Physical examination intelligent shunting method, device and equipment and storage medium
By constructing a time series data set and a machine learning model, the optimal physical examination path is generated, which solves the problems of long waiting time for physical examinations and irrational resource allocation, and realizes efficient and convenient physical examination services.
Patent Information
- Application Number
- CN202510478655.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-16
- Publication Date
- 2025-09-12
AI Technical Summary
The existing physical examination process lacks a scientific and reasonable diversion mechanism, resulting in long waiting times for physical examinations and irrational resource allocation, making it difficult to meet the needs of physical examination recipients for efficient and convenient services.
By collecting physical examination data, building a time series dataset and training a machine learning model, we can predict the number of examinees and their paths, generate the optimal physical examination path, and update the path in real time to optimize the physical examination process.
Reduce waiting time for physical examinations, allocate resources rationally, improve physical examination efficiency and resource utilization, and enhance the experience and satisfaction of physical examination recipients.
Smart Images
Figure CN120634074A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of data processing technology, and specifically relates to a physical examination intelligent diversion method, device, equipment and storage medium. Background Art
[0002] With the steady improvement of living standards, people are paying more attention to their health and becoming more aware of their health, which has led to a continuous increase in demand for the health checkup industry. During the health checkup process, some examination items require a long time to complete due to complex procedures and limited equipment, which often leads to long waiting lines for these items. This is especially true during group checkups for companies, where the large number of people undergoing examinations increases waiting times significantly, seriously affecting the efficiency of the examination and the patient experience.
[0003] The traditional medical examination process lacks a scientific and rational diversion mechanism. Patients often arbitrarily choose the order of examination items based on personal experience or on-site guidance. This can easily lead to excessive crowding of people at popular examination items, while other items remain idle, resulting in underutilized medical examination resources. Although some medical examination institutions have tried to alleviate the queue problem by using manual guidance, manual guidance not only consumes a lot of manpower costs, but its effectiveness is also greatly affected by staff experience and the complexities of the on-site situation. This makes it difficult to fundamentally solve the problems of long waiting times and irrational resource allocation.
[0004] In recent years, big data and artificial intelligence technologies have developed rapidly, achieving remarkable results in numerous fields. Although these technologies are gradually being applied in the medical industry, their application in health checkup triage is still in its exploratory stages. Currently, methods for intelligently triaging patients for different health checkup items based on big data and artificial intelligence models are not yet mature. This method fails to effectively reduce wait times or optimize the configuration of various examination items, making it difficult to meet patients' demand for efficient and convenient health checkup services. Summary of the Invention
[0005] In view of the problem that existing methods cannot effectively reduce the waiting time for physical examinations, optimize the configuration of various physical examination items, and are difficult to meet the physical examination subjects' needs for efficient and convenient physical examination services, the present invention provides a physical examination intelligent diversion method, device, equipment and storage medium.
[0006] In a first aspect, the technical solution of the present invention provides a physical examination intelligent triage method, comprising: Desensitized collection of medical examination item data, medical examination result data, appointment data, and basic registration information and historical medical examination data of medical examination institutions; Standardize the collected data to form initial sample data; Based on the initial sample data, the benchmark time for physical examination items and the time required to travel between physical examination items are calculated respectively; Based on the initial sample data, the set physical examination period is divided into time nodes according to the set intervals to construct a time series data set. The machine learning model is trained based on the time series data set to obtain a prediction model. Based on the prediction model, the number of physical examinations at the physical examination institution is predicted using the data from the previous day. Based on the calculation and prediction results, a physical examination triage model is constructed to predict the time nodes for the examinees to complete each physical examination item and generate the optimal physical examination path; Update the time required for subsequent physical examination items based on real-time data and optimize the physical examination path.
[0007] As a further limitation of the technical solution of the present invention, in the step of respectively calculating the reference time of the physical examination items and the time required for the distance between the physical examination items based on the initial sample data, the step of calculating the reference time of the physical examination items includes: Divide the physical examination data in the initial sample data into categorical data and numerical data; Encode the categorical data and map each category label to a dense vector space to obtain a category vector. The category vector is then concatenated with the numerical data to form new feature data. Input the new feature data into the multi-layer perceptron model to obtain the feature vector and generate a feature vector set; Perform cluster analysis on the eigenvectors and calculate the benchmark time of physical examination items.
[0008] As a further limitation of the technical solution of the present invention, the step of performing cluster analysis on the feature vectors and calculating the reference time of the physical examination items includes: Clustering feature vectors; After clustering, the actual time taken by all examinees to complete the physical examination items in each cluster is obtained, and the average of the actual time in each cluster is calculated as the benchmark time for the physical examination items corresponding to the cluster; Determine the cluster to which the feature vector of a new examinee belongs; The reference time of the physical examination item corresponding to the cluster is obtained as the reference time of the physical examination item of the new physical examination subject.
