Management method based on intelligent physical examination queuing calling system

Abstract

The invention relates to the technical field of medical informatization, and discloses a management method based on an intelligent physical examination queuing and calling system, which is characterized in that a physical examination queuing and calling management process is modeled into a Markov decision process by constructing an intelligent calling model, and a queuing strategy and a calling management scheme are dynamically adjusted. The model defines a multi-dimensional state space including physical examination person information, queuing conditions, physical examination equipment and calling equipment, and an action space including actions of adjusting the sequence of physical examination items, guiding physical examination persons, sending calling notifications and the like. By designing a reward function which comprehensively considers the satisfaction degree, the waiting time, the equipment utilization rate and the emergency response capability of physical examinees, a deep reinforcement learning algorithm is adopted to train a model, and a strategy is continuously optimized by taking maximization of accumulated rewards as a target. The physical examination efficiency and the service quality are improved, the waiting time of physical examinees is shortened, the coping ability of the system to emergencies is enhanced, and the physical examination system has remarkable technical advantages and application value.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a management method based on an intelligent physical examination queuing calling system. Background Art

[0002] With the rapid development of modern medical services, physical examinations, as an important part of preventive medicine, are increasingly attracting people's attention for their efficiency and service quality. In traditional physical examination queuing and calling systems, a simple electronic calling system is often relied on to coordinate the queuing and calling processes of physical examinees. However, these traditional methods have some significant defects when facing sudden situations such as large flow of people during peak physical examination periods, the complexity of multi-item physical examinations, and equipment failures.

[0003] The existing system usually follows a fixed queuing order and cannot be flexibly adjusted according to the actual situation of the examinee, resulting in some urgent or important items being delayed, reducing the examinee's satisfaction and experience.

[0004] During the physical examination, if a piece of physical examination equipment fails or the examinee fails to respond to the call in time, the traditional system often lacks an effective immediate response mechanism, which may lead to increased equipment idle time, reduced equipment utilization, and prolonged waiting time for other examinees.

[0005] The existing system is unable to analyze the data of the physical examination site in real time, such as the number of examinees, project duration, equipment status, etc., and therefore cannot dynamically adjust the queuing strategy and call management plan, which limits the system's ability to respond to emergencies and reduces the efficiency of the overall physical examination process. Summary of the invention

[0006] The purpose of the present invention is to provide a management method based on an intelligent physical examination queuing call system to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solution: a management method based on an intelligent physical examination queuing call system, the method comprising:

[0008] Construct an intelligent call model for the medical examination queue call system, model the medical examination queue and call management process as a Markov decision process, and dynamically adjust the call management strategy of the medical examination queue by analyzing the real-time data of the medical examination site;

[0009] The steps to build an intelligent call model include:

[0010] Define the state space, including the examinee information dimension, queuing status dimension, examination equipment dimension and call equipment dimension at the examination site;

[0011] Define the action space, including the possible actions that the agent can take when handling medical examination queues and call exceptions, including adjusting the order of medical examination items, guiding the examinee to an idle device, sending a call notification, and activating a backup call device;

[0012] Design a reward function to evaluate the results of actions based on examinee satisfaction, examinee waiting time, equipment utilization, equipment utilization, and emergency response capabilities, and give corresponding rewards or penalties;

[0013] Formulate strategies, rules for selecting actions in a given state, and continuously optimize them through machine learning algorithms to maximize cumulative rewards;

[0014] Use deep reinforcement learning algorithm to train the intelligent call model.

[0015] Preferably, the physical examinee information dimension includes the physical examinee's scheduled items, completed items and estimated completion time; the queuing situation dimension includes the real-time queue length, average waiting time and longest waiting time of each physical examination item; the physical examination equipment dimension includes the usage status, equipment efficiency and estimated available time of the physical examination equipment; the calling equipment dimension includes the working status of the pager, the time of the last call, the call success rate and the response time.

