A closed-loop management system for hospital information equipment

Abstract

The invention discloses a hospital information equipment closed-loop management system, which relates to the field of equipment closed-loop management. The technical scheme includes: collecting patient medical treatment data and equipment status data, extracting high-level value data categories related to detection time prediction through an ElasticNet regression model based on a lagged data set, and forming a correlation feature set; based on information data and the correlation feature set in a current time period, predicting the time conversion coefficient of a detection item in each detection equipment in the next time period; constructing a high-dimensional data set and a closed-loop fitting model, and training the closed-loop fitting model to obtain a trained closed-loop fitting model; importing the high-dimensional data set in the current time period into the trained closed-loop fitting model, predicting an adjustment plan for each patient in the next time period, and dynamically adjusting the patient's examination items and equipment according to the predicted adjustment plan, so as to realize efficient closed-loop management of hospital information equipment.

Description

Technical Field

[0001] The present invention relates to the field of equipment closed-loop management, and more specifically, to a hospital information equipment closed-loop management system. Background Art

[0002] As a place that provides continuous medical services, hospitals need to cope with the continuous influx of patients and their diverse medical testing needs. These needs usually rely on various special medical information equipment, such as X-ray machines, CT scanners, MRI machines and ultrasonic equipment. The closed-loop management of hospital information equipment refers to closely linking various information equipment in the hospital to form a complete and efficient management system. The purpose of closed-loop management is to optimize the operating efficiency of hospital information equipment and improve the management level, ensure the efficient use of medical resources and the smooth flow of patients, and achieve efficient, accurate and sustainable equipment management.

[0003] In the closed-loop management system of hospital information equipment, the hospital needs to handle continuous testing requests. Each patient may need to use different medical equipment for testing according to the doctor's diagnostic needs. The time when patients submit test appointment applications is also different, which leads to the complexity of testing time scheduling. In addition, each patient needs to undergo different tests, such as blood routine, urine analysis, electrocardiogram, imaging examination, etc., which increases the difficulty of scheduling. Therefore, there will be complexity in time management, equipment use conflicts, and patients have difficulty in obtaining adjustment information of test items in a timely manner, resulting in reduced equipment use efficiency and affecting the patient's medical treatment efficiency. In order to solve the above technical difficulties, an innovative design is made for the closed-loop management system of hospital information equipment. Summary of the invention

[0004] In view of the problems existing in the prior art, the purpose of the present invention is to provide a closed-loop management system for hospital information equipment to achieve efficient closed-loop management of hospital information equipment.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solution: a hospital information equipment closed-loop management system, the hospital information equipment closed-loop management system comprising:

[0006] Data preprocessing module: used to collect patient visit data and equipment status data, and extract high-level value data categories related to detection time prediction features to form a related feature set;

[0007] Time prediction module: used to predict the time conversion coefficient required for the detection items in the detection equipment in the next time period;

[0008] Scheduling optimization module: constructs a high-dimensional data set and a closed-loop fitting model, imports the high-dimensional data set in the current time period into the trained closed-loop fitting model, predicts the adjustment plan for each patient in the next time period, and dynamically adjusts the patient's examination items according to the predicted adjustment plan;

[0009] The modules are connected to each other via wired and / or wireless means.

[0010] Preferably, the patient consultation data includes basic information of the patient, emergency mark value and test items, and the equipment status data includes working status and failure rate of the equipment.

[0011] Preferably, the acquisition of the time conversion coefficient specifically includes:

[0012] Based on the information data in the current time period e, according to the prediction formula , predict the time conversion coefficient of the test items in each test device in the next time period e+1, where, , and represents the regression coefficient, Indicates the status of the device. represents the number of data categories in the associated feature set, Indicates The actual performance value of each data category, Indicates Performance values ​​for each data category standard;

[0013] Indicates the device In the test project The time conversion factor applied when the current time period is e is the time period before l. In the test project The detection time applied at the time of smoothing calculation is then obtained in, represents the smoothing factor in the current time period e, Indicates the detection items in the current time period e The detection time of the application, represents the smoothing factor in the first l time periods, Indicates the detection items in the time period in the first l time periods The detection time of the application, is the trend influencing parameter, is the seasonal influence parameter, is the trend correction term, is a seasonal adjustment item.

[0014] Preferably, extracting high-level valuable data categories related to detection time prediction to form a related feature set specifically includes:

[0015] Create a lagged data set: extract the historical information data of q days from the hospital database, take the information data of each day as the unit data set, extract the detection items of the equipment in each unit data set and the corresponding actual detection time and predicted detection time as the value data set, extract the data category in the information data associated with the value data set as the gain data set, perform numerical coding and standardization on the value data set and the gain data set to obtain q groups of lagged data sets;

[0016] Take the regression parameters corresponding to each data category as output, the value data set and the gain data set as input, and optimize the objective function ,in, For the The actual testing time for each test item, For the The detection time predicted by the detection item model is is the number of detection items in each set of lagged data, is the number of categories of information data in the gain data set, For the The regression parameters corresponding to the predictions of the data category model are is the L1 regularization coefficient, is the L2 regularization coefficient;

[0017] Adjusting the coefficients of ElasticNet using cross-validation and , find the best coefficient, use the best coefficient to build the ElasticNet regression model and train it, divide the q groups of lagged data sets into training sets and validation sets, input the lagged data sets in the training set into the ElasticNet regression model in turn, and use the validation set to calculate the constraint value , when the constraint value is the largest and does not change, the trained ElasticNet regression model is obtained, where Indicates The regression parameters actually corresponding to the data categories are express The average value of the actual regression parameters in the test items;

[0018] The actual detection items, detection time and related information data of the input data category of each device in the current time period e, as well as the detection time obtained by predicting the detection items in the previous time period e-1, are collected as the value data set in the current time period e, and input into the trained ElasticNet regression model to obtain the regression parameters corresponding to each data category, set the threshold r_w, identify and extract the data categories with regression parameters greater than the threshold r_w, and form the associated feature set with the extracted data categories.

