Medical self-service terminal based on AI

Abstract

The invention provides an AI-based medical self-service terminal, and relates to the field of data processing, the AI-based medical self-service terminal comprises a plurality of interaction nodes and a plurality of data processing nodes, and a cloud platform is used for distributing the corresponding data processing nodes to the interaction nodes based on distribution auxiliary information; the data processing node corresponding to the interaction node is used for generating registration recommendation information and performing appointment registration based on feedback of a user on the registration recommendation information; the interaction node is used for receiving a file query instruction initiated by a user, and the data processing node corresponding to the interaction node is used for obtaining a target medical file from the cloud platform based on the file query instruction and sending the target medical file to the interaction node; the interaction node is used for receiving an information query instruction initiated by a user, the data processing node corresponding to the interaction node is used for obtaining the target medical information from the cloud platform and sending the target medical information to the interaction node, and the method has the advantage of improving the quality of medical self-service.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to an AI-based medical self-service terminal. Background Art

[0002] With the rapid development of science and technology, artificial intelligence (AI) is gradually penetrating into various industries, injecting new vitality into the innovative development of enterprises. In the medical field, the application of AI technology is in full swing. From remote consultation to intelligent diagnosis, to the formulation of personalized treatment plans, AI is reshaping the landscape of the medical industry. In this context, AI-based medical self-service terminals have emerged, aiming to provide patients with more convenient and efficient medical services.

[0003] Existing medical self-service terminals lack natural language processing capabilities and cannot accurately understand patients' verbal instructions, further reducing the interactive experience. In addition, although AI technology has significant advantages in data processing and analysis, existing medical self-service terminals still have bottlenecks in data processing capabilities. On the one hand, the hardware configuration of the terminal may not be sufficient to support real-time processing of large-scale data; on the other hand, the software algorithm of the terminal may not be efficient enough, resulting in slow data processing speed and unable to meet the immediate needs of patients. The degree of intelligence of existing medical self-service terminals still needs to be improved. Some terminals lack advanced AI algorithm and model support and cannot provide accurate medical services.

[0004] Therefore, there is a need to provide an AI-based medical self-service terminal to improve the quality of medical self-service. Summary of the invention

[0005] The present invention provides an AI-based medical self-service terminal, comprising a plurality of interactive nodes and a plurality of data processing nodes, wherein the interactive nodes are used to interact with users, and the plurality of data processing nodes perform data interaction with a cloud platform, and the cloud platform is used to allocate corresponding data processing nodes to the interactive nodes based on allocation auxiliary information, wherein the allocation auxiliary information at least includes the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple current time periods: the interactive node is used to receive a registration instruction initiated by a user, and the data processing node corresponding to the interactive node is used to generate registration recommendation information based on the registration instruction, and make an appointment for registration based on the user's feedback on the registration recommendation information; the interactive node is used to receive a file query instruction initiated by a user, and the data processing node corresponding to the interactive node is used to obtain a target medical file from the cloud platform based on the file query instruction, and send the target medical file to the interactive node; the interactive node is used to receive an information query instruction initiated by a user, and the data processing node corresponding to the interactive node is used to obtain a target medical information from the cloud platform based on the information query instruction, and send the target medical information to the interactive node.

[0006] Furthermore, the cloud platform is used to allocate corresponding data processing nodes to the interactive nodes based on the allocation auxiliary information, including: obtaining the number of registration instructions, the number of file query instructions, and the number of information query instructions of the interactive nodes in multiple historical time periods; calculating the registration instruction number correlation coefficient, the file query instruction number correlation coefficient, and the information query instruction number correlation coefficient of any two interactive nodes based on the registration instruction number correlation coefficient, the file query instruction number correlation coefficient, and the information query instruction number correlation coefficient of any two interactive nodes; for each interactive node, determining the registration instruction number correlation coefficient, the file query instruction number correlation coefficient, and the information query instruction number correlation coefficient of any two interactive nodes based on the registration instruction number correlation coefficient, the file query instruction number correlation coefficient, and the information query instruction number correlation coefficient of any two interactive nodes. Positively associated interactive nodes and negatively associated interactive nodes of interactive nodes; obtaining the number of registration instructions, the number of file query instructions, and the number of information query instructions of the interactive nodes in multiple current time periods; based on the positively associated interactive nodes and negatively associated interactive nodes of each interactive node and the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple future time periods, predicting the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node; based on the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple future time periods, allocating corresponding data processing nodes to the interactive nodes.

[0007] Further, the cloud platform determines the positively associated interactive nodes and negatively associated interactive nodes of the interactive nodes based on the registration instruction quantity correlation coefficient, file query instruction quantity correlation coefficient and information query instruction quantity correlation coefficient of any two interactive nodes, including: taking a weighted sum of the registration instruction quantity correlation coefficient, file query instruction quantity correlation coefficient and information query instruction quantity correlation coefficient of any two interactive nodes to calculate the comprehensive correlation coefficient of any two interactive nodes; for each interactive node, judging whether there is an interactive node whose registration instruction quantity correlation coefficient, file query instruction quantity correlation coefficient or information query instruction quantity correlation coefficient is greater than a first correlation coefficient positive threshold, and whose registration instruction quantity correlation coefficient, file query instruction quantity correlation coefficient and information query instruction quantity correlation coefficient are all greater than a second correlation coefficient positive threshold; if so, as the positively associated interactive node of the interactive node, judging whether there is an interactive node whose comprehensive correlation coefficient is greater than the first comprehensive correlation coefficient threshold, If so, as a positively associated interactive node of the interactive node, wherein the first positive correlation coefficient threshold, the second positive correlation coefficient threshold and the first comprehensive correlation coefficient threshold are all greater than 0, and the first positive correlation coefficient threshold is greater than the second positive correlation coefficient threshold; for each interactive node, determine whether there is an interactive node whose at least one of the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient or the information query instruction quantity correlation coefficient is less than the first negative correlation coefficient threshold, and whose registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient and the information query instruction quantity correlation coefficient are all less than the second negative correlation coefficient threshold; if so, as a negatively associated interactive node of the interactive node, determine whether there is an interactive node whose comprehensive correlation coefficient is less than the second comprehensive correlation coefficient threshold; if so, as a negatively associated interactive node of the interactive node, wherein the first negative correlation coefficient threshold, the second negative correlation coefficient threshold and the second comprehensive correlation coefficient threshold are all less than 0, and the second negative correlation coefficient threshold is greater than the first negative correlation coefficient threshold.

