An outpatient process intelligent scheduling and resource optimization method and system

Abstract

The application discloses an outpatient process intelligent scheduling and resource optimization method and system, relates to the field of medical information intelligent scheduling and resource optimization, and comprises the following steps: collecting images of waiting areas of various diagnosis and treatment points, extracting personnel density features, combining a multi-target tracking algorithm to obtain personnel moving tracks, and constructing a diagnosis and treatment point people flow network topology graph; analyzing people flow convergence and dispersion features based on the topology graph, combining historical treatment data to predict future congestion trends of various diagnosis and treatment points; and then, according to congestion evolution and real-time reception capacity, establishing an optimization model with the minimum average waiting time as the target, generating an optimal shunt path combination, and planning a personalized navigation path for patients. The application realizes dynamic intelligent scheduling of the outpatient process, and effectively improves medical resource utilization and patient treatment experience.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical information intelligent scheduling and resource optimization, in particular to a method and system for outpatient process intelligent scheduling and resource optimization. BACKGROUND

[0002] Currently, there are problems of patient concentration, waiting congestion and uneven distribution of medical resources in the outpatient treatment process. Traditional scheduling methods rely on manual experience or simple queuing management, which is difficult to perceive the flow dynamics and waiting conditions of the diagnosis and treatment points in real time, resulting in overload in some diagnosis and treatment links, while other resources are idle, which not only reduces the overall efficiency of the hospital, but also increases the waiting time of patients and the burden of the treatment experience. With the development of video monitoring, image processing and artificial intelligence technology, the combination of real-time flow data and diagnosis and treatment resource optimization has become an important direction to improve the efficiency of outpatient operation. The present application provides a method for outpatient process intelligent scheduling and resource optimization, which can realize dynamic perception and optimal shunting of outpatient flow through image acquisition, personnel trajectory analysis and prediction model construction, so as to improve the utilization rate of medical resources and reduce the average waiting time. SUMMARY

[0003] The purpose of the present application is to provide a method and system for outpatient process intelligent scheduling and resource optimization, which realizes dynamic optimization and scheduling of outpatient process through intelligent flow monitoring and prediction model, shortens the waiting time of patients and improves the utilization efficiency of medical resources.

[0004] The method for outpatient process intelligent scheduling and resource optimization provided in the embodiment of the present application comprises the following steps:

[0005] Collecting the image of the waiting area of each diagnosis and treatment point, extracting the personnel density features in the waiting area, and combining the multi-target tracking algorithm to extract the personnel moving track;

[0006] According to the personnel density features and the personnel moving track features, a diagnosis and treatment point flow network topology graph is constructed;

[0007] Analyzing the flow convergence points and dispersion points in the diagnosis and treatment point flow network topology graph, and calculating the flow transfer probability between each diagnosis and treatment point;

[0008] Based on the flow transfer probability and the historical treatment data of the diagnosis and treatment points, the congestion evolution trend of each diagnosis and treatment point in the future time window is predicted;

[0009] According to the congestion evolution trend and the real-time treatment capacity of the diagnosis and treatment points, a path optimization model with the minimum average waiting time of patients as the objective function is established, and the optimal shunting path combination is calculated;

[0010] Generate specific shunt scheduling instructions according to the optimal shunt path combination, and plan a personalized navigation path for the patient based on the personnel movement trajectory features.

[0011] Further, the waiting area image of each medical treatment point is collected, the personnel density features in the waiting area are extracted, and the personnel movement trajectory is extracted by combining a multi-target tracking algorithm, including:

[0012] Adjust the collection parameters according to the illumination information and people flow information in the waiting area image;

[0013] Distortion correction and perspective transformation are performed on the waiting area image to obtain an overhead image, and standardization processing is performed on the overhead image to obtain a standardized image;

[0014] Based on the personnel distribution in the standardized image, a grid unit is divided, the number of personnel in the grid unit is counted to obtain a grid density value, a smoothing radius is determined according to the grid density value, and the grid density value is smoothed using the smoothing radius to obtain a continuous density distribution;

[0015] The time-space change rate of the continuous density distribution is calculated to obtain a density change feature, and the continuous density distribution and the density change feature are combined to construct a waiting area personnel density feature;

[0016] The feature vector of the personnel target in the standardized image is extracted, a multi-target tracking algorithm is used to calculate the similarity between consecutive frames based on the feature vector to obtain a correlation matrix, and the correlation matrix is used to simultaneously track multiple personnel targets to obtain a trajectory sequence, and the personnel movement trajectory is constructed according to the trajectory sequence.

[0017] Further, according to the personnel density feature and the personnel movement trajectory feature, a medical treatment point people flow network topology graph is constructed, including:

[0018] According to the personnel density feature, the density distribution value and the density change value are obtained, and the service capacity value of the network node is calculated;

[0019] Based on the personnel movement trajectory feature, the flow intensity value between the network nodes is calculated, and the ratio of the flow intensity value to the service capacity value is set as the edge weight value between the network nodes;

[0020] The medical treatment point is set as a network node, and the medical treatment point people flow network topology graph is constructed according to the service capacity value and the edge weight value.

[0021] Further, the people flow convergence point and the dispersion point in the medical treatment point people flow network topology graph are analyzed, and the people flow transfer probability between each medical treatment point is calculated, including:

[0022] Obtain the people flow net inflow rate of each network node in the medical treatment point people flow network topology graph in consecutive multiple time windows;

[0023] According to the change trend of the net inflow rate of people in a plurality of continuous time windows, the attributes of the network nodes are identified, the nodes meeting a first change trend are marked as people gathering points, and the nodes meeting a second change trend are marked as people dispersing points;

[0024] According to the marking results of the people gathering points and the people dispersing points and the net inflow rate of people of the target node in the next time window, the people transfer probability between the network nodes is calculated.

[0025] Further, based on the people transfer probability and the historical treatment data of the diagnosis and treatment points, the congestion evolution trend of each diagnosis and treatment point in a future time window is predicted, including:

[0026] Based on the people transfer probability, the transfer prediction treatment data at the target time is calculated in combination with the historical treatment data of each diagnosis and treatment point;

[0027] The periodic characteristics are obtained by performing time series analysis on the historical treatment data in a preset time window, and the time series prediction treatment data at the target time is calculated according to the periodic characteristics through a neural network;

[0028] The prediction errors of the transfer prediction treatment data and the time series prediction treatment data relative to the actual treatment data are calculated, and the dynamic weights are generated according to the prediction errors;

[0029] The transfer prediction treatment data and the time series prediction treatment data are weighted and fused by using the dynamic weights to obtain the prediction treatment data at the target time;

[0030] Based on the prediction treatment data, the capacity occupation degree and the treatment change trend are calculated, and the congestion evolution trend of the diagnosis and treatment point is determined.

[0031] Further, according to the congestion evolution trend and the real-time treatment capacity of the diagnosis and treatment point, a path optimization model is established with the minimum patient average waiting time as the objective function, and the optimal shunt path combination is calculated, including:

[0032] According to the congestion evolution trend, the congestion correlation relationship between adjacent road segments is determined, and the road congestion index in the target time window is generated;

[0033] The medical resource state data and the treatment efficiency data of the diagnosis and treatment point are obtained, the real-time treatment capacity of the diagnosis and treatment point is evaluated, and the real-time treatment amount of the diagnosis and treatment point is obtained;

[0034] According to the road congestion index, the patient path travel time is calculated, and based on the real-time treatment amount of the diagnosis and treatment point, the patient treatment waiting time is calculated, and the sum of the path travel time and the treatment waiting time is taken as the patient average waiting time;

[0035] A path optimization model is constructed, the congestion evolution trend and the real-time reception capacity of the diagnosis and treatment point are taken as state parameters, a patient distribution scheme is taken as a decision parameter, and a target function is constructed based on the average waiting time of the patient;

[0036] The path optimization model is iteratively calculated until an optimal shunt path combination satisfying the traffic condition and the reception capacity limit is obtained.

[0037] Further, generating specific shunt scheduling instructions according to the optimal shunt path combination, and planning a personalized navigation path for the patient based on the personnel movement trajectory feature include:

[0038] The regional patient flow distribution ratio is calculated according to the optimal shunt path combination, the regional transfer threshold is determined based on the road traffic capacity and the reception capacity of the diagnosis and treatment point, and the shunt scheduling instruction is generated;

[0039] The patient historical movement trajectory is collected, the movement behavior feature is extracted, and the patient is classified according to the movement ability;

[0040] The shunt scheduling instruction is split according to the movement ability classification, the candidate path combination is calculated combined with the road network structure, and the path combination is given a differentiated weight;

[0041] The intersection turning instruction is generated based on the path combination with the differentiated weight and the real-time road condition, the traffic time is calculated combined with the signal timing, and the personalized navigation path is formed.