[0019] Preferably, the step of performing numerical coding and standardization on the value data set and the gain data set comprises:

[0020] For data categories, one-hot encoding is used to convert them into binary strings; for numerical information data, the original numerical values ​​are used directly; for text information data, mapping conversion methods are used to convert them into numerical vector features;

[0021] The method of using a mapping conversion method to convert text information data into a numerical vector feature includes:

[0022] The information data corresponding to all texts are parsed through regular expressions, each text is parsed into a combination of o simplest words, a non-repeating string is defined for each different simplest word, and the simplest words and the corresponding strings are mapped to form pointing data;

[0023] Construct a blank dictionary, fill the pointing data into the blank dictionary, add a query index for each pointing data, and the query index is defined as the page number of the pointing data in the dictionary and the label of the page number;

[0024] For each text-type information data, identify the simplest prime words of the combination and find the pointing data, add and combine them according to the query index of each pointing data to form a string of digital expressions. The content of the string of digital expressions is the numerical vector feature corresponding to the text-type information data.

[0025] Preferably, the constructing of a high-dimensional data set specifically includes:

[0026] Collect the three-dimensional traffic planning map inside the hospital and extract the time conversion coefficient of each detection equipment in the current time period e ;

[0027] Calculate the characteristic deviation data of each patient based on the patient's historical medical data ,in, and represents the weight coefficient, which can be assigned through experimental data analysis or according to historical experience. G represents the number of visits of the patient. Indicates The total distance of the route traveled during each visit, Indicates the total time used when the patient's route moves. Indicates the standard unit value set. It indicates the poor mobility assessment value, which can be subjectively assessed by the doctor during the consultation. represents the standard movement evaluation value;

[0028] Calculate equipment performance value based on equipment status data , where, it is set that when the working status of the equipment is under maintenance, , when the device is in idle or in use, ;

[0029] The time conversion coefficients of various detection devices in the current time period e, the characteristic deviation data of each patient and the equipment performance value are combined to form a data set, and the patient's medical data is binary encoded as the index of the data set, thereby forming a high-dimensional data set in the current time period e.

[0030] Preferably, the adjustment plan includes the test items for each patient in the next period of time, the test time corresponding to the test items, the waiting time from the current moment to the start of the test item, and the guidance route to the next test item.

[0031] Preferably, the construction of the closed-loop fitting model specifically includes:

[0032] The closed-loop fitting model is based on the LSTM basic framework, and sets the input layer, LSTM layer, fully connected layer and output layer. The input of the input layer is set to a high-dimensional data set, the LSTM layer is set to two layers, each with 50 units, LSTM layer 1 is set to return a sequence, and LSTM layer 2 is set to not return a sequence;

[0033] The fully connected layer is set to two layers, each layer has 10 nodes and the activation function is ReLU. The fully connected layer 1 is used for regression tasks, and the fully connected layer 2 is used for classification tasks.

[0034] The output layer includes the regression task output layer, the classification task output layer 1 and the classification task output layer 2. The activation function is set to the softmax function. The regression task output layer is used to predict the detection time and waiting time, and one node is set; the classification task output layer 1 is used to predict the detection items, and the classification task output layer 2 is used to predict the guidance route;

[0035] Set the objective function and constraints;

[0036] The Adam optimization algorithm is used, the learning rate is set to 0.001, and the loss function is ,in, and is the uncertainty parameter for regression and classification tasks, and They represent the predicted detection time and waiting time respectively. and They represent the actual detection time and waiting time, respectively. is the sample index, represents the loss of the classification task, and ,in, represents the predicted probability of each patient's test item, represents the true label of each patient test item, represents the predicted probability of each patient's guidance route, Represents the true label of each patient’s guidance route.

[0037] Preferably, the training and use of the closed-loop fitting model includes:

[0038] S11, collect F groups of high-dimensional data sets and corresponding adjustment schemes to form a sample set, divide it into a training set and a validation set in a ratio of 8:2, input the training set into the constructed closed-loop fitting model in turn, and perform forward propagation in the closed-loop fitting model;

[0039] S12, obtaining an adjustment plan for the prediction output of the closed-loop fitting model, calculating the loss value using the loss function, and calculating and updating the parameters through back propagation and gradient descent algorithms;

[0040] S13, repeat S11 and S12 until all training sets are imported, input the verification set for verification, when the verification is successful, obtain the trained closed-loop fitting model, when the verification is unsuccessful, repeat S13;

[0041] S14. Input the high-dimensional data set of the patient in the current time period into the trained closed-loop fitting model, obtain the adjustment plan of the predicted output, and dynamically adjust the patient's test items.

