding to the patient terminal; deriving a plurality of predicted queries for which the calculated correlation is greater than a reference value; and deriving the priorities for each predicted query based on the number of appearances of the first query data or the second query data corresponding to each of the plurality of derived predicted queries, and dynamically changing the priorities for each predicted query based on the number of times the first query data or the second query data appears, after determining the number of times the patient was treated and the stage of the treatment of the patient corresponding to the patient terminal.
A method for providing customized health information to a patient based on an artificial intelligence chatbot, wherein at least one of the position, size, and color of each expected query displayed on the patient terminal changes according to the above priority, and the priority is automatically set based on the number of consultations and the treatment stage of the patient corresponding to the first query data and the second query data whenever the first query data and the second query data are generated.
delete delete delete In claim 6,
After the step of receiving inquiry content from the above patient terminal,
A method for providing personalized health information to a patient based on an artificial intelligence chatbot, further comprising a step of newly deriving a plurality of expected queries and a priority for each expected query based on the query content input from the patient terminal in the above expected query provision unit.