[0042] Further, the method further includes:

[0043] The real-time position of the patient is acquired, and when the position deviates from the personalized navigation path, a new path is selected from the candidate path combination for re-planning;

[0044] The actual traffic data of the patient is updated to the historical movement trajectory, the movement ability classification is re-performed, the differentiated weight is adjusted, and the personalized navigation path is updated.

[0045] An outpatient process intelligent scheduling and resource optimization system is provided in the embodiment of the application, and the system includes:

[0046] A data acquisition unit is configured to collect the waiting area image of each diagnosis and treatment point, extract the personnel density feature in the waiting area, and extract the personnel movement trajectory combined with a multi-target tracking algorithm;

[0047] A topology construction unit is configured to construct a diagnosis and treatment point people flow network topology graph according to the personnel density feature and the personnel movement trajectory feature;

[0048] A probability calculation unit is configured to analyze the people flow convergence point and the people flow dispersion point in the diagnosis and treatment point people flow network topology graph, and calculate the people flow transfer probability between the diagnosis and treatment points;

[0049] a trend prediction unit configured to predict congestion evolution trends of each diagnosis and treatment point in a future time window based on the people flow transfer probability and historical diagnosis and treatment data of the diagnosis and treatment points;

[0050] a path optimization unit configured to establish a path optimization model with a target function of minimizing the average waiting time of patients according to the congestion evolution trends and real-time reception capacities of the diagnosis and treatment points, and calculate an optimal shunt path combination;

[0051] a scheduling execution unit configured to generate specific shunt scheduling instructions according to the optimal shunt path combination, and plan individualized navigation paths for patients based on the personnel movement trajectory features.

[0052] A technical solution provided in an embodiment of the present application is an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and capable of running on the processor, and the processor implements the steps in any of the preceding methods when executing the computer program.

[0053] In the embodiment, the real-time people flow dynamics in the diagnosis and treatment points can be accurately reflected by collecting the waiting area images and extracting the personnel density and movement trajectory features. In combination with multi-target tracking and people flow network topology construction, the fine modeling of the crowd gathering, dispersion and flow rules is realized, so that the people flow transfer probability between the diagnosis and treatment points can be effectively calculated. In combination with the historical diagnosis data, the future congestion evolution trends can be predicted, so that the potential high-load areas can be identified in advance. By establishing the path optimization model with the target of minimizing the average waiting time of patients, the optimal shunt path combination can be dynamically generated, and the reasonable allocation of diagnosis and treatment resources can be realized. In combination with the individualized movement trajectory features, the precise navigation paths for patients can be planned, so that the patient diagnosis efficiency and experience can be improved, the local congestion risk can be reduced, and the overall reception capacity and operation efficiency of the medical institutions can be improved.

[0054] The above summary of the application is only a summary of the technical solutions of the present application. In order to more clearly understand the technical means of the present application, the present application can be implemented according to the content of the specification, and in order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application are described below. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creating any creative labor.

[0056] Figure 1A flowchart of a clinic process intelligent scheduling and resource optimization method provided for an embodiment of the present application;

[0057] Figure 2 A structural schematic diagram of a clinic process intelligent scheduling and resource optimization system provided for an embodiment of the present application. DETAILED DESCRIPTION

[0058] The exemplary embodiments will be described in detail herein with reference to the attached drawings. In the following description, the same numbers are used to represent the same elements throughout the several views. But it is understood that no limitation of the scope of the embodiments depicted in the figures is intended. Changes in the meaning of the numbers are understood by those of ordinary skill in the art.

[0059] The embodiments described in the following exemplary embodiments are not meant to represent all embodiments consistent with one or more embodiments of the present description. It is noted that the steps of the corresponding methods are not necessarily performed in the order shown and described in the present description in other embodiments. In some other embodiments, the steps included in the methods can be more or less than those described in the present description. In addition, a single step described in the present description can be divided into multiple steps for description in other embodiments, and multiple steps described in the present description can be combined into a single step for description in other embodiments.

[0060] As shown in Figure 1 , Figure 1 A flowchart of a clinic process intelligent scheduling and resource optimization method provided for an embodiment of the present application, the method comprising the following steps:

[0061] Collecting the waiting area image of each diagnosis and treatment point, extracting the personnel density feature in the waiting area, and combining the multi-target tracking algorithm to extract the personnel moving track;

[0062] According to the personnel density feature and the personnel moving track feature, a diagnosis and treatment point people flow network topology graph is constructed;

[0063] Analyzing the people flow convergence point and dispersion point in the diagnosis and treatment point people flow network topology graph, and calculating the people flow transfer probability between each diagnosis and treatment point;

[0064] Based on the people flow transfer probability and the historical diagnosis data of the diagnosis and treatment point, the congestion evolution trend of each diagnosis and treatment point in the future time window is predicted;

[0065] According to the congestion evolution trend and the real-time reception capacity of the diagnosis and treatment point, a path optimization model with the minimum patient average waiting time as the objective function is established, and the optimal shunt path combination is calculated;

[0066] According to the optimal shunt path combination, specific shunt scheduling instructions are generated, and individualized navigation paths are planned for patients based on the personnel moving track feature.

[0067] Further, the image of the waiting area of each medical treatment point is collected, the personnel density features in the waiting area are extracted, and a multi-target tracking algorithm is combined to extract the personnel moving track, including:

[0068] The collection parameters are adjusted according to the light information and the people flow information in the image of the waiting area.

[0069] The image of the waiting area is subjected to distortion correction and perspective transformation to obtain an overhead image, and the overhead image is subjected to standardization processing to obtain a standardized image.

[0070] Based on the personnel distribution in the standardized image, a grid unit is divided, the number of personnel in the grid unit is counted to obtain a grid density value, a smoothing radius is determined according to the grid density value, and the grid density value is subjected to smoothing processing by using the smoothing radius to obtain a continuous density distribution.

[0071] The time-space change rate of the continuous density distribution is calculated to obtain a density change feature, and the continuous density distribution and the density change feature are combined to construct the personnel density feature of the waiting area.

[0072] The feature vector of the personnel target in the standardized image is extracted, a multi-target tracking algorithm is used to calculate the similarity between continuous frames based on the feature vector to obtain a correlation matrix, the correlation matrix is used to simultaneously track multiple personnel targets to obtain a track sequence, and the personnel moving track is constructed according to the track sequence.

[0073] When the image of the waiting area of each medical treatment point is collected, the collection parameters can be adjusted according to the real-time scene. For the waiting area with large light change, an adaptive exposure technology is used to dynamically adjust the camera exposure parameters. In a bright environment, the exposure time is set to 1 / 200 seconds, the aperture value is f / 5.6, and the ISO value is 400. In a dark environment, the exposure time is automatically adjusted to 1 / 60 seconds, the aperture value is f / 3.5, and the ISO value is 1600. At the same time, the frame rate parameter is adjusted based on the people flow density. When it is detected that the personnel density of the waiting area exceeds 3 people per square meter, the frame rate of the camera equipment is increased to 25 frames per second to ensure that the details of the rapid movement of personnel are captured.

[0074] After the image of the waiting area is obtained, distortion correction processing is performed to eliminate lens distortion. A polynomial model is used to calculate the distortion parameters, and the image is corrected by the distortion center point and the distortion coefficient. For the four marker points in the waiting area, a perspective transformation matrix is used to convert the original image into an overhead image, and the conversion matrix is obtained by solving a linear equation set. Subsequently, the overhead image is subjected to standardization processing, including size normalization to 1280x720 pixel resolution, brightness equalization processing to maintain the average brightness value of pixels at 128, and contrast stretching to expand the dynamic range to 0-255, thereby obtaining a standardized image.

[0075] The standardized image is meshed, and the entire waiting area is divided into 10x10 mesh units, each of which actually corresponds to an area of about 2 square meters in the waiting area. The background subtraction and personnel detection algorithm is used to identify the personnel in each mesh unit. When the pixel point concentration in the unit exceeds the preset threshold 0.65 and the shape meets the personnel characteristics, it is determined that there is a personnel target. The number of personnel in each mesh unit is counted to obtain a mesh density value matrix. The smoothing radius is determined according to the mesh density value. A smaller smoothing radius such as 1.5 mesh unit widths is used in the area with a higher density value, and a larger smoothing radius such as 3 mesh unit widths is used in the area with a lower density value. The mesh density value is smoothed by an adaptive Gaussian kernel function to generate a continuous density distribution.