[0042] Preferably, the setting of the objective function and constraint conditions includes:

[0043] The objective function is set to minimize the total waiting time and device idle time, and the formula is expressed as ,in, The waiting time of patient h before the next test item, Indicates the idle time of device b;

[0044] Set the priority of the constraint conditions to condition 1> condition 2, condition 1 is , used to select the maximum value from the emergency mark values ​​of all selectable patients in the time period, representing the emergency mark value of the ath patient, condition 2 is , represents the waiting time of the ath patient, represents the start time of the appointment for the ath patient, It indicates the end time of the appointment for the ath patient. The same device cannot be used repeatedly in the same time period.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] In the present invention, by constructing and training an ElasticNet regression model, the data categories related to the detection time prediction in the hospital database are explored, and by setting a threshold r_w, the data categories with regression parameters greater than the threshold r_w are identified and high-value extraction is performed, thereby further reducing the amount of data analysis during detection time prediction. The time conversion coefficient of the detection item in each detection device is predicted through the high-value data category, and the detection time is converted into data expression, thereby achieving high-quality input for subsequent data in-depth analysis. By constructing a high-dimensional data set, the data related to the adjustment plan is highly concentrated, thereby improving the data acquisition and analysis of the closed-loop fitting model. The closed-loop fitting model implemented by this platform adds objective functions and constraints based on the LSTM basic framework, limits the range of the selection of detection items and patients, realizes dynamic adjustment and optimization design of equipment and patients, and then outputs a highly efficient solution design; the adaptive regulation and optimization mechanism continuously improves the overall efficiency of intelligent analysis, and multiple modules cooperate with each other to achieve comprehensive data collection, processing and analysis, and provide real-time and accurate analysis of data storage. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 The present invention proposes a structural schematic diagram of a closed-loop management system for hospital information equipment;

[0048] Figure 2 It is a schematic diagram of the method steps applied to the closed-loop management system of hospital information equipment in the present invention. DETAILED DESCRIPTION

[0049] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0050] Embodiment 1

[0051] Reference Figure 1 and Figure 2Embodiment 1 further illustrates a closed-loop management system for hospital information equipment proposed by the present invention.

[0052] As a place that provides continuous medical services, hospitals need to cope with the continuous influx of patients and their diverse medical testing needs. These needs usually rely on various special medical information equipment, such as X-ray machines, CT scanners, MRI machines and ultrasonic equipment. The closed-loop management of hospital information equipment refers to closely linking various information equipment in the hospital to form a complete and efficient management system. The purpose of closed-loop management is to optimize the operating efficiency of hospital information equipment and improve the management level, ensure the efficient use of medical resources and the smooth flow of patients, and achieve efficient, accurate and sustainable equipment management.

[0053] In the closed-loop management system of hospital information equipment, the hospital needs to handle continuous testing requests. Each patient may need to use different medical equipment for testing according to the doctor's diagnostic needs. The time when patients submit test appointment applications is also different, which leads to the complexity of testing time arrangement. In addition, each patient needs to undergo different tests, such as blood routine, urine analysis, electrocardiogram, imaging examination, etc., which increases the difficulty of scheduling. Therefore, there will be the following problems and difficulties:

[0054] A. Complexity of time management: Due to differences in factors such as the complexity of the disease, the speed of the doctor's diagnosis, and the patient's cooperation, the time it takes for each patient to be tested on the equipment is different and difficult to predict accurately, which will pose a challenge to the time arrangement of subsequent appointment applications.

[0055] B. Equipment usage conflict: It may happen that the test appointment time of a patient has arrived, but the test of the previous patient has not been completed, resulting in a long queue. Especially during the peak period of medical consultation, the backlog of personnel caused by unreasonable time arrangement not only affects the patient's medical experience, but also increases the maintenance pressure of medical staff and reduces the efficiency of hospital diagnosis operation.

[0056] C. Optimization of multi-item testing: While waiting for a certain test item, the patient may have the opportunity to undergo other tests. However, due to the uneven distribution of distances between the various testing departments, it is difficult for the patient to obtain information in a timely manner and it is difficult to grasp the timing. Therefore, it is necessary to provide timely reminders and notifications on the dynamic adjustments of the patient's various test items when the time is right.

[0057] D. Equipment status and operator proficiency: The status of different equipment and the operator proficiency of medical personnel affect the detection time. Different detection items use different detection equipment, resulting in different usage times. Secondly, even for the same type of detection equipment, the usage time varies due to the age, user proficiency, number and distribution of technical personnel. Therefore, how to grasp the time gaps caused by these differences and realize the closed-loop management of hospital information equipment when using them reasonably is an urgent need.

[0058] E. Priority of urgent patients: The conditions of patients vary, which may result in urgent patients having to jump the queue for testing, affecting the testing arrangements of other patients.

[0059] In order to solve the above technical difficulties, a reasonable design of the closed-loop management system of hospital information equipment is made as follows, including:

[0060] Data preprocessing module: used to collect patient visit data and equipment status data, and extract high-level value data categories related to detection time prediction features to form a related feature set;

[0061] Patient visit data includes basic patient information (such as name, age, gender, medical record number), emergency mark values ​​(such as normal, emergency, critical), and test items (such as blood routine, urine analysis, electrocardiogram, imaging examination). Equipment status data includes the working status of the equipment (such as idle, in use, under maintenance) and failure rate.

[0062] Patient appointment data can be collected through the hospital information management system (HIS), and equipment status data can be collected through the equipment management system (EMS) or real-time monitoring equipment (such as sensors and cameras).

