Intelligent dynamic triage system and method for emergency patients

Abstract

The invention discloses an intelligent dynamic triage system and method for emergency patients. The triage system comprises a data acquisition module used for acquiring basic information, physiological parameters and historical records of patients; the artificial intelligence analysis module generates a preliminary triage result; the Internet of Things connection module collects hospital resource data in real time; the central control processing unit is connected with each module to receive data; the dynamic grading decision-making module divides the patients into critical symptoms, severe symptoms, subcritical symptoms and common symptoms; the resource allocation module dynamically generates a resource allocation scheme according to the grading result; the diagnosis and treatment process management module tracks the whole process of the patient and updates state data; when the state of the patient changes, the system transmits data to the artificial intelligence analysis module again for evaluation, triggers the grading decision and resource allocation module, dynamically updates the triage priority and the resource allocation scheme, and achieves closed-loop feedback and continuous optimization.

Description

Technical Field

[0001] The present invention relates to the field of intelligent recommendation, and particularly to an intelligent dynamic triage system and method for emergency patients. Background Art

[0002] As an important part of the emergency department of a hospital, the emergency triage system is of crucial significance for improving the quality and efficiency of emergency medical services. With the continuous growth of the demand for medical services, the contradiction between relatively limited medical resources has become increasingly prominent. How to scientifically and reasonably allocate limited medical resources, timely treat critically ill patients, and at the same time ensure that ordinary patients can also receive appropriate medical care has become an important challenge faced by current emergency medicine. Traditional emergency triage systems mainly rely on the experience judgment of medical staff and fixed grading standards, and it is difficult to cope with the increasingly complex and changeable emergency treatment situations. Especially during the peak patient period, it often leads to problems such as inaccurate triage and unreasonable allocation of medical resources, affecting the quality of medical services and patient satisfaction.

[0003] Currently, common emergency triage technologies mainly include traditional triage methods based on fixed grading standards and semi-automated triage systems that partially introduce information technology. Traditional triage methods usually adopt a three-level or five-level triage standard (such as level 1 for critical illness, level 2 for severe illness, level 3 for sub-severe illness, and level 4 for ordinary symptoms). The receiving nurse grades the patient according to the patient's symptoms and signs, and then arranges the treatment order according to different levels. With the development of information technology, some hospitals have begun to apply computer-aided triage systems to initially grade patients through a preset assessment scale and determine the final triage result in combination with the professional judgment of medical staff. In addition, some hospitals have also tried to introduce mobile terminals and Internet technologies to allow patients to conduct self-symptom assessment and appointment registration through mobile applications to reduce the on-site waiting time.

[0004] However, the existing technologies have obvious deficiencies in practical applications: First, traditional triage systems lack intelligent analysis capabilities and are difficult to handle complex and changeable emergency situations. Especially when patient information is incomplete or symptom descriptions are unclear, it is easy to lead to triage errors. Second, existing systems generally lack a dynamic adjustment mechanism and cannot be intelligently allocated according to the real-time resource status of the hospital, patient flow, and overall emergency environment, resulting in low utilization efficiency of medical resources. Third, the phenomenon of information islands is serious, and it is difficult to effectively integrate patient historical medical data, medical equipment data, and the resource status of each department in the hospital, affecting the accuracy of triage decisions. Fourth, there is a lack of personalized triage strategies, and it is difficult to make targeted arrangements according to the special needs of different patients.

[0005] The above problems seriously restrict the improvement of the quality of emergency medical services. There is an urgent need to develop an intelligent dynamic triage system and method that integrates artificial intelligence, Internet of Things, and big data analysis technologies to improve triage accuracy and the efficiency of medical resource allocation. Summary of the Invention

[0006] The object of the present invention is to provide an intelligent dynamic triage system and method for emergency patients, which can improve the triage accuracy and the efficiency of medical resource allocation.

[0007] To achieve the above object, the present invention provides the following technical solution: An intelligent dynamic triage system for emergency patients, comprising:

[0008] A data acquisition module, configured to obtain basic patient information, physiological parameters, symptom descriptions, and historical medical records, and generate patient data;

[0009] An artificial intelligence analysis module, connected to the data acquisition module, receiving the patient data, and analyzing and processing the patient data according to a pre-trained classification model to generate a preliminary triage result;

[0010] An Internet of Things connection module, configured to collect hospital resource data in real time, where the hospital resource data includes the resource status of each department in the hospital, the availability of medical equipment, and the workload of medical staff;

[0011] A central control processing unit, respectively connected to the data acquisition module, the artificial intelligence analysis module, and the Internet of Things connection module, for receiving the preliminary triage result and the hospital resource data;

[0012] A dynamic grading decision module, connected to the central control processing unit, for receiving the preliminary triage result and the hospital resource data, classifying and grading the patients according to preset decision rules to generate a grading result, where the grading result includes classifying the patients into critically ill, severely ill, sub-severely ill, and ordinary symptoms, and transmitting the grading result to the central control processing unit;

[0013] A resource allocation module, connected to the central control processing unit, receiving the grading result and the hospital resource data, and dynamically generating a diagnosis and treatment resource allocation plan and a diagnosis and treatment order arrangement according to the grading result and the hospital resource data to form a resource allocation result; and

[0014] A diagnosis and treatment process management module, connected to the central control processing unit and the resource allocation module, receiving the resource allocation result, tracking the whole process of the patient from triage to treatment, collecting and updating the patient status data in real time, and when the patient status data changes, transmitting the patient status data to the central control processing unit, and the central control processing unit transmitting the patient status data to the artificial intelligence analysis module for re-evaluation to generate an updated preliminary triage result, and sequentially triggering the dynamic grading decision module and the resource allocation module to dynamically update the triage priority and the resource allocation plan.

[0015] Preferably, the data acquisition module includes:

[0016] A patient information input terminal for entering the patient's identity information and chief complaint symptoms;

[0017] A physiological parameter acquisition device for acquiring the patient's physiological parameters, where the physiological parameters include body temperature, blood pressure, heart rate, blood oxygen saturation, and respiratory rate;

[0018] A medical imaging interface for receiving the patient's medical images, where the medical images include X-ray images, CT images, and ultrasound images; and

[0019] An electronic medical record interface for obtaining the patient's historical medical records and past medical history data.

[0020] Preferably, the artificial intelligence analysis module includes:

[0021] A data preprocessing unit for performing format conversion, standardization, and missing value processing on the patient data to generate preprocessed standardized patient data;

[0022] A multi-modal feature extraction unit connected to the data preprocessing unit for receiving the standardized patient data, extracting key features from the text symptom description, physiological parameter data, and medical imaging data respectively, and integrating the key features into a feature vector;

[0023] A deep learning classification unit connected to the multi-modal feature extraction unit for receiving the feature vector and analyzing the feature vector using a pre-trained neural network model, where the neural network model includes a convolutional neural network CNN for processing medical imaging data and a recurrent neural network RNN for processing time series data, to generate a patient classification prediction result; and

[0024] A confidence evaluation unit connected to the deep learning classification unit for receiving the patient classification prediction result, calculating a reliability index of the analysis result, and determining whether manual intervention is required according to a preset confidence threshold, and combining the patient classification prediction result and the confidence evaluation result to form a preliminary triage result and transmitting it to the central control processing unit.

[0025] Preferably, the classification algorithm steps adopted by the deep learning classification unit include:

[0026] Step 1: Normalize the input feature vector X = {x1, x2,..., x n} to obtain a normalized feature vector X', where x1 to x n represent different feature parameters;

[0027] Step 2: Input the normalized feature vector X' into the pre-trained neural network model, and calculate the intermediate layer output H = σ(W1·X' + b1) through forward propagation, where W1 is the weight matrix, b1 is the bias vector, and σ is the activation function;

[0028] Step 3: Calculate the output layer result Y = softmax(W2·H + b2), where W2 is the output layer weight matrix, b2 is the output layer bias vector, and Y = {y1, y2, y3, y4} respectively represent the probability values that the patient belongs to critical illness, severe illness, sub-severe illness, and ordinary symptoms;

[0029] Step 4: Select the category with the highest probability as the preliminary classification result R = argmax(Y), where the argmax function returns the category index with the largest probability value;

[0030] Step 5: Calculate the confidence index C = -∑y i log(y i ), where y i is the i-th probability value in Y, ∑ represents the sum of all i from 1 to 4, and when C is lower than the preset threshold, it is marked as a case that requires manual review; and

[0031] Step 6: Combine the preliminary classification result R and the confidence index C to form a complete classification evaluation result, and transmit it to the confidence evaluation unit for further processing.

[0032] Preferably, the IoT connection module includes:

[0033] Department resource monitoring unit, which obtains the bed usage situation and the length of the waiting queue in each department in real time through the hospital information system API interface, and generates department resource status data;

[0034] Medical device status acquisition unit, which obtains the location and usage status of key medical devices through RFID and sensor networks, where RFID represents radio frequency identification technology, and generates device status data;

[0035] Medical staff workload monitoring unit, which collects the current workload, remaining available time, and professional skill matching degree of medical staff through the electronic scheduling system and the task management system, and generates personnel status data; and

[0036] Data integration unit, which is connected to the department resource monitoring unit, the medical device status acquisition unit, and the medical staff workload monitoring unit, receives and integrates the department resource status data, the device status data, and the personnel status data, standardizes and synchronizes the time of the collected different types of resource data, and integrates them into hospital resource data in a unified format, and transmits it to the central control processing unit.