[0008] Furthermore, the cloud platform predicts the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple future time periods based on the positively associated interactive nodes and the negatively associated interactive nodes of each interactive node and the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple current time periods, including: for each interactive node, based on the positively associated interactive nodes and the negatively associated interactive nodes of the interactive node, establishing a quantity prediction model and a quantity correction model corresponding to the interactive node; for each interactive node, through the quantity prediction model corresponding to the interactive node, based on the number of registration instructions, the number of file query instructions, and the number of information query instructions of the interactive node in multiple current time periods and the positively associated interactive nodes of the interactive node The number of registration instructions, file query instructions and information query instructions of the interactive nodes and negatively associated interactive nodes in multiple current time periods are used to predict the number of registration instructions, file query instructions and information query instructions of the interactive nodes in multiple future time periods; through the quantity correction model corresponding to each interactive node, based on the number of registration instructions, file query instructions and information query instructions of the positively associated interactive nodes and negatively associated interactive nodes of each interactive node in multiple future time periods, the number of registration instructions, file query instructions and information query instructions of the interactive nodes in multiple future time periods are iteratively corrected to generate the corrected number of registration instructions, file query instructions and information query instructions of each interactive node in multiple future time periods.

[0009] Furthermore, the cloud platform allocates corresponding data processing nodes to interactive nodes based on the number of registration instructions, the number of file query instructions and the number of information query instructions of each interactive node in multiple future time periods, including: establishing a node scheduling objective function; generating multiple sample node scheduling schemes; calculating the node scheduling objective function value of the sample node scheduling scheme based on the node scheduling objective function, the number of registration instructions, the number of file query instructions and the number of information query instructions of each interactive node in multiple future time periods, and the correlation coefficient of the number of registration instructions, the correlation coefficient of the number of file query instructions and the correlation coefficient of the number of information query instructions of any two interactive nodes; determining the target node scheduling scheme based on the node scheduling objective function value of each sample node scheduling scheme through a genetic algorithm; and allocating corresponding data processing nodes to the interactive nodes based on the target node scheduling scheme.

[0010] Furthermore, the data processing node corresponding to the interactive node generates registration recommendation information based on the registration instruction, including: reading the user's identity information based on the registration instruction; obtaining the user's electronic medical record from the cloud platform based on the user's identity information; obtaining the user's medical consultation information based on the disease feature knowledge graph and the user's electronic medical record through natural language processing; and generating registration recommendation information based on the user's medical consultation information.

[0011] Furthermore, the data processing node corresponding to the interactive node obtains the user's medical consultation information through natural language processing based on the disease characteristic knowledge graph and the user's electronic medical record, including: obtaining the electronic medical records of multiple historical users; determining the onset correlation parameters of any two diseases based on the electronic medical records of multiple historical users; obtaining the user's condition description information; sorting multiple diseases based on the user's condition description information, the disease characteristic knowledge graph and the onset correlation parameters of any two diseases, and generating disease sorting results; generating medical consultation questions based on the disease characteristic knowledge graph and the disease sorting results through natural language processing, and obtaining the user's feedback information on the medical consultation questions.

[0012] Furthermore, the data processing node corresponding to the interaction node is also used to: obtain reaction time characteristics of multiple sample users to basic interaction information; obtain reaction time characteristics of users to basic interaction information; and adjust the display font size of the interaction node based on the reaction time characteristics of multiple sample users to the basic interaction information and the reaction time characteristics of users to the basic interaction information.

[0013] Furthermore, the data processing node corresponding to the interactive node obtains the target medical file from the cloud platform based on the file query instruction, including: reading the user's identity information based on the file query instruction; obtaining keyword information of multiple medical files of the user from the cloud platform based on the user's identity information; generating file selection interaction information based on the keyword information of multiple medical files of the user; obtaining user feedback on the file selection interaction information through the interactive node; obtaining the encrypted target medical file from the cloud platform based on the user's feedback on the file selection interaction information; and decrypting the encrypted target medical file to generate the target medical file.

[0014] Furthermore, the data processing node corresponding to the interactive node obtains target medical information from the cloud platform based on the information query instruction, including: determining information search keywords based on the information query instruction; and obtaining target medical information from the cloud platform based on the information search keywords.

[0015] Compared with the prior art, the AI-based medical self-service terminal provided by the present invention has at least the following beneficial effects:

[0016] 1. Users can easily initiate registration instructions, file query instructions and information query instructions through interactive nodes without waiting in line at the manual window, which greatly saves time. The self-service terminal is open 24 hours a day, and users can operate it anytime and anywhere, meeting the medical needs of different time periods. The data processing node can generate registration recommendation information based on the user's registration instructions to help users choose the department and doctor that suits them, improving the pertinence and efficiency of medical treatment. The self-service terminal can automatically process instructions such as registration, file query and information query, reducing the work pressure of medical staff and enabling them to focus more on patient treatment and services. With the assistance of AI technology, the data processing node can quickly and accurately process large amounts of medical data, improving the overall efficiency of medical services.

