Internet hospital expert registration adaptation system based on artificial intelligence driving

Abstract

The invention relates to the technical field of medical consultation, in particular to an expert registration adaptation system of an internet hospital based on artificial intelligence driving, and the system comprises an artificial intelligence guiding module which automatically obtains the medical record of a patient, and deduces the disease of the patient and a corresponding department; the dynamic number source matching module is based on the real-time number source state, the disease of the patient and the corresponding department, when the system is used, the problem that the patient is difficult to accurately judge which department outpatient service should be applied to the disease due to lack of medical knowledge is effectively solved, accurate department recommendation is provided for the patient through multi-source data fusion analysis, and the patient experience is improved. The accuracy of first registration of a patient is greatly improved, time waste and medical resource mismatching caused by wrong department registration are reduced, meanwhile, through dynamic number source monitoring and an intelligent candidate mechanism and an automatic recommendation replacement expert function, the problem that repeated operation is needed when the patient encounters insufficient number sources when inquiring expert outpatient service information is solved, and the patient experience is improved. The registration time of the patient is greatly saved, and the registration efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of medical consultation technology, and in particular to an expert registration adaptation system for an Internet hospital driven by artificial intelligence. Background Art

[0002] Artificial Intelligence (AI) is an interdisciplinary field that aims to enable computer systems to possess human-like intelligent behaviors. It covers multiple fields such as computer science, mathematics, psychology, and linguistics. The expert registration adaptation of Internet hospitals refers to the use of technical architecture, process optimization, and user experience design to enable expert registration services to operate efficiently on the Internet platform and form a seamless connection with the offline medical system and user needs.

[0003] Explanation of terms: Hospital Information Management System (HIS); Laboratory Information Management System (LIS); Medical imaging system (PACS / RIS).

[0004] The patent publication number is CN108200176A, which states in its specification: "A smart hospital registration data processing method based on mobile Internet, involving a hospital registration system, the method comprising the following steps: Step 1, the patient purchases a medical card and a patient mobile terminal on the hospital platform, then inserts the medical card into the patient mobile terminal, uses the patient mobile terminal to read the information on the medical card, and enters the patient information into the medical card through the patient mobile terminal. By preserving personal information through the medical card, the security of personal information can be ensured, and the problem of personal information leakage can be effectively avoided; the cooperation between this method and the device can be used not only for ordinary numbers, but also for expert numbers, so that The nutritional range is wider; both the patient's mobile terminal and the doctor's mobile terminal are relatively small and easy to carry. Although the above technology achieves information security by physically binding the dedicated terminal to the diagnosis and treatment card, and combines the mobile Internet to synchronize doctor-patient data in real time, thus solving the information asymmetry problem of traditional registration, patients need to clarify which department of the outpatient clinic they need to see for their own illness when seeking medical treatment, and the existing technology does not provide specific guidance. At the same time, when registering with a specialist, the existing technology requires first querying the specialist clinic information. However, some specialist clinics may not have available numbers, causing patients to exit the expert interface and then query other specialist clinic information, resulting in repeated operations and not conducive to saving patient registration time.

[0005] To sum up, the development of an expert registration adaptation system for Internet hospitals driven by artificial intelligence is still a key issue that needs to be urgently addressed in the field of medical consulting technology. Summary of the Invention

[0006] The purpose of the present invention is to solve the problem in the prior art that patients need to clarify which department of the clinic they need to go to for their illness when seeking medical treatment, but the prior art does not provide specific guidance. At the same time, when registering with a specialist, the prior art needs to first query the specialist clinic information, but some specialist clinics may have no number sources, resulting in the patient needing to exit the specialist interface and then query other specialist clinic information, resulting in repeated operations, which is not conducive to saving patient registration time.

[0007] To achieve the above objectives, the present invention provides an artificial intelligence-driven expert registration adaptation system for Internet hospitals, comprising: The artificial intelligence guidance module automatically obtains the patient's medical records and infers the patient's disease and the corresponding department; The dynamic number source matching module executes the expert matching-candidate-feedback closed loop based on the real-time number source status, the patient's disease, and the corresponding department: (a) Based on the real-time appointment status, the target disease, and the target department, selecting matching experts from available experts and generating a recommendation list; (b) When the number of specialists selected by the patient is full, the dual-path operation is automatically performed: (b1) Add the patient to the specialist’s waiting list and monitor the availability of appointments in real time to allocate appointments in order; (b2) Selecting alternative experts who meet the similarity criteria from experts in the same target department and recommending them to the patient; (c) Dynamically optimize matching rules based on the patient's final consultation result feedback.