[0037] Preferably, the dynamic grading decision module grades the patients by using a weighted scoring algorithm, and the weighted scoring algorithm includes the following steps:

[0038] Step 1: Set the basic scoring item set E = {e1, e2,..., e m}, where e1, e2 up to e m respectively represent different evaluation indicators. The evaluation indicators are divided into four categories: physiological parameter indicators, clinical symptom indicators, medical history risk indicators, and special population identification indicators. The physiological parameter indicators include body temperature, blood pressure, heart rate, blood oxygen saturation, and respiratory rate; the clinical symptom indicators include pain level, consciousness state, bleeding condition, and degree of dyspnea; the medical history risk indicators include a history of cardiovascular and cerebrovascular diseases, immune function status, and recent surgical history; the special population identification indicators include pregnancy identification, elderly identification, and child identification;

[0039] Step 2: Assign weight values to each scoring item e1, e2 up to e m in Step 1, which are w1, w2 up to w m respectively, to form a weight vector W = {w1, w2,..., w m}, satisfying w1 + w2 +... + w m = 1. Among them, the weight values of the physiological parameter indicators are set in the range of 0.1 - 0.3, the weight values of the clinical symptom indicators are set in the range of 0.2 - 0.4, the weight values of the medical history risk indicators are set in the range of 0.05 - 0.15, and the weight values of the special population identification indicators are set in the range of 0.05 - 0.15;

[0040] Step 3: Calculate the comprehensive score value T of the patient according to the patient data obtained from the data acquisition module and the preset scoring criteria. Specifically, it includes: First, assign scores s1, s2 up to s m to each scoring item e1, e2 up to e m in Step 1 according to the deviation degree of its measured value from the standard value. The greater the deviation degree, the higher the score, and the value range is 0 - 10; then calculate the weighted total score according to the formula T = w1 × s1 + w2 × s2 +... + w m × s m , where w1, w2 up to w m are the weight values determined in Step 2;

[0041] Step 4: Calculate the resource load factor L based on the hospital resource data provided by the Internet of Things connection module, where L is determined by the following formula: L = C1 × (number of currently occupied beds / total number of beds) + C2 × (number of patients waiting for diagnosis / average processing capacity) + C3 × (workload of medical staff / normal workload), where C1 is the weight coefficient of bed resources, C2 is the weight coefficient of the patient queue, and C3 is the weight coefficient of medical staff; the value range of C1 is 0.3 - 0.5, the value range of C2 is 0.2 - 0.4, the value range of C3 is 0.2 - 0.4, and C1 + C2 + C3 = 1; the value range of L is 0 - 1;

[0042] Step 5: Adjust the triage thresholds th1, th2, and th3 according to the resource load factor L in Step 4, which are used for the determination of critical illness, severe illness, and sub-severe illness respectively. The adjustment formulas are as follows:

[0043] th1 = B1 + K1 × L;

[0044] th2 = B2 + K2 × L;

[0045] th3 = B3 + K3 × L;

[0046] where B1, B2, and B3 are the basic thresholds for critical illness, severe illness, and sub-severe illness respectively, K1, K2, and K3 are the corresponding threshold adjustment coefficients respectively. The value range of B1 is 7 - 8, the value range of B2 is 5 - 6, the value range of B3 is 3 - 4, and the value range of K1, K2, and K3 is 0.1 - 1.0;

[0047] Step 6: Compare the comprehensive patient score T in Step 3 with the adjusted thresholds th1, th2, and th3 in Step 5 to determine the grading result: If T ≥ th1, it is graded as critical illness, and the emergency treatment priority is set to level 1; if T < th1 and T ≥ th2, it is graded as severe illness, and the emergency treatment priority is set to level 2; if T < th2 and T ≥ th3, it is graded as sub-severe illness, and the emergency treatment priority is set to level 3; if T < th3, it is graded as ordinary symptoms, and the emergency treatment priority is set to level 4;

[0048] Step 7: Assign a maximum waiting time limit to each grading result in Step 6: The maximum waiting time limit for critical illness patients is 5 minutes; the maximum waiting time limit for severe illness patients is 15 minutes; the maximum waiting time limit for sub-severe illness patients is 30 minutes; the maximum waiting time limit for ordinary symptoms patients is 60 minutes;

[0049] Step 8: Package the grading result in Step 6, the processing priority, the maximum waiting time limit in Step 7, the original scoring data, and the threshold calculation process into a grading decision package, and transmit it to the central control processing unit. At the same time, record the complete data of this grading process in the system database for system self-learning and triage standard optimization.

[0050] Preferably, the resource allocation module uses a multi-objective optimization algorithm to generate a diagnosis and treatment resource allocation plan. The multi-objective optimization algorithm includes the following steps:

[0051] Step 1: Define the patient set P = {p1, p2,..., p u}, where p1, p2 up to p u represent different patients respectively. Each patient has an attribute set, and the attribute set includes the grading result, waiting time, and special needs. The grading result takes values of critical illness, severe illness, sub-severe illness, or ordinary symptoms. The waiting time represents the time length from triage to the current moment for the patient, and the special needs indicate whether the patient requires specific medical equipment or a specialist physician;

[0052] Step 2: Define the medical resource set R = {r1, r2,..., r v}, where r1, r2 up to r v represent different medical resources respectively. Each medical resource has an attribute set, and the attribute set includes the resource type, available status, and processing capacity. The resource type includes consulting rooms, physicians, nurses, and medical equipment. The available status takes values of idle or occupied, and the processing capacity represents the number of patients that the resource can handle per unit time;

[0053] Step 3: Construct a resource allocation matrix M. The element m ij in M represents the decision variable for patient p i to be allocated to resource r j . When m ij = 1, it means that patient p i is allocated to resource r j . When m ij = 0, it means not to allocate;

[0054] Step 4: Define the objective functions f1(M) and f2(M), where f1(M) represents the average waiting time weighted by the urgency level. The calculation formula is: f1(M) = Σ(w i ×t i ) / u, where w i is the urgency level weight of patient p i . The value for critical illness is 4, for severe illness is 3, for sub-severe illness is 2, and for ordinary symptoms is 1; t i is the waiting time of patient p iThe estimated total waiting time; u is the total number of patients; f2(M) represents the resource utilization rate, and the calculation formula is: f2(M) = Σ(occupied resource numbers) / v, where v is the total number of medical resources;

[0055] Step 5: Set the constraints:

[0056] Each patient can be assigned to and only assigned to one main resource;

[0057] The resource allocation conforms to the priority requirements of the patient classification results;

[0058] The resource allocation meets the special needs of patients;

[0059] Resources cannot exceed their processing capabilities;

[0060] Step 6: Apply the genetic algorithm to solve the multi-objective optimization problem min{f1(M), -f2(M)}, and generate the Pareto optimal solution set. The specific steps include: initializing the population, randomly generating a resource allocation plan that meets the constraints; calculating the fitness value of each plan, and the fitness value is jointly determined by f1(M) and f2(M); generating a new generation of population through selection, crossover, and mutation operations; repeating the above process until the preset number of iterations or convergence conditions are reached;

[0061] Step 7: Select the final solution from the Pareto optimal solution set according to the current hospital strategy. When the hospital is in the peak period, f1(M) is given priority, and when the hospital is in the off-peak period, f2(M) is given priority, and generate the resource allocation result;

[0062] Step 8: Transmit the resource allocation result to the diagnosis and treatment process management module to guide patient triage and medical resource arrangement, and at the same time record the complete data of this resource allocation process for system continuous optimization.

[0063] Preferably, the diagnosis and treatment process management module includes:

[0064] A patient tracking unit that real-time tracks the patient's location and status through the hospital positioning system and the electronic wristband. The hospital positioning system includes an indoor positioning sensor network and a data processing server. The electronic wristband has a unique identification code and is equipped with a physiological parameter acquisition sensor. The patient tracking unit integrates the collected patient location data and status data into the patient's real-time trajectory information;

[0065] A process monitoring unit, connected to the patient tracking unit, receives the patient's real-time trajectory information, monitors the stay time and transfer process of each patient at different diagnosis and treatment nodes. The diagnosis and treatment nodes include the triage desk, waiting area, consulting room, examination room, treatment room, and observation room. The process monitoring unit calculates the deviation between the actual stay time and the expected time of the patient at each node and generates process monitoring data;

[0066] Anomaly warning unit, connected to the patient tracking unit and the process monitoring unit, receives real-time patient trajectory information and process monitoring data, triggers an alarm when the patient status data shows abnormal changes or the waiting time exceeds the safety threshold, where the abnormal changes include physiological parameters exceeding the safe range or abnormal location, the safety threshold is set according to the patient grading result, and the anomaly warning unit generates a warning message and determines the warning level; and

[0067] Dynamic feedback unit, connected to the anomaly warning unit, receives the warning message and the warning level, when it detects that the patient's status changes and requires re-triage, transmits the patient status data to the central control processing unit, triggers the re-evaluation process, and at the same time sends a notice to the relevant medical staff.