[0009] As a further limitation of the technical solution of the present invention, the step of clustering the feature vectors includes: In the feature vector set of physical examination data, each feature vector is a data point, randomly selected K data points as the initial K The centroid of each cluster is used to divide the feature vector set into K clusters; For each data point, calculate its Euclidean distance to the K centroids respectively; assign the data point to the cluster with the nearest centroid based on the calculated result; For each cluster, calculate the mean of all data points in the cluster as the new centroid; The process of assigning data points to the cluster with the nearest centroid and updating the centroid is repeated until the change in the centroid is less than the preset threshold or the maximum number of iterations is reached, and clustering is completed; The new centroid of the jth cluster The calculation formula is:
[0010] in, is the jth cluster, is the number of data points in the jth cluster, for Data points within.
[0011] As a further limitation of the technical solution of the present invention, in the step of respectively calculating the reference time of the physical examination items and the time required for the distance between the physical examination items based on the initial sample data, the step of calculating the time required for the distance between the physical examination items includes: Obtain the start time of two adjacent physical examination items X and Y from the initial sample data, and determine the time range from item X to item Y; Obtain the actual distance between each physical examination item based on the physical examination center's floor plan; Obtain basic registration information of the examinees from the initial sample data; Obtain the physical examinee's stride length and cadence based on the physical examinee's basic registration information; Calculate the time required to travel from one physical examination item to another based on the actual distance between the physical examination items within the physical examination institution and the stride length and frequency of each examinee; The time required between each physical examination item calculated for each physical examination subject is counted, and the average value is calculated to obtain the time required between the physical examination items.
[0012] As a further limitation of the technical solution of the present invention, based on the initial sample data, the set physical examination time period is divided into time nodes according to set intervals to construct a time series data set; and a machine learning model is trained based on the time series data set to obtain a prediction model. The steps of predicting the number of physical examination recipients at the physical examination institution using the data of the physical examination institution from the previous day based on the prediction model include: According to the set physical examination time period, filter out the registration time data within the time period; Divide the set physical examination time period according to the set intervals to form multiple time nodes; Traverse the filtered data and count the number of registered physical examination candidates at each time point in the initial sample data, and use it as the number of people who have been examined at that time point; Combine each time node and its corresponding number of people who came for inspection into a time series dataset. Each sample in the time series dataset contains a time node and the number of people who came for inspection corresponding to that node, and the samples are arranged in chronological order. The GRU model is trained based on the time series dataset to obtain a trained prediction model; The registration time data of the physical examination participants on the previous day is collected and processed, and then input into the trained prediction model to obtain the predicted value of the number of people attending the examination at each time node on the day.
[0013] As a further limitation of the technical solution of the present invention, based on the calculation results and the prediction results, a physical examination triage model is constructed to predict the time nodes when the examinee completes each physical examination item, and the steps of generating the optimal physical examination path include: Each physical examination item is considered a node in the graph, and the connection between two physical examination items is considered an edge in the graph. The weight of the edge is the time required to travel between the physical examination items. A queuing model is introduced. According to the predicted number of people arriving for examination at different time nodes, the number of people in the queue and the expected waiting time for each physical examination item are dynamically adjusted to complete the construction of the physical examination diversion model. The queuing model is: ,in, Waiting time in queue, is the inspection rate calculated based on the predicted number of inspection personnel, It is the reciprocal of the benchmark time of the physical examination item; Through the constructed physical examination triage model, the time nodes for the examinees to complete each physical examination item are predicted, and the optimal physical examination path is generated; During the physical examination process, the actual number of people arriving for examination and the number of people in queue are obtained in real time. Based on the real-time data, the predicted number of people in queue and the node processing time (i.e., the benchmark time for physical examination items) are updated, and the optimal path is recalculated. In a second aspect, the technical solution of the present invention further provides a physical examination intelligent diversion device, comprising a data acquisition module, a preprocessing module, a first calculation module, a physical examination number prediction module, an optimal path generation module and a path update module; The data acquisition module is used to desensitize and collect the medical examination item data, medical examination result data, appointment data, and basic registration information and historical medical examination data of the medical examination institution; The preprocessing module is used to standardize the collected data to form initial sample data; The first calculation module is used to calculate the reference time of the physical examination items and the time required for the distance between the physical examination items based on the initial sample data; The module for predicting the number of physical examination participants is used to divide the set physical examination period into time nodes according to the set intervals based on the initial sample data, and construct a time series data set. The module also trains the machine learning model based on the time series data set to obtain a prediction model, and uses the data from the previous day of the physical examination institution to predict the number of physical examination participants based on the prediction model. The optimal path generation module is used to build a physical examination triage model based on the calculation results and prediction results, predict the time nodes when the examinee will complete each physical examination item, and generate the optimal physical examination path; The path update module is used to update the time required for subsequent physical examination items based on real-time data and optimize the physical examination path.
[0014] As a further limitation of the technical solution of the present invention, the first calculation module includes a benchmark time calculation unit, which is specifically used to divide the physical examination data in the initial sample data into categorical data and numerical data; encode the categorical data, and map each category label to a dense vector space to obtain a category vector, splicing the category vector with the numerical data to form new feature data; input the new feature data into a multi-layer perceptron model to obtain a feature vector, and generate a feature vector set; perform cluster analysis on the feature vector to calculate the benchmark time of the physical examination items.
