Emergency patient triage system based on artificial intelligence

Abstract

The invention discloses an emergency patient triage system based on artificial intelligence, which belongs to the technical field of medical treatment and comprises a hospital terminal, a patient terminal and a triage server, according to the system, emergency triage is carried out on hospital resources of the whole city, and a patient can quickly and accurately arrive at a proper hospital emergency treatment department for diagnosis and treatment.

Description

Technical Field

[0001] The present invention belongs to the field of medical technology, and in particular relates to an emergency patient triage system based on artificial intelligence. Background Art

[0002] Emergency patient triage is an important part of the medical service system. It refers to the rapid assessment and classification of patients according to the severity and urgency of their condition when they visit the emergency department, so as to reasonably arrange the order of patients' visits and medical resources, and ensure that critically ill patients can receive timely and effective treatment.

[0003] Chinese patent application No. 202410925172.3 discloses a method and system for hierarchical triage of emergency critically ill patients, comprising: obtaining multidimensional physical indicator data, historical clinical consultation records, and corresponding hierarchical triage situations of emergency patients during treatment from a historical treatment database established by a hospital; obtaining on-site clinical consultation records obtained through on-site clinical consultation of patients through cluster analysis of the correlation between the historical clinical consultation records and the multidimensional physical indicator data; obtaining next-step recommendation of medical questions corresponding to the multidimensional physical indicator data associated with the on-site clinical consultation records based on the correlation between the obtained on-site clinical consultation records and the multidimensional physical indicator data; until, based on the on-site clinical consultation records obtained through the medical questions, it is confirmed that multidimensional physical indicator data with a similarity that meets a preset value exists in the historical treatment database, the on-site clinical consultation is terminated, the obtained multidimensional physical indicator data with a similarity that meets the preset value is quantified, and the patients are hierarchically triaged based on the severity assessment.

[0004] The above technology has the following problems: the above-mentioned triage method for emergency critical patients is aimed at triage within a hospital, rather than emergency triage for hospitals in the entire city. It has a small coverage area and it is difficult for patients to quickly and accurately reach the appropriate hospital emergency department for diagnosis and treatment.

[0005] In view of this, an artificial intelligence-based emergency patient triage system is designed to solve the above problems. Summary of the Invention

[0006] To solve the problems raised in the above background technology, the present invention provides an artificial intelligence-based emergency patient triage system, which has the characteristics of triaging emergency cases based on the hospital resources of the entire city and enabling patients to quickly and accurately reach the appropriate hospital emergency department for diagnosis and treatment.

[0007] To achieve the above objectives, the present invention provides the following technical solutions: an artificial intelligence-based emergency patient triage system, comprising:

[0008] Hospital terminals: register and log in to the system, upload the hospital address, patient history data, and real-time emergency data. Historical data includes patient ID numbers, contact information, home addresses, and visit process information. Real-time emergency data includes the number of emergency patients in each emergency department.

[0009] Patient terminals register and log into the system, upload patient symptoms and travel mode data, and receive data on the hospital and emergency department of the hospital after the system triages the patient;

[0010] The triage server receives and saves the registration information of the hospital terminal and the patient terminal, performs terminal login verification based on the registration information, and after successful verification, the terminal enters the system to upload or receive data, saves the data uploaded by the hospital terminal, extracts the key information of the patient's symptom data uploaded by the patient terminal, and first judges the patient's emergency department based on the similarity with the saved patient treatment process information. When the similarity is lower than the preset threshold, the patient's emergency department is judged based on the constructed and trained patient symptom emergency department judgment model, based on the saved patient home and hospital address, as well as the patient's The time for patients to arrive at the emergency department of each hospital is judged based on the travel mode uploaded by the terminal. The number of patients visiting the emergency department of each hospital at that time is predicted based on the constructed and trained emergency number prediction model. Then, the number is multiplied by the preset single visit time to judge the waiting time for patients to arrive at the emergency department of each hospital. The sum of the time for patients to arrive at the emergency department of each hospital and the waiting time for the emergency department of each hospital is used as the judgment of the patient's visit time. The hospital with the shortest visit time and the emergency department of the hospital are recommended to the saved patient contact information.

[0011] Furthermore, the hospital terminal includes:

[0012] Hospital registration module, hospital registration system account, including registration number and password, where the registration number is the hospital code;

[0013] Hospital login module: the hospital logs in based on the registration number and password, and enters the system after successful verification;

[0014] Hospital address upload module, hospital upload address;

[0015] Patient historical medical data upload module. Before uploading each patient's historical medical data, the hospital must sign a patient historical medical data sharing agreement. Only after signing can the hospital upload each patient's historical medical data. The historical medical data includes the patient's ID number, contact information, home address, and medical process information;

[0016] The hospital's real-time emergency data upload module requires the hospital to sign a real-time emergency data sharing agreement before uploading real-time emergency data. Only after signing can the hospital upload real-time emergency data. The real-time emergency data includes the number of emergency patients in each emergency department.