[0076] The rate of change of the continuous density distribution in the time and space dimensions is calculated. The time rate of change is calculated by comparing the density distributions of adjacent time frames, and the difference between the two frames is divided by the time interval. The spatial rate of change is obtained by calculating the gradient of the density distribution in the horizontal and vertical directions. The time rate of change and the spatial rate of change are combined to construct a three-dimensional feature vector representing the density change characteristics. Finally, the continuous density distribution and the density change characteristics are combined to construct the personnel density characteristics of the waiting area, forming a comprehensive feature description including spatial distribution, time change and trend prediction.

[0077] When extracting the feature vector of the personnel target in the standardized image, first, the candidate region extraction algorithm is used to detect the position of the personnel in each frame of image, and the appearance features and motion features of each personnel target are extracted. The appearance features include color histogram features, texture features and shape features; the motion features include displacement vectors and motion directions. These features are combined to form a 128-dimensional feature vector representing each personnel target. A multi-target tracking algorithm is used to process consecutive video frames, calculate the feature similarity between each detected personnel target in the current frame and all personnel targets in the previous frame, and generate a correlation matrix. The correlation matrix element value ranges from 0 to 1. When the similarity exceeds 0.75, it is considered to be the same personnel target. The Hungarian algorithm can be used to perform optimal matching on the correlation matrix to solve the target assignment problem and achieve simultaneous tracking of multiple personnel targets.

[0078] During tracking, each personnel target is assigned a unique ID, and its position coordinates in consecutive frames are recorded to form a trajectory sequence. When a target disappears from the scene for more than 30 frames, the tracking of that ID is terminated. The obtained trajectory sequence is smoothed to eliminate noise, and a sliding window average method is used with a window size of 5 frames. Based on the smoothed trajectory sequence, personnel movement features are extracted, including moving speed, direction change and stopping points. Positions with a stay time of more than 15 seconds are marked as waiting hotspots by analyzing the stay time, and the main flow direction of the moving trajectory is determined to determine the main flow channel of the waiting area.

[0079] The above-mentioned technology realizes accurate extraction of the personnel density characteristics and the moving trajectory of the outpatient waiting area, effectively improving the scientificity of the intelligent scheduling and resource optimization of the outpatient process. The method can monitor the congestion of the waiting area of each diagnosis and treatment point in real time, and through the analysis of the personnel density distribution and the moving trajectory, it can predict the possible congestion of the waiting area, so that the medical institutions can dynamically adjust the allocation of medical resources, optimize the diagnosis and treatment order, improve the patient experience, and improve the medical efficiency.

[0080] Further, constructing the diagnosis and treatment point people flow network topology graph according to the personnel density characteristics and the personnel moving trajectory characteristics comprises:

[0081] According to the personnel density characteristics, the density distribution value and the density change value are obtained, and the service capacity value of the network node is calculated;

[0082] Based on the personnel moving trajectory characteristics, the flow intensity value between the network nodes is calculated, and the ratio of the flow intensity value to the service capacity value is set as the edge weight value between the network nodes;

[0083] The diagnosis and treatment point is set as the network node, and the diagnosis and treatment point people flow network topology graph is constructed according to the service capacity value and the edge weight value.

[0084] When constructing the diagnosis and treatment point people flow network topology graph, the personnel density characteristics need to be processed to obtain the density distribution value and the density change value. The density distribution value is extracted from the continuous density distribution, and is obtained by calculating the average personnel density in each diagnosis and treatment point waiting area, with the unit being person / m2. For a general comprehensive outpatient service, the density distribution value usually fluctuates between 0.2-2.5 person / m2. The density change value is obtained by calculating the change rate of the density distribution value per unit time, with the unit being person / (m2·min). During the morning outpatient peak period, the density change value can reach 0.15 person / (m2·min), and during the relatively flat period in the afternoon, it is about 0.03 person / (m2·min). The density distribution value and the density change value are smoothed by using the exponential weighted moving average method, with the weight factor being 0.8 and the smoothing time being 30 minutes, so as to eliminate the influence of short-term fluctuations.

[0085] The service capacity value is a key indicator representing the patient handling capacity of the diagnosis and treatment point, which is calculated by combining the density distribution value and the density change value. In the calculation process, the density distribution value is assigned a weight coefficient of 0.7, and the density change value is assigned a weight coefficient of 0.3, and the weighted sum is calculated. When the density distribution value is high and the density change value is negative, it indicates that the personnel of the diagnosis and treatment point are concentrated but slowly dispersed, and at this time the service capacity value is relatively high; when the density distribution value is high and the density change value is positive, it indicates that the personnel of the diagnosis and treatment point are continuously gathering, and the service capacity value is relatively low. Taking the internal medicine diagnosis and treatment point as an example, its service capacity value is about 8.5 under normal operating conditions, indicating that the number of patients that can be served per hour.

[0086] When calculating the flow intensity value between network nodes based on the trajectory feature of personnel movement, first analyze the trajectory sequence and identify the movement pattern of patients between diagnosis and treatment points. For any two diagnosis and treatment points A and B, count the number of patients moving from A to B within the observation time window, denoted as the direct flow number. Meanwhile, consider the indirect flow, that is, the case where patients travel from A to B through other diagnosis and treatment points, and assign a decay coefficient of 0.6 to the indirect flow. Add the direct flow number and the weighted indirect flow number to obtain the total flow number. The flow intensity value is defined as the total flow number per unit time, with the unit of person / hour. Taking the flow between the radiology department and the internal medicine department as an example, the peak flow intensity value can reach 12 person / hour, and the non-peak flow intensity value is about 4 person / hour.

[0087] The edge weight value is calculated by the ratio of the flow intensity value and the service capacity value, representing the flow intensity per unit service capacity. In specific calculation, the flow intensity value is divided by the service capacity value of the target diagnosis and treatment point. The larger the edge weight value, the higher the congestion degree of the corresponding path, and the longer the waiting time of patients on the path. To avoid the influence of extreme values, the calculated edge weight value is normalized to distribute between 0 and 1. Taking the edge weight value from the internal medicine department to the pharmacy as an example, it can reach 0.85 during the outpatient peak period, indicating that the path is highly congested; and it is about 0.25 during the non-peak period, indicating that the path is smooth.

[0088] When setting diagnosis and treatment points as network nodes, all diagnosis and treatment points in the hospital are numbered and assigned unique identifiers, such as N001 for the registration desk, N002 for the internal medicine clinic, N003 for the radiology department, N004 for the pharmacy, etc. Each network node contains attribute information such as location coordinates, service capacity value, and current service status. The connection relationship between nodes is determined according to the physical space layout and the actual movement possibility of patients, not all node pairs have direct connections. An adjacency matrix is constructed according to the edge weight value, and the element value in the matrix corresponds to the edge weight value between nodes. The element value of node pairs without connection is set to infinity. The topology structure represented by the adjacency matrix can intuitively reflect the flow relationship between diagnosis and treatment points.

[0089] The diagnosis and treatment point flow network topology graph is presented through a visual interface, with node size proportional to service capacity value, node color reflecting current congestion degree, and connection line thickness proportional to edge weight value. The topology graph also supports dynamic display in the time dimension, allowing for the retrieval of historical data or the prediction of network state changes in the future. The diagnosis and treatment point flow network topology graph is updated regularly, with an update frequency of 5 minutes, to reflect the latest flow status. At the same time, key nodes and bottleneck paths are marked in the topology graph, with key nodes defined as nodes with service capacity value less than 5 and connectivity greater than 3, and bottleneck paths defined as connections with edge weight value greater than 0.7.

[0090] The out-patient clinic flow network topology graph constructed by implementing the method can intuitively reflect the flow distribution and flow relationship between each diagnosis and treatment point, and provide data support for intelligent scheduling of out-patient clinic process and optimization of resources. The method can monitor the service capacity changes of each diagnosis and treatment point and the flow of people between diagnosis and treatment points in real time, dynamically identify the bottleneck link and key node in the out-patient clinic process, so that the hospital managers can reasonably allocate medical resources, the hospital can adjust resource allocation in advance to avoid congestion, and the patient's medical experience and satisfaction are greatly improved.

[0091] Further, the people flow gathering point and the people flow dispersing point in the diagnosis and treatment point people flow network topology graph are analyzed, and the people flow transfer probability between each diagnosis and treatment point is calculated, including:

[0092] Obtaining the people flow net inflow rate of each network node in the diagnosis and treatment point people flow network topology graph in continuous multiple time windows;

[0093] According to the change trend of the people flow net inflow rate in the continuous multiple time windows, the attribute of the network node is identified, the node meeting the first change trend is marked as a people flow gathering point, and the node meeting the second change trend is marked as a people flow dispersing point;

[0094] According to the marking results of the people flow gathering point and the people flow dispersing point and the people flow net inflow rate of the target node in the next time window, the people flow transfer probability between the network nodes is calculated.