[0016] Preferably, the indicators of the reward function include examinee satisfaction, examinee waiting time, equipment utilization rate and emergency response capability. The action results are evaluated by the following formula and corresponding rewards or penalties are given:

[0017] R=α·S satis faction -β·T wa iting +γ·U device +δ·R response

[0018] Among them, S satis faction It indicates the satisfaction of the examinee, obtained through questionnaire survey or feedback system, with a value range of [0,1], where 1 indicates complete satisfaction; T wa iting It represents the average waiting time of the examinee, in minutes, calculated through real-time data; U device Represents the equipment utilization rate, which is calculated by dividing the actual equipment usage time by the total equipment available time. The value range is [0,1]. responseIt represents the emergency response capability, which is determined by measuring the response speed and processing efficiency of the system under abnormal conditions. The value range is [0,1], where 1 represents the fastest response speed and the highest processing efficiency. α, β, γ, and δ are weight coefficients, which respectively represent the importance of each indicator in the reward function and satisfy α+β+γ+δ=1. The specific value is adjusted according to the actual application scenario and requirements.

[0019] Preferably, a deep Q network DQN algorithm is used to train the intelligent call model.

[0020] Preferably, the step of training the intelligent call model includes:

[0021] S1: Design the structure of a deep Q network to represent the strategy. The input of the network is the current state, and the output is the probability distribution of each action. Use the deep Q network to parameterize the strategy, denoted as Q(s, a; θ), where s represents the state, a represents the action, and θ represents the parameters of the neural network.

[0022] S2: Based on the defined state space and action space, construct the representation of the input layer and output layer of the neural network; the state space includes the dimensions of the examinee information, the queuing situation, the examination equipment and the calling equipment at the physical examination site, and the state information will be encoded as the input features of the neural network; the action space is the output of the neural network, representing the possible actions that the model may take;

[0023] S3: By simulating the physical examination queuing and calling process, the four-tuple (s, a, r, s) of state, action, reward and next state is collected as training data;

[0024] S4: Using the collected training data, the parameters of the neural network are updated through the back propagation algorithm and the gradient descent method. During the training process, the model continuously tries different actions and adjusts its strategy based on the feedback obtained from the reward function to maximize the cumulative reward; further including:

[0025] S401: For each training iteration, a batch of state transition samples (s, a, r, s) are randomly extracted from the experience replay memory, where s is the current state, a is the action taken, r is the reward, and s′ is the next state;

[0026] S402: According to the Bellman equation, calculate the loss function L(θ), which is L(θ) = (r + γmax a 'Q(s',a';θ')-Q(s,a;θ)) 2 , where γ is the discount factor, Q(s,a;θ) is the estimate of the state and action value function under the current strategy, and θ′ is the parameter of the target network;

[0027] S403: Using the gradient descent algorithm to update the neural network parameter θ so as to minimize the loss function L(θ);

[0028] S5: During the training process, the performance of the intelligent call model is regularly evaluated, including the examinee satisfaction, waiting time, equipment utilization, and emergency response capability indicators, and the weight coefficient of the reward function and the learning rate of the model are adjusted according to the evaluation results;

[0029] S6: Repeat steps S4 and S5 until a preset number of training rounds is reached.

[0030] Preferably, in step 403, the formula for updating the parameters of the gradient descent algorithm is:

[0031]

[0032] Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

[0033] Preferably, an ε-greedy strategy is adopted to dynamically adjust the exploration and utilization ratio of the model in the early and late stages of training.

[0034] Preferably, after the training of the intelligent call model is completed, the model is tested and evaluated, and a strategy combining simulation and field application is adopted to compare the performance of the queuing and calling solutions generated by the model with the traditional management methods in terms of examinee waiting time, call accuracy, system response speed and examination process efficiency.

[0035] Preferably, MATLAB / Simulink simulation software is used to construct a simulation environment for physical examination queuing and calling.

[0036] Preferably, in the constructed simulation environment, various physical examination scenarios are simulated, including the time difference of different physical examination items, the arrival distribution of physical examinees, and the parallel operation of multiple types of physical examination equipment; by adjusting the simulation parameters, a variety of test cases are generated to cover various situations encountered in actual operations.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The intelligent call model can analyze the data of the physical examination site in real time, such as the number of examinees, project duration, equipment status, etc., and dynamically adjust the queuing strategy and call management plan based on these data, avoiding the inefficiency caused by a fixed queuing order and ensuring the smooth progress of the physical examination process, thereby significantly improving the overall physical examination efficiency.