[0017] 2. Through historical data analysis and prediction models, the cloud platform can predict in advance the number of registration instructions, file query instructions, and information query instructions for each interactive node in the future time period. This predictive allocation mechanism ensures that the data processing nodes can be prepared in advance, and resources can be scheduled and configured according to the predicted needs, avoiding idle or overloaded resources. As time goes by, the cloud platform will continuously obtain new data and update the prediction model to achieve dynamic adjustment of resource allocation. This dynamic adjustment mechanism can ensure that the data processing nodes are always in the best working state to meet the changing needs of users. Through predictive allocation and dynamic adjustment, the cloud platform can ensure that the data processing nodes have sufficient resources during peak processing, thereby reducing the waiting time of users. Users can obtain registration recommendation information, medical documents, and information query results more quickly, improving medical efficiency and satisfaction.

[0018] 3. By weighted summing the correlation coefficients of registration instructions, file query instructions, and information query instructions, the cloud platform can more comprehensively evaluate the correlation between two interaction nodes. This comprehensive consideration method avoids the one-sidedness that may be caused by a single number of instructions and improves the accuracy of correlation analysis. By setting the first correlation coefficient positive threshold and the first comprehensive correlation coefficient threshold to determine positively correlated interaction nodes, and setting the first correlation coefficient negative threshold and the second comprehensive correlation coefficient threshold to determine negatively correlated interaction nodes, the cloud platform can flexibly adjust the thresholds according to different business needs and scenarios. This flexibility helps ensure the accuracy and effectiveness of correlation analysis. It is possible to comprehensively consider the correlation coefficients of multiple dimensions, thereby more accurately determining the positive and negatively correlated interaction nodes of interaction nodes.

[0019] 4. The cloud platform can predict resource requirements in advance and perform predictive scheduling by analyzing the number of instructions for each interactive node in multiple time periods in the future. This approach avoids the situation where resource allocation lags behind actual demand and improves the efficiency of resource scheduling. When determining the node scheduling plan, the cloud platform not only considers the number of instructions of the interactive node itself, but also considers the multiple correlation coefficients between any two interactive nodes. This comprehensive consideration makes resource scheduling more accurate and can better meet actual business needs. At the same time, it also avoids the surge in instructions of a certain interactive node in a short period of time, resulting in overload. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0021] Figure 1 It is a module schematic diagram of an AI-based medical self-service terminal according to some embodiments of this specification;

[0022] Figure 2 It is a flowchart of allocating corresponding data processing nodes to interactive nodes based on allocating auxiliary information according to some embodiments of this specification. DETAILED DESCRIPTION

[0023] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of this specification. For ordinary technicians in this field, this specification can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0024] Figure 1 is a module schematic diagram of an AI-based medical self-service terminal according to some embodiments of this specification, such as Figure 1 As shown, an AI-based medical self-service terminal includes multiple interactive nodes and multiple data processing nodes. The interactive nodes are used to interact with users, and the multiple data processing nodes interact with the cloud platform for data.

[0025] The cloud platform is used to allocate corresponding data processing nodes to the interactive nodes based on allocation auxiliary information, wherein the allocation auxiliary information includes at least the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple current time periods. The current time period may be a time period in the current cycle (e.g., the current day), or a time period whose time distance from the current time is less than a preset time distance threshold.

[0026] Figure 2 is a flow chart of allocating corresponding data processing nodes to interactive nodes based on allocation auxiliary information according to some embodiments of this specification, such as Figure 2 As shown, in some embodiments, the cloud platform is used to allocate a corresponding data processing node to the interactive node based on the allocation auxiliary information, including:

[0027] Obtain the number of registration instructions, file query instructions, and information query instructions of the interactive node in multiple historical time periods;

[0028] According to the number of registration instructions, the number of file query instructions and the number of information query instructions of any two interactive nodes in multiple historical time periods, the registration instruction number correlation coefficient, the file query instruction number correlation coefficient and the information query instruction number correlation coefficient of any two interactive nodes are calculated;

[0029] For each interactive node, based on the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient and the information query instruction quantity correlation coefficient of any two interactive nodes, the positively correlated interactive node and the negatively correlated interactive node of the interactive node are determined;

[0030] Obtain the number of registration instructions, file query instructions, and information query instructions of the interactive node in multiple current time periods;

[0031] Based on the positively associated interactive nodes and negatively associated interactive nodes of each interactive node and the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple current time periods, the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple future time periods are predicted;

[0032] Based on the number of registration instructions, the number of file query instructions and the number of information query instructions of each interactive node in multiple future time periods, a corresponding data processing node is allocated to the interactive node.

[0033] Specifically, the number of registration instructions refers to the number of registration instructions initiated by users completed by the interactive node within a certain period of time. The number of file query instructions refers to the number of file query instructions initiated by users completed by the interactive node within a certain period of time. The number of information query instructions refers to the number of information query instructions initiated by users completed by the interactive node within a certain period of time.

[0034] The correlation coefficient of the number of registration instructions of two interactive nodes can be calculated according to the following formula:

[0035]

[0036] Among them, r ((i,j),1) is the correlation coefficient of the number of registration instructions between the ith interaction node and the jth interaction node, M is the total number of sampled historical time periods, and N (i,m) is the number of registration instructions of the i-th interaction node in the m-th historical time period of sampling, N (j,m) is the number of registration instructions of the jth interaction node in the mth historical time period of sampling.

[0037] The calculation method of the correlation coefficient of the number of file query instructions and the correlation coefficient of the number of information query instructions of two interactive nodes is consistent with the calculation method of the correlation coefficient of the number of registration instructions, which will not be repeated here.