[0095] When analyzing the people flow gathering point and the people flow dispersing point in the diagnosis and treatment point people flow network topology graph, the people flow net inflow rate of each network node in continuous multiple time windows needs to be obtained. The people flow net inflow rate represents the ratio of the difference between the flow into a node and the flow out of the node per unit time and the capacity of the node. When calculating the people flow net inflow rate, first, the number of people entering the node and the number of people leaving the node in the time window are counted, and the difference between the two is the net inflow number. The net inflow number is divided by the length of the time window to obtain the net inflow rate. The setting of the time window is adjusted according to the actual situation of the out-patient clinic, and is generally set to 15 minutes. For example, in the time window of 8:00-8:15, 25 patients flow into the internal medicine diagnosis and treatment point, and 10 patients flow out, the net inflow number is 15, and the capacity of the diagnosis and treatment point is 30. Therefore, the people flow net inflow rate in the time window is 0.5.

[0096] To obtain the accurate trend of the people flow, data in multiple consecutive time windows are needed. Usually, data in 8 consecutive time windows, i.e., 2 hours, are selected for analysis. For each network node, the sequence of the net inflow rate of the people flow in the 8 consecutive time windows is recorded. To eliminate the influence of short-term fluctuations, the moving average method is used to smooth the sequence data, and the smoothing window size is set to 3. The smoothed net inflow rate sequence can better reflect the actual trend of the people flow change. For example, the original net inflow rate of the people flow of the radiology department in the 8 consecutive time windows is [-0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.4, 0.2], and the smoothed sequence is [0.0, 0.17, 0.3, 0.4, 0.5, 0.5, 0.4, 0.3].

[0097] The attributes of the network nodes are identified according to the change trend of the net inflow rate of the people flow in multiple consecutive time windows. The first change trend is defined as: in at least 4 consecutive time windows, the net inflow rate of the people flow is continuously positive and shows an upward trend, or gradually increases from negative to positive and continues to grow. The nodes meeting the first change trend are marked as people flow convergence points, indicating that the patients are gradually gathering at the treatment point. The second change trend is defined as: in at least 4 consecutive time windows, the net inflow rate of the people flow is continuously negative and shows a downward trend, or gradually decreases from positive to negative and continues to decrease. The nodes meeting the second change trend are marked as people flow dispersion points, indicating that the patients are gradually decreasing at the treatment point and transferring to other treatment points. For nodes that neither meet the first change trend nor the second change trend, they are marked as neutral nodes.

[0098] The node attribute marking is updated in real time in a sliding window manner. When new time window data is generated, the earliest time window data is removed, the new data is added, and the trend analysis and node attribute marking are performed again. In this way, the change of the people flow at each treatment point can be dynamically reflected. For example, during the observation period from 9:00 to 11:00, the registration office is initially marked as a people flow convergence point, and after 10:30, its attribute may change to a neutral node or a people flow dispersion point as the morning rush hour ends.

[0099] According to the marking results of the people flow convergence points and the people flow dispersion points and the net inflow rate of the people flow of the target node in the next time window, the people flow transfer probability between the network nodes is calculated. The people flow transfer probability represents the proportion of patients transferred from one treatment point to another. When calculating the transfer probability, the attributes of the source node and the target node, the physical distance between the two nodes, the historical transfer data, and the predicted net inflow rate of the people flow of the target node in the next time window are considered. Specifically, if the source node is marked as a people flow dispersion point, the target node is marked as a people flow convergence point, and there is a direct connection between them, the transfer probability between them is relatively high.

[0100] The calculation of the transfer probability adopts a weighted combination method, and a weight of 0.4 is given to a node attribute factor, that is, when a source node is a human flow dispersion point and a target node is a human flow convergence point, the contribution value of the factor is 0.4; otherwise, the contribution value is 0. A weight of 0.3 is given to a physical distance factor, the physical distance between two nodes is converted into a normalized distance factor, and the closer the distance, the closer the contribution value of the factor to 0.3. A weight of 0.2 is given to historical transfer data, the historical transfer probability is calculated according to the historical transfer data of the same period in the past week, and the contribution value is at most 0.2. A weight of 0.1 is given to a predicted net inflow rate, and the higher the predicted net inflow rate of the target node, the closer the contribution value of the factor to 0.1. The contribution values of the factors are added to obtain the final transfer probability.

[0101] In the implementation process, it is necessary to adjust the weight configuration of the transfer probability calculation for different types of diagnosis and treatment points. For example, for diagnosis and treatment points such as the laboratory department with relatively fixed service time, the weight of historical transfer data can be increased to 0.3; for diagnosis and treatment points such as the emergency department with strong temporality, the weight of the node attribute factor can be increased to 0.5. The transfer probability calculation result is stored in the form of a matrix, and the matrix element value represents the probability of the row corresponding node transferring to the column corresponding node. For example, the probability of the internal medicine diagnosis and treatment point transferring to the radiology department is 0.35, the probability of transferring to the laboratory is 0.25, and the probability of transferring to the pharmacy is 0.4.

[0102] The human flow convergence point, dispersion point and transfer probability analyzed by the above method provide a key basis for intelligent scheduling of outpatient flow. The method can identify the congestion hotspots and the human flow aggregation areas that will be generated in the outpatient department in real time, predict the human flow of each diagnosis and treatment point, calculate the optimal patient diversion path according to the transfer probability, guide the patients to choose the path with lower congestion degree to the next diagnosis and treatment point through the guidance system, effectively reduce the average waiting time of the patients between each diagnosis and treatment point, and is suitable for medical institutions of different scales, thereby providing strong technical support for the optimization of outpatient flow.

[0103] Further, based on the human flow transfer probability and the historical diagnosis data of the diagnosis and treatment points, the congestion evolution trend of each diagnosis and treatment point in a future time window is predicted, including:

[0104] Based on the human flow transfer probability, the transfer prediction diagnosis data of the target time is calculated in combination with the historical diagnosis data of each diagnosis and treatment point;

[0105] The periodic characteristics are obtained by time series analysis on the historical diagnosis data in a preset time window, and the time series prediction diagnosis data of the target time is calculated according to the periodic characteristics through a neural network;

[0106] The prediction errors of the transfer prediction diagnosis data and the time series prediction diagnosis data relative to the actual diagnosis data are calculated, and the dynamic weight is generated according to the prediction errors;

[0107] The transfer prediction visit data and the time series prediction visit data are weighted and fused by using dynamic weights to obtain prediction visit data of a target time;

[0108] The capacity occupancy degree and the visit change trend are calculated based on the prediction visit data, and the congestion evolution trend of the diagnosis and treatment point is determined.

[0109] First, the transfer prediction visit data of the target time is calculated by using the crowd transfer probability, and the calculation process considers the actual number of visits of each diagnosis and treatment point at the current time and the crowd transfer probability matrix. For each diagnosis and treatment point, the number of patients currently in the diagnosis and treatment point is counted, and the number of these patients distributed to each diagnosis and treatment point at the target time is predicted according to the crowd transfer probability matrix. For example, there are 30 patients in the internal medicine diagnosis and treatment point at the current time, according to the transfer probability matrix, it is estimated that 10 of them will transfer to the radiology department at the next time, 8 will transfer to the laboratory, 5 will transfer to the pharmacy, and the remaining 7 will still stay in the internal medicine department. At the same time, the number of patients transferred from other diagnosis and treatment points to the target diagnosis and treatment point is considered. For the radiology department, there may be 10 patients transferred from the internal medicine department, 5 patients transferred from the surgery department, 3 patients transferred from the pediatrics department, etc. The number of all transferred patients is added to the number of patients who are currently in the radiology department and are expected to still stay in the radiology department, and the transfer prediction visit data of the radiology department at the target time is obtained as 25.

[0110] The historical visit data in a preset time window is analyzed in time series to extract periodic characteristics, which is the basis of the second prediction method. The historical visit data usually has obvious periodicity, including intraday fluctuation, weekly fluctuation and seasonal fluctuation. The intraday fluctuation shows the morning and afternoon visit peaks; the weekly fluctuation shows that the visit quantity is larger on Monday and Friday; the seasonal fluctuation shows that the internal medicine visit quantity increases in the flu season. When extracting periodic characteristics, a method combining time domain analysis and frequency domain analysis is adopted. Time domain analysis identifies the periodicity of data by calculating autocorrelation function; frequency domain analysis converts time series data to frequency domain by fast Fourier transform to identify the main periodic components. For a general general hospital, the intraday period is usually 4 hours, the weekly period is 5 days, and the seasonal period is 12 weeks.