[0015] As a further limitation of the technical solution of the present invention, the reference time calculation unit is also used to cluster the feature vectors; obtain the actual time for all examinees in each cluster to complete the physical examination items after clustering, and calculate the average value of the actual time in each cluster as the physical examination item reference time corresponding to the cluster; determine the cluster to which the feature vector of a new examinee belongs; obtain the physical examination item reference time corresponding to the cluster as the physical examination item reference time of the new examinee. Specifically, in the feature vector set of the physical examination data, each feature feature vector is a data point, and a random selection is made. K data points as the initial K The feature vector set is divided into K clusters based on the centroids of the clusters. For each data point, the Euclidean distance to each of the K centroids is calculated. Based on the calculated distance, the data point is assigned to the cluster with the nearest centroid. For each cluster, the mean of all data points in the cluster is calculated as the new centroid. The process of assigning data points to the cluster with the nearest centroid and updating the centroid is repeated until the change in the centroid is less than a preset threshold or the maximum number of iterations is reached, completing the clustering. The new centroid of the jth cluster The calculation formula is:
[0016] in, is the jth cluster, is the number of data points in the jth cluster, for Data points within.
[0017] As a further limitation of the technical solution of the present invention, the first calculation module also includes a distance required time calculation unit, which is used to obtain the start time of two adjacent physical examination items X and physical examination item Y of the examinee in the initial sample data, and determine the time range spent from item X to item Y; obtain the actual travel distance between each physical examination item based on the floor plan of the physical examination center; obtain the basic registration information of the examinee in the initial sample data; obtain the examinee's stride length and cadence based on the examinee's basic registration information; calculate the time required to travel from one physical examination item to another based on the actual travel distance between each physical examination item in the physical examination institution and the stride length and cadence of each examinee; and statistically analyze the time required between each physical examination item calculated for each examinee, and calculate the average value to obtain the time required for the distance between the physical examination items.
[0018] As a further limitation of the technical solution of the present invention, the physical examination number prediction module is specifically used to filter out the registration time data within the set physical examination time period according to the set physical examination time period; divide the set physical examination time period according to the set interval to form multiple time nodes; traverse the filtered data, count the number of physical examination registrations within each time node in the initial sample data, and use it as the number of people who arrived for examination at the time node; combine each time node and its corresponding group of people who arrived for examination into a time series data set; each sample in the time series data set contains a time node and the number of people who arrived for examination corresponding to the node, and the samples are arranged in chronological order; train the GRU model based on the time series data set to obtain a trained prediction model; collect the physical examination registration time data of the previous day, process it, and then input it into the trained prediction model to obtain the predicted value of the number of people who arrived for examination at each time node on the day.
[0019] As a further limitation of the technical solution of the present invention, the optimal path generation module is specifically used to regard each physical examination item as a node in the graph, and the connection between two physical examination items as an edge in the graph. The weight of the edge is the time required for the distance between the physical examination items. A queuing model is introduced to dynamically adjust the number of people in the queue and the expected waiting time for each physical examination item based on the predicted number of people arriving for examination at different time nodes, thereby completing the construction of the physical examination diversion model; the queuing model is: ,in, Waiting time in queue, is the inspection rate calculated based on the predicted number of inspection personnel, It is the inverse of the benchmark time of the physical examination items; through the constructed physical examination diversion model, the time nodes for the examinees to complete each physical examination item are predicted, and the optimal physical examination path is generated; during the physical examination process, the actual number of people arriving for examination and the number of people in the queue are obtained in real time, and the predicted value of the number of people in the queue and the node processing time, that is, the benchmark time of the physical examination items, are updated according to the real-time data, and the optimal path is recalculated.
[0020] In a third aspect, the technical solution of the present invention also provides an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores computer program instructions that can be executed by the at least one processor, and the computer program instructions are executed by the at least one processor so that the at least one processor can execute the intelligent diversion method for physical examination as described in the first aspect.
[0021] In a fourth aspect, the technical solution of the present invention further provides a non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the intelligent physical examination triage method as described in the first aspect.
[0022] The beneficial effect of the technical solution of the present invention is that it can plan a reasonable physical examination route for the examinee based on the historical data and real-time situation of the physical examination institution. The route fully takes into account the benchmark time of each physical examination item, the time required for the distance between items, and the predicted number of people to be examined, avoiding the situation where the examinees are concentrated in queuing for a certain physical examination item, reducing unnecessary waiting time, making the entire physical examination process smoother, and significantly improving the efficiency of the physical examination. It can reasonably allocate physical examination resources, avoid the idleness or excessive use of equipment and medical personnel due to excessive concentration of personnel in some physical examination items, and at the same time allow other physical examination items to be fully utilized, optimize the resource allocation of the physical examination institution, improve resource utilization, and reduce operating costs. It can provide the examinee with an estimated completion time for each physical examination item, so that the examinee can better plan his time, avoid anxiety due to long waiting times, and improve the examinee's experience and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] In order to more clearly illustrate the technical solution of the present invention, the following is a brief introduction to the drawings required for the description. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0024] Figure 1 A flowchart of a method provided in an embodiment of the present invention.