[0039] By defining a rich action space, including adjusting the order of physical examination items, guiding the examinee to idle equipment, sending call notifications and starting backup call equipment, the intelligent call model can quickly respond to emergencies such as equipment failure and the examinee's failure to respond to the call in time, effectively reducing equipment idle time, improving equipment utilization, and reducing the waiting time for other examinees.

[0040] The intelligent call model fully considers the actual situation of the examinee, such as the scheduled items, completed items and expected completion time, and can arrange the order of physical examinations more flexibly to ensure that urgent or important items are handled in a timely manner. At the same time, by optimizing call management, call confusion and missed calls are reduced, and the satisfaction and experience of the examinee are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A processing step diagram of a management method based on an intelligent physical examination queuing call system according to the present invention;

[0042] Figure 2 Schematic diagram of the design of the reward function;

[0043] Figure 3 This is the training flowchart of the intelligent calling model. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0045] See also Figure 1-3 The present invention provides a technical solution: a management method based on an intelligent physical examination queue call system, which models the physical examination queue and call management process as a Markov decision process, and trains the intelligent call model through a deep reinforcement learning algorithm to dynamically adjust the call management strategy of the physical examination queue. The following is a specific implementation method of the method:

[0046] First, the intelligent call model is constructed, which includes the following steps:

[0047] Define the state space: The state space covers multiple dimensions of the physical examination site, including:

[0048] Physical examinee information dimension: includes the physical examinee's scheduled items, completed items and estimated completion time. This information is used to understand the needs and progress of each physical examinee.

[0049] Queuing situation dimension: including the real-time queue length, average waiting time and longest waiting time of each physical examination item. These data reflect the degree of congestion and service efficiency at the physical examination site.

[0050] Physical examination equipment dimension: includes the usage status, equipment efficiency, and estimated available time of physical examination equipment, which helps to understand the operating status and availability of the equipment.

[0051] Call device dimensions: including the working status of the pager, the last call time, call success rate and response time. This information is crucial for monitoring the performance and reliability of the call system.

[0052] Define the action space: The action space contains various actions that the agent may take when handling medical examination queues and call exceptions, including:

[0053] Adjust the order of physical examination items: Intelligently adjust the order of physical examination items according to real-time conditions to optimize the process.

[0054] Guide examinees to idle equipment: Guide examinees to idle equipment to reduce waiting time.

[0055] Send call notification: notify the examinee to go to the designated examination area or equipment through the calling system.

[0056] Start the backup calling device: When the main calling device fails, start the backup device to ensure the continuity of the system.

[0057] Design reward function: The reward function is used to evaluate the results of the actions taken by the agent and give corresponding rewards or penalties based on the examinee's satisfaction, examinee's waiting time, equipment utilization, and emergency response capabilities. Specific indicators include:

[0058] Physical examinee satisfaction: calculated based on the physical examinee's feedback and evaluation.

[0059] Physical examinee waiting time: The shorter the waiting time, the higher the reward.

[0060] Equipment utilization: The more efficiently the equipment is used, the greater the reward.

[0061] Emergency response capability: In abnormal situations, the speed and effectiveness of the agent’s response will also be taken into account in the reward.

[0062] Formulate strategies: The strategy is the rule for selecting actions in a given state. The strategy is continuously optimized through machine learning algorithms (especially deep reinforcement learning algorithms) to maximize the cumulative reward. The agent will choose the best action to achieve the long-term goal based on the current state.

[0063] Using deep reinforcement learning algorithm to train intelligent call model:

[0064] The intelligent call model is trained using a deep reinforcement learning algorithm. By simulating the actual scenarios of physical examination queues and calls, the model parameters are continuously iterated and optimized, so that the intelligent agent can make the best decision in various situations.

[0065] This outbreak is further described below in conjunction with Examples 1 to 3:

[0066] Embodiment 1:

[0067] The reward function is used to evaluate the performance of the agent in the medical examination queue and call management process, and to give corresponding rewards or penalties accordingly to optimize the management strategy. This embodiment is used to illustrate the design method of the reward function.