[0063] Time prediction module: used to predict the time conversion coefficient required for the detection items in the detection equipment in the next time period;

[0064] Based on the information data in the current time period e, according to the prediction formula , predict the time conversion coefficient of the test items in each test device in the next time period e+1, where, , and represents the regression coefficient, Indicates the status of the device (such as idle, in use, under maintenance), represents the number of data categories in the associated feature set, Indicates The actual performance value of each data category, Indicates Performance values ​​for each data category standard;

[0065] Indicates the device In the test project The time conversion factor applied when the current time period is e is the time period before l. In the test project The detection time applied at the time of smoothing calculation is then obtained ,in, represents the smoothing factor in the current time period e, Indicates the detection items in the current time period e The detection time of the application, represents the smoothing factor in the first l time periods, Indicates the detection items in the time period in the first l time periods The detection time of the application, is the trend influencing parameter, is the seasonal influence parameter, is the trend correction term, is a seasonal adjustment item.

[0066] Trend influence parameters can be obtained by managers through linear fitting based on historical data, seasonal influence parameters can also be obtained by managers based on seasonal change rate statistics, and the values ​​of trend correction items and seasonal adjustment items can be set by managers or obtained through analysis of experimental data.

[0067] Extract high-level valuable data categories related to detection time prediction to form a set of associated features, including:

[0068] Create a lagged data set: extract the historical information data of q days from the hospital database, take the information data of each day as the unit data set, extract the detection items of the equipment in each unit data set and the corresponding actual detection time and predicted detection time as the value data set, extract the data category in the information data associated with the value data set as the gain data set, perform numerical coding and standardization on the value data set and the gain data set to obtain q groups of lagged data sets;

[0069] Take the regression parameters corresponding to each data category as output, the value data set and the gain data set as input, and optimize the objective function ,in, For the The actual testing time for each test item, For the The detection time predicted by the detection item model is is the number of detection items in each set of lagged data, is the number of categories of information data in the gain data set, For the The regression parameters corresponding to the predictions of the data category model are is the L1 regularization coefficient, is the L2 regularization coefficient;

[0070] Adjusting the coefficients of ElasticNet using cross-validation and , find the best coefficient, use the best coefficient to build the ElasticNet regression model and train it, divide the q groups of lagged data sets into training sets and validation sets, input the lagged data sets in the training set into the ElasticNet regression model in turn, and use the validation set to calculate the constraint value , when the constraint value is the largest and does not change, the trained ElasticNet regression model is obtained, where Indicates The regression parameters actually corresponding to the data categories are express The average value of the actual regression parameters in the test items;

[0071] The actual detection items, detection time and related information data of the input data category of each device in the current time period e, as well as the detection time obtained by predicting the detection items in the previous time period e-1, are collected as the value data set in the current time period e, and input into the trained ElasticNet regression model to obtain the regression parameters corresponding to each data category, set the threshold r_w, identify and extract the data categories with regression parameters greater than the threshold r_w, and form the associated feature set with the extracted data categories.

[0072] For example, suppose we extract information data from the hospital database for the past q = 5 days, use the data for each day as a unit data set, train the ElasticNet regression model, and use the cross-validation method to find the best coefficients and , get the trained ElasticNet regression model, assuming we have the following data:

[0073] The gain data categories are data category 1, data category 2, and data category 3, and the numerical codes are 0, 1, and 2. The numerical codes of the test items are 000 and 001. Then the test item data of two devices in the current time period are:

[0074] Device 1: detection item

[000] , actual detection time

[10] , predicted detection time [9], gain data category [0,2];

[0075] Device 2: detection item

[001] , actual detection time

[12] , predicted detection time

[11] , gain data category [0, 1, 2];

[0076] Assume that the standardized data is:

[0077] Device 1: detection item

[000] , actual detection time [0.5], predicted detection time [0.4], gain data category [0,2];

[0078] Device 2: detection item

[001] , actual detection time [0.7], predicted detection time [0.6], gain data category [0, 1, 2];

[0079] The trained ElasticNet regression model gives the following regression parameters:

[0080] Device 1: Regression parameters for class 0 ; Regression parameters for category 1 ; Regression parameters for category 2 ;

[0081] Device 2: Regression parameters for class 0 ; Regression parameters for category 1 ; Regression parameters for category 2 ;

[0082] Set the threshold r_w=0.5, identify and extract the data categories whose regression parameters are greater than the threshold, and then obtain: the data category [0,1] gained by device 1 in the detection item

[000] ; the data category [2] gained by device 2 in the detection item

[001] . Therefore, extract the data of these categories to form an associated feature set.

[0083] Through the above steps and specific values, important data categories can be identified and optimization of deep data processing can be achieved.

[0084] The hospital database is the hospital's information center, including but not limited to various equipment information, personnel information and patient information in the hospital. Equipment information may include the following data categories: equipment type, equipment model, equipment interface type, equipment location information, equipment service life, equipment inspection and maintenance information, equipment operator information, equipment inspection items, the time used for equipment inspection of each item, etc.; personnel information may include the following data categories: basic information of medical staff (such as name, age, gender, department and specialty), consultation hours, patients, prescription information and medical staff's operational proficiency, etc.; patient information may include the following data categories: basic information of patients (such as name, age, gender, medical record number), appointment time, consultation time, appointment department, test items, etc., as well as data included in other information centers.