[0017] Furthermore, the patient terminal includes:

[0018] Patient registration module, patient registration system account, including registration number and password, where the registration number is the ID number;

[0019] Patient login module: patients log in based on their registration number and password, and enter the system after successful verification;

[0020] Patient symptom upload module, where patients describe and upload symptoms that require emergency treatment, including one or more of text descriptions, image descriptions, and video descriptions;

[0021] Patient travel mode upload module, where patients select and upload their travel mode according to their needs, including walking, cycling, public transportation or driving;

[0022] The emergency treatment information receiving module receives the hospital and the emergency treatment department of the hospital after the system triage.

[0023] Furthermore, the triage server includes:

[0024] Registration information saving module, which saves the registration information of the hospital registration module and the patient registration module;

[0025] The login verification module verifies whether the login information of the hospital login module and the patient login module is correct based on the saved hospital and patient registration information. If the verification is correct, the verification is successful;

[0026] Hospital address data storage module, which stores the addresses of each hospital uploaded by the hospital address upload module;

[0027] Patient history medical data integration module, which integrates the patient history medical data uploaded by the patient history medical data upload module;

[0028] The patient history medical treatment data storage module saves the patient history medical treatment information integrated by the patient history medical treatment data integration module;

[0029] The hospital real-time emergency data storage module stores the hospital real-time emergency data uploaded by the hospital real-time emergency data upload module;

[0030] The patient symptom medical information judgment module builds and trains a patient symptom emergency room judgment model, extracts key information describing symptoms from the patient symptom upload module, and compares it with the patient medical process information stored in the patient historical medical data storage module. If the similarity between the two is lower than the preset threshold, the patient's emergency room is judged based on the patient symptom emergency room judgment model. If the similarity between the two is higher than the preset threshold, the patient's emergency room is judged based on the medical process information.

[0031] A patient consultation duration judgment module is used to construct and train an emergency department number prediction model. The module takes the patient's home address saved in the patient history consultation data storage module as the starting point and the addresses of each hospital saved in the hospital address data storage module as the end point. The module uses the travel mode described in the patient travel mode upload module to calculate the time it takes for the patient to arrive at the emergency department of each hospital. The module predicts the number of patients visiting the emergency department of each hospital at that time based on the emergency department number prediction model. The module multiplies the predicted number of patients visiting the emergency department of each hospital by the preset visit time of a single patient to calculate the waiting time for the patient to arrive at the emergency department of each hospital. The sum of the time the patient arrives at the emergency department of each hospital and the waiting time for the visit to the emergency department of each hospital is used as the determined patient consultation duration.

[0032] The patient treatment recommendation module recommends the hospital with the shortest patient treatment time as the patient's emergency treatment hospital based on the patient treatment time of each hospital's emergency treatment department determined by the patient treatment time determination module, and sends the emergency treatment hospital and the emergency treatment department of the hospital to the patient contact information saved by the patient history treatment data storage module.

[0033] Furthermore, the specific steps of the integration of the patient history medical data integration module include: classifying the patient history medical data according to the ID number saved by the patient history medical data upload module, summarizing the patient history medical data with the same ID number, classifying the summarized patient history medical data with the same ID number according to the ID number, contact information, home address and medical process information, retaining only one data for each of the ID number, contact information and home address, and calculating the similarity of the medical process information at the same time. If the similarity is higher than a preset threshold, the medical process information of the latest date is retained. If the similarity is lower than the preset threshold, both medical process information are retained at the same time, and the integration is completed. The calculation expression of the similarity is:

[0034]

[0035] Where: A=(a1,a2,…,a n ) is represented as the bag-of-words vector of the medical process information A, B=(b1,b2,…,b n) represents the bag-of-words vector of the medical process information B, n represents the vocabulary size, a i It is represented by the frequency of the i-th word in the medical process information A, b i It is represented by the frequency of the i-th word appearing in the medical process information B.

[0036] Furthermore, the specific steps of the patient symptom medical information judgment module extracting key information describing symptoms from the patient symptom upload module include:

[0037] If the symptom description data uploaded by the patient is in text format, the key information of the symptom description data uploaded by the patient is extracted based on the TF-IDF algorithm;

[0038] If the symptom description data uploaded by the patient is in image format, the key information of the symptom description data uploaded by the patient is extracted based on the SIFT algorithm;

[0039] If the symptom description data uploaded by the patient is in video format, the key frame extraction algorithm based on optical flow is used to extract key information of the symptom description data uploaded by the patient.