[0008] Furthermore, the operation process of the artificial intelligence guidance module includes: The artificial intelligence guidance module includes a multi-source data acquisition unit, a semantic parsing model and a knowledge graph association unit. The multi-source data acquisition unit is provided with a front-end data acquisition interface adapted to a variety of Internet hospital platforms, including but not limited to web pages, mobile apps, and WeChat mini-program interfaces, and integrates a voice input plug-in in the patient symptom description area. The voice input plug-in is used to convert the patient's voice input into text in real time. The expression:

[0009] Where, A function representing speech-to-text conversion, Mathematical representation space for speech signals is the set of real numbers It is the dimension, Represents text space, Indicates time The collected speech signal feature vector, Expressing the moment The speech feature vector Perform speech-to-text conversion operations, Is the mathematical symbol for the independent variable corresponding to the maximum value, Represents a text candidate set The text candidates, Indicates that in a given speech feature vector Time text candidate The conditional probability of occurrence, The parameters representing the speech recognition model are then securely connected to the hospital's HIS system, LIS system, and PACS system through a standardized data interface protocol to automatically obtain patient medical records, which include but are not limited to historical medical record data and past registration behavior data, and the patient medical records are encrypted and stored in a data warehouse.

[0010] Furthermore, the operation process of the artificial intelligence guidance module includes: The semantic parsing model is pre-trained using medical text data including but not limited to medical textbooks, medical records, and medical papers. The expression:

[0011] Where, Represents the model parameters Find the minimum value, Is the expectation symbol, Represents the dataset Sample pairs are obtained by sampling , Understood as input medical text, Indicates the label corresponding to the input medical text, Represents the medical field text dataset used for pre-training, which is the weight coefficient used to control different loss terms The proportion of total losses in is the masked language model loss function, is the named entity recognition loss function, It is a sentence classification related loss function. The pre-training of the semantic parsing model is carried out by minimizing the multi-task loss function training. Multiple automatically acquired patient medical records are input into the semantic parsing model, and the semantic parsing model outputs patient symptom information.

[0012] Furthermore, the operation process of the artificial intelligence guidance module includes: The knowledge graph association unit uses the knowledge graph construction tool to construct the medical knowledge system including but not limited to diseases, symptoms, and departments into a medical knowledge graph.

[0013]

[0014] Where, It is a medical knowledge graph. is a collection of entities, It is a relationship set, which is stored and managed through a graph database. It realizes search and reasoning in the medical knowledge graph based on the patient's symptom information, and uses the probabilistic graph reasoning method to establish a probability propagation model from symptoms to diseases. Similarly, through the disease-department relationship edge in the medical knowledge graph, the corresponding department is identified, and the patient's disease and the corresponding department are output. The expression is:

[0015] Where, Indicates that when the symptom set is observed Temporary Disease The posterior probability of occurrence, Indicates the diseases, represents the patient's symptom set, represents the normalization factor, Indicates a collection of symptoms The number of symptoms included in Represents a collection of symptoms The Symptoms, Indicates disease Symptoms when they occur The conditional probability of occurrence, Indicates disease The prior probability in the population, From the disease set The disease with the largest posterior probability is selected from Is the mathematical symbol for the independent variable corresponding to the maximum value, Represents a collection of diseases, Indicates the department corresponding to the inferred patient disease, Represented in the medical knowledge graph The input of the function for traversal reasoning is from the disease set The disease with the highest posterior probability selected .

[0016] Furthermore, the operation process of the dynamic number source matching module includes: The dynamic number source matching module includes a number source monitoring unit, an artificial intelligence matching unit, a candidate and recommendation mechanism, and a closed-loop optimization mechanism. The number source monitoring unit develops an interface program for real-time docking with the hospital number source management system. The message queue technology used is Kafka's data transmission message queue technology to achieve real-time data transmission and monitor number source changes and expert scheduling status. The expression is:

[0017] Where, Indicates The collection of information related to the number source and expert scheduling captured by the time number source monitoring unit, is a collection The element in is a two-tuple, Indicates the message content. Indicates that the timestamp is used to record the corresponding message The specific time point captured is connected to the hospital's medical quality assessment system to obtain expert historical diagnosis and treatment data.

[0018] Furthermore, the operation process of the dynamic number source matching module includes: The expert's historical diagnosis and treatment data includes verified data on cure rate and patient satisfaction. The hospital's number source management system is logged in in real time to query number source information changes and update the local number source database in real time. The expression is:

[0019] Where, express The number source information set obtained at all times, express Moment The number source status data of the experts, express The number of experts included in the moment number source information, express The status of the local number source database after constant update, Is the database update function, express The old status of the local number source database at the moment, Indicates A collection of messages related to number sources and expert scheduling captured by the time number source monitoring unit.