[0085] The numerical encoding and standardization of the value dataset and the gain dataset include:

[0086] For data categories, one-hot encoding is used to convert them into binary strings; for numerical information data, the original numerical values ​​are used directly; for text information data, mapping conversion methods are used to convert them into numerical vector features;

[0087] The method of using a mapping conversion method to convert text information data into a numerical vector feature includes:

[0088] The information data corresponding to all texts are parsed through regular expressions, each text is parsed into a combination of o simplest words, a non-repeating string is defined for each different simplest word, and the simplest words and the corresponding strings are mapped to form pointing data;

[0089] Construct a blank dictionary, fill the pointing data into the blank dictionary, add a query index for each pointing data, and the query index is defined as the page number of the pointing data in the dictionary and the label of the page number;

[0090] For each text-type information data, identify the simplest prime words of the combination and find the pointing data, add and combine them according to the query index of each pointing data to form a string of digital expressions. The content of the string of digital expressions is the numerical vector feature corresponding to the text-type information data.

[0091] For example, assuming that a text information data is "a nodule is transmitted at the lower right part of the left lobe of the lung", the form of parsing the nodule at the lower right part of the left lobe of the lung into o simplest word combinations is "lung + left lobe + lower right part + conduct + nodule + transmittance";

[0092] Among them, the simplest words in the present invention refer to the simplest form of words in medical terms, that is, the minimum words that cannot be split or fixedly expressed. This simplest definition can be set by doctors or managers as long as it conforms to the meaning expressed in the present invention.

[0093] Assuming that a non-repeating string is defined for each of the different simplest words, the string definitions for "lung + left lobe + lower right + conduction + nodule + transmission" are:

[0094] Lung: 123; Left lobe: 234; Lower right: 345; Progression: 12; Nodule: 789; Transmission: 678; The resulting directional data is:

[0095] Lung→123; left lobe→234; right lower→345; progression→12; nodule→789; radiography→678;

[0096] Assume that the query index for each data point is: (lung→123:11,2); (left lobe→234:1,2); (lower right→345:13,7); (transverse→12:4,22); (nodule→789:15,2); (transmission→678:19,2);

[0097] The obtained string of numbers is expressed as 11212137422152192, that is, the numerical vector feature of "nodule transmission in the lower right part of the left lobe of the lung" is "11212137422152192".

[0098] Scheduling optimization module: construct a high-dimensional data set and a closed-loop fitting model, import the high-dimensional data set into the trained closed-loop fitting model, predict the adjustment plan for each patient in the next time period, and dynamically adjust the patient's examination items according to the predicted adjustment plan.

[0099] Construct a high-dimensional dataset, including:

[0100] Collect the three-dimensional traffic planning map inside the hospital and extract the time conversion coefficient of each detection equipment in the current time period e ;

[0101] Calculate the characteristic deviation data of each patient based on the patient's historical medical data ,in, and represents the weight coefficient, which can be assigned through experimental data analysis or according to historical experience. G represents the number of visits of the patient. Indicates The total distance of the route traveled during each visit, Indicates the total time used when the patient's route moves. Indicates the standard unit value set. It indicates the poor mobility assessment value, which can be subjectively assessed by the doctor during the consultation. represents the standard movement evaluation value;

[0102] Calculate equipment performance value based on equipment status data , where, it is set that when the working status of the equipment is under maintenance, , when the device is in idle or in use, ;

[0103] The time conversion coefficients of various detection devices in the current time period e, the characteristic deviation data of each patient and the equipment performance value are combined to form a data set, and the patient's medical data is binary encoded as the index of the data set, thereby forming a high-dimensional data set in the current time period e.

[0104] Historical medical data refers to the information data left by patients during their previous visits to the hospital, including but not limited to the patient's visit time, stay time, walking route, stay location and all other information data that can be saved in the hospital. This information data can be collected and obtained through the hospital's information center and monitoring equipment.

[0105] The three-dimensional traffic planning map inside the hospital refers to the three-dimensional display of each building in the hospital, which can be obtained through the hospital's architectural design drawings or data information database. The three-dimensional traffic planning map inside the hospital includes the three-dimensional coordinates of all buildings inside the hospital and the passable roads. The three-dimensional coordinates are mainly used to represent the specific surface of each building (room or department, chair, plant, etc., so the items displayed inside the hospital). For example, the three-dimensional coordinates of the Radiology Department 002 can be set to (x1, x2, x3), x1 can be used to represent the horizontal unit distance on the coordinate axis with the main gate of the hospital as the origin, x2 can be used to represent the vertical unit distance, and x3 can be used to represent the height unit distance. The three-dimensional traffic planning map inside the hospital can also be refined and developed based on Beidou or Gaode map to realize route navigation inside the hospital.

[0106] The adjustment plan includes the test items for each patient in the next period of time, the test time corresponding to the test items, the waiting time from the current moment to the start of the test item, and the guidance route to the next test item.

[0107] The construction of the closed-loop fitting model specifically includes:

[0108] The closed-loop fitting model is based on the LSTM basic framework, and sets the input layer, LSTM layer, fully connected layer and output layer. The input of the input layer is set to a high-dimensional data set, the LSTM layer is set to two layers, each with 50 units, LSTM layer 1 is set to return a sequence, and LSTM layer 2 is set to not return a sequence;

[0109] The fully connected layer is set to two layers, each layer has 10 nodes and the activation function is ReLU. The fully connected layer 1 is used for regression tasks, and the fully connected layer 2 is used for classification tasks.