[0068] Preferably, it further includes a human-computer interaction module, connected to the central control processing unit, and the human-computer interaction module includes:

[0069] Medical staff workstation, equipped with a high-resolution display screen and a touch operation interface, used to display triage decision suggestions and resource allocation plans, display the patient distribution status and resource usage in real time, and allow medical staff to perform necessary manual interventions and adjustments after passing the permission authentication. The medical staff workstation, as the operation center, distributes relevant instructions to the patient information display terminal, the voice interaction system, and the mobile application interface;

[0070] Patient information display terminal, connected to the medical staff workstation, installed in the waiting area and the diagnosis and treatment area, adopts privacy protection display technology, displays personal triage results, estimated waiting time, and treatment process through the patient's unique identification code, and at the same time provides hospital layout navigation and treatment progress query functions. The patient information display terminal sends the patient's query request back to the medical staff workstation for processing;

[0071] Voice interaction system, connected to the medical staff workstation, equipped with a noise reduction microphone array and a natural language processing unit, supports medical staff to quickly query and update patient information through voice commands, execute resource allocation instructions, and respond to emergency calls. The voice interaction system transmits the processed voice instructions to the medical staff workstation for execution; and

[0072] Mobile application interface, connected to the medical staff workstation, supports medical staff to remotely access the system through hospital-authorized mobile devices and receive key reminders, realizes real-time monitoring of patient status and remote consultation. The mobile application interface adopts an encrypted communication protocol to ensure data transmission security, and sets up a multi-level access permission management mechanism. The remote operation instructions received by the mobile application interface are synchronized to the patient information display terminal and the voice interaction system through the medical staff workstation.

[0073] An intelligent dynamic triage method for emergency patients, based on an intelligent dynamic triage system for emergency patients, includes the following steps:

[0074] Step 1: Collect patient data, including obtaining the patient's basic information, physiological parameters, symptom descriptions, and historical medical records, and generating patient data;

[0075] Step 2: Perform artificial intelligence analysis, receive the patient data in Step 1, and analyze and process the patient data according to a pre-trained classification model to generate a preliminary triage result;

[0076] Step 3: Monitor hospital resources, and collect hospital resource data in real time. The hospital resource data includes the resource status of each department in the hospital, the availability of medical equipment, and the workload of medical staff;

[0077] Step 4: Perform dynamic grading decisions, receive the preliminary triage result in Step 2 and the hospital resource data in Step 3, classify and grade the patients according to preset decision rules, and generate a grading result. The grading result includes classifying the patients into critically ill, severe, sub-severe, and ordinary symptoms;

[0078] Step 5: Conduct resource allocation, receive the grading result in Step 4 and the hospital resource data in Step 3, and dynamically generate a diagnosis and treatment resource allocation plan and a diagnosis and treatment order arrangement according to the grading result and the hospital resource data to form a resource allocation result;

[0079] Step 6: Manage the diagnosis and treatment process, receive the resource allocation result in Step 5, track the whole process of the patient from triage to treatment, and collect and update the patient status data in real time;

[0080] Step 7: Perform dynamic adjustment. When the patient status data in Step 6 changes, return the updated patient status data to Step 2 for re-evaluation, generate an updated preliminary triage result, and sequentially trigger the dynamic grading decision in Step 4 and the resource allocation in Step 5 to form a closed-loop feedback system, dynamically updating the triage priority and the resource allocation plan.

[0081] Compared with the prior art, the advantages of the present invention are as follows: The system comprehensively collects patient information through the data acquisition module to establish a data foundation; subsequently, the artificial intelligence analysis module uses a pre-trained classification model to intelligently process this data to form a preliminary triage judgment; at the same time, the Internet of Things connection module monitors the status of hospital resources in real time to provide environmental parameters for subsequent decision-making; the core of the system lies in the central control processing unit as an information hub to coordinate the data flow between modules; the dynamic grading decision module combines the preliminary triage results and the real-time hospital resource status, and accurately grades patients into four levels from critically ill to common symptoms according to preset rules; the resource allocation module then scientifically allocates medical resources and arranges the diagnosis and treatment order based on the grading results and resource data; the diagnosis and treatment process management module continuously tracks the patient's status. When the patient's condition changes, it immediately triggers a re-evaluation process, automatically adjusts the triage priority and resource allocation, ensuring that the system can dynamically adapt to the real-time changes in the emergency environment, significantly improving the utilization efficiency of medical resources and the timeliness of treating critically ill patients, and solving the problem of the lack of flexibility in traditional fixed triage criteria. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0083] Figure 1 is the principle block diagram of the present invention;

[0084] Figure 2 is the working flow chart of the artificial intelligence analysis module in the present invention;

[0085] Figure 3 is the working flow chart of the present invention;

[0086] In the figure, 1, data acquisition module; 2, artificial intelligence analysis module; 3, Internet of Things connection module; 4, central control processing unit; 5, dynamic grading decision module; 6, resource allocation module; 7, diagnosis and treatment process management module; 8, human-computer interaction module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0087] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0088] Embodiment 1: As shown in the figure, an intelligent dynamic triage system for emergency patients includes:

[0089] A data acquisition module, which is used to obtain the patient's basic information, physiological parameters, symptom descriptions, and historical medical records, and generate patient data;

[0090] An artificial intelligence analysis module, connected to the data acquisition module, receiving patient data, analyzing and processing the patient data according to a pre-trained classification model, and generating a preliminary triage result;

[0091] An Internet of Things connection module, which is used to collect hospital resource data in real time. The hospital resource data includes the resource status of each department in the hospital, the availability of medical equipment, and the workload of medical staff;

[0092] A central control processing unit, respectively connected to the data acquisition module, the artificial intelligence analysis module, and the Internet of Things connection module, for receiving the preliminary triage result and the hospital resource data;

[0093] A dynamic grading decision module, connected to the central control processing unit, for receiving the preliminary triage result and the hospital resource data, classifying and grading the patient according to preset decision rules, generating a grading result. The grading result includes classifying the patient into critically ill, severely ill, sub-severely ill, and general symptoms, and transmitting the grading result to the central control processing unit;

[0094] A resource allocation module, connected to the central control processing unit, receiving the grading result and the hospital resource data, dynamically generating a diagnosis and treatment resource allocation plan and a diagnosis and treatment order arrangement according to the grading result and the hospital resource data, forming a resource allocation result; and

[0095] A diagnosis and treatment process management module, connected to the central control processing unit and the resource allocation module, receiving the resource allocation result, tracking the whole process of the patient from triage to treatment, collecting and updating the patient status data in real time, and when the patient status data changes, transmitting the patient status data to the central control processing unit, and the central control processing unit transmits the patient status data to the artificial intelligence analysis module for re-evaluation, generating an updated preliminary triage result, and sequentially triggering the dynamic grading decision module and the resource allocation module to dynamically update the triage priority and the resource allocation plan.

[0096] In this embodiment, the data acquisition module includes:

[0097] A patient information input terminal, which is used to input the patient's identity information and the main complaint symptoms;

[0098] A physiological parameter acquisition device for acquiring physiological parameters of a patient, where the physiological parameters include body temperature, blood pressure, heart rate, blood oxygen saturation, and respiratory rate;

[0099] A medical imaging interface for receiving medical images of a patient, where the medical images include X-ray images, CT images, and ultrasound images; and

[0100] An electronic medical record interface for obtaining the patient's historical medical records and past medical history data.

[0101] This module ensures the accuracy and comprehensiveness of system decisions through a multi-dimensional and all-round patient information acquisition mechanism. The module integrates four key sub-units: The patient information input terminal is responsible for collecting the patient's basic identity information and chief complaint symptoms, and establishing the patient's initial file; The physiological parameter acquisition device continuously monitors key vital signs such as body temperature, blood pressure, heart rate, blood oxygen saturation, and respiratory rate, providing objective data support for the urgency assessment; The medical imaging interface receives and processes medical images such as X-rays, CTs, and ultrasounds, enabling the system to obtain intuitive information on the status of the patient's internal organs and tissues; The electronic medical record interface, through docking with the hospital information system, obtains the patient's historical medical records and past medical history, providing important background information for the analysis of current symptoms.

[0102] This diversified data acquisition architecture not only improves the accuracy of triage, but also reduces waste of medical resources and diagnostic delays. At the same time, through a standardized data acquisition process, it ensures the consistency and comparability of different patients' data, providing high-quality input for subsequent artificial intelligence analysis. And this module can also add new data acquisition channels according to the actual situation of the hospital, such as wearable device interfaces or remote monitoring systems, further enhancing the adaptability and practical value of the system.

[0103] In this embodiment, the artificial intelligence analysis module includes:

[0104] A data preprocessing unit for performing format conversion, standardization, and missing value processing on the patient data to generate preprocessed standardized patient data; Its workflow can be further refined as follows: First, data cleaning is performed to identify and process outliers. For example, body temperature data outside the physiological possible range (such as above 40.5°C) will be marked as potential errors. Then, data standardization is performed to convert physiological parameters with different dimensions (such as blood pressure in mmHg and body temperature in °C) to a unified scale (such as the 0-1 interval), ensuring that each parameter has a comparable weight in subsequent analysis; For missing value processing, the system adopts various strategies according to the data type. For example, for continuous physiological parameters, the mean value can be used for filling or forward filling, while for categorical data such as symptom descriptions, the most frequent value or a special "missing" category marker can be used. In actual implementation, these preprocessing parameters can be customized and adjusted according to the specific situation of the hospital and continuously optimized as data accumulates.

[0105] A multi-modal feature extraction unit, connected to the data preprocessing unit, is used to receive the standardized patient data, extract key features from the text symptom description, physiological parameter data, and medical image data respectively, and integrate the key features into a feature vector. The specific implementation of the multi-modal feature extraction unit includes: using a medical-specific word vector model (such as Medical-BERT) to extract semantic features from the text symptom description, identifying key symptom words and their severity descriptions, not only extracting static values from the physiological parameter data, but also calculating time series features such as change rate, volatility, and trend, and using transfer learning methods for medical images, using a deep convolutional network pre-trained on a large-scale medical image dataset to extract features; in the feature fusion stage, an attention mechanism is adopted to dynamically adjust the weights of various features according to different disease types. For example, for respiratory diseases, the system will give priority to paying attention to respiratory rate, blood oxygen saturation, and lung image features.