[0040] Furthermore, the specific steps of the patient symptom medical information judgment module for performing similarity comparison with the patient medical process information stored in the patient historical medical data storage module are the same as the specific steps of the patient medical process information similarity calculation in the patient historical medical data integration module.

[0041] Furthermore, the specific steps of the patient symptom consultation information judgment module for judging the patient's emergency consultation clinic based on the patient symptom emergency consultation clinic judgment model include:

[0042] The patient symptom emergency room visit judgment model includes a text judgment sub-model, an image judgment sub-model, a video judgment sub-model, and a comprehensive judgment model;

[0043] The symptom description data uploaded by the patient terminal is input into each sub-model according to the corresponding data format. The sub-model makes an emergency room judgment on the symptom description data uploaded by the patient. If the patient only inputs symptom description data in one format, the judgment result of the sub-model is directly output as the emergency room judgment result. If the patient inputs symptom description data in two or three formats, the sub-model judgment result is output to the comprehensive judgment model. The comprehensive judgment model makes an emergency room judgment according to the preset weights of each sub-model, and the judgment result of the comprehensive model is used as the emergency room judgment result.

[0044] Furthermore, the emergency number prediction model of the patient consultation duration judgment module is an LSTM model.

[0045] Furthermore, the patient consultation duration determination module calculates the time it takes for patients to arrive at the emergency departments of various hospitals based on the Dijkstra algorithm.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] 1. The present invention consists of a hospital terminal, a patient terminal and a triage server. The hospital terminal uploads the hospital address, the patient's historical medical data and the hospital's real-time emergency data. The patient's historical medical data includes the patient's ID number, contact information, home address and medical process information. The hospital's real-time emergency data includes the number of emergency patients in each emergency department. The patient terminal uploads symptom descriptions and travel mode descriptions. The triage server first determines the patient's emergency department based on the similarity between the symptom description uploaded by the patient terminal and the patient's historical medical data uploaded by the hospital terminal and the constructed patient symptom emergency department judgment model, and then determines the patient's emergency department based on the hospital terminal. The uploaded hospital address, patient home address, and travel mode uploaded by the patient terminal are used to predict the time it takes for the patient to arrive at the emergency department of each hospital. The number of patients visiting the emergency department of each hospital at that time is then predicted based on the constructed emergency number prediction model. The waiting time for patients to arrive at the emergency department of each hospital is then predicted. Finally, the hospital with the shortest sum of the time it takes for patients to arrive at the emergency department of each hospital and the waiting time for patients to arrive at the emergency department of each hospital is used as the triage recommendation result. The system performs emergency triage on the hospital resources of the entire city, and can enable patients to quickly and accurately reach the appropriate hospital emergency department for diagnosis and treatment.

[0048] 2. In the process of the triage server of the present invention judging the patient's emergency treatment department, it first makes a judgment based on the similarity between the symptom description uploaded by the patient terminal and the patient's historical treatment data uploaded by the hospital terminal. After the similarity is lower than the preset threshold, it makes a judgment based on the constructed patient symptom emergency treatment department judgment model. Compared with the model judgment, the similarity judgment has less data processing volume and faster judgment speed. Compared with the similarity judgment, the model has a large data processing volume and slow judgment speed. The combination of the two can not only ensure the judgment accuracy but also improve the system's judgment speed as much as possible, that is, it can improve the system's triage speed.

[0049] 3. In the process of judging the visiting time of each hospital's emergency department, the triage server of the present invention comprehensively considers the time when the patient arrives at the emergency department of each hospital and the waiting time when the patient arrives at the emergency department of each hospital. Compared with simply considering the waiting time when the patient arrives at the emergency department of each hospital, the triage accuracy of the system can be improved.

[0050] 4. When the triage server of the present invention determines the patient's emergency department and the length of time the patient visits the emergency department of each hospital, the patient terminal only needs to upload a description of symptoms and travel mode. Other data are directly retrieved from the triage server, which can reduce the amount of data input by the patient and speed up the triage processing speed of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is the overall framework diagram of the system of the present invention;

[0052] Figure 2 This is the overall framework diagram of the hospital terminal of the present invention;

[0053] Figure 3 This is the overall framework diagram of the patient terminal of the present invention;

[0054] Figure 4 This is the overall framework diagram of the triage server of the present invention;

[0055] Figure 5 This is a framework diagram of the association between the hospital terminal, patient terminal and triage server of the present invention;

[0056] In the figure: 1. Hospital terminal; 101. Hospital registration module; 102. Hospital login module; 103. Hospital address upload module; 104. Patient history medical data upload module; 105. Hospital real-time emergency data upload module;

[0057] 2. Patient terminal; 201. Patient registration module; 202. Patient login module; 203. Patient symptom upload module; 204. Patient travel mode upload module; 205. Emergency treatment information receiving module;

[0058] 3. Triage server; 301. Registration information storage module; 302. Login verification module; 303. Hospital address data storage module; 304. Patient history medical data integration module; 305. Patient history medical data storage module; 306. Hospital real-time emergency data storage module; 307. Patient symptom medical information judgment module; 308. Patient medical treatment duration judgment module; 309. Patient medical treatment recommendation module. DETAILED DESCRIPTION

[0059] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments 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 are within the scope of protection of the present invention.