[0110] The output layer includes the regression task output layer, the classification task output layer 1 and the classification task output layer 2. The activation function is set to the softmax function. The regression task output layer is used to predict the detection time and waiting time, and one node is set; the classification task output layer 1 is used to predict the detection items, and the classification task output layer 2 is used to predict the guidance route;

[0111] Set the objective function and constraints;

[0112] The Adam optimization algorithm is used, the learning rate is set to 0.001, and the loss function is ,in, and is the uncertainty parameter for regression and classification tasks, and They represent the predicted detection time and waiting time respectively. and They represent the actual detection time and waiting time, respectively. is the sample index, represents the loss of the classification task, and ,in, represents the predicted probability of each patient test item, represents the true label of each patient test item, represents the predicted probability of each patient's guidance route, represents the true label of each patient's guidance route;

[0113] The training and use of the closed-loop fitting model includes:

[0114] S11, collect F groups of high-dimensional data sets and corresponding adjustment schemes to form a sample set, divide it into a training set and a validation set in a ratio of 8:2, input the training set into the constructed closed-loop fitting model in turn, and perform forward propagation in the closed-loop fitting model;

[0115] S12, obtaining an adjustment plan for the prediction output of the closed-loop fitting model, calculating the loss value using the loss function, and calculating and updating the parameters through back propagation and gradient descent algorithms;

[0116] S13, repeat S11 and S12 until all training sets are imported, input the verification set for verification, when the verification is successful, obtain the trained closed-loop fitting model, when the verification is unsuccessful, repeat S13;

[0117] S14. Input the high-dimensional data set of the patient in the current time period into the trained closed-loop fitting model, obtain the adjustment plan of the predicted output, and dynamically adjust the patient's test items.

[0118] The setting of objective function and constraints includes:

[0119] The objective function is set to minimize the total waiting time and device idle time, and the formula is expressed as ,in, The waiting time of patient h before the next test item, Indicates the idle time of device b;

[0120] Set the priority of the constraint conditions to condition 1> condition 2, condition 1 is , used to select the maximum value from the emergency mark values ​​of all selectable patients in the time period, representing the emergency mark value of the ath patient, condition 2 is , represents the waiting time of the ath patient, represents the start time of the appointment for the ath patient, It indicates the end time of the appointment for the ath patient. The same device cannot be used repeatedly in the same time period.

[0121] Assume that we have two devices E1E_1E1 and E2E_2E2, and four patients P1P_1P1, P2P_2P2, P3P_3P3, P4P_4P4. The acquired patient data and device data are as follows:

[0122] P1P_1P1: urgency mark value 5, appointment time [9:00,9:30];

[0123] P2P_2P2: urgency mark value 3, appointment time [9:15,9:45];

[0124] P3P_3P3: urgency mark value 7, appointment time [9:30,10:00];

[0125] P4P_4P4: emergency mark value 4, appointment time [9:45,10:15];

[0126] E1E_1E1: currently idle;

[0127] E2E_2E2: currently idle;

[0128] The scheduling is as follows:

[0129] First time period [9:00,9:30]:

[0130] Optional patient: P1P_1P1 (urgent tag value 5);

[0131] Select P1P_1P1 for testing because this is the only available patient;

[0132] Device E1E_1E1 is assigned to P1P_1P1.

[0133] Second time period [9:15,9:45]:

[0134] Optional Patient: P2P_2P2 (urgent tag value 3);

[0135] Select P2P_2P2 for testing because this is the only available patient.

[0136] Device E2E_2E2 is assigned to P2P_2P2.

[0137] The third time period [9:30,10:00]:

[0138] Optional patient: P3P_3P3 (urgent tag value 7);

[0139] Select P3P_3P3 for testing because this is the only available patient.

[0140] Device E1E_1E1 is assigned to P3P_3P3 (device E1E_1E1 was idle at 9:30).

[0141] Fourth time period [9:45,10:15]:

[0142] Optional patient: P4P_4P4 (urgent tag value 4);

[0143] Select P4P_4P4 for testing because this is the only available patient.

[0144] Device E2E_2E2 is assigned to P4P_4P4 (device E2E_2E2 was idle at 9:45).

[0145] Objective function: Minimization of total waiting time and device idle time.

[0146] Condition 1: Patients with the highest urgency mark value are given priority.

[0147] Condition 2: The patient waiting time is calculated as the difference between the appointment start time and the current time.

[0148] The optimization results are as follows:

[0149] Patient P1P_1P1 started at 9:00 and had a waiting time of 0 minutes.

[0150] Patient P2P_2P2 started at 9:15 with a waiting time of 0 minutes.

[0151] Patient P3P_3P3 started at 9:30 with a waiting time of 0 minutes.

[0152] Patient P4P_4P4 started at 9:45 with a waiting time of 0 minutes.

[0153] By setting the objective function and constraints, we can effectively schedule equipment, minimize patient waiting time and equipment idle time, and ensure that emergency situations are treated first. This optimal scheduling solution can improve hospital equipment utilization and patient circulation efficiency.

[0154] In this embodiment, the present invention constructs an ElasticNet regression model and performs training to explore the data categories related to the detection time prediction in the hospital database, and identifies the data categories with regression parameters greater than the threshold r_w by setting a threshold r_w and performs high-value extraction, thereby further reducing the amount of data analysis during detection time prediction, predicting the time conversion coefficient of the detection item in each detection device through the high-value data category, converting the detection time into data expression, and realizing high-quality input for subsequent data in-depth analysis. By constructing a high-dimensional data set, the data related to the adjustment plan is highly concentrated, and the data acquisition and analysis of the closed-loop fitting model are improved. The closed-loop fitting model implemented by this platform adds objective functions and constraints based on the LSTM basic framework, limits the range of the selection of detection items and patients, realizes dynamic adjustment and optimization design of equipment and patients, and then outputs a highly efficient solution design; the adaptive regulation and optimization mechanism continuously improves the overall efficiency of intelligent analysis, and multiple modules collaborate with each other to realize comprehensive data collection, processing and analysis, and provide real-time and accurate analysis of data storage.