[0106] A deep learning classification unit, connected to the multi-modal feature extraction unit, is used to receive the feature vector and analyze the feature vector using a pre-trained neural network model. The neural network model includes a convolutional neural network CNN for processing medical image data and a recurrent neural network RNN for processing time series data, generating a patient classification prediction result; the network architecture of the deep learning classification unit adopts a hybrid model, with the backbone network responsible for general feature processing, supplemented by multiple expert sub-networks for discriminant of specific disease categories. The training process adopts a phased strategy, first pre-training with a large-scale standard dataset, and then fine-tuning with the historical data of a specific hospital. To improve the recognition rate of rare but critical situations, special training techniques for imbalanced data are adopted, such as focal loss function or oversampling technique; in the model deployment stage, the system can also adopt model distillation technology to transfer the knowledge of complex models to lightweight models, improving the inference speed to meet the real-time requirements of the emergency environment.

[0107] The confidence evaluation unit, connected to the deep learning classification unit, is used to receive the patient classification prediction result, calculate the reliability index of the analysis result, determine whether manual intervention is required according to the preset confidence threshold, combine the patient classification prediction result and the confidence evaluation result to form a preliminary triage result, and transmit it to the central control processing unit. The confidence evaluation unit adopts a variety of uncertainty quantification methods, including prediction probability entropy, consistency measure of the ensemble model, and uncertainty estimation of the Bayesian neural network. It sets dynamic confidence thresholds according to different classification results, adopts a more stringent threshold standard for critical illness prediction, and also introduces interpretive algorithms such as SHAP values or attention heatmaps to help medical staff understand the basis of AI decisions; for cases identified as low confidence, the system automatically generates a list of supplementary information to be obtained to guide medical staff for further examination. In addition, the confidence evaluation result is also used for the continuous improvement of the system. Cases with low confidence and misclassification will be preferentially included in the training data for subsequent model updates.

[0108] In actual deployment, this module can also be customized according to the specific needs of the hospital. For example, regional disease characteristics can be reflected through model fine-tuning; during the epidemic period of seasonal diseases (such as the flu season), the weights of relevant symptoms can be dynamically adjusted; for special populations (such as the elderly, children, or pregnant women), a dedicated set of model parameters can be enabled. The system can also be configured to support progressive analysis, giving a preliminary analysis result based on the initial data, and continuously updating and refining the triage suggestions as more data (such as test results) are obtained, thus accelerating the emergency process.

[0109] In this embodiment, the classification algorithm steps adopted by the deep learning classification unit include:

[0110] Step 1: Normalize the input feature vector X = {x1, x2,..., x n} to obtain the normalized feature vector X', where x1 to x n represent different feature parameters;

[0111] Step 2: Input the normalized feature vector X' into the pre-trained neural network model, and calculate the intermediate layer output H = σ(W1·X' + b1) through forward propagation, where W1 is the weight matrix, b1 is the bias vector, and σ is the activation function;

[0112] Step 3: Calculate the output layer result Y = softmax(W2·H + b2), where W2 is the output layer weight matrix, b2 is the output layer bias vector, and Y = {y1, y2, y3, y4} respectively represent the probability values of the patient belonging to critical illness, severe illness, sub-severe illness, and normal symptoms;

[0113] Step 4: Select the category with the highest probability as the preliminary classification result R = argmax(Y), where the argmax function returns the index of the category with the largest probability value;

[0114] Step 5: Calculate the confidence index C = -∑y i log(y i ), where y i is the i-th probability value in Y, and ∑ represents the sum over all i from 1 to 4. When C is below the preset threshold, it is marked as a case that requires manual review; and

[0115] Step 6: Combine the preliminary classification result R with the confidence index C to form a complete classification evaluation result, and transmit it to the confidence evaluation unit for further processing.

[0116] This algorithm realizes a complete processing chain from data input to result output through six steps. First, in Step 1, the input feature vectors are normalized to ensure that features with different dimensions (such as blood pressure, body temperature, etc.) are compared on the same numerical scale, avoiding certain features with large numerical values from dominating the model decision-making; in Step 2, the normalized features are input into a pre-trained neural network, and the intermediate layer representations are calculated through weight matrices, bias vectors, and activation functions to capture the complex non-linear relationships between features; in Step 3, the softmax function is used to convert the output of the neural network into a probability distribution, intuitively representing the likelihood that a patient belongs to four categories of emergency levels (critical illness, severe illness, sub-severe illness, and common symptoms); in Step 4, the preliminary classification result of the patient is determined based on the maximum probability principle; in Step 5, the certainty of the classification decision is quantified by calculating information-theoretic metrics such as entropy value, providing a reference for the credibility of the AI judgment for clinicians; finally, in Step 6, the classification result is combined with the confidence index to form a complete evaluation result.

[0117] This algorithm design not only pursues classification accuracy but also focuses on the reliability evaluation of the results. It not only gives play to the advantages of AI in pattern recognition but also reserves the necessary space for manual intervention through the confidence mechanism, which is very suitable for high-risk decision-making scenarios such as medical treatment.

[0118] In this embodiment, the Internet of Things connection module includes:

[0119] The department resource monitoring unit obtains the bed usage situation and the length of the waiting queue in each department in real time through the hospital information system API interface, and generates department resource status data;

[0120] The medical device status acquisition unit obtains the location and usage status of key medical devices through RFID and sensor networks, where RFID represents radio frequency identification technology, and generates device status data;

[0121] The medical staff workload monitoring unit collects the current workload, remaining available time, and professional skill matching degree of medical staff through the electronic scheduling system and task management system, and generates personnel status data; and

[0122] The data integration unit is connected to the department resource monitoring unit, the medical equipment status acquisition unit, and the medical staff workload monitoring unit, receives and integrates the department resource status data, equipment status data, and personnel status data, standardizes the collected different types of resource data and synchronizes the time, and then integrates them into hospital resource data in a unified format and transmits it to the central control processing unit.

[0123] The above modules construct an all-round hospital resource monitoring network through four sub-units. The department resource monitoring unit obtains the bed usage situation and the length of the waiting queue in real time through the hospital information system API interface, and provides a view of resource availability at the macro level; the medical equipment status acquisition unit innovatively combines RFID technology and sensor networks to achieve precise tracking of the location and usage status of key medical equipment, making the allocation of equipment resources more efficient; the medical staff workload monitoring unit integrates the data of the electronic scheduling system and the task management system, comprehensively evaluates the current workload, available time, and professional matching degree of medical staff, and provides data support for human resource optimization; the data integration unit standardizes and synchronizes the time of the above three types of heterogeneous data to form hospital resource data in a unified format, providing a comprehensive and consistent resource information basis for subsequent triage decisions.

[0124] Among them, the method of docking with the hospital's existing information system through the API interface minimizes the additional infrastructure investment and improves the economy and compatibility of system deployment; the application of RFID and sensor networks realizes the automated and contactless monitoring of equipment resources, greatly improving the timeliness and accuracy of data compared with manual registration. The monitoring of the workload of medical staff not only considers the quantitative index of workload, but also includes the qualitative factor of professional skill matching degree, making the allocation of human resources more accurate.

[0125] In this embodiment, the dynamic grading decision module grades patients using a weighted scoring algorithm, and the weighted scoring algorithm includes the following steps:

[0126] Step 1: Set the set of basic scoring items E = {e1, e2,..., e m}, where e1, e2 up to e mrespectively represent different evaluation indicators, which are divided into four categories: physiological parameter indicators, clinical symptom indicators, medical history risk indicators, and special population identification indicators. The physiological parameter indicators include body temperature, blood pressure, heart rate, blood oxygen saturation, and respiratory rate; the clinical symptom indicators include pain level, state of consciousness, bleeding condition, and degree of dyspnea; the medical history risk indicators include previous cardiovascular and cerebrovascular disease history, immune function status, and recent surgical history; the special population identification indicators include pregnant woman identification, elderly identification, and child identification;

[0127] Step 2: Assign weight values to each scoring item e1, e2, up to e m in Step 1, which are w1, w2, up to w m , to form a weight vector W = {w1, w2,..., w m}, satisfying w1 + w2 +... + w m = 1, where the weight value of the physiological parameter indicators is set within the range of 0.1 - 0.3, the weight value of the clinical symptom indicators is set within the range of 0.2 - 0.4, the weight value of the medical history risk indicators is set within the range of 0.05 - 0.15, and the weight value of the special population identification indicators is set within the range of 0.05 - 0.15;

[0128] Step 3: Calculate the comprehensive score value T of the patient according to the patient data obtained from the data acquisition module and the preset scoring criteria, specifically including: First, for each scoring item e1, e2, up to e m in Step 1, assign scores s1, s2, up to s m according to the deviation degree of its measured value from the standard value. The greater the deviation degree, the higher the score, and the value range is 0 - 10; then calculate the weighted total score according to the formula T = w1×s1 + w2×s2 +... + w m ×s m , where w1, w2, up to w m are the weight values determined in Step 2;

[0129] Step 4: Calculate the resource load factor L based on the hospital resource data provided by the Internet of Things connection module, where L is determined by the following formula: L = C1×(current number of occupied beds / total number of beds) + C2×(number of patients waiting for diagnosis / average processing capacity) + C3×(workload of medical staff / normal working load), where C1 is the bed resource weight coefficient, C2 is the patient queue weight coefficient, and C3 is the medical staff weight coefficient; the value range of C1 is 0.3 - 0.5, the value range of C2 is 0.2 - 0.4, the value range of C3 is 0.2 - 0.4, and C1 + C2 + C3 = 1; the value range of L is 0 - 1;