[0060] The present invention provides the following technical solutions: Figure 1, an artificial intelligence-based emergency patient triage system, comprising:

[0061] Hospital terminal 1 registers and logs into the system, and uploads the hospital address, patient history data, and real-time emergency data. Historical data includes patient ID numbers, contact information, home addresses, and visit information. Real-time emergency data includes the number of emergency patients in each emergency department.

[0062] Patient Terminal 2 registers and logs into the system, uploads patient symptoms and travel mode data, and receives data on the hospital and emergency department of the hospital triaged by the system;

[0063] The triage server 3 receives and saves the registration information of the hospital terminal 1 and the patient terminal 2, performs terminal login verification based on the registration information, and after successful verification, the terminal enters the system to upload or receive data, saves the data uploaded by the hospital terminal 1, extracts the key information of the patient symptom data uploaded by the patient terminal 2, and first judges the patient's emergency department based on the similarity between the saved patient treatment process information and the saved patient treatment process information. When the similarity is lower than the preset threshold, the patient's emergency department is judged based on the constructed and trained patient symptom emergency department judgment model, and based on the saved patient home and hospital addresses, The time for patients to arrive at the emergency department of each hospital is determined based on the travel mode uploaded by the patient terminal 2. The number of patients visiting the emergency department of each hospital at that time is predicted based on the constructed and trained emergency number prediction model. Then, the number is multiplied by the preset single visit time to determine the waiting time for patients to arrive at the emergency department of each hospital. The sum of the time for patients to arrive at the emergency department of each hospital and the waiting time for the emergency department of each hospital is used as the judgment of the patient's visit time. The hospital with the shortest patient visit time and the emergency department of the hospital are recommended to the saved patient contact information.

[0064] Specifically, see the attached Figure 2 , hospital terminal 1 includes:

[0065] Hospital registration module 101, hospital registration system account, including registration number and password, where the registration number is the hospital code;

[0066] Hospital code refers to a set of codes compiled by relevant departments according to certain rules to facilitate hospital management and identification. It is unique and can effectively distinguish hospitals.

[0067] Hospital login module 102, the hospital logs in based on the registration number and password during registration, and enters the system after successful verification;

[0068] Hospital address uploading module 103, hospital upload address;

[0069] Patient history medical data uploading module 104: Before uploading each patient's history medical data, the hospital signs a patient history medical data sharing agreement. Only after signing can the hospital upload each patient's history medical data. The history medical data includes the patient's ID number, contact information, home address, and medical process information;

[0070] The hospital real-time emergency data uploading module 105, before uploading real-time emergency data, the hospital signs a real-time emergency data sharing agreement. Only after signing can the hospital upload real-time emergency data, wherein the real-time emergency data includes the number of emergency patients in each emergency department.

[0071] Specifically, the patient terminal 2 includes:

[0072] Patient registration module 201, patient registration system account, including registration number and password, where the registration number is the ID number;

[0073] The ID number is the abbreviation of the Citizen Identity Number of the People's Republic of China. It is a unique and lifelong identity code for every citizen. It is compiled by the public security organs in accordance with the national standard for citizen identity numbers. The ID number is used as the registration number to accurately link to the patient's relevant information in subsequent steps;

[0074] Patient login module 202, where the patient logs in based on the registration number and password used during registration, and enters the system after successful verification;

[0075] Patient symptom uploading module 203, where the patient describes and uploads symptoms that require emergency treatment, including one or more of text description, image description, and video description;

[0076] Patient travel mode upload module 204, where patients select and upload their travel mode according to their needs, including walking, cycling, public transportation, or driving;

[0077] The emergency treatment information receiving module 205 receives the hospital and the emergency treatment department of the hospital after the system triage.

[0078] Specifically, the triage server 3 includes:

[0079] The registration information storage module 301 stores the registration information of the hospital registration module 101 and the patient registration module 201;

[0080] The login verification module 302 verifies whether the login information of the hospital login module 102 and the patient login module 202 is correct based on the stored hospital and patient registration information. If the verification is correct, the verification is successful.