[0020] Furthermore, the operation process of the dynamic number source matching module includes: The artificial intelligence matching unit is provided with an artificial intelligence matching algorithm model. Based on the patient's disease and the corresponding department, it analyzes the patient's disease type and severity, and in combination with the patient's preference settings for experts, screens experts that match the patient's disease and the corresponding department from the local number source database, and calculates and ranks the matching scores for each expert based on the expert's historical diagnosis and treatment data and current consultation status, and recommends the top-ranked experts to the patient. The expression is:

[0021] Where, Indicates expert With patients The total matching score of Representing individual experts, Represents individual patients, is the weight coefficient, Indicates expert With patients Disease depth fitness function, It is for experts Treatment results of factors related to visit time, is the Sigmoid activation function, Representative experts Characteristics of visit time, Indicates expert Expert attribute ratings, Is a patient The preference vector of is an expert The eigenvector of is the patient preference vector and expert feature vector The dot product operation, is an expert The number source is available status mark, is the slope parameter of the Sigmoid function, is the mean of the time characteristics.

[0022] Furthermore, the operation process of the dynamic number source matching module includes: The waiting and recommendation mechanism automatically adds the patient to the waiting queue of the expert when the number of expert numbers selected by the patient is full, and models the waiting queue as a priority queue. Rearrange the priority, expression:

[0023] Where, Indicates patient In the experts The priority score in the waiting queue, is the weight coefficient, represents the time when the patient enters the waiting queue, The minimum value is usually 0.001. Representative patients A quantitative value of the urgency of the condition, Indicates expert With patients The total matching score of Representing individual experts, Indicates that the patient with the highest priority selected from the waiting list is the one who is selected by the specialist The patients who are notified and assigned a number first when a number is released, Indicates expert The corresponding waiting patient set records the patient's condition information, preference settings and waiting time.

[0024] Furthermore, the operation process of the dynamic number source matching module includes: The candidate and recommendation mechanism is based on the urgency of the patient's condition and the expert matching degree, and selects experts from the same department to recommend , select alternative experts with high similarity scores and show them to patients, and define the similarity of disease treatment between experts. The expression is:

[0025] Where, Indicates expert and experts Similarity in disease treatment between Represents different expert individuals from the same department expert set Experts selected from is an expert The corresponding disease treatment feature vector, is a vector and The dot product operation, They are vectors The L2 norm of Indicates the final screened items to be shown to the patient A collection of alternative experts, It is a collection of experts from the same department from the hospital expert collection The expert group selected from the above and consistent with the patient's target department, This represents the target specialist that the patient originally wanted to see but whose appointments were full. is the similarity threshold, It is an indicator variable that monitors the changes in expert appointments in real time during the waiting process. When appointments are released, patients are notified and assigned appointments in order of waiting time. When a patient successfully registers with an alternative expert during the waiting period, the patient's registration information is automatically deleted from the waiting queue.

[0026] Furthermore, the operation process of the dynamic number source matching module includes: After the patient's consultation, the closed-loop optimization mechanism pushes a satisfaction questionnaire to the patient through the Internet hospital platform to collect the patient's evaluation data on the consultation. At the same time, the expert's diagnosis and treatment feedback information on the patient's condition is obtained from the hospital HIS system. The expert's diagnosis and treatment feedback information on the patient's condition is used as training data and input into the artificial intelligence matching algorithm model again. The artificial intelligence matching algorithm model is retrained and optimized through the policy gradient method. The expression is:

[0027] Where, is the gradient operator for the parameter Find the partial derivative, The objective function is an indicator to measure the performance of the artificial intelligence matching algorithm model by adjusting To optimize , Indicates that in the current strategy The expected value under It is based on parameters The policy function, Indicates at time Recommended experts, Indicates at time The patient's symptoms, Representation Policy Model In the parameters Time in state Next select action The probability of It's a strategy In state Take action The logarithmic probability of No. The total reward value of the registration match, Three weight parameters, Is a patient satisfaction rating, Is a patient Matching success indicator, is an expert The quality score of the diagnosis and treatment is calculated, and the weights of various factors and matching rules in the artificial intelligence matching algorithm model are adjusted.