[0130] Step 5: Adjust the triage thresholds th1, th2, and th3 according to the resource load factor L in Step 4, which are used for the determination of critical illness, severe illness, and sub-severe illness respectively. The adjustment formulas are as follows:

[0131] th1 = B1 + K1 × L;

[0132] th2 = B2 + K2 × L;

[0133] th3 = B3 + K3 × L;

[0134] Where B1, B2, and B3 are the basic thresholds for critical illness, severe illness, and sub-severe illness respectively, and K1, K2, and K3 are the corresponding threshold adjustment coefficients. The value range of B1 is 7 - 8, the value range of B2 is 5 - 6, the value range of B3 is 3 - 4, and the value range of K1, K2, and K3 is 0.1 - 1.0;

[0135] Step 6: Compare the comprehensive patient score T in Step 3 with the adjusted thresholds th1, th2, and th3 in Step 5 to determine the grading result: If T ≥ th1, it is graded as critical illness, and the emergency treatment priority is set to level 1; If T < th1 and T ≥ th2, it is graded as severe illness, and the emergency treatment priority is set to level 2; If T < th2 and T ≥ th3, it is graded as sub-severe illness, and the emergency treatment priority is set to level 3; If T < th3, it is graded as ordinary symptoms, and the emergency treatment priority is set to level 4;

[0136] Step 7: Assign the maximum waiting time limit for each grading result in Step 6: The maximum waiting time limit for critical illness patients is 5 minutes; The maximum waiting time limit for severe illness patients is 15 minutes; The maximum waiting time limit for sub-severe illness patients is 30 minutes; The maximum waiting time limit for ordinary symptoms patients is 60 minutes;

[0137] Step 8: Package the grading result in Step 6, the treatment priority, the maximum waiting time limit in Step 7, together with the original score data and the threshold calculation process into a grading decision package, and transmit it to the central control processing unit. At the same time, record the complete data of this grading process into the system database for system self-learning and triage standard optimization.

[0138] In the specific algorithm steps of the above-mentioned dynamic hierarchical decision module, the actual implementation of the scoring item set can be implemented in a modular configuration manner. A hierarchical scoring table can be established for physiological parameter indicators, such as body temperature <35℃ or >39℃ with a score of 8-10 points, 35-36℃ or 38-39℃ with a score of 4-7 points, and 36-38℃ with a score of 0-3 points; blood pressure is calculated using MAP (mean arterial pressure) and an age-related threshold is set; heart rate is set with upper and lower limits, such as adults with a higher score of <50 or >120; blood oxygen saturation <90% is a high score, 90-94% is a medium score, and >94% is a low score; respiratory rate is also set with an age-related interval. In the clinical symptom assessment, the pain level uses the standardized NRS pain scale (0-10 points); the consciousness state uses the AVPU scale (awake, responsive to language, responsive to pain, unresponsive) or the GCS scale; bleeding is graded according to the site and estimated blood loss; dyspnea is assessed using the modified Borg scale.

[0139] In the actual operation of weight allocation, a customized parameter table for the hospital is used. When the system is initially deployed, the weight configuration that best suits the patient characteristics of the hospital is determined by retrospectively analyzing the hospital's historical emergency data. For example, for hospitals with cardiovascular centers, the weights of cardiovascular-related indicators can be slightly higher; children's hospitals can increase the weights of child-specific indicators. Hospitals can also adjust weights according to seasonal changes, such as increasing the weight of respiratory symptoms during the flu season.

[0140] The resource load factor calculation adopts the sliding time window method, and the data is updated every 15-30 minutes to avoid decision shocks caused by instantaneous fluctuations. Bed resources consider the differentiated weights of different types of beds (such as intensive care beds and ordinary beds); in addition to considering the quantity ratio, the workload of medical staff also needs to be combined with the complexity coefficient, such as the number of critically ill patients currently managed by each doctor.

[0141] Threshold adjustment is implemented using a piecewise linear function, such as th1=7.5+0.5×L, when L<0.4; th1=7.5+0.8×L, when 0.4≤L<0.7; th1=7.5+1.0×L, when L≥0.7. This ensures that the threshold adjustment is more sensitive when resources are extremely tight.

[0142] The system sets up fast-track rules for specific diseases. For example, for time-sensitive diseases such as chest pain and stroke, even if the comprehensive score does not reach the critical threshold, specific key symptom combinations will trigger the priority treatment mechanism. At the same time, the system should achieve an explainable display of grading results, display the key factors leading to high scores at the triage terminal, and help medical staff understand the basis for grading.

[0143] This algorithm constructs an adaptable decision-making system through multiple steps. Steps 1 and 2 set the scoring item set and assign weights, covering four major categories of indicators: physiological parameters, clinical symptoms, medical history risks, and special population identifiers, ensuring the comprehensiveness of the evaluation; Step 3 calculates the comprehensive score based on the actual patient data, quantifies the deviation degree of each scoring item, and accumulates them with weights; Step 4 introduces a resource load factor, taking into account the bed utilization rate, the number of patients waiting for diagnosis, and the workload of medical staff, realizing the dynamic association between triage and resource status; Step 5 adjusts the determination thresholds for critical, severe, and sub-severe cases according to the resource load factor, making the triage criteria adaptively change with the hospital load status; Step 6 determines the final classification result and sets the processing priority by comparing the patient's comprehensive score with the adjusted threshold; Step 7 assigns the maximum waiting time limit for patients at different levels, establishing a time management framework; Step 8 records and archives the complete decision-making process to support system self-learning and continuous optimization.

[0144] The advantage of this algorithm lies in its dynamic response ability to the medical resource status. Traditional triage systems usually adopt fixed criteria and are difficult to cope with the challenges brought by hospital resource fluctuations. However, this module dynamically adjusts the triage threshold through the resource load factor, raising the classification standard when resources are scarce to ensure that the most urgent patients can obtain resources first; and appropriately relaxing the standard when resources are sufficient to improve resource utilization. The weight ranges set in the algorithm (such as the weight of physiological parameter indicators being 0.1 - 0.3, and the weight of clinical symptom indicators being 0.2 - 0.4) also leave room for adjustment and can be customized according to the specialty characteristics of different hospitals or the regional disease spectrum.

[0145] In this embodiment, the resource allocation module uses a multi-objective optimization algorithm to generate a diagnosis and treatment resource allocation plan. The multi-objective optimization algorithm includes the following steps:

[0146] Step 1: Define the patient set P = {p1, p2,..., p u}, where p1, p2 up to p u respectively represent different patients. Each patient has an attribute set, and the attribute set includes the classification result, waiting time, and special needs. The classification result takes values of critical, severe, sub-severe, or ordinary symptoms. The waiting time represents the time length from triage to the current moment for the patient, and the special needs indicate whether the patient requires specific medical equipment or specialist physicians;

[0147] Step 2: Define the medical resource set R = {r1, r2,..., r v}, where r1, r2 up to r vThey respectively represent different medical resources. Each medical resource has a set of attributes, and the set of attributes includes resource type, available status, and processing capacity. The resource type includes consulting rooms, physicians, nurses, and medical equipment. The available status takes values of idle or occupied, and the processing capacity represents the number of patients that the resource can handle per unit time;

[0148] Step 3: Construct a resource allocation matrix M. The element m in M ij represents the decision variable for patient p i to be allocated to resource r j . When m ij = 1, it means that patient p i is allocated to resource r j . When m ij = 0, it means no allocation;

[0149] Step 4: Define the objective functions f1(M) and f2(M), where f1(M) represents the average waiting time weighted by the urgency level. The calculation formula is: f1(M) = Σ(w i × t i ) / u, where w i is the urgency level weight of patient p i . For critically ill patients, the value is 4; for severe patients, the value is 3; for sub-severe patients, the value is 2; for ordinary symptoms, the value is 1; t i is the estimated total waiting time of patient p i ; u is the total number of patients; f2(M) represents the resource utilization rate, and the calculation formula is: f2(M) = Σ(number of occupied resources) / v, where v is the total number of medical resources;

[0150] Step 5: Set the constraint conditions:

[0151] Each patient can be allocated to and only allocated to one main resource;

[0152] The resource allocation conforms to the priority requirements of the patient grading results;

[0153] The resource allocation meets the special needs of patients;

[0154] The resources cannot exceed their processing capacity;

[0155] Step 6: Apply the genetic algorithm to solve the multi-objective optimization problem min{f1(M), -f2(M)} and generate the Pareto optimal solution set. The specific steps include: initializing the population and randomly generating a resource allocation plan that meets the constraint conditions; calculating the fitness value of each plan, and the fitness value is jointly determined by f1(M) and f2(M); generating a new generation of population through selection, crossover, and mutation operations; repeating the above process until the preset number of iterations or convergence conditions are reached;

[0156] Step 7: Select the final solution from the Pareto optimal solution set according to the current hospital strategy. When the hospital is in the peak period, give priority to f1(M); when the hospital is in the off-peak period, give priority to f2(M), and generate the resource allocation result;

[0157] Step 8: Transmit the resource allocation result to the diagnosis and treatment process management module to guide patient triage and medical resource arrangement, and at the same time record the complete data of this resource allocation process for system continuous optimization.