[0081] The hospital address data storage module 303 stores the addresses of the hospitals uploaded by the hospital address upload module 103;

[0082] The patient history medical data integration module 304 integrates the patient history medical data uploaded by the patient history medical data uploading module 104;

[0083] The patient's historical medical treatment data is classified by the ID number saved by the patient's historical medical treatment data upload module 104, and the historical medical treatment data of patients with the same ID number are summarized. The summarized historical medical treatment data of patients with the same ID number are classified by the ID number, contact information, home address and medical treatment process information. Only one data of the ID number, contact information and home address is retained. At the same time, the similarity of the medical treatment process information is calculated. If the similarity is higher than the preset threshold, the medical treatment process information of the latest date is retained. If the similarity is lower than the preset threshold, both medical treatment process information are retained and the integration is completed. The calculation expression of the similarity is:

[0084]

[0085] Where: A=(a1,a2,…,a n ) is represented as the bag-of-words vector of the medical process information A, B=(b1,b2,…,b n ) represents the bag-of-words vector of the medical process information B, n represents the vocabulary size, a i It is represented by the frequency of the i-th word in the medical process information A, b i It is represented by the frequency of the i-th word in the medical process information B;

[0086] The patient history medical treatment data storage module 305 stores the patient history medical treatment information integrated by the patient history medical treatment data integration module 304;

[0087] The hospital real-time emergency data storage module 306 stores the hospital real-time emergency data uploaded by the hospital real-time emergency data upload module 105;

[0088] The patient symptom medical information judgment module 307 constructs and trains a patient symptom emergency room judgment model, extracts key information describing the symptoms from the patient symptom upload module 203, and compares it with the patient medical process information stored in the patient historical medical data storage module 305 for similarity. If the similarity between the two is lower than a preset threshold, the patient's emergency room is judged based on the patient symptom emergency room judgment model. If the similarity between the two is higher than the preset threshold, the emergency room is judged based on the medical process information.

[0089] The patient symptom emergency room visit judgment model includes a text judgment sub-model, an image judgment sub-model, a video judgment sub-model and a comprehensive judgment model;

[0090] If the symptom description data uploaded by the patient is in text format, the key information of the symptom description data uploaded by the patient is extracted based on the TF-IDF algorithm;

[0091] The TF-IDF algorithm, or term frequency-inverse document frequency algorithm, is a commonly used weighting technique for information retrieval and text mining;

[0092] TF refers to the frequency of a word in a text. It is calculated by dividing the number of times the word appears in the text by the total number of words in the text. For example, in a text containing 10 words, the word "leg pain" appears 5 times, so the word frequency of "leg pain" is 5 ÷ 10 = 0.5. The higher the word frequency, the more important the word is in the text.

[0093] IDF is an adjustment to word frequency, which is used to measure the general importance of a word in the entire text collection. It is calculated by dividing the total number of texts by the number of texts containing the word, and then taking the logarithm. For example, in a collection containing two texts (i.e., the main text and the supplementary text), the word "leg pain" appears in one document, then the inverse document frequency of "leg pain" is log(2÷1)=log(2)≈0.3. If a word appears in most texts, then its inverse document frequency will be low, indicating that the word's discrimination is not high. On the contrary, the inverse document frequency will be high, indicating that the word has a high discrimination.

[0094] When the TF-IDF algorithm is used to extract keywords from a text, words with higher TF-IDF values ​​are usually important words in the text and can be used as keywords for the text;

[0095] If the symptom description data uploaded by the patient is in image format, the key information of the symptom description data uploaded by the patient is extracted based on the SIFT algorithm;

[0096] SIFT algorithm, namely scale-invariant feature transform algorithm, is a classic algorithm for image feature extraction and matching;

[0097] First, the original image is Gaussian blurred at different scales to obtain a series of images at different scales. Then, the local extreme values ​​between adjacent scale images are detected using the differential Gaussian operator. The precise position and scale of the key points are determined by fitting a three-dimensional quadratic function. Low-contrast key points and unstable edge response points are removed. The main direction is determined by calculating the gradient direction histogram within the neighborhood of the key point. With the key point as the center, the gradient direction and amplitude are calculated within its neighborhood to construct a feature descriptor. This feature descriptor contains the gradient information of the area around the key point and can better describe the local characteristics of the key point.

[0098] If the symptom description data uploaded by the patient is in video format, the key frame extraction algorithm based on optical flow is used to extract the key information of the symptom description data uploaded by the patient;

[0099] Optical flow refers to the displacement of pixels between adjacent frames caused by the movement of objects in an image. The key frame extraction algorithm based on optical flow mainly determines which frames are key frames based on the changes in optical flow in the video.

[0100] There are many methods for calculating optical flow. Common ones include gradient-based methods, which assume that the optical flow is constant in a small neighborhood and calculate the optical flow of pixels by solving a set of linear equations. There are also energy-based methods, which calculate the optical flow by minimizing an energy function that contains an optical flow smoothness constraint. After obtaining the optical flow field, the amplitude and direction of the optical flow are usually calculated. The degree of change of the optical flow can be measured by calculating the average value, variance and other statistics of the optical flow amplitude of the entire frame image. When these statistics exceed a certain threshold, the frame is considered to be a key frame.