[0025] Figure 2 This is a connection block diagram of the device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0026] In order to make the purpose, features, and advantages of the present invention more obvious and easy to understand, the technical solutions of the present invention will be clearly and completely described below in conjunction with the drawings in the specific embodiments. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0027] like Figure 1 As shown, an embodiment of the present invention provides a physical examination intelligent diversion method, including: S1: Desensitized collection of medical examination item data, medical examination result data, appointment data, and basic registration information and historical medical examination data of medical examination institutions; It should be noted that to minimize the impact on physical examination services, a collection task is created to periodically collect physical examination data from physical examination institutions and add it to the model library when no patients are undergoing examinations. When creating the collection task, sensitive information such as name, ID number, and mobile phone number is desensitized to prevent personal information leakage. During the desensitization process, it is necessary to ensure that the unique identifier of each examinee remains consistent and to ensure the correlation between the physical examination item data and the physical examination result data.
[0028] S2: Standardize the collected data to form initial sample data; Due to the differences in the business systems of various physical examination institutions, it is necessary to standardize the data of physical examination institutions. By building a governance program, configuring the mapping relationship between the two, creating collection tasks, connecting the model library and the physical examination institution business library, desensitizing the physical examination data collected by the physical examination institutions to the model library, removing invalid or abnormal data, and obtaining the initial sample data required by the model.
[0029] S3: Based on the initial sample data, calculate the benchmark time of the physical examination items and the time required to travel between physical examination items; This step includes S31: calculating the reference time of the physical examination items; It should be noted that based on the data of examinees of various physical examination items in the physical examination institutions in recent years, they are grouped by medical staff or large medical equipment and sorted by time. The start time of the first examinee of the group on that day is used as the starting time, and the difference in the start time of the physical examination item of the two adjacent examinees A and B is calculated as the physical examination time of the previous examinee for that item. The time set of the item is obtained in turn, and divided into the baseline time combination, positive time set and major positive set according to the physical examination results being normal, positive and major positive. Embedding + MLP is used to process data containing multiple physical examination indicators. The advantage of using Embedding is that for physical examination data with rich categories (such as gender, smoking or not), the embedding layer (Embedding Layer) can effectively capture the similarities between categories and reduce the problems caused by high-dimensional sparsity by mapping the labels of each category to a dense vector space. Then the numerical data and the embedded category vector are spliced and input into the MLP model for processing to obtain the feature vector. Specifically including: S311: Divide the physical examination data in the initial sample data into categorical data and numerical data; S312: Encode the categorical data and map each category label to a dense vector space to obtain a category vector. Concatenate the category vector with the numerical data to form new feature data. S313: Input the new feature data into the multilayer perceptron model to obtain a feature vector and generate a feature vector set; S314: clustering the eigenvectors; specifically, in the eigenvector set of the physical examination data, each eigenvector is a data point, randomly selecting K data points as the initial K The centroid of each cluster is used to divide the feature vector set into K clusters; For each data point, calculate its Euclidean distance to the K centroids respectively; assign the data point to the cluster with the nearest centroid based on the calculated result; For each cluster, calculate the mean of all data points in the cluster as the new centroid; The process of assigning data points to the cluster with the nearest centroid and updating the centroid is repeated until the change in the centroid is less than the preset threshold or the maximum number of iterations is reached, and clustering is completed; The new centroid of the jth cluster The calculation formula is:
[0030] in, is the jth cluster, is the number of data points in the jth cluster, for Data points within.
[0031] S315: Obtain the actual time taken by all examinees to complete the physical examination items in each cluster after clustering, and calculate the average of the actual time in each cluster as the reference time for the physical examination items corresponding to the cluster; S316: Determine the cluster to which the feature vector of the new examinee belongs; S317: Obtain the reference time of the physical examination items corresponding to the cluster as the reference time of the physical examination items of the new examinee.
[0032] The method further includes S32: calculating the time required for the distance between the physical examination items; specifically, the method includes: S321: Obtain the start time of two adjacent physical examination items X and Y of the examinee in the initial sample data, and determine the time range spent from item X to item Y; S322: Obtaining the actual travel distances between various physical examination items based on the floor plan of the physical examination center; S323: Obtaining basic registration information of the examinee from the initial sample data; S324: Obtaining the stride length and stride frequency of the examinee based on the examinee's registered basic information; S325: Calculate the time required to travel from one physical examination item to another based on the actual distance between the physical examination items in the physical examination institution and the stride length and stride frequency of each examinee; S326: Counting the time required between each physical examination item calculated for each examinee, and calculating the average value to obtain the time required between the physical examination items.
[0033] It should be noted that the initial sample data is based on the start time of two adjacent physical examination items X and physical examination item Y, as well as the physical examination items that completed physical examination item Y within the above time period. The basic model is constructed based on the distribution of each physical examination item of the physical examination institution and the stride length and stride frequency. The calculation formula is as follows:
[0034] Stride length is generally proportional to height, and the empirical formula is as follows: Stride length = 0.413 × height; Weight and age affect athletic ability, which in turn affects cadence. A multivariate regression model was constructed based on height and age to capture the nonlinear relationship between these characteristics. The regression coefficients (a, b, c) were obtained through analysis. The formula is as follows: Cadence = a + b ×weight+ c × Age. Different regression coefficients are set here according to age and weight. In this application, only the corresponding coefficients are selected based on age and weight to calculate the cadence.