[0038] In some embodiments, the cloud platform determines the positively associated interactive nodes and the negatively associated interactive nodes of the interactive nodes based on the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient, and the information query instruction quantity correlation coefficient of any two interactive nodes, including:

[0039] The weighted sum of the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient and the information query instruction quantity correlation coefficient of any two interactive nodes is calculated to calculate the comprehensive correlation coefficient of any two interactive nodes;

[0040] For each interactive node, determine whether there is an interactive node whose registration instruction quantity correlation coefficient, file query instruction quantity correlation coefficient or information query instruction quantity correlation coefficient is greater than the first correlation coefficient positive threshold, and whose registration instruction quantity correlation coefficient, file query instruction quantity correlation coefficient and information query instruction quantity correlation coefficient are all greater than the second correlation coefficient positive threshold. If so, act as a positively associated interactive node of the interactive node. Determine whether there is an interactive node whose comprehensive correlation coefficient is greater than the first comprehensive correlation coefficient threshold. If so, act as a positively associated interactive node of the interactive node, wherein the first correlation coefficient positive threshold, the second correlation coefficient positive threshold and the first comprehensive correlation coefficient threshold are all greater than 0, and the first correlation coefficient positive threshold is greater than the second correlation coefficient positive threshold.

[0041] For each interactive node, determine whether there is an interactive node whose at least one of the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient or the information query instruction quantity correlation coefficient is less than the first correlation coefficient negative threshold, and whose registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient and the information query instruction quantity correlation coefficient are all less than the second correlation coefficient negative threshold. If so, act as a negatively associated interactive node of the interactive node. Determine whether there is an interactive node whose comprehensive correlation coefficient is less than the second comprehensive correlation coefficient threshold. If so, act as a negatively associated interactive node of the interactive node, wherein the first correlation coefficient negative threshold, the second correlation coefficient negative threshold and the second comprehensive correlation coefficient threshold are all less than 0, and the second correlation coefficient negative threshold is greater than the first correlation coefficient negative threshold.

[0042] Specifically, the weights of the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient, and the information query instruction quantity correlation coefficient may be determined based on the following process:

[0043]

[0044] Among them, w1 is the weight of the registration instruction number correlation coefficient, w2 is the weight of the file query instruction number correlation coefficient, w3 is the weight of the information query instruction number correlation coefficient, γ1 is the fluctuation parameter of the registration instruction number correlation coefficient, γ2 is the fluctuation parameter of the file query instruction number correlation coefficient, γ3 is the fluctuation parameter of the information query instruction number correlation coefficient, and K is the total number of interactive nodes. The calculation method of the fluctuation parameter of the file query instruction number correlation coefficient and the fluctuation parameter of the information query instruction number correlation coefficient is consistent with the calculation method of the fluctuation parameter of the registration instruction number correlation coefficient, which will not be repeated here.

[0045] As an example only, the weight of the correlation coefficient of the number of registration instructions is 0.5, the weight of the correlation coefficient of the number of file query instructions is 0.3, and the weight of the correlation coefficient of the number of information query instructions is 0.2. For any two interactive nodes A and B, if their correlation coefficient of the number of registration instructions is 0.8, the correlation coefficient of the number of file query instructions is 0.6, and the correlation coefficient of the number of information query instructions is 0.4, then the comprehensive correlation coefficient is 0.5×0.8+0.3×0.6+0.2×0.4=0.66. For any two interactive nodes A and B, the cloud platform has calculated the correlation coefficient of the number of registration instructions, the correlation coefficient of the number of file query instructions, and the correlation coefficient of the number of information query instructions between them, which are respectively recorded as R_AB (registration), F_AB (file query), and I_AB (information query). Check whether there is at least one correlation coefficient (any one of R_AB, F_AB, and I_AB) greater than the first correlation coefficient positive threshold (for example, 0.5). If so, it is considered that B is a positively associated interactive node of A, and A is also a positively associated interactive node of B. If there are two interactive nodes A and C, the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient or the information query instruction quantity correlation coefficient are all less than or equal to the first correlation coefficient positive threshold, and the comprehensive correlation coefficient of the interactive nodes A and C is greater than the first comprehensive correlation coefficient threshold, then C is considered to be a positively correlated interactive node of A, and A is also a positively correlated interactive node of C.

[0046] In some embodiments, the cloud platform predicts the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple future time periods based on the positively associated interactive nodes and the negatively associated interactive nodes of each interactive node and the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple current time periods, including:

[0047] For each interaction node, based on the positively associated interaction nodes and negatively associated interaction nodes of the interaction node, a quantity prediction model and a quantity correction model corresponding to the interaction node are established, wherein the quantity prediction model and the quantity correction model may be a multiple linear regression model;

[0048] For each interactive node, using the quantity prediction model corresponding to the interactive node, based on the number of registration instructions, the number of file query instructions, and the number of information query instructions of the interactive node in multiple current time periods, and the number of registration instructions, the number of file query instructions, and the number of information query instructions of the positively associated interactive nodes and the negatively associated interactive nodes of the interactive node in multiple current time periods, predict the number of registration instructions, the number of file query instructions, and the number of information query instructions of the interactive node in multiple future time periods;

[0049] Through the quantity correction model corresponding to each interactive node, based on the number of registration instructions, file query instructions and information query instructions of the positively associated interactive nodes and negatively associated interactive nodes of each interactive node in multiple future time periods, the number of registration instructions, file query instructions and information query instructions of the interactive node in multiple future time periods are iteratively corrected to generate the corrected number of registration instructions, file query instructions and information query instructions for each interactive node in multiple future time periods.