[0101] The calculation expression for the similarity between the uploaded symptom description and the patient's medical process information is:

[0102]

[0103] Where: A=(a1,a2,…,a n ) represents the bag-of-words vector describing the patient’s symptoms, B=(b1,b2,…,b n ) represents the bag-of-words vector of the patient's medical process information, n represents the vocabulary size, a i It is represented by the frequency of the i-th word in the patient's symptom description, b i It is represented by the frequency of the i-th word appearing in the patient's medical process information;

[0104] The symptom description data uploaded by the patient terminal 2 is input into each sub-model according to the corresponding data format. The sub-model makes an emergency room judgment on the symptom description data uploaded by the patient. If the patient only inputs symptom description data in one format, the judgment result of the sub-model is directly output as the emergency room judgment result. If the patient inputs symptom description data in two or three formats, the sub-model judgment result is output to the comprehensive judgment model. The comprehensive judgment model makes an emergency room judgment according to the preset weights of each sub-model, and the judgment result of the comprehensive model is used as the emergency room judgment result.

[0105] The patient consultation duration judgment module 308 constructs and trains an emergency department visit prediction model, using the patient's home address stored in the patient history consultation data storage module 305 as the starting point and the addresses of each hospital stored in the hospital address data storage module 303 as the end point. The patient's arrival time at the emergency department of each hospital is calculated based on the travel mode described by the patient travel mode upload module 204. The number of patients visiting the emergency department of each hospital at that time is predicted based on the emergency department visit prediction model. The predicted number of patients visiting the emergency department of each hospital is multiplied by the preset visit time of a single patient to calculate the waiting time for the patient to arrive at the emergency department of each hospital. The sum of the patient's arrival time at the emergency department of each hospital and the waiting time at the emergency department of each hospital is used as the determined patient consultation duration.

[0106] The emergency room visit prediction model is an LSTM model;

[0107] The LSTM model, or long short-term memory network, is a special recurrent neural network. LSTM cells are mainly composed of an input gate, a forget gate, an output gate, and a cell state.

[0108] Forget gate: determines how much information in the cell state at the previous moment needs to be forgotten. It receives the hidden state at the previous moment and the input at the current moment, and outputs a value between 0 and 1 through a Sigmoid function. The closer the value is to 1, the more information is retained, and the closer it is to 0, the more information is forgotten.

[0109] Input gate: determines how much information in the current input needs to be added to the cell state. It consists of two parts: one is to determine which values ​​need to be updated through the Sigmoid function, and the other is to create a new candidate value vector through the tanh function;

[0110] Cell state update: According to the output of the forget gate and the input gate, the cell state is updated by multiplying the cell state at the previous moment by the output of the forget gate, and then adding the product of the output of the input gate and the candidate value vector;

[0111] Output gate: determines how much information in the current cell state needs to be output to the hidden state. It receives the hidden state of the previous moment and the input of the current moment, outputs a value between 0 and 1 through the Sigmoid function, then processes the cell state through the tanh function, and then multiplies it with the output of the output gate to obtain the hidden state of the current moment;

[0112] When predicting the number of emergency room visits, the LSTM model learns the trends and patterns in time series data and predicts future values;

[0113] The calculation of the time when the patient arrives at the emergency department of each hospital by the patient visit duration judgment module 308 is based on the Dijkstra algorithm;

[0114] The Dijkstra algorithm is an algorithm used to calculate the shortest path or shortest duration from a given source vertex to all other vertices in a weighted directed graph or undirected graph;

[0115] Let the starting point be S, set the time from the starting point to itself as 0, that is, d[S]=0. For all other vertices v, set d[v]=∞, indicating that initially the time from the starting point to these vertices is unknown. Create two sets, the set S of vertices with the shortest time path determined, initially S = {S}, that is, only contains the starting point, and the set V-S of vertices with the shortest time path not determined, which contains all other vertices except the starting point. Find the vertex with the current shortest time in the set V-S that has not been determined, and find the vertex u with the smallest time value d[v]. This vertex is the vertex that can be reached fastest from the starting point currently. Add it to the determined set. Move the vertex u from V-S to the determined set S, indicating that the shortest time path from the starting point to the vertex u has been found. Update the time of adjacent vertices. For each vertex v adjacent to the vertex u, calculate the time t from the starting point through u to v as t = d[u]+w(u,v), where w(u,v) is the weight of the edge (u,v), representing the time required from u to v. If t < d[v], then update d[v]=t, which means that a shorter time path from the starting point to v has been found, passing through the vertex u for transfer. Repeat the steps of finding the vertex with the current shortest time, adding it to the determined set, and updating the time of adjacent vertices until the set V-S that has not been determined is empty. At this time, for each vertex in the graph, the value stored in d[v] is the time value of the shortest time path from the starting point S to this vertex, that is, the time when the patient arrives at the emergency department of each hospital;

[0116] The patient visit recommendation module 309 recommends the hospital with the shortest patient visit duration as the patient's emergency visit hospital based on the patient visit durations of the emergency departments of each hospital judged by the patient visit duration judgment module 308, and sends the emergency visit hospital and the emergency department of this hospital to the contact information of the patient saved by the patient historical visit data storage module 305.