[0035] Considering that weight and age affect cadence, a multivariate regression model was constructed: "Cadence = a + b × weight + c × age." Here, a, b, and c are coefficients to be determined, which can be obtained through regression analysis of the actual walking data of a large number of physical examination subjects. For example, weight, age, and actual cadence data for 1,000 physical examination subjects were collected, and the values of coefficients a, b, and c were fitted using methods such as least squares.
[0036] Calculate each examinee's cadence: Substitute each examinee's weight and age into the above multiple regression model to calculate their cadence. Cadence is usually measured in steps per minute.
[0037] S4: Based on the initial sample data, the set physical examination period is divided into time nodes according to the set intervals to construct a time series data set; the machine learning model is trained based on the time series data set to obtain a prediction model, and the number of physical examinations at the physical examination institution is predicted based on the prediction model using the data of the physical examination institution from the previous day; This step specifically includes S41: according to the set physical examination time period, filtering out the registration time data within the time period; S42: Divide the set physical examination time period according to the set intervals to form multiple time nodes; S43: Traverse the filtered data, count the number of registered physical examination candidates at each time point in the initial sample data, and use it as the number of people who have been examined at that time point; S44: Combining each time node and its corresponding number of people who came for inspection into a time series data set; each sample in the time series data set contains a time node and the number of people who came for inspection corresponding to the node, and the samples are arranged in chronological order; S45: Train the GRU model based on the time series dataset to obtain a trained prediction model; It should be noted that the steps for training a GRU model based on a time series dataset include: S451: Divide the time series dataset into an input sequence X and a corresponding target value y according to a certain time step. For the task of predicting the number of people arriving for inspection at each time node on the day, the input sequence can be the number of people arriving for inspection at the previous n time nodes, and the target value is the number of people arriving for inspection at the next time node. S452: Initialize the weight matrix and bias vector of the GRU model; S453: For input sequence X For each sample in, the update gate is calculated in sequence according to the time step 、 Reset the gate, candidate hidden state, and current hidden state to get the predicted output. This step is calculated according to the specific formula of the GRU model. S454: Select mean square error as the loss function to measure the difference between the predicted output and the target value; S45: The registration time data of the physical examination participants on the previous day is collected and processed, and then input into the trained prediction model to obtain the predicted value of the number of people who will be examined at each time point on the day.
[0038] S456: Use the backpropagation algorithm to calculate the gradient of the loss function with respect to the model parameters; S457: Use the Adam optimization algorithm to update the model parameters according to the calculated gradient.
[0039] S5: Based on the calculation and prediction results, a physical examination triage model is constructed to predict the time nodes for the examinees to complete each physical examination item and generate the optimal physical examination path; Objective function , T Total time range, physical examination items Dynamic processing time It is the sum of the baseline time and the queue waiting time. is a binary variable. If the examinee is at time t From the physical examination items i Transfer to physical examination items j ,but =1, otherwise =0; edge weight For physical examination items i To the physical examination items j The time required.
[0040] S6: Update the time required for subsequent physical examination items based on real-time data and optimize the physical examination path.
[0041] In some embodiments, based on the calculation results and the prediction results, a physical examination triage model is constructed to predict the time nodes when the examinee will complete each physical examination item. The steps of generating the optimal physical examination path include: S51: Consider each medical examination item as a node in the graph, and the connection between two medical examination items as an edge in the graph. The weight of the edge is the time required to travel the distance between the medical examination items. A queuing model is introduced. Based on the predicted number of people arriving for examination at different time nodes, the number of people in the queue and the estimated waiting time for each medical examination item are dynamically adjusted to complete the construction of the medical examination diversion model. The queuing model is: ,in, Waiting time in queue, is the inspection rate calculated based on the predicted number of inspection personnel, It is the reciprocal of the benchmark time of the physical examination item; S52: Using the constructed physical examination triage model, the time nodes for the examinee to complete each physical examination item are predicted, and the optimal physical examination path is generated; S53: During the physical examination process, the actual number of people arriving for examination and the number of people in queue are obtained in real time. The predicted number of people in queue and the node processing time (i.e., the benchmark time of the physical examination item) are updated based on the real-time data, and the optimal path is recalculated. Specifically, the following steps are performed: For each person undergoing physical examination, record their arrival time and a list of physical examination items ， Assume that the first physical examination item is , the time to arrive at the physical examination item ,in From the starting point to the physical examination items The time required to travel the distance.
[0042] Calculation of physical examination items Waiting time in queue (Calculated based on the number of people in the queue at the current time and the queue model), physical examination items Start time , end time ,in It is a physical examination item The base time; Calculate the physical examination items for the physical examination items ( ), the time to arrive at the physical examination item , From the physical examination items To the physical examination items Time required to cover the distance; Same calculation of physical examination items Waiting time in queue , physical examination items Start time , end time ,in It is a physical examination item The base time; Let the starting node be s, initialize the distance of all nodes to infinity, the distance of the starting node d(s) = 0, maintain a priority queue Mq, add all nodes to the queue, and sort them by distance from small to large. Take the node u with the smallest distance from the priority queue Mq. u All adjacent nodes of v , calculate from the starting node through u arrive v Total time , It's the edge The weight (i.e. u arrive v time required to travel the distance), is a node vThe waiting time in queues, is a node v The benchmark time for physical examination items, From the starting node to the node u The total time of the currently known shortest path, if , then update , and record v The predecessor node is u , starting from the end node, backtracking to the start node according to the predecessor node information to obtain the optimal path.