[0028] Beneficial effects Compared with the known public technology, the technical solution provided by the present invention has the following beneficial effects: When used, the present invention effectively solves the problem that patients have difficulty in accurately judging which department of the outpatient clinic they should go to for their own diseases due to lack of medical knowledge. Through multi-source data fusion analysis, it provides patients with accurate department recommendations, greatly improves the accuracy of patients' first registration, and reduces time waste and medical resource mismatch caused by registering with the wrong department. At the same time, through dynamic number source monitoring and intelligent standby mechanism, as well as automatic recommendation of alternative experts, it solves the problem of patients having to repeat operations due to insufficient numbers when inquiring about expert outpatient information, greatly saves patients' registration time, and improves registration efficiency. The intelligent expert matching algorithm based on the patient's condition and preferences ensures that patients can receive expert diagnosis and treatment services that are most suitable for their condition, thereby improving the quality and effectiveness of medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a system diagram of an expert registration adaptation system for an Internet hospital driven by artificial intelligence according to the present invention. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0032] The present invention is described in further detail below with reference to the accompanying drawings: Example: like Figure 1As shown, the present invention provides an expert registration adaptation system for an Internet hospital based on artificial intelligence drive, comprising: The artificial intelligence guidance module automatically obtains the patient's medical records and infers the patient's disease and the corresponding department; Furthermore, the operation process of the artificial intelligence guidance module includes: The artificial intelligence guidance module includes a multi-source data acquisition unit, a semantic parsing model and a knowledge graph association unit. The multi-source data acquisition unit is provided with a front-end data acquisition interface adapted to a variety of Internet hospital platforms, including but not limited to web pages, mobile apps, and WeChat mini-program interfaces, and integrates a voice input plug-in in the patient symptom description area. The voice input plug-in is used to convert the patient's voice input into text in real time. The expression:

[0033] Where, A function representing speech-to-text conversion, Mathematical representation space for speech signals is the set of real numbers It is the dimension, Represents text space, Indicates time The collected speech signal feature vector, Expressing the moment The speech feature vector Perform speech-to-text conversion operations, Is the mathematical symbol for the independent variable corresponding to the maximum value, Represents a text candidate set The text candidates, Indicates that in a given speech feature vector Time text candidate The conditional probability of occurrence, The parameters representing the speech recognition model are then securely connected to the hospital's HIS system, LIS system, and PACS system through a standardized data interface protocol to automatically obtain patient medical records, which include but are not limited to historical medical record data and past registration behavior data, and the patient medical records are encrypted and stored in a data warehouse.

[0034] Furthermore, the operation process of the artificial intelligence guidance module includes: The semantic parsing model is pre-trained using medical text data including but not limited to medical textbooks, medical records, and medical papers. The expression:

[0035] Where, Represents the model parameters Find the minimum value, Is the expectation symbol, Represents the dataset Sample pairs are obtained by sampling , Understood as input medical text, Indicates the label corresponding to the input medical text, Represents the medical field text dataset used for pre-training, which is the weight coefficient used to control different loss terms The proportion of total losses in is the masked language model loss function, is the named entity recognition loss function, It is a sentence classification related loss function. The pre-training of the semantic parsing model is carried out by minimizing the multi-task loss function training. Multiple automatically acquired patient medical records are input into the semantic parsing model, and the semantic parsing model outputs patient symptom information.

[0036] Furthermore, the operation process of the artificial intelligence guidance module includes: The knowledge graph association unit uses the knowledge graph construction tool to construct the medical knowledge system including but not limited to diseases, symptoms, and departments into a medical knowledge graph.

[0037]

[0038] Where, It is a medical knowledge graph. is a collection of entities, It is a relationship set, which is stored and managed through a graph database. It realizes search and reasoning in the medical knowledge graph based on the patient's symptom information, and uses the probabilistic graph reasoning method to establish a probability propagation model from symptoms to diseases. Similarly, through the disease-department relationship edge in the medical knowledge graph, the corresponding department is identified, and the patient's disease and the corresponding department are output. The expression is:

[0039] Where, Indicates that when the symptom set is observed Temporary Disease The posterior probability of occurrence, Indicates the diseases, represents the patient's symptom set, represents the normalization factor, Represents a collection of symptoms The number of symptoms included in Represents a collection of symptoms The Symptoms, Indicates disease Symptoms when they occur The conditional probability of occurrence, Indicates disease The prior probability in the population, From the disease set The disease with the largest posterior probability is selected from Is the mathematical symbol for the independent variable corresponding to the maximum value, Represents a collection of diseases, Indicates the department corresponding to the inferred patient disease, Represented in the medical knowledge graph The input of the function for traversal reasoning is from the disease set The disease with the highest posterior probability selected .