[0158] This module first defines the patient set, establishes an attribute profile for each patient, including the grading result, waiting time, and special needs. Then it defines the medical resource set, clarifies the types, status, and processing capabilities of various resources, constructs a resource allocation matrix as a decision variable, and sets two objective functions to measure the average waiting time weighted by urgency and resource utilization rate respectively. On the premise of meeting a series of constraints (such as patients must be and can only be assigned to one main resource, and resource allocation must conform to the grading priority, etc.), apply the genetic algorithm to solve this multi-objective optimization problem. Finally, select the final solution from the Pareto optimal solution set according to the current operation strategy of the hospital to realize the generation and application of the resource allocation result.

[0159] In actual operation, the definition of the patient set and the resource set can be implemented through a unified data structure, such as JSON or XML format, which is convenient for internal data exchange within the system. The grading result in the patient attributes can be directly docked with the four-level classification result in Claim 6. The waiting time can be calculated in real time and color warnings can be set (such as turning yellow and red when exceeding 50% and 75% of the set threshold). Special needs can be designed as a standardized coding set, such as the need for isolation, the need for specific equipment, the need for specific language translation services, etc.

[0160] The generation and maintenance of the resource allocation matrix is the core operation process of the system. During the peak period, the system can recalculate the allocation plan every 3 - 5 minutes, and during the off-peak period, the frequency can be reduced to once every 15 - 30 minutes to reduce the consumption of computing resources. The weight coefficients of the objective functions can be adjusted according to the hospital type and service orientation. For example, a tertiary emergency center may focus more on the waiting time weighted by urgency, while a community hospital may be more concerned about resource utilization rate.

[0161] The selection strategy of the Pareto optimal solution set in Step 6 is set to a dynamic adjustment mode. The system monitors the number of patients in the waiting area, the proportion of critically ill patients, and the workload of medical staff. When these indicators exceed specific thresholds, it automatically tends to select the solution that optimizes the waiting time; when the indicators are within the normal range, it tends to select the solution that balances the waiting time and resource utilization rate.

[0162] In this embodiment, the diagnosis and treatment process management module includes:

[0163] Patient tracking unit, which real-time tracks the location and status of patients through the hospital positioning system and electronic wristbands. The hospital positioning system includes an indoor positioning sensor network and a data processing server. The electronic wristband has a unique identification code and is equipped with physiological parameter acquisition sensors. The patient tracking unit integrates the collected patient location data and status data into patient real-time trajectory information;

[0164] Process monitoring unit, connected to the patient tracking unit, receives patient real-time trajectory information, monitors the stay time and transfer process of each patient at different diagnosis and treatment nodes. The diagnosis and treatment nodes include triage desks, waiting areas, consulting rooms, examination rooms, treatment rooms and observation rooms. The process monitoring unit calculates the deviation between the actual stay time and the expected time of patients at each node and generates process monitoring data;

[0165] Abnormal warning unit, connected to the patient tracking unit and the process monitoring unit, receives patient real-time trajectory information and process monitoring data, triggers an alarm when the patient status data shows abnormal changes or the waiting time exceeds the safety threshold. The abnormal changes include physiological parameters exceeding the safety range or abnormal location. The safety threshold is set according to the patient grading results. The abnormal warning unit generates warning information and determines the warning level; and

[0166] Dynamic feedback unit, connected to the abnormal warning unit, receives warning information and warning level. When it detects that the patient's status changes and requires re-triage, it transmits the patient status data to the central control processing unit, triggers the re-evaluation process, and sends a notice to the relevant medical staff at the same time.

[0167] The design advantage of this module is to achieve seamless tracking and dynamic monitoring of emergency patients from admission to the completion of the diagnosis and treatment process. The patient tracking unit realizes the real-time and accurate monitoring of the patient's location and physiological status through the combination of the indoor positioning sensor network and the electronic wristband, solving the problems of lag and discontinuity of traditional manual inspections; the process monitoring unit quantitatively monitors the stay time and transfer process of patients at each diagnosis and treatment node, calculates the deviation between the actual and expected time, and provides a data basis for process optimization.

[0168] The abnormal warning unit timely identifies the situations that need to be intervened through a dual-trigger mechanism (abnormal changes in patient status or waiting time exceeding the limit), and sets different safety thresholds according to the patient grading results, realizing the precision and personalization of early warnings. The dynamic feedback unit completes the closed-loop design. When it detects a status change that requires re-triage, it immediately feeds back the information to the central control system, triggers the re-evaluation process, and ensures that the triage decision is consistent with the patient's current status.

[0169] The greatest advantage of this design lies in transforming patient monitoring from static node checks to dynamic continuous monitoring, significantly enhancing the sensitivity and response speed to the deterioration of the patient's condition. Through real-time location tracking, the system can also identify abnormal patient movement patterns, such as staying in non-treatment areas for a long time or not moving along the expected path, thus detecting potential problems in advance. In addition, the detailed process data generated by this module is not only used for immediate monitoring but also supports long-term process optimization, such as identifying common bottleneck links or areas of inefficiency.

[0170] In practical applications, the electronic wristband can integrate multiple sensors, which not only monitor basic physiological parameters but also detect fall events or abnormal activity patterns, further enhancing patient safety. The accuracy of the indoor positioning system can reach the meter level, sufficient to distinguish whether the patient is waiting in the clinic or in the corridor. The warning information can adopt a hierarchical push strategy according to the urgency level, from general prompts to emergency alerts, and be transmitted to relevant medical staff through different channels (workstation display, mobile device push, voice broadcast, etc.), providing safer and more timely medical protection for emergency patients and also providing detailed operation data for hospital managers.

[0171] In this embodiment, it further includes a human-computer interaction module, which is connected to the central control processing unit. The human-computer interaction module includes:

[0172] A medical staff workstation, equipped with a high-resolution display screen and a touch operation interface, is used to display triage decision suggestions and resource allocation plans, display the patient distribution status and resource usage in real time, and allow medical staff to perform necessary manual interventions and adjustments after permission authentication. The medical staff workstation, as the operation center, distributes relevant instructions to the patient information display terminal, the voice interaction system, and the mobile application interface;

[0173] A patient information display terminal, connected to the medical staff workstation, is installed in the waiting area and the diagnosis and treatment area. It adopts privacy protection display technology and displays personal triage results, estimated waiting time, and treatment process through the patient's unique identification code. At the same time, it provides hospital layout navigation and treatment progress query functions. The patient information display terminal sends the patient's query request back to the medical staff workstation for processing;

[0174] A voice interaction system, connected to the medical staff workstation, is equipped with a noise reduction microphone array and a natural language processing unit, supporting medical staff to quickly query and update patient information through voice commands, execute resource allocation instructions, and respond to emergency calls. The voice interaction system transmits the processed voice instructions to the medical staff workstation for execution; and

[0175] The mobile application interface is connected to the medical staff workstation, supporting medical staff to remotely access the system through hospital - authorized mobile devices and receive critical reminders, enabling real - time monitoring of patient status and remote consultation. The mobile application interface adopts an encrypted communication protocol to ensure data transmission security and sets up a multi - level access permission management mechanism. The remote operation instructions received by the mobile application interface are synchronized to the patient information display terminal and the voice interaction system through the medical staff workstation.

[0176] This module constructs a multi - level and all - round interaction system through four functionally distinct sub - units. The medical staff workstation, as the operation center, provides a high - resolution visualization interface and a permission authentication mechanism, enabling medical staff to intuitively grasp triage decision suggestions and resource allocation situations, while retaining the necessary manual intervention ability, achieving a balance between automation and professional judgment.

[0177] The privacy - protection display technology and unique identification code mechanism adopted by the patient information display terminal solve the contradiction between information display in public areas and personal privacy protection, allowing patients to conveniently obtain their personal triage results, estimated waiting times, and treatment processes, while preventing the leakage of sensitive information. This design significantly reduces patients' anxiety and inquiries due to unclear information, improving patient satisfaction and the order in the waiting area.

[0178] The voice interaction system, through a noise - canceling microphone array and natural language processing technology, realizes hands - free information query and instruction execution, greatly improving the response speed and operation convenience in emergency situations. This design not only reduces the operation burden on medical staff but also shortens the time for obtaining critical information and executing commands, which is particularly important for time - sensitive emergency environments.

[0179] The design of the mobile application interface extends the accessibility of the system, enabling authorized medical staff to remotely monitor patient status, receive critical reminders, and participate in remote consultations through mobile devices, breaking the physical limitations of traditional systems. Through the encrypted communication protocol and multi - level access permission management, it ensures the security and compliance of remote operations, while improving the utilization efficiency of expert resources, especially during periods of tight medical resources such as at night or on holidays.

[0180] The overall design advantage of this module lies in its accurate grasp of different user needs and differential response. By concentrating the instruction distribution mechanism in the medical staff workstation, it realizes the collaborative work and information synchronization of each subsystem, avoiding information inconsistency problems among multiple systems. In addition, this design fully considers the particularity of the emergency environment and provides multiple complementary interaction methods to ensure efficient information exchange in various situations (including emergencies, high - noise environments, remote consultation requirements, etc.), significantly enhancing the usability and practical value of the emergency triage system.