[0068] The indicators for designing the reward function include examinee satisfaction, examinee waiting time, equipment utilization, and emergency response capability. The action results are evaluated using the following formula and corresponding rewards or penalties are given:

[0069] R=α·S satis faction -β·T wa iting +γ·U device +δ·R response

[0070] Among them, S satis faction It indicates the satisfaction of the examinee, which is obtained through questionnaire survey or feedback system, and the value range is [0,1], where 1 indicates complete satisfaction. This indicator reflects the examinee's overall evaluation of the queuing and call management service. In practical applications, the feedback data of the examinees can be collected regularly and the average value can be calculated as S satis faction The value of .

[0071] T wa iting It indicates the average waiting time of the examinees in minutes, which is calculated through real-time data. It reflects the waiting situation of the examinees in the queue. In order to reduce the waiting time of the examinees, the system should optimize the queuing strategy as much as possible to improve service efficiency.

[0072] U device Indicates the equipment utilization rate, which is calculated by dividing the actual equipment usage time by the total equipment available time, and the value range is [0,1]. This indicator reflects the efficiency of equipment use, and improving equipment utilization helps reduce resource waste and improve overall service efficiency. The calculation formula is:

[0073]

[0074] Rresponse It represents the emergency response capability, which is determined by measuring the system's response speed and processing efficiency under abnormal conditions. The value range is [0,1], where 1 indicates the fastest response speed and the highest processing efficiency. This indicator can be evaluated by simulating the system response time and processing results under abnormal conditions.

[0075] α, β, γ, and δ are weight coefficients, representing the importance of each indicator in the reward function, and satisfying α+β+γ+δ=1. The specific values ​​are adjusted according to the actual application scenario and needs. For example, if more attention is paid to the satisfaction and waiting time of the examinee, the values ​​of α and β can be appropriately increased; if more attention is paid to equipment utilization and emergency response capabilities, the values ​​of γ and δ can be appropriately increased.

[0076] Assume that during a physical examination queue and call management process, the system obtains the following data:

[0077] The satisfaction of the examinee satis faction =0.85;

[0078] Average waiting time for physical examination patients wa iting = 15 minutes;

[0079] Equipment utilization rate U device =0.75;

[0080] Emergency Response Capability response =0.90;

[0081] According to the reward function formula, the reward value of this management can be calculated:

[0082] R=0.3·0.85-0.25·15+0.2·0.75+0.25·0.90=0.255-3.75+0.15+0.225=-2.92

[0083] Since the reward value is negative, it indicates that the management strategy needs to be optimized. The system can adjust the strategy based on the evaluation results to increase the future reward value.

[0084] Embodiment 2:

[0085] This embodiment is used to describe the training process of the intelligent call model.

[0086] The deep Q network DQN algorithm is used to train the intelligent call model. The steps of training the intelligent call model include:

[0087] S1: Design a deep Q network, which is used to represent the strategy. Its input is the current state and its output is the probability distribution of each action. The strategy is parameterized using the deep Q network, denoted as Q(s,a;θ), where s represents the state, a represents the action, and θ represents the parameters of the neural network. The deep Q network consists of a multi-layer neural network, including an input layer, a hidden layer, and an output layer.

[0088] S2: Construct the input layer and output layer of the neural network according to the defined state space and action space. The state space contains multi-dimensional data of the examinee information at the physical examination site (such as the number of examinees, physical examination item requirements, etc.), queuing conditions (such as the length of each queue, waiting time, etc.), physical examination equipment (such as equipment status, number of equipment, etc.) and calling equipment (such as call frequency, call content, etc.). These data will be encoded as input feature vectors of the neural network. The action space corresponds to the output layer of the neural network, which represents various actions that the intelligent agent may take, such as adjusting the order of physical examination items, guiding the examinee to an idle device, etc. Each action corresponds to a neuron in the output layer, and the output value is the value estimate of the action.

[0089] S3: By simulating the physical examination queuing and calling process, the four-tuple (s, a, r, s) of state, action, reward and next state is collected as training data. In the simulation environment, the agent selects an action based on the current state, and the environment feedbacks the reward and next state based on the action. The training data is used to update the parameters of the neural network so that it gradually approaches the optimal Q function.