[0040] Specifically, it receives user input through front-end platforms such as web pages, mobile apps, and WeChat applets, and supports automatic conversion of voice descriptions into structured text, achieving full coverage of different user habits. By connecting with hospital HIS, LIS, PACS and other systems, it retrieves patient medical records and registration records, and relies on medical semantic parsing models to perform named entity recognition and semantic classification on the input symptom text, extract standardized disease-related information, and accurately identify symptoms, disease entities and their relationships. In the reasoning stage, this system uses the medical knowledge graph built by the graph database to use the symptom-to-disease boundary The Yes probability propagation method realizes the joint reasoning of multiple symptoms of possible diseases, and further automatically determines the most appropriate registration department based on the disease-department mapping relationship in the graph, realizing intelligent matching "from symptoms to departments". Patients do not need to judge the department by themselves or repeatedly register the wrong number. This system can intelligently and accurately guide them to the most relevant specialist, reducing the number of misregistrations in the initial diagnosis. Voice recognition and multi-platform access lower the usage threshold, which is especially friendly to elderly users and patients with low cultural level. Through the combination of knowledge graph and artificial intelligence model, it can provide rapid triage judgment in high-concurrency scenarios, alleviating the pressure on physical hospital outpatient clinics.

[0041] The dynamic number source matching module executes the expert matching-candidate-feedback closed loop based on the real-time number source status, the patient's disease, and the corresponding department: (a) Based on the real-time appointment status, the target disease, and the target department, selecting matching experts from available experts and generating a recommendation list; (b) When the number of specialists selected by the patient is full, the dual-path operation is automatically performed: (b1) Add the patient to the specialist’s waiting list and monitor the availability of appointments in real time to allocate appointments in order; (b2) Selecting alternative experts who meet the similarity criteria from experts in the same target department and recommending them to the patient; (c) Dynamically optimize matching rules based on the patient's final consultation result feedback.

[0042] Furthermore, the operation process of the dynamic number source matching module includes: The dynamic number source matching module includes a number source monitoring unit, an artificial intelligence matching unit, a candidate and recommendation mechanism, and a closed-loop optimization mechanism. The number source monitoring unit develops an interface program for real-time docking with the hospital number source management system. The message queue technology used is Kafka's data transmission message queue technology to achieve real-time data transmission and monitor number source changes and expert scheduling status. The expression is:

[0043] Where, Indicates The collection of information related to the number source and expert scheduling captured by the time number source monitoring unit, is a collection The element in is a two-tuple, Indicates the message content. Indicates that the timestamp is used to record the corresponding message The specific time point captured is connected to the hospital's medical quality assessment system to obtain expert historical diagnosis and treatment data.

[0044] Furthermore, the operation process of the dynamic number source matching module includes: The expert's historical diagnosis and treatment data includes verified data on cure rate and patient satisfaction. The hospital's number source management system is logged in in real time to query number source information changes and update the local number source database in real time. The expression is:

[0045] Where, express The number source information set obtained at all times, express Moment The number source status data of the experts, express The number of experts included in the moment number source information, express The status of the local number source database after constant update, Is the database update function, express The old status of the local number source database at the moment, Indicates A collection of messages related to number sources and expert scheduling captured by the time number source monitoring unit.

[0046] Furthermore, the operation process of the dynamic number source matching module includes: The artificial intelligence matching unit is provided with an artificial intelligence matching algorithm model. Based on the patient's disease and the corresponding department, it analyzes the patient's disease type and severity, and in combination with the patient's preference settings for experts, screens experts that match the patient's disease and the corresponding department from the local number source database, and calculates and ranks the matching scores for each expert based on the expert's historical diagnosis and treatment data and current consultation status, and recommends the top-ranked experts to the patient. The expression is:

[0047] Where, Indicates expert With patients The total matching score of Representing individual experts, Represents individual patients, is the weight coefficient, Indicates expert With patients Disease depth fitness function, It is for experts Treatment results of factors related to visit time, is the Sigmoid activation function, Representative experts Characteristics of visit time, Indicates expert Expert attribute ratings, Is a patient The preference vector of is an expert The eigenvector of is the patient preference vector and expert feature vector The dot product operation, is an expert The number source is available status mark, is the slope parameter of the Sigmoid function, is the mean of the time characteristics.

[0048] Furthermore, the operation process of the dynamic number source matching module includes: The waiting and recommendation mechanism automatically adds the patient to the waiting queue of the expert when the number of expert numbers selected by the patient is full, and models the waiting queue as a priority queue. Rearrange the priority, expression:

[0049] Where, Indicates patient In the experts The priority score in the waiting queue, is the weight coefficient, represents the time when the patient enters the waiting queue, The minimum value is usually 0.001. Representative patients A quantitative value of the urgency of the condition, Indicates expert With patients The total matching score of Representing individual experts, Indicates that the patient with the highest priority selected from the waiting list is the one who is selected by the specialist The patients who are notified and assigned a number first when a number is released, Indicates expert The corresponding waiting patient set records the patient's condition information, preference settings and waiting time.