[0181] Embodiment 2: As shown in the figure, an intelligent dynamic triage method for emergency patients, based on the intelligent dynamic triage system for emergency patients in Embodiment 1, includes the following steps:

[0182] Step 1: Collect patient data, including obtaining the patient's basic information, physiological parameters, symptom descriptions, and historical medical records, and generating patient data;

[0183] Step 2: Perform artificial intelligence analysis, receive the patient data in Step 1, and analyze and process the patient data according to a pre-trained classification model to generate a preliminary triage result;

[0184] Step 3: Monitor hospital resources, and collect hospital resource data in real time. The hospital resource data includes the resource status of each department in the hospital, the availability of medical equipment, and the workload of medical staff;

[0185] Step 4: Perform dynamic grading decision-making, receive the preliminary triage result in Step 2 and the hospital resource data in Step 3, classify and grade the patient according to a preset decision rule, and generate a grading result. The grading result includes classifying the patient into critically ill, severe, sub-severe, and ordinary symptoms;

[0186] Step 5: Conduct resource allocation, receive the grading result in Step 4 and the hospital resource data in Step 3, and dynamically generate a diagnosis and treatment resource allocation plan and a diagnosis and treatment order arrangement according to the grading result and the hospital resource data to form a resource allocation result;

[0187] Step 6: Manage the diagnosis and treatment process, receive the resource allocation result in Step 5, track the whole process of the patient from triage to treatment, and collect and update the patient status data in real time;

[0188] Step 7: Perform dynamic adjustment. When the patient status data in Step 6 changes, return the updated patient status data to Step 2 for re-evaluation, generate an updated preliminary triage result, and sequentially trigger the dynamic grading decision-making in Step 4 and the resource allocation in Step 5 to form a closed-loop feedback system, and dynamically update the triage priority and resource allocation plan.

[0189] This method forms a closed-loop workflow from patient data collection, AI analysis, resource monitoring, grading decision-making, resource allocation, process management to dynamic adjustment, ensuring that the whole process of the patient from admission to final disposal is within the monitoring and optimization scope of the system. The data flow between each step is clear and definite, facilitating system implementation and hospital process integration. The dynamic nature of the method is reflected in the mechanism of continuously monitoring the change of the patient's status and triggering re-evaluation, so that the triage priority and resource allocation plan always reflect the latest situation of the patient, especially adapting to the characteristics of the rapid change of the patient's condition in the emergency environment.

[0190] The above are only the embodiments of the present application, and do not thus limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made by using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, shall be similarly included in the patent protection scope of the present application.

Claims

1. An intelligent dynamic triage system for emergency patients, characterized in that: include: Data collection module, used to obtain basic patient information, physiological parameters, symptom descriptions and historical medical records to generate patient data; An artificial intelligence analysis module is connected to the data acquisition module, receives patient data, analyzes and processes the patient data according to a pre-trained classification model, and generates a preliminary triage result; An IoT connection module, used to collect hospital resource data in real time, including the resource status of each hospital department, the availability of medical equipment, and the workload of medical staff; A central control processing unit, connected to the data acquisition module, the artificial intelligence analysis module and the Internet of Things connection module, respectively, for receiving the preliminary triage results and the hospital resource data; A dynamic grading decision module is connected to the central control processing unit, and is used to receive the preliminary triage result and the hospital resource data, grade and classify the patients according to the preset decision rules, generate a grading result, and the grading result includes grading the patients into critical, severe, sub-severe and common symptoms, and transmit the grading result to the central control processing unit; A resource allocation module is connected to the central control processing unit, receives the classification result and the hospital resource data, and dynamically generates a diagnosis and treatment resource allocation plan and a diagnosis and treatment sequence arrangement according to the classification result and the hospital resource data to form a resource allocation result; as well as The diagnosis and treatment process management module is connected to the central control processing unit and the resource allocation module, receives the resource allocation result, tracks the entire process of the patient from triage to medical treatment, collects and updates the patient status data in real time, and transmits the patient status data to the central control processing unit when the patient status data changes. The central control processing unit transmits the patient status data to the artificial intelligence analysis module for re-evaluation, generates an updated preliminary triage result, and triggers the dynamic hierarchical decision module and the resource allocation module in turn to dynamically update the triage priority and resource allocation plan.

2. The intelligent dynamic triage system for emergency patients according to claim 1 is characterized in that: The data acquisition module comprises: Patient information input terminal, used to input the patient's identity information and main symptoms; A physiological parameter acquisition device is used to acquire the patient's physiological parameters, including body temperature, blood pressure, heart rate, blood oxygen saturation and respiratory rate; A medical imaging interface, for receiving medical images of a patient, wherein the medical images include X-ray images, CT images and ultrasound images; and Electronic medical record interface, used to obtain patients' historical medical records and medical history data.

3. The intelligent dynamic triage system for emergency patients according to claim 1 is characterized in that: The artificial intelligence analysis module includes: A data preprocessing unit, used to perform format conversion, standardization and missing value processing on the patient data to generate preprocessed standardized patient data; a multimodal feature extraction unit, connected to the data preprocessing unit, for receiving the standardized patient data, extracting key features from text symptom descriptions, physiological parameter data, and medical imaging data, respectively, and integrating the key features into a feature vector; a deep learning classification unit, connected to the multimodal feature extraction unit, configured to receive the feature vector, analyze the feature vector using a pre-trained neural network model, wherein the neural network model includes a convolutional neural network (CNN) for processing medical imaging data and a recurrent neural network (RNN) for processing time series data, and generate a patient classification prediction result; and The confidence assessment unit is connected to the deep learning classification unit, and is used to receive the patient classification prediction result, calculate the reliability index of the analysis result, and determine whether manual intervention is required according to a preset confidence threshold, and combine the patient classification prediction result and the confidence assessment result to form a preliminary triage result, which is transmitted to the central control processing unit.

4. The intelligent dynamic triage system for emergency patients according to claim 3 is characterized in that: The classification algorithm steps adopted by the deep learning classification unit include: Step 1: For the input feature vector X = {x1, x2, ..., x n } is normalized to obtain the normalized feature vector X', where x1 to x n Indicates different characteristic parameters; Step 2: Input the normalized feature vector X' into the pre-trained neural network model, and calculate the intermediate layer output H = σ(W1·X'+b1) through forward propagation, where W1 is the weight matrix, b1 is the bias vector, and σ is the activation function; Step 3: Calculate the output layer result Y = softmax(W2 H + b2), where W2 is the output layer weight matrix, b2 is the output layer bias vector, and Y = {y1, y2, y3, y4} represents the probability values ​​of the patient being critical, severe, sub-severe, and common symptoms respectively; Step 4: Select the category with the highest probability as the preliminary classification result R = argmax(Y), where the argmax function returns the index of the category with the largest probability value; Step 5: Calculate the confidence index C = -∑y i log(y i ), where y i is the i-th probability value in Y, ∑ represents the sum of all i from 1 to 4. When C is lower than the preset threshold, it is marked as a case that requires manual review; and Step 6: Combine the preliminary classification result R with the confidence index C to form a complete classification evaluation result, which is transmitted to the confidence evaluation unit for further processing.

5. The intelligent dynamic triage system for emergency patients according to claim 1 is characterized in that: The Internet of Things connection module includes: The department resource monitoring unit obtains the bed usage and consultation queue length of each department in real time through the hospital information system API interface, and generates department resource status data; Medical equipment status acquisition unit, which obtains the location and usage status of key medical equipment through RFID and sensor networks, where RFID stands for radio frequency identification technology, and generates equipment status data; The medical staff workload monitoring unit collects the current workload, remaining available time and professional skill matching of medical staff through the electronic scheduling system and task management system to generate personnel status data; and The data integration unit is connected to the department resource monitoring unit, the medical equipment status collection unit and the medical staff workload monitoring unit, receives and integrates the department resource status data, the equipment status data and the personnel status data, standardizes the collected different types of resource data, synchronizes them with time, and integrates them into a unified format of hospital resource data, which is then transmitted to the central control processing unit.

6. The intelligent dynamic triage system for emergency patients according to claim 1 is characterized in that: The dynamic grading decision module grades the patients using a weighted scoring algorithm, and the weighted scoring algorithm includes the following steps: Step 1: Set the basic scoring item set E = {e1, e2, ..., e m }, where e1, e2, and so on m They represent different evaluation indicators, which are divided into four categories: physiological parameter indicators, clinical symptom indicators, medical history risk indicators and special population identification indicators. The physiological parameter indicators include body temperature, blood pressure, heart rate, blood oxygen saturation and respiratory rate; the clinical symptom indicators include pain level, consciousness state, bleeding situation and dyspnea level; the medical history risk indicators include previous history of cardiovascular and cerebrovascular diseases, immune function status and recent surgical history; the special population identification indicators include pregnant women identification, elderly identification and children identification; Step 2: For each scoring item e1, e2, and so on in step 1, m Assign weight values, w1, w2, and so on to w m , forming a weight vector W = {w1,w2,...,w m }, satisfying w1+w2+...+w m =1, where the weight value of the physiological parameter index is set in the range of 0.1-0.3, the weight value of the clinical symptom index is set in the range of 0.2-0.4, the weight value of the medical history risk index is set in the range of 0.05-0.15, and the weight value of the special population identification index is set in the range of 0.05-0.15; Step 3: Calculate the patient comprehensive score T based on the patient data obtained from the data acquisition module and the preset scoring criteria, specifically including: first, for each scoring item e1, e2 to e m According to the degree of deviation between the measured value and the standard value, the score is assigned from s1 to s2 up to s m The greater the deviation, the higher the score, ranging from 0 to 10; then according to the formula T = w1×s1+w2×s2+...+w m ×s m Calculate the weighted total score, where w1, w2, and so on m is the weight value determined in step 2; Step 4: Calculate the resource load factor L based on the hospital resource data provided by the Internet of Things connection module, where L is determined by the following formula: L = C1×(number of currently occupied beds / total number of beds) + C2×(number of patients waiting for diagnosis / average processing capacity) + C3×(workload of medical staff / normal workload), where C1 is the weight coefficient of bed resources, C2 is the weight coefficient of the patient queue, and C3 is the weight coefficient of medical staff; the value range of C1 is 0.3 - 0.5, the value range of C2 is 0.2 - 0.4, the value range of C3 is 0.2 - 0.4, and C1 + C2 + C3 = 1; the value range of L is 0 - 1; Step 5: Adjust the triage thresholds th1, th2, and th3 according to the resource load factor L in Step 4, which are used for the determination of critical illness, severe illness, and sub-severe illness respectively. The adjustment formulas are as follows: th1 = B1 + K1×L; th2 = B2 + K2×L; th3 = B3 + K3×L; where B1, B2, and B3 are the basic thresholds for critical illness, severe illness, and sub-severe illness respectively, and K1, K2, and K3 are the corresponding threshold adjustment coefficients. The value range of B1 is 7 - 8, the value range of B2 is 5 - 6, the value range of B3 is 3 - 4, and the value range of K1, K2, and K3 is 0.1 - 1.0; Step 6: Compare the comprehensive patient score T in Step 3 with the adjusted thresholds th1, th2, and th3 in Step 5 to determine the grading result: If T ≥ th1, it is graded as critical illness, and the emergency treatment priority is set to level 1; if T < th1 and T ≥ th2, it is graded as severe illness, and the emergency treatment priority is set to level 2; if T < th2 and T ≥ th3, it is graded as sub-severe illness, and the emergency treatment priority is set to level 3; if T < th3, it is graded as ordinary symptoms, and the emergency treatment priority is set to level 4; Step 7: Assign a maximum waiting time limit to each grading result in Step 6: The maximum waiting time limit for critical illness patients is 5 minutes; the maximum waiting time limit for severe illness patients is 15 minutes; the maximum waiting time limit for sub-severe illness patients is 30 minutes; the maximum waiting time limit for ordinary symptoms patients is 60 minutes; Step 8: Package the grading result in Step 6, the treatment priority, the maximum waiting time limit in Step 7, together with the original score data and the threshold calculation process into a grading decision package, and transmit it to the central control processing unit. At the same time, record the complete data of this grading process into the system database for system self-learning and triage standard optimization.