[0090] S4: Using the collected training data, the parameters of the neural network are updated through the back propagation algorithm and the gradient descent method. During the training process, the model continuously tries different actions and adjusts its strategy based on the feedback obtained from the reward function to maximize the cumulative reward; further including:

[0091] S401: For each training iteration, a batch of state transition samples (s, a, r, s) are randomly extracted from the experience replay memory, where s is the current state, a is the action taken, r is the reward, and s′ is the next state; the experience replay memory is a buffer of limited size used to store recent state transition samples to break the correlation between samples and improve training stability.

[0092] S402: According to the Bellman equation, calculate the loss function L(θ), which is L(θ) = (r + γmax a ' Q(s',a';θ')-Q(s,a;θ)) 2, where γ is the discount factor, Q(s,a;θ) is the estimate of the state and action value function under the current strategy, and θ′ is the parameter of the target network; the target network is a stable network that periodically copies parameters from the main network to calculate the target value to stabilize the training process.

[0093] S403: Update the neural network parameters θ using the gradient descent algorithm to minimize the loss function L(θ); the formula for updating the parameters using the gradient descent algorithm is:

[0094]

[0095] Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

[0096] S5: During the training process, the performance of the intelligent call model is regularly evaluated, including the indicators of patient satisfaction, waiting time, equipment utilization, and emergency response capability. According to the evaluation results, the weight coefficient of the reward function and the learning rate of the model are adjusted to optimize the model performance. The design of the reward function should comprehensively consider multiple goals such as patient satisfaction, equipment utilization, and emergency response capability to guide the model to learn a reasonable call strategy.

[0097] S6: Repeat steps S4 and S5 until the preset number of training rounds is reached. During the training process, the model will gradually learn the optimal calling strategy to maximize the cumulative reward while meeting the requirements of the examinee's satisfaction, equipment utilization, and emergency response capabilities. Through continuous iterative training, the performance of the intelligent calling model will gradually improve, and ultimately achieve efficient and intelligent physical examination calling services.

[0098] In the early stages of training, the model needs to explore different actions to find a better strategy; in the later stages of training, it should make more use of the knowledge it has learned to make decisions. To achieve this balance, methods such as the ε-greedy strategy or soft strategy are used to dynamically adjust the ratio of exploration and utilization to improve training efficiency and model performance.

[0099] The following is a specific example showing the application of the above intelligent call model in a physical examination center:

[0100] In the physical examination center, a large number of people need to undergo multiple examinations every day. In order to optimize the physical examination process and improve the satisfaction of the examinees and the utilization rate of the equipment, the physical examination center introduced an intelligent call model based on the deep Q network (DQN) algorithm. The model first constructs the input feature vector of the neural network based on the multi-dimensional data of the physical examination site, including the physical examination's scheduled items, completed items, and estimated completion time, as well as the real-time queue length, average waiting time, equipment usage status, and working status of the calling equipment for each physical examination item. Then, the model learns a set of optimal call strategies through training.

[0101] In actual applications, when the examinee arrives at the physical examination center and registers, the intelligent calling model will automatically assign the best examination order to the examinee according to the current status, and notify the examinee to go to the designated examination area through the calling device. If the queue for a certain examination item is too long, the model will adjust the strategy in time to guide the examinee to do other items first to balance the load of each examination area. At the same time, if a certain equipment fails or needs maintenance, the model can also respond quickly and adjust the calling strategy to ensure that the physical examination process is not affected. By introducing the intelligent calling model, the physical examination center has successfully optimized the physical examination process. The waiting time of the examinee has been significantly shortened, the equipment utilization rate has been improved, and the satisfaction of the examinee has also increased. In addition, the model also has emergency response capabilities, and can make rapid adjustments in emergencies such as equipment failures, ensuring the smooth progress of the physical examination.