[0050] Furthermore, the operation process of the dynamic number source matching module includes: The candidate and recommendation mechanism is based on the urgency of the patient's condition and the expert matching degree, and selects experts from the same department to recommend , select alternative experts with high similarity scores and show them to patients, and define the similarity of disease treatment between experts. The expression is:

[0051] Where, Indicates expert and experts Similarity in disease treatment between Represents different expert individuals from the same department expert set Experts selected from is an expert The corresponding disease treatment feature vector, is a vector and The dot product operation, They are vectors The L2 norm of Indicates the final screened items to be shown to the patient A collection of alternative experts, It is a collection of experts from the same department from the hospital expert collection The expert group selected from the above and consistent with the patient's target department, This represents the target specialist that the patient originally wanted to see but whose appointments were full. is the similarity threshold, It is an indicator variable that monitors the changes in expert appointments in real time during the waiting process. When appointments are released, patients are notified and assigned appointments in order of waiting time. When a patient successfully registers with an alternative expert during the waiting period, the patient's registration information is automatically deleted from the waiting queue.

[0052] Furthermore, the operation process of the dynamic number source matching module includes: After the patient's consultation, the closed-loop optimization mechanism pushes a satisfaction questionnaire to the patient through the Internet hospital platform to collect the patient's evaluation data on the consultation. At the same time, the expert's diagnosis and treatment feedback information on the patient's condition is obtained from the hospital HIS system. The expert's diagnosis and treatment feedback information on the patient's condition is used as training data and input into the artificial intelligence matching algorithm model again. The artificial intelligence matching algorithm model is retrained and optimized through the policy gradient method. The expression is:

[0053] Where, is the gradient operator for the parameter Find the partial derivative, The objective function is an indicator to measure the performance of the artificial intelligence matching algorithm model by adjusting To optimize , Indicates that in the current strategy The expected value under It is based on parameters The policy function, Indicates at time Recommended experts, Indicates at time The patient's symptoms, Representation Policy Model In the parameters Time in state Next select action The probability of It's a strategy In state Take action The logarithmic probability of No. The total reward value of the registration match, Three weight parameters, Is a patient satisfaction rating, Is a patient Matching success indicator, is an expert The quality score of the diagnosis and treatment is calculated, and the weights of various factors and matching rules in the artificial intelligence matching algorithm model are adjusted.

[0054] Specifically, through the dynamic number source matching module, a complete "expert matching-standby-feedback closed loop" system is built around the patient's medical process. It monitors the hospital's number source management system at all times, captures number source changes and expert scheduling status, obtains expert historical diagnosis and treatment data from the medical quality assessment system, and updates the local number source database in real time. Through the artificial intelligence matching unit, based on the patient's disease and the corresponding department, the disease type and severity are first analyzed, and then combined with the patient's preference for experts, experts are screened from the local number source database. The expert's historical diagnosis and treatment data and current consultation situation are also combined to calculate the matching score and rank recommendation for the expert. When the patient's favorite expert number is full, on the one hand, the patient is added to the waiting queue, and a priority queue is modeled according to factors such as the urgency of the disease, waiting time, and matching degree. On the other hand, substitutes with high similarity in disease treatment are screened from the same department. Substitute experts are provided for patients to choose from, and the number of appointments is monitored in real time on the waiting list. When a number is available, the patient will be notified in order. Once the patient successfully registers for the substitute expert, he / she will automatically exit the original queue to avoid waste of resources. By utilizing the closed-loop optimization mechanism, satisfaction surveys and expert diagnosis and treatment feedback are collected after the patient's consultation. These data are used to retrain and optimize the artificial intelligence matching algorithm model. For patients, the cost of trial and error in registration is greatly reduced, and there is no need to repeat operations due to wrong registration or full number of appointments. The appropriate expert can be matched faster. In urgent cases, medical treatment can be obtained in time through the priority queue, which improves the efficiency and quality of medical services and increases patient satisfaction. The registration adaptation system is continuously polished through closed-loop optimization, making the expert registration process more intelligent, efficient and humane.

[0055] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An artificial intelligence-driven expert registration adaptation system for Internet hospitals, characterized by: include: The artificial intelligence guidance module automatically obtains the patient's medical records and infers the patient's disease and the corresponding department; The dynamic number source matching module is used to perform the following closed-loop operations based on the real-time number source status, the patient's disease, and the corresponding department: (a) Based on the real-time appointment status, the target disease, and the target department, selecting matching experts from available experts and generating a recommendation list; (b) When the number of specialists selected by the patient is full, the dual-path operation is automatically performed: (b1) Add the patient to the specialist’s waiting list and monitor the availability of appointments in real time to allocate appointments in order; (b2) Selecting alternative experts who meet the similarity criteria from experts in the same target department and recommending them to the patient; (c) Dynamically optimize matching rules based on the patient's final consultation result feedback.