[0050] Specifically, iterative correction can be performed according to the following process:

[0051] S11. For each interactive node, the number of registration instructions, the number of file query instructions, and the number of information query instructions of the interactive node in multiple future time periods are corrected through the number correction model corresponding to the interactive node, based on the number of registration instructions, the number of file query instructions, and the number of information query instructions of the positively associated interactive nodes and the negatively associated interactive nodes of the interactive node corresponding to the t-1th iteration in multiple future time periods, to generate the corrected number of registration instructions, the number of file query instructions, and the number of information query instructions of the interactive node in multiple future time periods corresponding to the t-1th iteration, wherein the number of registration instructions, the number of file query instructions, and the number of information query instructions of the positively associated interactive nodes and the negatively associated interactive nodes of the interactive node corresponding to the t-1th iteration used in the first iteration are the number of registration instructions, the number of file query instructions, and the number of information query instructions output by the quantity prediction model;

[0052] S12, based on the revised number of registration instructions, file query instructions, and information query instructions of the interactive node in multiple future time periods corresponding to the t-th iteration and the revised number of registration instructions, file query instructions, and information query instructions of the interactive node in multiple future time periods corresponding to the t-1-th iteration, calculate the global difference corresponding to the t-th iteration;

[0053] S13, determine whether t is greater than 3, if so, execute S14, if not, set t=t+1, execute S11;

[0054] S14, determine whether the global difference corresponding to the t-th iteration, the global difference corresponding to the t-1-th iteration, and the global difference corresponding to the t-2-th iteration are all less than the global difference threshold. If so, complete the correction; if not, set t=t+1 and execute S11.

[0055] The global difference can be calculated according to the following formula:

[0056]

[0057] Among them, △ tis the global difference corresponding to the tth iteration, N (k,1,t,e) is the number of registration instructions of the kth interaction node in the eth future time period corresponding to the tth iteration, N (k,1,t-1,e) is the number of registration instructions of the kth interaction node in the eth future time period corresponding to the t-1th iteration, N (k,2,t,e) is the number of file query instructions of the kth interactive node in the eth future time period corresponding to the tth iteration, N (k,2,t-1,e) is the number of file query instructions of the kth interactive node in the eth future time period corresponding to the t-1th iteration, N (k,3,t,e) is the number of information query instructions of the kth interactive node in the eth future time period corresponding to the tth iteration, N (k,3,t-1,e) is the number of information query instructions of the kth interactive node in the eth future time period corresponding to the t-1th iteration, K is the total number of interactive nodes, and E is the total number of future time periods.

[0058] In some embodiments, the cloud platform allocates corresponding data processing nodes to the interactive nodes based on the number of registration instructions, file query instructions, and information query instructions of each interactive node in multiple future time periods, including:

[0059] Establish node scheduling objective function;

[0060] Generate multiple sample node scheduling schemes;

[0061] Calculate the node scheduling objective function value of the sample node scheduling solution based on the node scheduling objective function, the number of registration instructions, the number of file query instructions and the number of information query instructions of each interactive node in multiple future time periods, and the correlation coefficient of the number of registration instructions, the correlation coefficient of the number of file query instructions and the correlation coefficient of the number of information query instructions between any two interactive nodes;

[0062] Determine the target node scheduling scheme based on the node scheduling objective function value of each sample node scheduling scheme through genetic algorithm;

[0063] Based on the target node scheduling scheme, corresponding data processing nodes are allocated to the interactive nodes.

[0064] Specifically, the node scheduling objective function is a mathematical model that evaluates the pros and cons of different node scheduling schemes. It calculates the total value or score of each scheduling scheme based on multiple factors. The node scheduling objective function considers the number of instructions (registration, file query, information query) of the interactive node and the correlation between these instructions (represented by the correlation coefficient).

[0065] As an example only, the node scheduling objective function can be:

[0066]

[0067] Where F is the node scheduling objective function, w4, w5 and w6 are preset weights, w4, w5 and w6 are greater than 0, w4+w5+w6=1, P1, P2 and P3 are normalized parameters, P1, P2 and P3 are greater than 0, N1 is the number of interactive nodes whose corresponding data processing nodes change, L (g,e) is the computing load of the g-th data processing node in the e-th future time period, E is the total number of future time periods, G is the total number of data processing nodes, and r (i,j) is the comprehensive correlation coefficient between the i-th interaction node and the j-th interaction node corresponding to the g-th data processing node, Q g is the total number of interaction nodes corresponding to the g-th data processing node.

[0068] Genetic algorithm is an optimization algorithm that simulates natural selection and genetic mechanism. It gradually approaches the optimal solution by iteratively selecting, crossover and mutating sample scheduling schemes. The genetic algorithm screens and evolves according to the node scheduling objective function value of the sample scheduling scheme, and finally determines the optimal target node scheduling scheme.

[0069] The interactive node is used to receive the registration instruction initiated by the user, and the data processing node corresponding to the interactive node is used to generate registration recommendation information based on the registration instruction, and make an appointment for registration based on the user's feedback on the registration recommendation information.

[0070] Specifically include:

[0071] Based on the registration instruction, read the user's identity information, such as medical insurance card number, ID card number, etc.;

[0072] Based on the user's identity information, the user's electronic medical record is obtained from the cloud platform. Among them, EMR (Electronic Medical Record), also known as a computerized medical record system or a computer-based patient record (CPR), refers to the medical records that can be stored, managed, transmitted and reproduced by medical personnel using information systems to generate digital information such as text, symbols, charts, graphics, numbers, and images during medical activities. It is a form of medical record, including outpatient (emergency) medical records and inpatient medical records;

[0073] Through natural language processing, based on the disease characteristic knowledge graph and the user's electronic medical record, the user's medical consultation information is obtained. The construction of the disease characteristic knowledge graph is based on a graph data structure, which consists of nodes (representing entities) and edges (representing relationships). In the disease characteristic knowledge graph, nodes can represent diseases and symptoms. The disease characteristic knowledge graph can systematically record a variety of diseases and their corresponding symptoms. For example, for allergic rhinitis, the graph can record its typical symptoms such as paroxysmal sneezing, clear watery nasal discharge, nasal congestion and nasal itching, as well as symptoms such as hyposmia that may occur in some patients. Similarly, for hyperthyroidism, the graph can list in detail its facial features (such as shocked face, widened palpebral fissures, bulging eyeballs, etc.) and psychological symptoms such as excitement, restlessness and irritability;

[0074] Generate registration recommendation information based on the user's medical consultation information.