[0043] Each time it is taken out from the priority queue Mq The node with the smallest value u This is because we believe that the current Smallest node u It is the closest to the starting node. Continuing to explore other nodes from this node may find a shorter path.
[0044] For the removed nodes u All adjacent nodes of v , calculate from the starting node through u arrive v Total time .if Less than the current , indicating that a shorter route to the node has been found v The path will be updated for l, and record v The predecessor node is u By continuously performing such updating operations, the shortest path from the starting node to all other nodes can be found eventually.
[0045] like Figure 2 As shown, an embodiment of the present invention provides a physical examination intelligent diversion device, including a data acquisition module, a preprocessing module, a first calculation module, a physical examination number prediction module, an optimal path generation module and a path update module; The data acquisition module is used to desensitize and collect the medical examination item data, medical examination result data, appointment data, and basic registration information and historical medical examination data of the medical examination institution; The preprocessing module is used to standardize the collected data to form initial sample data; The first calculation module is used to calculate the reference time of the physical examination items and the time required for the distance between the physical examination items based on the initial sample data; The module for predicting the number of physical examination participants is used to divide the set physical examination period into time nodes according to the set intervals based on the initial sample data, and construct a time series data set. The module also trains the machine learning model based on the time series data set to obtain a prediction model, and uses the data from the previous day of the physical examination institution to predict the number of physical examination participants based on the prediction model. The optimal path generation module is used to build a physical examination triage model based on the calculation results and prediction results, predict the time nodes when the examinee will complete each physical examination item, and generate the optimal physical examination path; The path update module is used to update the time required for subsequent physical examination items based on real-time data and optimize the physical examination path.
[0046] In some embodiments, the first calculation module includes a reference time calculation unit, which is specifically used to divide the physical examination data in the initial sample data into categorical data and numerical data; encode the categorical data, and map each category label to a dense vector space to obtain a category vector, and splice the category vector with the numerical data to form new feature data; input the new feature data into a multi-layer perceptron model to obtain a feature vector, and generate a feature vector set; perform cluster analysis on the feature vector to calculate the reference time of the physical examination items.
[0047] In some embodiments, the reference time calculation unit is also used to cluster the feature vectors; obtain the actual time for all examinees in each cluster to complete the physical examination items after clustering, and calculate the average value of the actual time in each cluster as the physical examination item reference time corresponding to the cluster; determine the cluster to which the feature vector of a new examinee belongs; obtain the physical examination item reference time corresponding to the cluster as the physical examination item reference time of the new examinee. Specifically, in the feature vector set of the physical examination data, each feature feature vector is a data point, and a random selection is made. K data points as the initial K The feature vector set is divided into K clusters based on the centroids of the clusters. For each data point, the Euclidean distance to each of the K centroids is calculated. Based on the calculated distance, the data point is assigned to the cluster with the nearest centroid. For each cluster, the mean of all data points in the cluster is calculated as the new centroid. The process of assigning data points to the cluster with the nearest centroid and updating the centroid is repeated until the change in the centroid is less than a preset threshold or the maximum number of iterations is reached, completing the clustering. The new centroid of the jth cluster The calculation formula is:
[0048] in, is the jth cluster, is the number of data points in the jth cluster, for Data points within.
[0049] In some embodiments, the first calculation module further includes a distance required time calculation unit, which is configured to obtain the start time of two adjacent physical examination items X and Y of the examinee in the initial sample data, and determine the time range spent from item X to item Y; obtain the actual travel distance between each physical examination item based on the floor plan of the physical examination center; obtain the examinee's basic registration information in the initial sample data; obtain the examinee's stride length and cadence based on the examinee's basic registration information; calculate the time required to travel from one physical examination item to another based on the actual travel distance between each physical examination item in the physical examination institution and the stride length and cadence of each examinee; and statistically analyze the time required between each physical examination item calculated for each examinee, and calculate the average value to obtain the time required to travel between the physical examination items.
[0050] In some embodiments, the physical examination number prediction module is specifically used to filter out the registration time data within the set physical examination time period according to the set physical examination time period; divide the set physical examination time period according to the set interval to form multiple time nodes; traverse the filtered data, count the number of physical examination registrations within each time node in the initial sample data, and use it as the number of people who arrive for examination at the time node; combine each time node and its corresponding group of people who arrive for examination into a time series data set; each sample in the time series data set contains a time node and the number of people who arrive for examination corresponding to the node, and the samples are arranged in chronological order; train the GRU model based on the time series data set to obtain a trained prediction model; collect the physical examination registration time data of the previous day, process it, and input it into the trained prediction model to obtain the predicted value of the number of people who arrive for examination at each time node on that day.