2. The expert registration adaptation system for an Internet hospital based on artificial intelligence drive according to claim 1 is characterized in that: The operation process of the artificial intelligence guidance module includes: The artificial intelligence guidance module includes a multi-source data acquisition unit, a semantic parsing model and a knowledge graph association unit. The multi-source data acquisition unit is provided with a front-end data acquisition interface adapted to a variety of Internet hospital platforms, including web pages, mobile apps and WeChat mini-program interfaces, and integrates a voice input plug-in in the patient symptom description area. The voice input plug-in is used to convert the patient's voice input into text in real time. The expression: Where, A function representing speech-to-text conversion, Mathematical representation space for speech signals is the set of real numbers It is the dimension, Represents text space, Indicates time The collected speech signal feature vector, Expressing the moment The speech feature vector Perform speech-to-text conversion operations, Is the mathematical symbol for the independent variable corresponding to the maximum value, Represents a text candidate set The text candidates, Indicates that in a given speech feature vector Time text candidate The conditional probability of occurrence, The parameters of the speech recognition model are represented, and then securely connected to the hospital HIS system, LIS system and PACS system through a standardized data interface protocol to automatically obtain patient medical records, which include historical medical record data and past registration behavior data, and encrypt and store the patient medical records in a data warehouse.

3. The expert registration adaptation system for an Internet hospital based on artificial intelligence drive according to claim 2 is characterized in that: The operation process of the artificial intelligence guidance module includes: The semantic parsing model is pre-trained using medical text data including medical textbooks, medical records, and medical papers. The expression is: Where, Represents the model parameters Find the minimum value, Is the expectation symbol, Represents the dataset Sample pairs are obtained by sampling , Understood as input medical text, Indicates the label corresponding to the input medical text, Represents the medical field text dataset used for pre-training, which is the weight coefficient used to control different loss terms The proportion of total losses in is the masked language model loss function, is the named entity recognition loss function, It is a sentence classification related loss function. The pre-training of the semantic parsing model is carried out by minimizing the multi-task loss function training. Multiple automatically acquired patient medical records are input into the semantic parsing model, and the semantic parsing model outputs patient symptom information.

4. The expert registration adaptation system for an Internet hospital based on artificial intelligence drive according to claim 3 is characterized in that: The operation process of the artificial intelligence guidance module includes: The knowledge graph association unit uses the knowledge graph construction tool to construct the information of diseases, symptoms, and departments in the medical knowledge system into a medical knowledge graph. Where, It is a medical knowledge graph. is a collection of entities, It is a relationship set, which is stored and managed through a graph database. It realizes search and reasoning in the medical knowledge graph based on the patient's symptom information, and uses the probabilistic graph reasoning method to establish a probability propagation model from symptoms to diseases. Similarly, through the disease-department relationship edge in the medical knowledge graph, the corresponding department is identified, and the patient's disease and the corresponding department are output. The expression is: Where, Indicates that when the symptom set is observed Temporary Disease The posterior probability of occurrence, Indicates the diseases, represents the patient's symptom set, represents the normalization factor, Represents a collection of symptoms The number of symptoms included in Represents a collection of symptoms The Symptoms, Indicates disease Symptoms when they occur The conditional probability of occurrence, Indicates disease The prior probability in the population, Represents a disease set The disease with the largest posterior probability is selected from Is the mathematical symbol for the independent variable corresponding to the maximum value, Represents a collection of diseases, Indicates the department corresponding to the inferred patient disease, Represented in the medical knowledge graph The input of the function for traversal reasoning is from the disease set The disease with the highest posterior probability selected .

5. The artificial intelligence-driven internet hospital expert registration adaptation system according to claim 4 is characterized in that: The operation process of the dynamic source matching module includes: The dynamic number source matching module includes a number source monitoring unit, an artificial intelligence matching unit, a candidate and recommendation mechanism, and a closed-loop optimization mechanism. The number source monitoring unit develops an interface program for real-time docking with the hospital number source management system. The message queue technology used is Kafka's data transmission message queue technology to achieve real-time data transmission and monitor number source changes and expert scheduling status. The expression is: Where, Indicates The collection of information related to the number source and expert scheduling captured by the time number source monitoring unit, is a collection The element in is a two-tuple, Indicates the message content. Indicates that the timestamp is used to record the corresponding message The specific time point captured is connected to the hospital's medical quality assessment system to obtain expert historical diagnosis and treatment data.

6. The expert registration adaptation system for an Internet hospital based on artificial intelligence drive according to claim 5 is characterized in that: The operation process of the dynamic source matching module includes: The expert's historical diagnosis and treatment data includes verified data on cure rate and patient satisfaction. The hospital's number source management system is logged in in real time to query number source information changes and update the local number source database in real time. The expression is: Where, express The number source information set obtained at all times, express Moment The number source status data of the experts, express The number of experts included in the moment number source information, express The status of the local number source database after constant update, Is the database update function, express The old status of the local number source database at the moment, Indicates A collection of messages related to number sources and expert scheduling captured by the time number source monitoring unit.