[0155] Embodiment 2

[0156] Reference Figure 1 and Figure 2 , Example 2 further illustrates a closed-loop management system for hospital information equipment proposed by the present invention.

[0157] The connector intelligent monitoring and control method is applied to a hospital information equipment closed-loop management system, and includes the following steps:

[0158] Step S1: Collect patient visit data and equipment status data, and extract high-level value data categories related to detection time prediction through the ElasticNet regression model based on the lagged data set to form a related feature set;

[0159] Step S2: Based on the information data and the associated feature set in the current time period, predict the time conversion coefficient of the detection item in each detection device in the next time period;

[0160] Step S3: constructing a high-dimensional data set and a closed-loop fitting model, and training the closed-loop fitting model to obtain a trained closed-loop fitting model;

[0161] Step S4: The high-dimensional data set in the current time period is imported into the trained closed-loop fitting model to predict the adjustment plan for each patient in the next time period, and the patient's examination items and equipment are dynamically adjusted according to the predicted adjustment plan.

[0162] The hospital information equipment closed-loop management system comprises:

[0163] Data preprocessing module: used to collect patient visit data and equipment status data, and extract high-level value data categories related to detection time prediction features to form a related feature set;

[0164] Time prediction module: used to predict the time conversion coefficient required for the detection items in the detection equipment in the next time period;

[0165] Scheduling optimization module: constructs a high-dimensional data set and a closed-loop fitting model, imports the high-dimensional data set in the current time period into the trained closed-loop fitting model, predicts the adjustment plan for each patient in the next time period, and dynamically adjusts the patient's examination items according to the predicted adjustment plan;

[0166] The modules are connected to each other via wired and / or wireless means.

[0167] In addition, according to the implementation of the present application, the process described in the attached figure of a hospital information equipment closed-loop management system can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, which stores machine-readable instructions, and the machine-readable instructions can be executed by a processor to execute instructions corresponding to the method steps provided in the present application. Of course, the architecture shown in the attached figure of a hospital information equipment closed-loop management system is only exemplary. When implementing different devices, adaptive selection or adjustment can be performed according to actual needs.

[0168] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0169] The above is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A closed-loop management system for hospital information equipment, characterized in that: The hospital information equipment closed-loop management system includes: Data preprocessing module: used to collect patient visit data and equipment status data, and extract high-level value data categories related to detection time prediction features to form a related feature set; The extracting of high-level value data categories related to detection time prediction to form a related feature set specifically includes: Create a lagged data set: extract the historical information data of q days from the hospital database, take the information data of each day as the unit data set, extract the detection items of the equipment in each unit data set and the corresponding actual detection time and predicted detection time as the value data set, extract the data category in the information data associated with the value data set as the gain data set, perform numerical coding and standardization on the value data set and the gain data set to obtain q groups of lagged data sets; Take the regression parameters corresponding to each data category as output, the value data set and the gain data set as input, and optimize the objective function ,in, For the The actual testing time for each test item, For the The detection time predicted by the detection item model is is the number of detection items in each set of lagged data, is the number of categories of information data in the gain data set, For the The regression parameters corresponding to the predictions of the data category model are for Regularization coefficient, is the L2 regularization coefficient; Adjusting the coefficients of ElasticNet using cross-validation and , find the best coefficient, use the best coefficient to build the ElasticNet regression model and train it, divide the q groups of lagged data sets into training sets and validation sets, input the lagged data sets in the training set into the ElasticNet regression model in turn, and use the validation set to calculate the constraint value , when the constraint value is the largest and does not change, the trained ElasticNet regression model is obtained, where Indicates The regression parameters actually corresponding to the data categories are express The average value of the actual regression parameters in the test items; Collect the associated information data of the actual detection items, detection time and input data categories of each device in the current time period e, as well as the detection time predicted for the detection items in the previous time period e-1, record them as the value data set in the current time period e, and input them into the trained ElasticNet regression model to obtain the regression parameters corresponding to each data category, set the threshold r_w, identify and extract the data categories with regression parameters greater than the threshold r_w, and form the associated feature set with the extracted data categories; Time prediction module: used to predict the time conversion coefficient required for the detection items in the detection equipment in the next time period; The acquisition of the time conversion coefficient specifically includes: Based on the information data in the current time period e, according to the prediction formula , predict the time conversion coefficient of the test items in each test device in the next time period e+1, where, , and represents the regression coefficient, Indicates the status of the device (such as idle, in use, under maintenance), represents the number of data categories in the associated feature set, Indicates The actual performance value of each data category, Indicates Performance values ​​for each data category standard; Indicates the device In the test project The time conversion factor applied when the current time period is e is the time period before l. In the test project The detection time applied at the time of smoothing calculation is then obtained ,in, represents the smoothing factor in the current time period e, Indicates the detection items in the current time period e The detection time of the application, represents the smoothing factor in the first l time periods, Indicates the detection items in the time period in the first l time periods The detection time of the application, is the trend influencing parameter, is the seasonal influence parameter, is the trend correction term, is the seasonal adjustment item; Scheduling optimization module: constructs a high-dimensional data set and a closed-loop fitting model, imports the high-dimensional data set in the current time period into the trained closed-loop fitting model, predicts the adjustment plan for each patient in the next time period, and dynamically adjusts the patient's examination items according to the predicted adjustment plan; The constructing of the high-dimensional data set specifically includes: Collect the three-dimensional traffic planning map inside the hospital and extract the time conversion coefficient of each detection equipment in the current time period e ; Calculate the characteristic deviation data of each patient based on the patient's historical medical data ,in, and represents the weight coefficient, which can be assigned through experimental data analysis or according to historical experience. G represents the number of visits of the patient. Indicates The total distance of the route traveled during each visit, Indicates the total time used when the patient's route moves. Indicates The total distance of the route traveled during each visit, Indicates the total time used when the patient's route moves. Indicates the standard unit value set. It indicates the poor mobility assessment value, which can be subjectively assessed by the doctor during the consultation. represents the standard movement evaluation value; Calculate equipment performance value based on equipment status data , where, it is set that when the working status of the equipment is under maintenance, , when the device is in idle or in use, ; The time conversion coefficients of each detection device in the current time period e, the characteristic deviation data of each patient, and the device performance value are combined to form a data set, and the patient's medical data is binary-encoded as the index of the data set, thereby forming a high-dimensional data set in the current time period e; The construction of the closed-loop fitting model specifically includes: The closed-loop fitting model is based on the LSTM basic framework, and sets the input layer, LSTM layer, fully connected layer and output layer. The input of the input layer is set to a high-dimensional data set, the LSTM layer is set to two layers, each with 50 units, LSTM layer 1 is set to return a sequence, and LSTM layer 2 is set to not return a sequence; The fully connected layer is set to two layers, each layer has 10 nodes and the activation function is ReLU. The fully connected layer 1 is used for regression tasks, and the fully connected layer 2 is used for classification tasks. The output layer includes the regression task output layer, the classification task output layer 1 and the classification task output layer 2. The activation function is set to the softmax function. The regression task output layer is used to predict the detection time and waiting time, and one node is set; the classification task output layer 1 is used to predict the detection items, and the classification task output layer 2 is used to predict the guidance route; Set the objective function and constraints; The Adam optimization algorithm is used, the learning rate is set to 0.001, and the loss function is ,in, and is the uncertainty parameter for regression and classification tasks, and They represent the predicted detection time and waiting time respectively. and They represent the actual detection time and waiting time, respectively. is the sample index, represents the loss of the classification task, and ,in, represents the predicted probability of each patient test item, represents the true label of each patient test item, represents the predicted probability of each patient's guidance route, represents the true label of each patient's guidance route; The modules are connected to each other via wired and / or wireless means.