[0051] In some embodiments, the optimal path generation module is specifically configured to treat each medical examination item as a node in a graph, and the connection between two medical examination items as an edge in the graph. The weight of the edge is the time required to travel the distance between the medical examination items. A queuing model is introduced to dynamically adjust the number of people in the queue and the estimated waiting time for each medical examination item based on the predicted number of people arriving for examination at different time nodes, thereby completing the construction of the medical examination triage model. The queuing model is: ,in, Waiting time in queue, is the inspection rate calculated based on the predicted number of inspection personnel, The inverse of the baseline time for the physical examination items; through the constructed physical examination triage model, the time nodes for the examinees to complete each physical examination item are predicted, and the optimal physical examination path is generated; during the physical examination process, the actual number of examinees and queues are obtained in real time, and the predicted number of queues and node processing time (i.e., the baseline time for the physical examination items) are updated based on the real-time data, and the optimal path is recalculated; An embodiment of the present invention further provides an electronic device comprising: a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus. The communication bus can be used to transmit information between the electronic device and a sensor. The processor can call the logic instructions in the memory to execute the following method: S1: desensitizing and collecting the physical examination item data, physical examination result data, appointment data, and basic registration information and historical physical examination data of the physical examination institution; S2: standardizing the collected data to form initial sample data; S3: based on the initial sample data, respectively calculating the benchmark time of the physical examination items and the time required for the distance between the physical examination items; S4: based on the initial sample data, dividing the set physical examination time period into time nodes according to the set intervals, and constructing a time series data set; and training the machine learning model based on the time series data set to obtain a trained prediction model, and using the data of the previous day to predict the number of people arriving for examination at each time node on the day, and predicting the number of people undergoing physical examinations at the physical examination institution; S5: based on the calculation results and the prediction results, constructing a physical examination diversion model, predicting the time nodes when the physical examination person completes each physical examination item, and generating the optimal physical examination path; S6: updating the time required for subsequent physical examination items based on real-time data, and optimizing the physical examination path.
[0052] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage media include various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.
[0053] An embodiment of the present invention provides a non-transitory computer-readable storage medium, which stores computer instructions, and the computer instructions enable the computer to execute the method provided by the above method embodiment, for example, including: S1: desensitizing the collection of physical examination item data, physical examination result data, appointment data, and basic registration information and historical physical examination data of the physical examination institution; S2: standardizing the collected data to form initial sample data; S3: based on the initial sample data, respectively calculating the benchmark time of the physical examination items and the time required for the distance between the physical examination items; S4: based on the initial sample data, dividing the set physical examination time period into time nodes according to the set intervals, and constructing a time series data set; and training the machine learning model based on the time series data set to obtain a trained prediction model, and using the data of the previous day to predict the number of people arriving for examination at each time node on the day, and predicting the number of people undergoing physical examinations at the physical examination institution; S5: based on the calculation results and the prediction results, constructing a physical examination diversion model, predicting the time nodes when the physical examination person completes each physical examination item, and generating the optimal physical examination path; S6: updating the time required for subsequent physical examination items based on real-time data, and optimizing the physical examination path.
[0054] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A physical examination intelligent triage method, characterized in that: include: Desensitized collection of medical examination item data, medical examination result data, appointment data, and basic registration information and historical medical examination data of medical examination institutions; Standardize the collected data to form initial sample data; Based on the initial sample data, the benchmark time for physical examination items and the time required to travel between physical examination items are calculated respectively; Based on the initial sample data, the set physical examination period is divided into time nodes according to the set intervals to construct a time series data set. The machine learning model is trained based on the time series data set to obtain a prediction model. Based on the prediction model, the number of physical examinations at the physical examination institution is predicted using the data from the previous day. Based on the calculation and prediction results, a physical examination triage model is constructed to predict the time nodes for the examinees to complete each physical examination item and generate the optimal physical examination path; Update the time required for subsequent physical examination items based on real-time data and optimize the physical examination path.
2. The intelligent medical examination triage method according to claim 1, characterized in that: In the step of respectively calculating the reference time of the physical examination items and the time required for the distance between the physical examination items based on the initial sample data, the step of calculating the reference time of the physical examination items includes: Divide the physical examination data in the initial sample data into categorical data and numerical data; Encode the categorical data and map each category label to a dense vector space to obtain a category vector. The category vector is then concatenated with the numerical data to form new feature data. Input the new feature data into the multi-layer perceptron model to obtain the feature vector and generate a feature vector set; Perform cluster analysis on the eigenvectors and calculate the benchmark time of physical examination items.
3. The intelligent medical examination triage method according to claim 2, characterized in that: The steps of performing cluster analysis on the feature vectors and calculating the benchmark time of the physical examination items include: Clustering feature vectors; After clustering, the actual time taken by all examinees to complete the physical examination items in each cluster is obtained, and the average of the actual time in each cluster is calculated as the benchmark time for the physical examination items corresponding to the cluster; The cluster to which the feature vector of the new physical examinee belongs is determined, and the physical examination item reference time corresponding to the cluster is obtained as the physical examination item reference time of the new physical examinee.