[0117] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An artificial intelligence-based emergency patient triage system, characterized in that: include: Hospital terminal (1), register and log in to the system, upload the hospital address, patient history visit data and hospital real-time emergency data, among which the history visit data includes the patient's ID number, contact information, home address and visit process information, and the real-time emergency data includes the number of emergency patients in each emergency department; Patient terminal (2) registers and logs into the system, uploads patient symptoms and travel mode data, and receives data on the hospital and emergency department of the hospital after the system triages the patient; The triage server (3) receives and saves the registration information of the hospital terminal (1) and the patient terminal (2), performs terminal login verification based on the registration information, and after successful verification, the terminal enters the system to upload or receive data, saves the data uploaded by the hospital terminal (1), extracts the key information of the patient symptom data uploaded by the patient terminal (2), first judges the patient's emergency department based on the similarity between the saved patient treatment process information, and when the similarity is lower than the preset threshold, judges the patient's emergency department based on the constructed and trained patient symptom emergency department judgment model, based on the saved patient family and hospital The patient's address and the travel mode uploaded by the patient terminal (2) are used to determine the time when the patient arrives at the emergency department of each hospital. The number of patients in the emergency department of each hospital at that time is predicted based on the emergency number prediction model built and trained. Then, the number is multiplied by the preset single visit time to determine the waiting time for the patient to arrive at the emergency department of each hospital. The sum of the time when the patient arrives at the emergency department of each hospital and the waiting time for the visit to the emergency department of each hospital is used as the judgment of the patient's visit time. The hospital with the shortest visit time and the emergency department of the hospital are recommended to the saved patient contact information.

2. The artificial intelligence-based emergency patient triage system according to claim 1, characterized in that: The hospital terminal (1) comprises: Hospital registration module (101), hospital registration system account, including registration number and password, wherein the registration number is the hospital code; Hospital login module (102), the hospital logs in based on the registration number and password at the time of registration, and enters the system after successful verification; Hospital address upload module (103), hospital upload address; Patient history medical data uploading module (104), before the hospital uploads each patient's history medical data, it signs the patient history medical data sharing agreement, after which the hospital can upload each patient's history medical data, wherein the history medical data includes the patient's ID number, contact information, home address and medical process information; The hospital's real-time emergency data uploading module (105) requires the hospital to sign a real-time emergency data sharing agreement before uploading the real-time emergency data. Only after signing the agreement can the hospital upload the real-time emergency data, wherein the real-time emergency data includes the number of emergency patients in each emergency department.

3. The artificial intelligence-based emergency patient triage system according to claim 2, characterized in that: The patient terminal (2) comprises: Patient registration module (201), patient registration system account, including registration number and password, wherein the registration number is the ID number; Patient login module (202), where the patient logs in based on the registration number and password used during registration, and enters the system after successful verification; Patient symptom uploading module (203), where patients describe and upload symptoms that require emergency treatment, including one or more of text description, image description and video description; Patient travel mode uploading module (204), where patients select and upload travel modes according to their needs, including walking, cycling, public transportation or driving; The emergency treatment information receiving module (205) receives the hospital triaged by the system and the emergency treatment department of the hospital.