[0111] Based on the extracted periodic features, a neural network model is constructed to calculate the time series prediction of the target time. The neural network adopts a long short-term memory network structure, which can effectively capture the long-term dependence of time series data. The number of input layer nodes is 48, corresponding to the first 48 time window of the visit data; the hidden layer adopts a double-layer structure, each layer contains 64 neurons; the number of output layer nodes is the number of prediction time windows, usually set to 6, corresponding to the prediction of the future 1.5 hours. During model training, the historical visit data of the past 3 months is used, the mean square error is used as the loss function, the Adam optimizer is used for parameter optimization, the learning rate is set to 0.001, and the training iteration number is 1000. After training, the first 48 time window of the visit data at the current time is input, and the model outputs the prediction of the visit data of the next 6 time windows. For example, the time series prediction of the radiology department at the target time is 28 people.

[0112] The prediction error of the transfer prediction visit data and the time series prediction visit data relative to the actual visit data is calculated to generate a dynamic weight. The prediction error is calculated by comparing the difference between the historical prediction value and the actual value. Specifically, the prediction data and the actual data of the last 10 time windows are taken, and the average absolute error of the transfer prediction method and the time series prediction method is calculated. The average absolute error of the transfer prediction method is 3.2 people, and the average absolute error of the time series prediction method is 2.1 people. The generation of the dynamic weight is based on the inverse of the prediction error, that is, the smaller the error, the greater the weight of the corresponding method. Through normalization processing, it is ensured that the sum of the weights of the two methods is 1. For the radiology department, the weight of the transfer prediction method is 0.4, and the weight of the time series prediction method is 0.6.

[0113] The transfer prediction visit data and the time series prediction visit data are weighted and fused using the dynamic weight to obtain the prediction visit data at the target time. The weighted fusion process is to multiply the transfer prediction visit data by its corresponding weight, multiply the time series prediction visit data by its corresponding weight, and then add them together. Taking the radiology department as an example, the transfer prediction visit data is 25 people, and the weight is 0.4; the time series prediction visit data is 28 people, and the weight is 0.6; the weighted fusion prediction visit data is 25x0.4+28x0.6=26.8 people, rounded to 27 people. This weighted fusion method can comprehensively utilize the advantages of the two prediction methods, and improve the accuracy and stability of the prediction.

[0114] The congestion evolution trend of the diagnosis and treatment point is determined based on the predicted visit data, the capacity occupation degree, and the visit change trend. The capacity occupation degree is defined as the ratio of the predicted visit data to the maximum capacity of the diagnosis and treatment point. The maximum capacity is determined according to the physical space, the number of medical equipment, and the number of medical staff of the diagnosis and treatment point. Taking the radiology department as an example, the maximum capacity is 35 people, the predicted visit data is 27 people, and the capacity occupation degree is 27 / 35 = 0.77. The visit change trend is calculated by comparing the predicted visit data of multiple consecutive time windows. If the predicted visit data of the next three time windows shows an upward trend, such as 27 people, 30 people, and 32 people, it is determined that the visit change trend is upward; if it shows a downward trend, such as 27 people, 25 people, and 22 people, it is determined that the visit change trend is downward; if it fluctuates slightly, such as 27 people, 28 people, and 27 people, it is determined that the visit change trend is stable.

[0115] According to the capacity occupation degree and the visit change trend, the congestion evolution trend of the diagnosis and treatment point is divided into five categories: severe congestion aggravation, mild congestion aggravation, stable congestion, congestion relief, and smooth. When the capacity occupation degree is greater than 0.85 and the visit change trend is upward, it is determined to be severe congestion aggravation; when the capacity occupation degree is between 0.7 and 0.85 and the visit change trend is upward, it is determined to be mild congestion aggravation; when the capacity occupation degree is greater than 0.7 and the visit change trend is stable, it is determined to be stable congestion; when the capacity occupation degree is greater than 0.7 and the visit change trend is downward, it is determined to be congestion relief; when the capacity occupation degree is less than 0.7, it is determined to be smooth. For the radiology department, the capacity occupation degree is 0.77, and the visit change trend is upward, so it is determined that the congestion evolution trend is mild congestion aggravation.

[0116] By fusing transfer prediction and time series prediction, accurate prediction of future congestion status of each diagnosis and treatment point in the outpatient department is realized, providing strong support for intelligent scheduling of outpatient processes. According to the prediction results, the hospital can adjust the allocation of medical resources in advance, such as increasing the number of medical staff or temporarily opening spare examination rooms in the diagnosis and treatment point where congestion is predicted to occur. At the same time, patient diversion suggestions are generated according to the prediction results, guiding patients to avoid congested diagnosis and treatment points and choose relatively smooth paths. The model continuously optimizes parameters through continuous learning of historical data, adapts to changes in outpatient visit patterns, and maintains the stability and reliability of the prediction.

[0117] Further, according to the congestion evolution trend and the real-time reception capacity of the diagnosis and treatment point, a path optimization model is established with the objective function of minimizing the average waiting time of patients, and the optimal diversion path combination is calculated, including:

[0118] According to the congestion evolution trend, the congestion correlation between adjacent road segments is determined, and the road segment congestion index in the target time window is generated;

[0119] Obtain medical resource state data and reception efficiency data of the diagnosis and treatment point, evaluate real-time reception capacity of the diagnosis and treatment point, and obtain real-time reception volume of the diagnosis and treatment point;

[0120] Calculate patient path travel time according to the road congestion index, and calculate patient diagnosis and treatment waiting time based on the real-time reception volume of the diagnosis and treatment point, and take the sum of the path travel time and the diagnosis and treatment waiting time as the average waiting time of the patient;

[0121] Construct a path optimization model, take the congestion evolution trend and the real-time reception volume of the diagnosis and treatment point as state parameters, take the patient distribution scheme as a decision parameter, and construct a target function based on the average waiting time of the patient;

[0122] Iteratively calculate the path optimization model until an optimal shunt path combination that meets the traffic condition and the reception volume limit is obtained.

[0123] When determining the congestion correlation between adjacent road segments according to the congestion evolution trend, historical congestion data and predicted congestion state of the connecting road segments between the diagnosis and treatment points need to be analyzed. The congestion correlation represents the influence degree of the congestion state of one road segment on adjacent road segments. In the calculation process, first, the congestion state of each road segment is quantified, and the congestion evolution trend is converted into a numerical index: 5 for severe congestion aggravation, 4 for mild congestion aggravation, 3 for stable congestion, 2 for congestion alleviation, and 1 for smoothness. For example, the current congestion evolution trend of the road segment from the radiology department to the internal medicine department is mild congestion aggravation, and the congestion state value is 4. Second, the correlation of the congestion state of adjacent road segments is analyzed, and the Pearson correlation coefficient is used to calculate the correlation degree of the congestion state of adjacent road segments. The higher the correlation degree, the more likely the congestion change of one road segment will affect adjacent road segments. For example, the correlation degree of the road segment from the internal medicine department to the laboratory and the road segment from the radiology department to the internal medicine department is 0.75, indicating that the congestion states of the two road segments have strong correlation.

[0124] When generating the road congestion index within the target time window, the current congestion state, the congestion evolution trend, and the congestion correlation are considered comprehensively. The road congestion index is represented in the range of 0 to 10, and the larger the value, the more serious the congestion. The calculation formula considers the basic congestion value, the trend factor, and the correlation influence. The basic congestion value is determined by the current congestion state, for example, the basic congestion value corresponding to mild congestion is 5; the trend factor is determined by the congestion evolution trend, the congestion aggravation trend corresponds to a positive trend factor, and the congestion alleviation trend corresponds to a negative trend factor; the correlation influence is determined by the congestion state and the correlation degree of adjacent road segments. Taking the road segment from the internal medicine department to the radiology department as an example, the current congestion state is mild congestion, and the basic congestion value is 5; the congestion evolution trend is stable congestion, and the trend factor is 0; the contribution value affected by adjacent road segments is 1.2; and the finally calculated road congestion index is 6.2.