[0102] Embodiment 3:

[0103] After the intelligent call model training is completed, in order to ensure the effectiveness and practicality of the model, it needs to be accurately tested and evaluated. The test evaluation adopts a strategy that combines simulation and field application to compare the performance of the queuing and calling solutions generated by the model with the traditional management methods in terms of the waiting time of the examinees, call accuracy, system response speed and efficiency of the physical examination process. The following is a specific implementation method:

[0104] Conduct simulation test:

[0105] Build a simulation environment: Use MATLAB / Simulink simulation software to build a simulation environment for physical examination queuing and calling. This environment can simulate various situations at the physical examination site, including the arrival of the examinee, queuing, physical examination process, and call notification. In the constructed simulation environment, simulate various physical examination scenarios to comprehensively evaluate the performance of the model. These scenarios include but are not limited to: ① Differences in the duration of different physical examination items: According to the time requirements of the actual physical examination items, different physical examination durations are set to simulate the complexity of different items. ② Arrival distribution of physical examinees: Based on historical data or assumptions, simulate the arrival time distribution of physical examinees, such as uniform distribution, Poisson distribution, etc., to reflect the arrival situation in actual physical examinations. ③ Parallel operation of multiple types of physical examination equipment: Simulate the parallel operation of multiple physical examination equipment (such as B-ultrasound, X-ray, blood testing, etc.), including the use status, efficiency, and maintenance time of the equipment.

[0106] Generate test cases: Generate diverse test cases by adjusting simulation parameters, such as the arrival rate of examinees, the duration of examination items, the number of devices, etc. These cases should cover various situations that may be encountered in actual operations, such as congestion during peak hours, equipment failures, etc., to ensure the stability and adaptability of the model under different circumstances.

[0107] Performance evaluation: Run the intelligent call model in a simulation environment and record the queue and call plans it generates. At the same time, run the traditional management method for comparison. By collecting and analyzing data, evaluate the performance of the two methods in terms of patient waiting time, call accuracy, system response speed, and examination process efficiency.

[0108] Conduct field application tests:

[0109] Select the test site: Select a representative physical examination center or medical institution as the test site to ensure the authenticity and diversity of the test environment.

[0110] Deploy a medical examination queuing call system based on the intelligent call model at the test site and ensure the stable operation of the system. Collect data in actual operation, including the waiting time of the examinee, call records, system response time, etc. By comparing and analyzing these data, evaluate the performance and effect of the intelligent call system in actual application. According to the results of the field application test, collect user feedback and opinions, and make necessary adjustments and optimizations to the intelligent call model. At the same time, traditional management methods can also be improved to improve the efficiency and service quality of the overall medical examination process.

[0111] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0112] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A management method based on an intelligent physical examination queuing call system, characterized in that: The method comprises: Construct an intelligent call model for the medical examination queue call system, model the medical examination queue and call management process as a Markov decision process, and dynamically adjust the call management strategy of the medical examination queue by analyzing the real-time data of the medical examination site; The steps to build an intelligent call model include: Define the state space, including the examinee information dimension, queuing status dimension, examination equipment dimension, and call equipment dimension at the examination site; Define the action space, including the possible actions that the agent can take when handling medical examination queues and call exceptions, including adjusting the order of medical examination items, guiding the examinee to an idle device, sending a call notification, and activating a backup call device; Design a reward function to evaluate the results of actions based on examinee satisfaction, examinee waiting time, equipment utilization, equipment utilization, and emergency response capabilities, and give corresponding rewards or penalties; Formulate strategies, rules for selecting actions in a given state, and continuously optimize them through machine learning algorithms to maximize cumulative rewards; Use deep reinforcement learning algorithm to train the intelligent call model.

2. A management method based on the intelligent physical examination queuing call system according to claim 1, characterized in that: The physical examinee information dimension includes the physical examinee's scheduled items, completed items and estimated completion time; the queuing status dimension includes the real-time queue length, average waiting time and longest waiting time of each physical examination item; the physical examination equipment dimension includes the usage status, equipment efficiency and estimated available time of the physical examination equipment; the calling equipment dimension includes the working status of the pager, the last call time, the call success rate and the response time.

3. A management method based on the intelligent physical examination queuing call system according to claim 2, characterized in that: The indicators for designing the reward function include examinee satisfaction, examinee waiting time, equipment utilization, and emergency response capability. The action results are evaluated using the following formula and corresponding rewards or penalties are given: R=α·S satis faction -β·T wa iting +γ·U device +δ·R response Among them, S satis faction It indicates the satisfaction of the examinee, obtained through questionnaire survey or feedback system, with a value range of [0,1], where 1 indicates complete satisfaction; T wa iting It represents the average waiting time of the examinee, in minutes, calculated through real-time data; U device Represents the equipment utilization rate, which is calculated by dividing the actual equipment usage time by the total equipment available time. The value range is [0,1]. response It represents the emergency response capability, which is determined by measuring the response speed and processing efficiency of the system under abnormal conditions. The value range is [0,1], where 1 represents the fastest response speed and the highest processing efficiency. α, β, γ, and δ are weight coefficients, which respectively represent the importance of each indicator in the reward function and satisfy α+β+γ+δ=1. The specific value is adjusted according to the actual application scenario and requirements.