[0075] In some embodiments, the data processing node corresponding to the interactive node obtains the user's consultation information through natural language processing based on the disease feature knowledge graph and the user's electronic medical record, including:

[0076] Access electronic medical records of multiple historical users;

[0077] Based on the electronic medical records of multiple historical users, determine the onset correlation parameters of any two diseases. Specifically, this can be achieved through statistical methods, machine learning algorithms or deep learning models. For example, through analysis, it can be found that the appearance of a certain symptom or disease A often indicates an increased possibility of another disease B, thereby calculating the onset correlation parameters between A and B;

[0078] Obtain the user's condition description information. Specifically, the user can enter his or her condition description through the interactive node, which usually includes symptoms, duration, severity, etc. This information is an important basis for the subsequent generation of medical questions and disease ranking;

[0079] Based on the user's condition description information, disease feature knowledge graph and the incidence correlation parameters of any two diseases, multiple diseases are ranked and disease ranking results are generated;

[0080] Through natural language processing, based on the disease feature knowledge graph and disease ranking results, we generate medical questions and obtain user feedback on the questions. The questions are intended to further clarify the user's condition or verify previous disease assumptions. Questions may involve symptom details, lifestyle habits, family medical history, etc. User feedback on these questions will be used to further refine the disease ranking or adjust the diagnosis direction.

[0081] Specifically, the disease identification model can be used to determine the possible disease of the user based on the user's condition description information and the disease feature knowledge graph, wherein the disease identification model can be a convolutional neural network model. According to the onset association parameters of any two diseases, the associated diseases that the user may have are determined, and the associated diseases that the user may have are sorted according to the onset association parameters.

[0082] In the process of generating medical questions, questions are asked based on the symptoms of diseases that the user may have as recorded in the disease feature knowledge graph first, and then questions are asked based on the sorting results.

[0083] The interactive node is used to receive file query instructions initiated by users, and the data processing node corresponding to the interactive node is used to obtain target medical files from the cloud platform based on the file query instructions, and send the target medical files to the interactive node.

[0084] Specifically include:

[0085] Read the user's identity information based on the file query instruction;

[0086] Based on the user's identity information, keyword information of multiple medical files (e.g., blood test, urine analysis, imaging examination, etc.) of the user is obtained from the cloud platform;

[0087] Generate document selection interaction information based on keyword information of multiple medical documents of the user;

[0088] Obtain user feedback on file selection interaction information through interactive nodes;

[0089] Based on the user's feedback on the file selection interaction information, the encrypted target medical file is obtained from the cloud platform;

[0090] The encrypted target medical file is decrypted to generate the target medical file.

[0091] As an example only, a cloud platform is a server cluster that stores a large amount of medical data, which contains multiple medical files of a user. Based on the user's identity information, keyword information of all medical files related to the user is retrieved from the cloud platform. These keyword information may be the name, type, creation date or content summary of the file, etc., which are used to help users quickly identify the files they need. Based on the keyword information obtained from the cloud platform, a file selection interaction information is generated. This interaction information is usually presented to the user in the form of a list, grid or tree structure, so that the user can intuitively see which medical files they have available. There may be some additional information next to each file, such as the size, creator or modification date of the file, to help the user make a selection. The user will view the file selection interaction information and select one or more medical files according to their needs. Their selection will be fed back through an interactive node (such as a touch screen, keyboard, mouse or voice input device). This feedback is a direct reflection of the user's intention, and it can be used to determine which medical files the user wants to view. Based on the user's feedback, the corresponding medical files are obtained from the cloud platform. However, due to the sensitivity of medical files, these files are usually stored encrypted on the cloud platform. Therefore, the data processing node corresponding to the interactive node will obtain the encrypted target medical file. The purpose of this step is to ensure that even if the data is intercepted during the file transmission process, it cannot be read by unauthorized persons. Using the appropriate decryption algorithm and key, the encrypted target medical file is decrypted. The decryption process is the process of converting the encrypted data back to its original form. Once the decryption is successful, the data processing node corresponding to the interactive node will generate a readable version of the target medical file and display it to the user or store it on the user's device.

[0092] The interactive node is used to receive information query instructions initiated by users, and the data processing node corresponding to the interactive node is used to obtain target medical information from the cloud platform based on the information query instructions, and send the target medical information to the interactive node.

[0093] Specifically include:

[0094] Based on the information query instruction, determine the information search keywords;

[0095] Based on information search keywords, target medical information is obtained from the cloud platform.

[0096] As an example only, the user enters an information query instruction through the interactive node, such as "query the cause and treatment of diabetes". After receiving the instruction, the data processing node will first parse the instruction and extract key information, that is, the information search keyword. In this example, the keywords may be "diabetes", "cause" and "treatment method". Based on the extracted keywords, the data processing node will construct a query request for the cloud platform. This request usually contains information such as keywords, query scope (such as medical field), sorting requirements, etc. The data processing node sends the query request to the cloud platform. The cloud platform will search for relevant information in its huge medical information database based on the keywords and scope in the request. The search results may include various types of medical information such as medical papers, clinical guidelines, and expert advice. After the cloud platform returns the search results, the data processing node will filter and organize the information. It will sort and filter the information based on factors such as authority, relevance, and timeliness to ensure that the user obtains the most accurate and useful information. The target medical information that has been filtered and organized will be sent back to the interactive node. The interactive node will display this information in a user-friendly manner, such as in the form of a list, paragraph, or chart. The user views the acquired target medical information through the interactive node. If users have questions about certain information or need further explanation, they can initiate query instructions again through the interactive node to interact.