[0125] The medical resource status data and the reception efficiency data are obtained to evaluate the real-time reception capacity of the diagnosis and treatment point. The medical resource status data includes the number of medical staff, the number of examination rooms, the available state of medical equipment, etc. The reception efficiency data includes the average single diagnosis time, the work saturation of medical staff, etc. By combining these data, the real-time reception capacity of the diagnosis and treatment point, i.e. the number of patients that can be handled per unit time, is calculated. In the calculation process, the basic reception capacity is multiplied by the resource adjustment factor and the efficiency adjustment factor. The basic reception capacity is determined by the number of medical staff and the number of examination rooms. The resource adjustment factor reflects the influence of the available state of medical equipment on the reception capacity. The efficiency adjustment factor reflects the influence of the work saturation of medical staff on the reception efficiency. For example, the basic reception capacity of the internal medicine department is 15 people per hour, 1 of the 2 diagnostic devices is in maintenance state, the resource adjustment factor is 0.7, the medical staff has been working for 4 hours, and the efficiency adjustment factor is 0.85. The calculated real-time reception capacity is 8.9 people per hour.

[0126] The patient path travel time is calculated according to the road congestion index, and the patient diagnosis and treatment waiting time is calculated based on the real-time reception capacity of the diagnosis and treatment point. The path travel time is the time required for a patient to move from one diagnosis and treatment point to another, which is affected by the road length and the congestion index. In the calculation, the basic travel time of the road in the non-congestion state is first determined, and then multiplied by the congestion factor to obtain the actual travel time. The congestion factor is a function of the road congestion index, and the higher the congestion index, the larger the congestion factor. For example, the road length from the internal medicine department to the radiology department is 50 meters, the basic travel time is 1 minute, the current congestion index is 6.2, the corresponding congestion factor is 1.8, and the calculated actual travel time is 1.8 minutes. The diagnosis and treatment waiting time is the time that a patient waits for reception at a diagnosis and treatment point, which is affected by the number of patients in the queue and the real-time reception capacity of the diagnosis and treatment point. In the calculation, the number of patients in the queue is divided by the real-time reception capacity of the diagnosis and treatment point. For example, the current number of patients in the queue of the radiology department is 18, the real-time reception capacity is 12 people per hour, i.e. 1 person per 5 minutes, and the calculated diagnosis and treatment waiting time is 90 minutes. The average waiting time of the patient is the sum of the path travel time and the diagnosis and treatment waiting time.

[0127] In constructing the path optimization model, the congestion evolution trend and the real-time reception capacity of the diagnosis and treatment points are taken as state parameters, and the patient allocation scheme is taken as the decision parameter. The patient allocation scheme represents the proportion of patients allocated to each available path. The objective function is to minimize the average waiting time of patients, i.e., the total waiting time of all patients divided by the total number of patients. The model constraints include: the sum of the proportions of patients allocated to each path is 1; the actual reception capacity of each diagnosis and treatment point does not exceed its maximum reception capacity; and the actual traffic volume of each road segment does not exceed its maximum traffic volume. To solve this optimization problem, a genetic algorithm is used for iterative calculation. The population size is set to 100, the crossover probability is 0.8, the mutation probability is 0.1, and the maximum number of iterations is 1000. In each iteration, a new allocation scheme is generated, the corresponding average waiting time of patients is calculated, and the scheme with the shortest waiting time is selected to be passed to the next generation.

[0128] The path optimization model is iteratively calculated until the optimal diversion path combination that satisfies the traffic conditions and reception capacity constraints is obtained. During the iteration process, the patient allocation scheme is continuously adjusted, and its impact on the average waiting time of patients is evaluated. When the change rate of the average waiting time of patients is less than 0.001 for 50 consecutive iterations, the algorithm is considered to have converged, and the current optimal diversion path combination is output. The optimal diversion path combination includes the proportion of patients allocated between each diagnosis and treatment point. For example, among the patients from the registration office to the internal medicine department, 70% are recommended to go directly to the internal medicine department, and 30% are recommended to go to the pre-examination and diagnosis first and then to the internal medicine department; among the patients from the internal medicine department to the radiology department, 60% are recommended to choose the first floor corridor path, and 40% are recommended to choose the second floor corridor path. The optimal diversion path combination also includes the recommended reception order and priority of each diagnosis and treatment point to balance the load of each diagnosis and treatment point.

[0129] To ensure the real-time effectiveness of the diversion path, the optimal diversion path combination is recalculated every 10 minutes, and the updated diversion recommendations are pushed to the guidance system and the patient mobile application. In the event of an emergency, such as the sudden closure of a diagnosis and treatment point or the emergency deployment of medical staff, the recalculation is triggered immediately to ensure that the diversion path is always the optimal solution under the current conditions. At the same time, the system records the diversion effect data, including the difference between the actual waiting time and the predicted waiting time, as a basis for subsequent optimization.

[0130] This path optimization method achieves efficient diversion of outpatient patients through accurate calculation and dynamic adjustment, significantly reducing the waiting time of patients. Compared with traditional fixed diversion schemes, this method can flexibly adjust according to real-time congestion conditions and reception capacity, making the load of each diagnosis and treatment point more balanced and reducing the situation of excessive congestion at some diagnosis and treatment points while others are idle. Hospital managers can monitor the operation status of each diagnosis and treatment point in real time through the system monitoring panel and adjust medical resource allocation in a timely manner. Patients can obtain personalized treatment path recommendations through the mobile application, improving the medical experience.

[0131] Further, generating specific shunt scheduling instructions according to the optimal shunt path combination, and planning a personalized navigation path for the patient based on the characteristics of the personnel movement trajectory includes:

[0132] Calculating the regional patient flow distribution ratio according to the optimal shunt path combination, determining the regional transfer threshold based on the road traffic capacity and the diagnosis and treatment point reception capacity, and generating the shunt scheduling instruction;

[0133] Collecting patient historical movement trajectories, extracting movement behavior characteristics, and classifying patients according to movement ability;

[0134] Splitting the shunt scheduling instructions according to the movement ability classification, calculating the candidate path combination combined with the road network structure, and assigning different weights to the path combination;

[0135] Generating intersection turning instructions based on the path combination with different weights and real-time road conditions, calculating the travel time combined with signal timing, and forming a personalized navigation path.

[0136] When calculating the regional patient flow distribution ratio according to the optimal shunt path combination, the hospital space layout and the traffic capacity of different regions need to be considered. In the specific implementation process, the hospital outpatient area is divided into several functional blocks, such as registration area, waiting area, diagnosis and treatment area, examination area and medicine taking area, etc. For each block, the proportion of patients passing through the block in the optimal shunt path combination is calculated. The patient proportions on each path are added to obtain the total flow distribution ratio of each region. At the same time, according to the physical space layout of the hospital, the road traffic capacity of each region is calculated, that is, the maximum number of patients that can pass through per unit time. The traffic capacity is affected by factors such as channel width, obstacle distribution and safety distance, which is determined by field measurement and flow simulation.

[0137] Determine the regional transfer threshold based on the road traffic capacity and the diagnosis and treatment point reception capacity. The regional transfer threshold refers to the threshold when the regional patient flow exceeds the threshold, and part of the patients need to be guided to the alternative path to start the shunt mechanism. The determination of the transfer threshold needs to balance the traffic efficiency and the diagnosis and treatment point reception capacity. Usually, the transfer threshold is set to a certain proportion of the traffic capacity to reserve enough buffer space to cope with flow fluctuations. At the same time, considering the reception capacity limit of each diagnosis and treatment point, ensure that the patient inflow of each diagnosis and treatment point after shunting does not exceed its maximum reception capacity. When generating the shunt scheduling instruction, compare the predicted flow with the transfer threshold, when the predicted flow exceeds the transfer threshold, calculate the proportion of patients that need to be shunted and the recommended alternative path. The shunt scheduling instruction contains information such as region identifier, time period, shunt proportion and alternative path.

[0138] Collect patient historical movement trajectories and extract movement behavior features. Patient movement trajectories are collected through the hospital's internal positioning system, recording the sequence of location changes within the hospital. The collected trajectory data is cleaned and smoothed to remove outliers and jitter interference. Movement behavior features are extracted from the processed trajectory data, including average movement speed, speed variation rate, stay frequency, turning flexibility, and path selection preference. Average movement speed represents the general speed of the patient walking within the hospital; speed variation rate reflects the stability of the patient's speed; stay frequency represents the frequency of the patient stopping to rest or waiting during movement; turning flexibility reflects the ability and frequency of the patient changing direction; path selection preference represents the degree of preference for different types of paths, such as whether to prefer elevators over stairs.

[0139] Classify patient movement ability based on extracted movement behavior features, and use clustering algorithms to divide patients into different movement ability categories. Generally, patients are divided into three categories: high, medium, and low movement ability. High movement ability patients have faster average movement speed, small speed variation rate, low stay frequency, high turning flexibility, and high acceptance of stairs and long-distance paths; medium movement ability patients have moderate average movement speed, moderate speed variation rate, moderate stay frequency, moderate turning flexibility, and high acceptance of elevators and medium-distance paths; low movement ability patients have slow average movement speed, large speed variation rate, high stay frequency, low turning flexibility, and high demand for wheelchair access and short-distance paths. Movement ability classification results are stored in the patient's electronic file and dynamically updated based on subsequent treatment conditions.