4. The artificial intelligence-based emergency patient triage system according to claim 3, characterized in that: The triage server (3) comprises: A registration information storage module (301) stores the registration information of the hospital registration module (101) and the patient registration module (201); A login verification module (302) verifies whether the login information of the hospital login module (102) and the patient login module (202) is correct based on the stored hospital and patient registration information, and the verification is successful if the verification is correct; The hospital address data storage module (303) stores the addresses of each hospital uploaded by the hospital address upload module (103); A patient history medical treatment data integration module (304) is configured to integrate the patient history medical treatment data uploaded by the patient history medical treatment data uploading module (104); A patient history medical treatment data storage module (305) stores the patient history medical treatment information integrated by the patient history medical treatment data integration module (304); The hospital real-time emergency data storage module (306) stores the hospital real-time emergency data uploaded by the hospital real-time emergency data upload module (105); The patient symptom medical information judgment module (307) constructs and trains a patient symptom emergency room judgment model, extracts key information describing the symptoms from the patient symptom upload module (203), and compares the similarity with the patient medical process information stored in the patient historical medical data storage module (305). If the similarity between the two is lower than a preset threshold, the patient's emergency room is judged based on the patient symptom emergency room judgment model. If the similarity between the two is higher than the preset threshold, the patient's emergency room is judged based on the medical process information. The patient consultation duration judgment module (308) constructs and trains an emergency number prediction model, takes the patient's home address stored in the patient history consultation data storage module (305) as the starting point, takes the addresses of each hospital stored in the hospital address data storage module (303) as the end point, calculates the time for the patient to arrive at the emergency department of each hospital based on the travel mode described by the patient travel mode upload module (204), predicts the number of patients visiting the emergency department of each hospital at that time based on the emergency number prediction model, multiplies the predicted number of patients visiting the emergency department of each hospital by the preset consultation time of a single patient, calculates the waiting time for the patient to arrive at the emergency department of each hospital, and takes the sum of the time for the patient to arrive at the emergency department of each hospital and the waiting time for the patient to arrive at the emergency department of each hospital as the determined patient consultation duration; The patient consultation recommendation module (309) recommends the hospital with the shortest patient consultation time as the patient's emergency consultation hospital based on the patient consultation time of each hospital's emergency consultation department determined by the patient consultation time determination module (308), and sends the emergency consultation hospital and the emergency consultation department of the hospital to the patient contact information stored in the patient history consultation data storage module (305).

5. The artificial intelligence-based emergency patient triage system according to claim 4, characterized in that: The specific steps of the integration of the patient history medical data integration module (304) include: classifying the patient history medical data by the ID card number stored in the patient history medical data uploading module (104), aggregating the patient history medical data with the same ID card number, classifying the aggregating patient history medical data with the same ID card number by the ID card number, contact information, home address and medical process information, retaining only one data of each of the ID card number, contact information and home address, and calculating the similarity of the medical process information. If the similarity is higher than a preset threshold, the medical process information of the latest date is retained. If the similarity is lower than the preset threshold, both medical process information are retained, and the integration is completed. The calculation expression of the similarity is: Where: A=(a1,a2,…,a n ) is represented as the bag-of-words vector of the medical process information A, B=(b1,b2,…,b n ) represents the bag-of-words vector of the medical process information B, n represents the vocabulary size, a i It is represented by the frequency of the i-th word in the medical process information A, b i It is represented by the frequency of the i-th word appearing in the medical process information B.

6. The artificial intelligence-based emergency patient triage system according to claim 5, characterized in that: The specific steps of the patient symptom medical information judgment module (307) extracting the key information describing the symptoms from the patient symptom upload module (203) include: If the symptom description data uploaded by the patient is in text format, the key information of the symptom description data uploaded by the patient is extracted based on the TF-IDF algorithm; If the symptom description data uploaded by the patient is in image format, the key information of the symptom description data uploaded by the patient is extracted based on the SIFT algorithm; If the symptom description data uploaded by the patient is in video format, the key frame extraction algorithm based on optical flow is used to extract key information of the symptom description data uploaded by the patient.

7. The artificial intelligence-based emergency patient triage system according to claim 6, characterized in that: The specific steps of the patient symptom medical information judgment module (307) for performing similarity comparison with the patient medical process information stored in the patient historical medical data storage module (305) are the same as the specific steps of the medical process information similarity calculation in the patient historical medical data integration module (304).

8. The artificial intelligence-based emergency patient triage system according to claim 7, characterized in that: The specific steps of the patient symptom consultation information judgment module (307) for judging the patient's emergency consultation clinic based on the patient symptom emergency consultation clinic judgment model include: The patient symptom emergency room visit judgment model includes a text judgment sub-model, an image judgment sub-model, a video judgment sub-model and a comprehensive judgment model; The symptom description data uploaded by the patient terminal (2) is input into each sub-model according to the corresponding data format. The sub-model makes an emergency room judgment on the symptom description data uploaded by the patient. If the patient only inputs symptom description data in one format, the judgment result of the sub-model is directly output as the emergency room judgment result. If the patient inputs symptom description data in two or three formats, the sub-model judgment result is output to the comprehensive judgment model. The comprehensive judgment model makes an emergency room judgment according to the preset weights of each sub-model, and the judgment result of the comprehensive model is used as the emergency room judgment result.

9. The artificial intelligence-based emergency patient triage system according to claim 8, characterized in that: The emergency department visit number prediction model of the patient visit duration judgment module (308) is an LSTM model.

10. The artificial intelligence-based emergency patient triage system according to claim 9, characterized in that: The patient consultation duration determination module (308) calculates the time it takes for patients to arrive at the emergency department of each hospital based on the Dijkstra algorithm.

Images