7. The intelligent dynamic triage system for emergency patients according to claim 1 is characterized in that: The resource allocation module uses a multi-objective optimization algorithm to generate a diagnosis and treatment resource allocation plan. The multi-objective optimization algorithm includes the following steps: Step 1: Define the patient set P = {p1, p2, ..., p u }, where p1, p2, and p u They represent different patients, each with an attribute set, including grading results, waiting time, and special needs. The grading results are critical, severe, sub-severe, or common symptoms. The waiting time indicates the length of time from triage to the current moment. Special needs indicate whether the patient needs specific medical equipment or specialists. Step 2: Define the medical resource set R = {r1, r2, ..., r v }, where r1, r2, and r v They represent different medical resources, each of which has an attribute set, including resource type, available status, and processing capacity. Resource types include consulting rooms, physicians, nurses, and medical equipment. Available status values ​​are idle or occupied. Processing capacity represents the number of patients that can be processed by the resource per unit time. Step 3: Construct the resource allocation matrix M, where element m in M ​​is ij Indicates patient p i Assign to resource r j The decision variables, when m ij =1 means that the patient p i Assign to resource r j , when m ij =0 means no allocation; Step 4: Define the objective functions f1(M) and f2(M), where f1(M) represents the average waiting time weighted by urgency, and the calculation formula is: f1(M) = Σ(w i ×t i ) / u, where w i For patients i The urgency weight is 4 for critical illness, 3 for severe illness, 2 for sub-severe illness, and 1 for common symptoms; i For patients i The estimated total waiting time; u is the total number of patients; f2(M) represents the resource utilization rate, and the calculation formula is: f2(M) = Σ(number of occupied resources) / v, where v is the total number of medical resources; Step 5: Set the constraint conditions: Each patient can be assigned to only one main resource; Resource allocation meets the priority requirements of the patient grading result; Resource allocation meets the special needs of patients; Resources cannot exceed their processing capacity; Step 6: Apply genetic algorithm to solve the multi-objective optimization problem min{f1(M),-f2(M)} and generate the Pareto optimal solution set. The specific steps include: initializing the population and randomly generating resource allocation schemes that meet the constraints; calculating the fitness value of each scheme, which is determined by f1(M) and f2(M); generating a new generation of population through selection, crossover and mutation operations; repeating the above process until the preset number of iterations or convergence condition is reached; Step 7: Select the final solution from the Pareto optimal solution set according to the current hospital strategy. When the hospital is in peak period, give priority to f1(M), and when the hospital is in off-peak period, give priority to f2(M), and generate the resource allocation result; Step 8: Transmit the resource allocation results to the diagnosis and treatment process management module to guide patient diversion and medical resource arrangement, and record the complete data of this resource allocation process for continuous system optimization.

8. The intelligent dynamic triage system for emergency patients according to claim 1 is characterized in that: The diagnosis and treatment process management module includes: A patient tracking unit that tracks the patient's location and status in real time through a hospital positioning system and an electronic wristband, wherein the hospital positioning system includes an indoor positioning sensor network and a data processing server, the electronic wristband has a unique identification code and is equipped with a physiological parameter collection sensor, and the patient tracking unit integrates the collected patient location data and status data into real-time patient trajectory information; A process monitoring unit is connected to the patient tracking unit, receives the patient's real-time trajectory information, monitors the residence time and flow process of each patient at different diagnosis and treatment nodes, and the diagnosis and treatment nodes include the triage desk, waiting area, consulting room, examination room, treatment room and observation room. The process monitoring unit calculates the deviation between the actual residence time of the patient at each node and the expected time, and generates process monitoring data; an abnormal warning unit, connected to the patient tracking unit and the process monitoring unit, receiving the patient's real-time trajectory information and process monitoring data, and triggering an alarm when an abnormal change occurs in the patient's status data or the waiting time exceeds a safety threshold, wherein the abnormal change includes a physiological parameter exceeding a safety range or an abnormal position, and the safety threshold is set according to the patient classification result, and the abnormal warning unit generates warning information and determines the warning level; and The dynamic feedback unit is connected to the abnormal warning unit to receive warning information and warning levels. When a change in patient status is detected and re-triage is required, the patient status data is transmitted to the central control processing unit to trigger a re-evaluation process and send a notification to relevant medical staff.

9. The intelligent dynamic triage system for emergency patients according to claim 1, characterized in that: It also includes a human-computer interaction module, which is connected to the central control processing unit, and the human-computer interaction module includes: Medical staff workstations are equipped with high-resolution display screens and touch operation interfaces, which are used to display triage decision suggestions and resource allocation plans, show patient distribution status and resource usage in real time, and allow medical staff to make necessary manual interventions and adjustments after authorization authentication. The medical staff workstations serve as operation centers, distributing relevant instructions to patient information display terminals, voice interaction systems, and mobile application interfaces; The patient information display terminal is connected to the medical staff workstation and installed in the waiting area and the diagnosis and treatment area. It uses privacy protection display technology to display personal triage results, estimated waiting time and treatment process through the patient's unique identification code, and provides hospital layout navigation and diagnosis and treatment progress query functions. The patient information display terminal transmits the patient query request back to the medical staff workstation for processing; A voice interaction system, connected to the medical staff workstation, equipped with a noise reduction microphone array and a natural language processing unit, supports medical staff to quickly query and update patient information, execute resource allocation instructions, and respond to emergency calls through voice commands. The voice interaction system transmits the processed voice instructions to the medical staff workstation for execution; and The mobile application interface is connected to the medical staff workstation, supporting medical staff to remotely access the system and receive key reminders through mobile devices authorized by the hospital, thereby realizing real-time monitoring of patient status and remote consultation. The mobile application interface adopts an encrypted communication protocol to ensure data transmission security, and sets up a multi-level access permission management mechanism. The remote operation instructions received by the mobile application interface are synchronized to the patient information display terminal and voice interaction system through the medical staff workstation.

10. An intelligent dynamic triage method for emergency patients, based on the intelligent dynamic triage system for emergency patients according to any one of claims 1 to 9, characterized in that: The steps include: Step 1: Collect patient data, including obtaining basic patient information, physiological parameters, symptom descriptions, and historical medical records to generate patient data; Step 2: Perform AI analysis, receive the patient data in step 1, analyze and process the patient data according to the pre-trained classification model, and generate preliminary triage results; Step 3: Monitor hospital resources and collect hospital resource data in real time, including the resource status of each department of the hospital, the availability of medical equipment, and the workload of medical staff; Step 4: Execute dynamic grading decision, receive the preliminary triage result in step 2 and the hospital resource data in step 3, grade and classify the patients according to the preset decision rules, and generate a grading result, wherein the grading result includes grading the patients into critical, severe, sub-severe and common symptoms; Step 5: Perform resource allocation, receive the classification result in step 4 and the hospital resource data in step 3, and dynamically generate a diagnosis and treatment resource allocation plan and a diagnosis and treatment sequence arrangement according to the classification result and the hospital resource data to form a resource allocation result; Step 6: Manage the diagnosis and treatment process, receive the resource allocation results in step 5, track the entire process of patients from triage to treatment, and collect and update patient status data in real time; Step 7: Perform dynamic adjustments. When the patient status data in step 6 changes, the updated patient status data is returned to step 2 for re-evaluation to generate an updated preliminary triage result, and the dynamic grading decision in step 4 and the resource allocation in step 5 are triggered in turn to form a closed-loop feedback system to dynamically update the triage priority and resource allocation plan.

Images