4. The intelligent medical examination triage method according to claim 3, characterized in that: The steps to cluster feature vectors include: In the feature vector set of physical examination data, each feature vector is a data point, randomly selected K data points as the initial K The centroid of each cluster is used to divide the feature vector set into K clusters; For each data point, calculate its Euclidean distance to the K centroids respectively; assign the data point to the cluster with the nearest centroid based on the calculated result; For each cluster, calculate the mean of all data points in the cluster as the new centroid; The process of assigning data points to the cluster with the nearest centroid and updating the centroid is repeated until the change in the centroid is less than the preset threshold or the maximum number of iterations is reached, and clustering is completed; The new centroid of the jth cluster The calculation formula is: in, is the jth cluster, is the number of data points in the jth cluster, for Data points within.
5. The intelligent medical examination triage method according to claim 4, characterized in that: In the step of respectively calculating the benchmark time of the physical examination items and the time required for the distance between the physical examination items based on the initial sample data, the step of calculating the time required for the distance between the physical examination items includes: Obtain the start time of two adjacent physical examination items X and Y of the examinee in the initial sample data, and determine the time range from physical examination item X to physical examination item Y; Obtain the actual distance between each physical examination item based on the physical examination center's floor plan; Obtain basic registration information of the examinees from the initial sample data; Obtain the physical examinee's stride length and cadence based on the physical examinee's basic registration information; Calculate the time required to travel from one physical examination item to another based on the actual distance between the physical examination items within the physical examination institution and the stride length and frequency of each examinee; The time required between each physical examination item calculated for each physical examination subject is counted, and the average value is calculated to obtain the time required between the physical examination items.
6. The intelligent medical examination triage method according to claim 5, characterized in that: Based on the initial sample data, the set physical examination period is divided into time nodes according to set intervals to construct a time series data set; and a machine learning model is trained based on the time series data set to obtain a prediction model. The steps of predicting the number of physical examination recipients at the physical examination institution based on the prediction model using the data of the physical examination institution from the previous day include: According to the set physical examination time period, filter out the registration time data within the time period; Divide the set physical examination time period according to the set intervals to form multiple time nodes; Traverse the filtered data and count the number of registered physical examination candidates at each time point in the initial sample data, and use it as the number of people who have been examined at that time point; Combine each time node and its corresponding number of people who came for inspection into a time series dataset. Each sample in the time series dataset contains a time node and the number of people who came for inspection corresponding to that node, and the samples are arranged in chronological order. The GRU model is trained based on the time series dataset to obtain a trained prediction model; The registration time data of the physical examination participants on the previous day is collected and processed, and then input into the trained prediction model to obtain the predicted value of the number of people attending the examination at each time node on the day.
7. The intelligent medical examination triage method according to claim 6, characterized in that: Based on the calculation and prediction results, a physical examination triage model is constructed to predict the time nodes when the examinee will complete each physical examination item. The steps to generate the optimal physical examination path include: Each physical examination item is considered a node in the graph, and the connection between two physical examination items is considered an edge in the graph. The weight of the edge is the time required to travel between the physical examination items. A queuing model is introduced. According to the predicted number of people arriving for examination at different time nodes, the number of people in the queue and the expected waiting time for each physical examination item are dynamically adjusted to complete the construction of the physical examination diversion model. The queuing model is: ,in, Waiting time in queue, is the inspection rate calculated based on the predicted number of inspection personnel, It is the reciprocal of the benchmark time of the physical examination item; Through the constructed physical examination triage model, the time nodes for the examinees to complete each physical examination item are predicted, and the optimal physical examination path is generated; During the physical examination process, the actual number of people arriving for examination and the number of people queuing are obtained in real time. The predicted number of people queuing and the node processing time, i.e. the benchmark time of the physical examination items, are updated based on the real-time data, and the optimal path is recalculated.
8. A physical examination intelligent diversion device, characterized in that: It includes a data acquisition module, a preprocessing module, a first calculation module, a physical examination number prediction module, an optimal path generation module and a path update module; The data acquisition module is used to desensitize and collect the medical examination item data, medical examination result data, appointment data, and basic registration information and historical medical examination data of the medical examination institution; The preprocessing module is used to standardize the collected data to form initial sample data; The first calculation module is used to calculate the reference time of the physical examination items and the time required for the distance between the physical examination items based on the initial sample data; The module for predicting the number of physical examination participants is used to divide the set physical examination period into time nodes according to the set intervals based on the initial sample data, and construct a time series data set. The module also trains the machine learning model based on the time series data set to obtain a prediction model, and uses the data from the previous day of the physical examination institution to predict the number of physical examination participants based on the prediction model. The optimal path generation module is used to build a physical examination triage model based on the calculation results and prediction results, predict the time nodes when the examinee will complete each physical examination item, and generate the optimal physical examination path; The path update module is used to update the time required for subsequent physical examination items based on real-time data and optimize the physical examination path.
9. An electronic device, characterized in that: The electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; the memory stores computer program instructions that can be executed by the at least one processor, and the computer program instructions are executed by the at least one processor to enable the at least one processor to execute the intelligent physical examination triage method as described in any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium, characterized in that The non-transitory computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the intelligent physical examination triage method according to any one of claims 1 to 7.