[0097] In some embodiments, the data processing node corresponding to the interaction node is further used to:

[0098] Obtaining the reaction time characteristics of multiple sample users to basic interaction information;

[0099] Obtain the user's reaction time characteristics to basic interactive information;

[0100] Based on the reaction time characteristics of multiple sample users to the basic interaction information and the reaction time characteristics of the users to the basic interaction information, the display font size of the interaction node is adjusted.

[0101] Specifically, the design of the interaction node not only focuses on the accurate transmission of information, but also strives to improve the user experience and ensure that information can be presented to users in the most efficient and comfortable way. To achieve this goal, the data processing node corresponding to the interaction node adopts a dynamic adjustment mechanism based on the user's reaction time characteristics to optimize the display settings of the interaction node, such as adjusting the display font size. A certain number of sample users (these users may represent a wide range of characteristics such as age, gender, and vision conditions) will be selected and presented with a series of basic interaction information. This basic interaction information may include simple text prompts, button labels, menu options, etc. The reaction time of each sample user to this information is recorded, that is, the time required from the information presentation to the user's response (such as click, reading completion, etc.). Through statistical analysis, the average reaction time of the sample users to the basic interaction information, the reaction time distribution and other characteristics can be obtained. When the user starts using the system, the interaction node also records the user's reaction time characteristics to the basic interaction information. These characteristics will be compared with the previously obtained sample user data to evaluate whether the target user's reaction speed is within the normal range or whether there are certain specific needs or restrictions. With the above two sets of data, the data processing node corresponding to the interactive node can start to adjust the display settings of the interactive node. Specifically, if the target user's reaction time is significantly longer than the average level of the sample users, this may mean that the user has difficulty reading or identifying information. In order to improve the user experience, the data processing node corresponding to the interactive node can automatically increase the display font size in the interactive node to make the information clearer and easier to read. On the contrary, if the target user's reaction time is short, the data processing node corresponding to the interactive node can consider reducing the font size to save screen space and provide a more compact information layout.

[0102] An AI-based medical self-service terminal can also be used to achieve other functions. For example, an AI-based medical self-service terminal supports multiple payment methods, including cash, bank cards, mobile payments (such as WeChat, Alipay), etc. Patients can complete the payment of various fees such as registration fees, examination fees, and drug fees on the terminal. During the payment process, patients can browse the fee details to ensure transparent consumption. For another example, an AI-based medical self-service terminal can provide self-service card issuance services for patients who visit the doctor for the first time. Patients only need to enter personal basic information, such as name, ID number, etc., and the device can automatically issue a medical card. For another example, an AI-based medical self-service terminal can measure basic vital signs such as blood pressure, blood sugar, body temperature, height, and weight. The test results can be automatically uploaded to the hospital information system for doctors' reference. For another example, patients can handle medical insurance services such as medical insurance inquiries, medical insurance reimbursement, and medical insurance payments on the terminal. Supports multiple identity authentication methods such as medical insurance cards and ID cards.

[0103] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. An AI-based medical self-service terminal, characterized in that: It includes multiple interactive nodes and multiple data processing nodes, wherein the interactive nodes are used to interact with users, and the multiple data processing nodes interact with the cloud platform for data, and the cloud platform is used to allocate corresponding data processing nodes to the interactive nodes based on allocation auxiliary information, wherein the allocation auxiliary information at least includes the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple current time periods: The interactive node is used to receive a registration instruction initiated by a user, and the data processing node corresponding to the interactive node is used to generate registration recommendation information based on the registration instruction, and make an appointment for registration based on the user's feedback on the registration recommendation information; The interactive node is used to receive a file query instruction initiated by a user, and the data processing node corresponding to the interactive node is used to obtain a target medical file from the cloud platform based on the file query instruction, and send the target medical file to the interactive node; The interactive node is used to receive information query instructions initiated by users, and the data processing node corresponding to the interactive node is used to obtain target medical information from the cloud platform based on the information query instructions, and send the target medical information to the interactive node.

2. The AI-based medical self-service terminal according to claim 1, characterized in that: The cloud platform is used to allocate corresponding data processing nodes to the interactive nodes based on the allocation auxiliary information, including: Obtain the number of registration instructions, file query instructions, and information query instructions of the interactive node in multiple historical time periods; According to the number of registration instructions, the number of file query instructions and the number of information query instructions of any two interactive nodes in multiple historical time periods, the registration instruction number correlation coefficient, the file query instruction number correlation coefficient and the information query instruction number correlation coefficient of any two interactive nodes are calculated; For each interactive node, based on the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient and the information query instruction quantity correlation coefficient of any two interactive nodes, the positively correlated interactive node and the negatively correlated interactive node of the interactive node are determined; Obtain the number of registration instructions, file query instructions, and information query instructions of the interactive node in multiple current time periods; Based on the positively associated interactive nodes and negatively associated interactive nodes of each interactive node and the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple current time periods, the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple future time periods are predicted; Based on the number of registration instructions, the number of file query instructions and the number of information query instructions of each interactive node in multiple future time periods, a corresponding data processing node is allocated to the interactive node.