[0140] Split the diversion scheduling instructions according to the movement ability classification, and develop differentiated diversion strategies for patients in different movement ability categories. For high movement ability patients, longer but less congested alternative paths can be allocated; for medium movement ability patients, paths that balance distance and congestion level are allocated; for low movement ability patients, paths with shorter distance and perfect barrier-free facilities are preferred, even if the path has slightly higher congestion level. Combine the hospital road network structure to calculate candidate path combinations for each movement ability category. The road network structure includes the connection relationship between nodes (intersections, treatment points) and edges (corridors, passages). Utilize multi-source shortest path algorithms to consider factors such as distance, congestion level, and facility conditions to generate multiple candidate paths.

[0141] Differential weights are assigned to path combinations, and the weight values of each candidate path are determined according to the patient's mobility and path characteristics. The weight calculation considers factors such as path length, congestion level, slope change, number of turns, and facility conditions. For high-mobility patients, the weight ratio of path length and congestion level is different; for medium-mobility patients, it is more balanced; for low-mobility patients, it emphasizes the accessibility of the path. For low-mobility patients, the weight of facility conditions is increased to ensure that the recommended path has sufficient accessibility facilities. According to the calculated weight values, the path that best suits the patient is selected as the recommended path.

[0142] Based on the differential weights of path combinations and real-time traffic, intersection turning instructions are generated. At each intersection or decision point, the optimal turning direction is determined according to the weight of each candidate path from the current location to the destination, combined with real-time traffic information. Real-time traffic information is obtained through the hospital's internal monitoring system and is updated regularly. Intersection turning instructions are presented in a simple and clear form, making it easy for patients to understand and execute. Combined with the timing information of the hospital's internal signal system, the travel time is calculated. For example, if a patient needs to pass through a channel with an automatic door, the estimated time for the patient to arrive at the automatic door will be calculated, and the travel delay will be estimated accordingly. The final personalized navigation path includes complete path description, estimated travel time for each segment of the path, total distance and total time, and key landmarks along the way.

[0143] To ensure the dynamic adaptability of the navigation path, the path recommendation is continuously adjusted based on real-time traffic and patient feedback. When a sudden congestion in a certain area is detected or the patient significantly deviates from the recommended path, the navigation suggestion will be recalculated and updated. At the same time, for patients with special needs, such as the disabled and visually impaired, more detailed and targeted navigation instructions are provided, including the location of accessible channels, the use of ramps and elevators, etc.

[0144] A collaborative optimization mechanism for group diversion and individual navigation is also provided. When multiple patients simultaneously travel from similar starting points to similar destinations, the overall diversion effect and individual needs are considered to avoid directing too many patients to the same alternative path and causing new congestion points. This collaborative optimization is achieved through iterative calculation, ensuring the overall diversion effect while maximizing the navigation needs of individual patients.

[0145] This method significantly optimizes the efficiency of outpatient flow and the experience of medical treatment through fine diversion scheduling and personalized navigation. Differentiated path planning for patients with different mobility ensures that all types of patients can find the most suitable treatment route, reducing the time spent wandering and detouring in the hospital. This alleviates the congestion within the hospital, significantly shortening the average travel time of patients. It avoids the situation where some areas are overcrowded while others are idle. The accuracy of the personalized navigation path is high, and the deviation rate of patients following the recommended path is significantly reduced.

[0146] Further, the method further comprises:

[0147] acquiring real-time position of the patient, selecting a new path from the candidate path combination to re-plan when the position deviates from the personalized navigation path;

[0148] updating the actual passing data of the patient to the historical moving track, re-classifying the moving ability, adjusting the differentiated weight and updating the personalized navigation path.

[0149] In this embodiment, the positioning system adopts a combination of Bluetooth beacons and WiFi signals. Beacon devices are deployed in major areas of the hospital to form a positioning network covering the entire outpatient area. Patients carry positioning cards issued by the hospital or install smart devices with the hospital mobile application to communicate with the positioning network in real time. The system regularly collects patient position information and transmits it to the process scheduling module through the security channel of the hospital information system.

[0150] When determining whether the patient's position deviates from the personalized navigation path, the offset distance between the actual position and the planned path needs to be calculated. When the offset distance exceeds the preset threshold, it is considered that the patient has deviated from the planned path. The preset threshold is set according to the spatial characteristics of different areas of the hospital. In open areas, the threshold can be set to a larger value, and in narrow corridors, the threshold can be appropriately reduced. At the same time, the moving direction of the patient's continuous multiple position points is analyzed. When the angle between the moving direction of the continuous multiple position points and the direction of the planned path exceeds a certain angle, even if the offset distance does not exceed the threshold, it is also determined that the path deviates.

[0151] When selecting a new path from the candidate path combination to re-plan, the current position of the patient is first determined as the new starting point, and the original destination is kept unchanged. Based on the patient's moving ability classification and current position, the candidate path combination calculated in advance is screened for feasible paths. The candidate paths selected are applied with differentiated weight calculation, and the path with the highest weight is selected as the new navigation path. Considering that the patient may deviate from the original path for specific reasons, the weight of the candidate path consistent with the patient's current moving direction is appropriately increased to increase the likelihood of the new path being accepted.

[0152] The re-planned path is immediately pushed to the patient and updated on the navigation interface. To avoid frequent updates disturbing the patient, a path update cooling time is set. During the cooling time, even if a slight deviation is detected, path re-planning will not be triggered. For serious deviations, the cooling time limit is ignored, and path re-planning is immediately performed.

[0153] Updating the patient's actual travel data to the historical movement trajectory is a crucial step in ensuring continuous system optimization. Actual travel data includes a complete location sequence, timestamps, stops, and the actual walking path. It records the patient's entire movement from origin to destination, including the selected path, walking speed, stop locations and durations, and whether elevators or stairs were used. To ensure data quality, the collected trajectory data undergoes preprocessing, including noise reduction, smoothing, and outlier detection. The processed trajectory data, along with the patient's basic information, is stored in the historical movement trajectory database.

[0154] When reclassifying mobility, a comprehensive analysis of the patient's recent travel history is conducted. Feature indicators are extracted from the processed trajectory data, including average movement speed, speed standard deviation, dwell frequency, and path selection preferences. Based on these features, a clustering algorithm is used to classify mobility. To improve classification accuracy, a time decay factor is introduced, giving higher weight to recent movement data. The classification results are updated in the patient's electronic record as a basis for subsequent navigation route planning.

[0155] Adjusting differentiated weights and updating personalized navigation paths involves dynamically adjusting the weight parameters for path selection based on mobility capability classification results. According to the latest mobility capability classification, the weighting of factors such as path length, congestion level, and facility conditions is adjusted accordingly. Simultaneously, personalized preference factors are adjusted based on the patient's actual path selection behavior. Based on the adjusted differentiated weights, the comprehensive score of candidate paths is recalculated, and the path with the highest score is selected as the updated personalized navigation path.

[0156] This dynamic adjustment and replanning method significantly improves the adaptability and user experience of the outpatient intelligent navigation system. Through real-time location monitoring and path deviation detection, it can promptly identify abnormal patient walking patterns and provide path adjustments, effectively preventing patients from getting lost or arriving at the wrong areas within the hospital. Dynamically updated personalized navigation significantly reduces the average walking time for patients within the hospital, decreasing anxiety caused by searching for examination rooms. Simultaneously, the continuously learning mobility classification mechanism accurately captures changes in patient mobility, optimizing hospital spatial layout and signage systems, and improving overall operational efficiency. This method provides strong support for intelligent scheduling and resource optimization of outpatient processes.

[0157] like Figure 2 As shown, Figure 2 This is a schematic diagram of the structure of an intelligent scheduling and resource optimization system for outpatient processes provided in an embodiment of the present invention. The system includes:

[0158] The data acquisition unit is used to collect images of the waiting areas of each clinic, extract the personnel density characteristics of the waiting areas, and combine them with a multi-target tracking algorithm to extract the personnel movement trajectory.