7. The artificial intelligence-driven internet hospital expert registration adaptation system according to claim 6 is characterized in that: The operation process of the dynamic source matching module includes: The artificial intelligence matching unit is provided with an artificial intelligence matching algorithm model. Based on the patient's disease and the corresponding department, it analyzes the patient's disease type and severity, and in combination with the patient's preference settings for experts, screens experts that match the patient's disease and the corresponding department from the local number source database, and calculates and ranks the matching scores for each expert based on the expert's historical diagnosis and treatment data and current consultation status, and recommends the top-ranked experts to the patient. The expression is: Where, Indicates expert With patients The total matching score of Representing individual experts, Represents individual patients, is the weight coefficient, Indicates expert With patients Disease depth fitness function, It is for experts Treatment results of factors related to visit time, is the Sigmoid activation function, Representative experts Characteristics of visit time, Indicates expert Expert attribute ratings, Is a patient The preference vector of is an expert The eigenvector of is the patient preference vector and expert feature vector The dot product operation, is an expert The number source is available status mark, is the slope parameter of the Sigmoid function, is the mean of the time characteristics.

8. The artificial intelligence-driven internet hospital expert registration adaptation system according to claim 7 is characterized in that: The operation process of the dynamic source matching module includes: The waiting and recommendation mechanism automatically adds the patient to the waiting queue of the expert when the number of expert numbers selected by the patient is full, and models the waiting queue as a priority queue. Rearrange the priority, expression: Where, Indicates patient In the experts The priority score in the waiting queue, is the weight coefficient, represents the time when the patient enters the waiting queue, The minimum value is usually 0.

001. Representative patients A quantitative value of the urgency of the condition, Indicates expert With patients The total matching score of Representing individual experts, Indicates that the patient with the highest priority selected from the waiting list is the patient who is selected by the specialist The patients who are notified and assigned a number first when a number is released, Indicates expert The corresponding waiting patient set records the patient's condition information, preference settings and waiting time.

9. The artificial intelligence-driven internet hospital expert registration adaptation system according to claim 8 is characterized in that: The operation process of the dynamic source matching module includes: The candidate and recommendation mechanism is based on the urgency of the patient's condition and the expert matching degree, and selects experts from the same department to recommend , select alternative experts with high similarity scores and show them to patients, and define the similarity of disease treatment between experts. The expression is: Where, Indicates expert and experts Similarity in disease treatment between Represents different expert individuals from the same department expert set Experts selected from is an expert The corresponding disease treatment feature vector, is a vector and The dot product operation, They are vectors The L2 norm of Indicates the final screened items to be shown to the patient A collection of alternative experts, It is a collection of experts from the same department from the hospital expert collection The expert group selected from the above and consistent with the patient's target department, This represents the target specialist that the patient originally wanted to see but whose appointments were full. is the similarity threshold, It is an indicator variable that monitors the changes in expert appointments in real time during the waiting process. When appointments are released, patients are notified and assigned appointments in order of waiting time. When a patient successfully registers with an alternative expert during the waiting period, the patient's registration information is automatically deleted from the waiting queue.

10. The artificial intelligence-driven internet hospital expert registration adaptation system according to claim 9 is characterized in that: The operation process of the dynamic source matching module includes: After the patient's consultation, the closed-loop optimization mechanism pushes a satisfaction questionnaire to the patient through the Internet hospital platform to collect the patient's evaluation data on the consultation. At the same time, the expert's diagnosis and treatment feedback information on the patient's condition is obtained from the hospital HIS system. The expert's diagnosis and treatment feedback information on the patient's condition is used as training data and input into the artificial intelligence matching algorithm model again. The artificial intelligence matching algorithm model is retrained and optimized through the policy gradient method. The expression is: Where, is the gradient operator for the parameter Find the partial derivative, The objective function is an indicator to measure the performance of the artificial intelligence matching algorithm model by adjusting To optimize , Indicates that in the current strategy The expected value under It is based on parameters The policy function, Indicates at time Recommended experts, Indicates at time The patient's symptoms, Representation Policy Model In the parameters Time in state Select Action The probability of It's a strategy In state Take action The logarithmic probability of No. The total reward value of the registration match, Three weight parameters, Is a patient satisfaction rating, Is a patient Matching success indicator, is an expert The quality score of the diagnosis and treatment is calculated, and the weights of various factors and matching rules in the artificial intelligence matching algorithm model are adjusted.

Images