3. The AI-based medical self-service terminal according to claim 2, characterized in that: The cloud platform determines the positively associated interactive nodes and the negatively associated interactive nodes of the interactive nodes based on the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient and the information query instruction quantity correlation coefficient of any two interactive nodes, including: The weighted sum of the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient and the information query instruction quantity correlation coefficient of any two interactive nodes is calculated to calculate the comprehensive correlation coefficient of any two interactive nodes; For each interactive node, determine whether there is an interactive node whose at least one of the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient or the information query instruction quantity correlation coefficient is greater than the first correlation coefficient positive threshold, and whose registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient and the information query instruction quantity correlation coefficient are all greater than the second correlation coefficient positive threshold. If so, serve as a positively associated interactive node of the interactive node. Determine whether there is an interactive node whose comprehensive correlation coefficient is greater than the first comprehensive correlation coefficient threshold. If so, serve as a positively associated interactive node of the interactive node, wherein the first correlation coefficient positive threshold, the second correlation coefficient positive threshold and the first comprehensive correlation coefficient threshold are all greater than 0, and the first correlation coefficient positive threshold is greater than the second correlation coefficient positive threshold. For each interactive node, determine whether there is an interactive node whose at least one of the registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient or the information query instruction quantity correlation coefficient is less than the first correlation coefficient negative threshold, and whose registration instruction quantity correlation coefficient, the file query instruction quantity correlation coefficient and the information query instruction quantity correlation coefficient are all less than the second correlation coefficient negative threshold. If so, act as a negatively associated interactive node of the interactive node. Determine whether there is an interactive node whose comprehensive correlation coefficient is less than the second comprehensive correlation coefficient threshold. If so, act as a negatively associated interactive node of the interactive node, wherein the first correlation coefficient negative threshold, the second correlation coefficient negative threshold and the second comprehensive correlation coefficient threshold are all less than 0, and the second correlation coefficient negative threshold is greater than the first correlation coefficient negative threshold.

4. The AI-based medical self-service terminal according to claim 3, characterized in that: The cloud platform predicts the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple future time periods based on the positively associated interactive nodes and the negatively associated interactive nodes of each interactive node and the number of registration instructions, the number of file query instructions, and the number of information query instructions of each interactive node in multiple current time periods, including: For each interaction node, based on the positively associated interaction nodes and negatively associated interaction nodes of the interaction node, a quantity prediction model and a quantity correction model corresponding to the interaction node are established; For each interactive node, using the quantity prediction model corresponding to the interactive node, based on the number of registration instructions, the number of file query instructions, and the number of information query instructions of the interactive node in multiple current time periods, and the number of registration instructions, the number of file query instructions, and the number of information query instructions of the positively associated interactive nodes and the negatively associated interactive nodes of the interactive node in multiple current time periods, predict the number of registration instructions, the number of file query instructions, and the number of information query instructions of the interactive node in multiple future time periods; Through the quantity correction model corresponding to each interactive node, based on the number of registration instructions, file query instructions and information query instructions of the positively associated interactive nodes and negatively associated interactive nodes of each interactive node in multiple future time periods, the number of registration instructions, file query instructions and information query instructions of the interactive node in multiple future time periods are iteratively corrected to generate the corrected number of registration instructions, file query instructions and information query instructions for each interactive node in multiple future time periods.

5. The AI-based medical self-service terminal according to claim 4, characterized in that: The cloud platform allocates corresponding data processing nodes to the interactive nodes based on the number of registration instructions, file query instructions, and information query instructions of each interactive node in multiple future time periods, including: Establish node scheduling objective function; Generate multiple sample node scheduling schemes; Calculate the node scheduling objective function value of the sample node scheduling solution based on the node scheduling objective function, the number of registration instructions, the number of file query instructions and the number of information query instructions of each interactive node in multiple future time periods, and the correlation coefficient of the number of registration instructions, the correlation coefficient of the number of file query instructions and the correlation coefficient of the number of information query instructions between any two interactive nodes; Determine the target node scheduling scheme based on the node scheduling objective function value of each sample node scheduling scheme through genetic algorithm; Based on the target node scheduling scheme, corresponding data processing nodes are allocated to the interactive nodes.

6. An AI-based medical self-service terminal according to any one of claims 1 to 5, characterized in that: The data processing node corresponding to the interactive node generates registration recommendation information based on the registration instruction, including: Based on the registration instruction, read the user's identity information; Based on the user's identity information, obtain the user's electronic medical record from the cloud platform; Through natural language processing, based on the disease feature knowledge graph and the user's electronic medical record, the user's consultation information is obtained; Generate registration recommendation information based on the user's medical consultation information.

7. The AI-based medical self-service terminal according to claim 6, characterized in that: The data processing node corresponding to the interaction node obtains the user's consultation information through natural language processing based on the disease feature knowledge graph and the user's electronic medical record, including: Access electronic medical records of multiple historical users; Determine the incidence correlation parameters of any two diseases based on the electronic medical records of multiple historical users; Get the user's medical condition description information; Based on the user's condition description information, disease feature knowledge graph and the incidence correlation parameters of any two diseases, multiple diseases are ranked and disease ranking results are generated; Through natural language processing, based on the disease feature knowledge graph and disease ranking results, medical questions are generated to obtain user feedback on the medical questions.

8. The AI-based medical self-service terminal according to claim 6, characterized in that: The data processing node corresponding to the interaction node is also used for: Obtaining the reaction time characteristics of multiple sample users to basic interaction information; Obtain the user's reaction time characteristics to basic interactive information; Based on the reaction time characteristics of multiple sample users to the basic interaction information and the reaction time characteristics of the users to the basic interaction information, the display font size of the interaction node is adjusted.

9. An AI-based medical self-service terminal according to any one of claims 1 to 5, characterized in that: The data processing node corresponding to the interactive node obtains the target medical file from the cloud platform based on the file query instruction, including: Read the user's identity information based on the file query instruction; Based on the user's identity information, keyword information of multiple medical files of the user is obtained from the cloud platform; Generate document selection interaction information based on keyword information of multiple medical documents of the user; Obtain user feedback on file selection interaction information through interactive nodes; Based on the user's feedback on the file selection interaction information, the encrypted target medical file is obtained from the cloud platform; The encrypted target medical file is decrypted to generate the target medical file.

10. An AI-based medical self-service terminal according to any one of claims 1 to 5, characterized in that: The data processing node corresponding to the interactive node obtains target medical information from the cloud platform based on the information query instruction, including: Based on the information query instruction, determine the information search keywords; Based on information search keywords, target medical information is obtained from the cloud platform.

Images