[0159] a topology construction unit configured to construct a network topology map of the flow of people in the medical treatment points according to the characteristics of the density of people and the trajectories of the movement of people;

[0160] a probability calculation unit configured to analyze the converging points and the dispersing points of the flow of people in the network topology map of the flow of people in the medical treatment points, and calculate the probabilities of the transfer of the flow of people between the medical treatment points;

[0161] a trend prediction unit configured to predict the congestion evolution trends of the medical treatment points in a future time window based on the probabilities of the transfer of the flow of people and the historical treatment data of the medical treatment points;

[0162] a path optimization unit configured to establish a path optimization model with a target function of minimizing the average waiting time of patients according to the congestion evolution trends and the real-time treatment capacities of the medical treatment points, and calculate an optimal combination of the paths of diversion;

[0163] a scheduling execution unit configured to generate specific scheduling instructions of diversion according to the optimal combination of the paths of diversion, and plan individualized navigation paths for the patients based on the characteristics of the trajectories of the movement of people.

[0164] A technical solution provided in the embodiments of the present application is an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and capable of running on the processor, and the processor implements the steps in any of the methods when executing the computer program.

[0165] The specific embodiments described above are the preferred embodiments of the present application, and do not limit the specific implementation range of the present application. The range of the present application includes but is not limited to the specific embodiments, and equivalent changes made according to the shape and structure of the present application are within the protection scope of the present application.

Claims

1. An outpatient process intelligent scheduling and resource optimization method, characterized in that, The method comprises the following steps: Collecting images of waiting areas of each medical treatment point, extracting personnel density features in the waiting areas, and combining a multi-target tracking algorithm to extract personnel moving tracks; Obtaining density distribution values and density change values according to the personnel density features, and calculating service capacity values of network nodes; Calculating flow intensity values between network nodes based on personnel moving track features, and setting a ratio of the flow intensity values to the service capacity values as edge weight values between the network nodes; Setting the medical treatment points as the network nodes, and constructing a medical treatment point human flow network topology graph according to the service capacity values and the edge weight values; Obtaining human flow net inflow rates of each network node in the medical treatment point human flow network topology graph in continuous multiple time windows; Identifying attributes of the network nodes according to change trends of the human flow net inflow rates in the continuous multiple time windows, marking nodes meeting a first change trend as human flow converging points, and marking nodes meeting a second change trend as human flow dispersing points; Calculating human flow transfer probabilities between the network nodes according to the marking results of the human flow converging points and the human flow dispersing points and human flow net inflow rates of target nodes in a next time window; Based on the human flow transfer probabilities and historical treatment data of the medical treatment points, predicting congestion evolution trends of each medical treatment point in a future time window; According to the congestion evolution trends and real-time treatment capacities of the medical treatment points, establishing a path optimization model with a target function of minimizing average waiting time of patients, and calculating an optimal shunt path combination; Generating specific shunt scheduling instructions according to the optimal shunt path combination, and planning individualized navigation paths for the patients based on the personnel moving track features.

2. The method of claim 1, wherein, The collecting images of waiting areas of each medical treatment point, extracting personnel density features in the waiting areas, and combining a multi-target tracking algorithm to extract personnel moving tracks comprises: Adjusting collection parameters according to light information and human flow information in the images of the waiting areas; Performing distortion correction and perspective transformation on the images of the waiting areas to obtain overhead images, and performing standardization processing on the overhead images to obtain standardized images; Dividing grid units based on personnel distribution in the standardized images, counting personnel numbers in the grid units to obtain grid density values, determining a smoothing radius according to the grid density values, and performing smoothing processing on the grid density values by using the smoothing radius to obtain continuous density distribution; Calculating a time-space change rate of the continuous density distribution to obtain density change features, and combining the continuous density distribution and the density change features to construct personnel density features of the waiting areas; Extracting feature vectors of personnel targets in the standardized images, calculating a similarity between continuous frames based on the feature vectors by using a multi-target tracking algorithm to obtain a correlation matrix, simultaneously tracking multiple personnel targets by using the correlation matrix to obtain track sequences, and constructing personnel moving tracks according to the track sequences.

3. The method of claim 1, wherein, The predicting congestion evolution trends of each medical treatment point in a future time window based on the human flow transfer probabilities and historical treatment data of the medical treatment points comprises: Calculating transfer prediction treatment data at a target time based on the human flow transfer probabilities and combining historical treatment data of each medical treatment point; Performing time series analysis on historical treatment data in a preset time window to obtain periodic features, and calculating time series prediction treatment data at the target time according to the periodic features by using a neural network. calculating prediction errors of the transfer prediction visit data and the timing prediction visit data relative to actual visit data, generating dynamic weights according to the prediction errors; weighting and fusing the transfer prediction visit data and the timing prediction visit data according to the dynamic weights to obtain prediction visit data at a target time; calculating a capacity occupancy degree and a visit change trend based on the prediction visit data, and determining a congestion evolution trend of the diagnosis and treatment point.

4. The method of claim 1, wherein, establishing a path optimization model with a target function of minimizing the average waiting time of patients according to the congestion evolution trend and the real-time reception capacity of the diagnosis and treatment point, and calculating an optimal shunt path combination including: determining a congestion correlation relationship between adjacent road segments according to the congestion evolution trend, and generating a road congestion index within a target time window; obtaining medical resource state data and reception efficiency data of the diagnosis and treatment point, and evaluating the real-time reception capacity of the diagnosis and treatment point to obtain a real-time reception amount of the diagnosis and treatment point; calculating a patient path travel time according to the road congestion index, and calculating a patient diagnosis and treatment waiting time based on the real-time reception amount of the diagnosis and treatment point, and taking a sum of the path travel time and the diagnosis and treatment waiting time as the average waiting time of the patients; constructing a path optimization model, taking the congestion evolution trend and the real-time reception amount of the diagnosis and treatment point as state parameters, and taking a patient distribution scheme as a decision parameter, and constructing a target function based on the average waiting time of the patients; iteratively calculating the path optimization model until an optimal shunt path combination that meets the traffic conditions and the reception amount limit is obtained.

5. The method of claim 1, wherein, generating specific shunt scheduling instructions according to the optimal shunt path combination, and planning personalized navigation paths for patients based on personnel movement trajectory features including: calculating a regional patient flow distribution ratio according to the optimal shunt path combination, determining a regional transfer threshold based on road traffic capacity and diagnosis and treatment point reception amount, and generating shunt scheduling instructions; collecting patient historical movement trajectories, extracting movement behavior features, and classifying patients according to movement ability; splitting the shunt scheduling instructions according to the movement ability classification, calculating candidate path combinations based on the road network structure, and assigning differential weights to the path combinations; generating intersection turning instructions based on the path combinations with differential weights and real-time road conditions, calculating travel times combined with signal timing, and forming personalized navigation paths.

6. The method of claim 5, wherein, The method further includes: obtaining a real-time position of the patient, and re-planning a new path from the candidate path combinations when the position deviates from the personalized navigation path; updating the actual travel data of the patient to the historical movement trajectory, re-classifying the movement ability, adjusting the differential weights, and updating the personalized navigation path.

7. An outpatient process intelligent scheduling and resource optimization system for implementing the method of any one of claims 1-6, characterized in that, The system includes: a data acquisition unit configured to collect images of a waiting area of each diagnosis and treatment point, extract personnel density features in the waiting area, and extract personnel movement trajectories combined with a multi-target tracking algorithm; a topology construction unit configured to obtain density distribution values and density change values according to the personnel density features, and calculate service capacity values of network nodes; calculating flow intensity values between network nodes based on the personnel movement trajectory features, and setting a ratio of the flow intensity values and the service capacity values as edge weight values between the network nodes; The medical treatment point is set as a network node, and a medical treatment point people flow network topology graph is constructed according to the service capability value and the edge weight value; A probability calculation unit is configured to acquire a people flow net inflow rate of each network node in the medical treatment point people flow network topology graph in continuous multiple time windows; According to a change trend of the people flow net inflow rate in the continuous multiple time windows, the attribute of the network node is identified, a node meeting a first change trend is marked as a people flow converging point, and a node meeting a second change trend is marked as a people flow dispersing point; According to the marking result of the people flow converging point and the people flow dispersing point and the people flow net inflow rate of a target node in a next time window, a people flow transfer probability between network nodes is calculated; A trend prediction unit is configured to predict a congestion evolution trend of each medical treatment point in a future time window based on the people flow transfer probability and historical medical treatment data of the medical treatment point; A path optimization unit is configured to establish a path optimization model with a target function of minimizing an average waiting time of patients according to the congestion evolution trend and real-time reception capacity of the medical treatment point, and to calculate an optimal shunt path combination; A scheduling execution unit is configured to generate a specific shunt scheduling instruction according to the optimal shunt path combination, and to plan a personalized navigation path for the patient based on a personnel movement trajectory feature.

8. An electronic device, comprising: The computer program product comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps in the method according to any one of claims 1 to 6 when executing the computer program. The computer program product comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps in the method according to any one of claims 1 to 6 when executing the computer program.

Images