2. A hospital information equipment closed-loop management system according to claim 1, characterized in that: The patient visit data includes the patient's basic information, emergency mark value and test items, and the equipment status data includes the working status and failure rate of the equipment.

3. A closed-loop management system for hospital information equipment according to claim 2, characterized in that: The numerical encoding and standardization of the value data set and the gain data set includes: For data categories, one-hot encoding is used to convert them into binary strings; for numerical information data, the original numerical values ​​are used directly; for text information data, mapping conversion methods are used to convert them into numerical vector features; The method of using a mapping conversion method to convert text information data into a numerical vector feature includes: The information data corresponding to all texts are parsed through regular expressions, each text is parsed into a combination of o simplest words, a non-repeating string is defined for each different simplest word, and the simplest words and the corresponding strings are mapped to form pointing data; Construct a blank dictionary, fill the pointing data into the blank dictionary, add a query index for each pointing data, and the query index is defined as the page number of the pointing data in the dictionary and the label of the page number; For each text-type information data, identify the simplest prime words of the combination and find the pointing data, add and combine them according to the query index of each pointing data to form a string of digital expressions, and then the digital string is the numerical vector feature corresponding to the text-type information data.

4. A hospital information equipment closed-loop management system according to claim 3, characterized in that: The adjustment plan includes the test items for each patient in the next period of time, the test time corresponding to the test items, the waiting time from the current moment to the start of the test items, and the guidance route to the next test item.

5. A hospital information equipment closed-loop management system according to claim 4, characterized in that: The training and use of the closed-loop fitting model includes: S11, collect F groups of high-dimensional data sets and corresponding adjustment schemes to form a sample set, divide it into a training set and a validation set in a ratio of 8:2, input the training set into the constructed closed-loop fitting model in sequence, and perform forward propagation in the closed-loop fitting model; S12, obtaining an adjustment plan for the prediction output of the closed-loop fitting model, calculating the loss value using the loss function, and calculating and updating the parameters through back propagation and gradient descent algorithms; S13, repeat S11 and S12 until all training sets are imported, input the verification set for verification, when the verification is successful, obtain the trained closed-loop fitting model, when the verification is unsuccessful, repeat S13; S14. Input the high-dimensional data set of the patient in the current time period into the trained closed-loop fitting model, obtain the adjustment plan of the predicted output, and dynamically adjust the patient's test items.

6. A hospital information equipment closed-loop management system according to claim 5, characterized in that: The setting of the objective function and constraint conditions includes: The objective function is set to minimize the total waiting time and device idle time, and the formula is expressed as ,in, The waiting time of patient h before the next test item, Indicates the idle time of device b; Set the priority of the constraint conditions to condition 1> condition 2, condition 1 is , used to select the maximum value from the emergency mark values ​​of all selectable patients in the time period, representing the emergency mark value of the ath patient, condition 2 is , represents the waiting time of the ath patient, represents the start time of the appointment for the ath patient, It indicates the end time of the appointment for the ath patient. The same device cannot be used repeatedly in the same time period.

Images