4. A management method based on the intelligent physical examination queuing call system according to claim 1, characterized in that: The deep Q network DQN algorithm is used to train the intelligent calling model.

5. A management method based on the intelligent physical examination queuing call system according to claim 4, characterized in that: The steps to train the smart call model include: S1: Design the structure of a deep Q network to represent the strategy. The input of the network is the current state, and the output is the probability distribution of each action. Use the deep Q network to parameterize the strategy, denoted as Q(s, a; θ), where s represents the state, a represents the action, and θ represents the parameters of the neural network. S2: Based on the defined state space and action space, construct the representation of the input layer and output layer of the neural network; the state space includes the dimensions of the examinee information, the queuing situation, the examination equipment and the calling equipment at the physical examination site, and the state information will be encoded as the input features of the neural network; the action space is the output of the neural network, representing the possible actions that the model may take; S3: By simulating the physical examination queuing and calling process, the four-tuple (s, a, r, s,) of state, action, reward and next state is collected as training data; S4: Using the collected training data, the parameters of the neural network are updated through the back propagation algorithm and the gradient descent method. During the training process, the model continuously tries different actions and adjusts its strategy based on the feedback obtained from the reward function to maximize the cumulative reward; further including: S401: For each training iteration, a batch of state transition samples (s, a, r, s,) are randomly extracted from the experience replay memory, where s is the current state, a is the action taken, r is the reward, and s' is the next state; S402: According to the Bellman equation, calculate the loss function L(θ), which is L(θ) = (r + γmax a' Q(s',a';θ')-Q(s,a;θ)) 2 , where γ is the discount factor, Q(s,a;θ) is the estimate of the state and action value function under the current strategy, and θ ' are the parameters of the target network; S403: Using the gradient descent algorithm to update the neural network parameter θ so as to minimize the loss function L(θ); S5: During the training process, the performance of the intelligent call model is regularly evaluated, including the examinee satisfaction, waiting time, equipment utilization, and emergency response capability indicators, and the weight coefficient of the reward function and the learning rate of the model are adjusted according to the evaluation results; S6: Repeat steps S4 and S5 until a preset number of training rounds is reached.

6. A management method based on the intelligent physical examination queuing call system according to claim 5, characterized in that: In step 403, the formula for updating the parameters of the gradient descent algorithm is: Among them, θ represents the model parameters, including all weights and biases that the model needs to learn; ω represents the learning rate, which is used to control the step size of parameter update; represents the gradient of the loss function J(θ) with respect to the parameter θ, is a vector pointing to the direction where the loss function grows fastest; := represents the assignment operation, that is, updating the value of the parameter θ.

7. A management method based on the intelligent medical examination queuing call system according to claim 6, characterized in that: The ε-greedy strategy is adopted to dynamically adjust the exploration and utilization ratio of the model in the early and late stages of training.

8. The management method based on the intelligent physical examination queuing call system according to claim 1 is characterized by: After the training of the intelligent call model is completed, the model is tested and evaluated. A strategy combining simulation and field application is adopted to compare the performance of the queuing and calling solutions generated by the model with the traditional management methods in terms of examinee waiting time, call accuracy, system response speed and examination process efficiency.

9. A management method based on the intelligent physical examination queuing call system according to claim 8, characterized in that: MATLAB / Simulink simulation software is used to build a simulation environment for physical examination queuing and calling.

10. A management method based on the intelligent physical examination queuing call system according to claim 9, characterized in that: In the constructed simulation environment, various physical examination scenarios are simulated, including the time difference of different physical examination items, the arrival distribution of physical examination subjects, and the parallel operation of multiple types of physical examination equipment; By adjusting simulation parameters, diverse test cases are generated to cover various situations encountered in actual operations.

Images