Gastrointestinal patient postoperative recovery management system based on big data

Abstract

The invention discloses a gastrointestinal patient postoperative recovery management system based on big data. The system comprises a data acquisition module, a postoperative recovery guidance model establishment module, a patient state monitoring model establishment module, a monitoring optimization module and a comprehensive management module. The invention relates to the technical field of postoperative recovery management, in particular to a gastrointestinal patient postoperative recovery management system based on big data, according to the scheme, by designing a local density function, a clustering center updating formula, a data point distribution rule and a target function, the convergence of a model and the stability of a clustering result are improved; by designing an energy function and a self-adaptive weight adjustment mechanism, the activation capability of a hidden unit is enhanced, and the adaptability of the model and the monitoring accuracy are improved; through design of adaptive parameters, coevolution search and elite reservation, initial parameter points are generated in combination with randomness and target guidance, and stability, adaptability, convergence speed and result quality of parameter search are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of postoperative recovery management, and in particular to a postoperative recovery management system for gastrointestinal patients based on big data. Background Art

[0002] The gastrointestinal patient postoperative recovery management system based on big data is an innovative medical technology. It uses big data technology and artificial intelligence algorithms to build a postoperative recovery guidance model and a patient status monitoring model, realizes intelligent management of the postoperative recovery process of gastrointestinal surgery patients, and monitors the patient's physiological indicators and recovery progress in real time. The system can detect abnormal situations in time and provide early warnings.

[0003] The traditional clustering method has problems in the postoperative recovery guidance model, such as uneven distribution of feature weights, easy neglect of local data, and inaccurate updating of cluster centers; the traditional patient status monitoring model has problems such as single model architecture, weak activation ability of hidden units, insufficient feature extraction ability and low monitoring accuracy; the traditional parameter search method has problems such as low search efficiency, poor stability and low quality of search results in the monitoring optimization module. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a gastrointestinal patient postoperative recovery management system based on big data. In view of the problems of uneven feature weight distribution, easy neglect of local data, and inaccurate update of cluster centers in the postoperative recovery guidance model of the traditional clustering method, this solution defines feature discrimination factors and feature weights, designs local density functions, designs cluster center update formulas, defines composite distribution distance functions, designs data point distribution rules and objective functions, so as to improve the model's ability to capture local features and balance global and local features, and improve the convergence of the model and the stability of clustering results. In view of the problems of single model architecture, weak activation ability of hidden units, insufficient feature extraction ability and low monitoring accuracy in the traditional patient status monitoring model, This scheme enhances the activation ability of hidden units and the model's feature extraction ability for data by designing model architecture, multi-layer nonlinear transformation, energy function and adaptive weight adjustment mechanism, thereby improving the model's adaptability to patients' postoperative recovery status and the accuracy of real-time monitoring. In response to the problems of low search efficiency, poor stability and low quality of search results in the monitoring optimization module of traditional parameter search methods, this scheme generates parameter search points, designs adaptive parameters, co-evolutionary search, elite retention and sets search rules, combines randomness and goal orientation to generate initial parameter points, improves the starting point quality and convergence speed of parameter search, retains the optimal parameter position in each search, improves the stability of parameter search and the quality of results, and improves the flexibility and adaptability of parameter search.

[0005] The technical solution adopted by the present invention is as follows: The gastrointestinal patient postoperative recovery management system based on big data provided by the present invention includes a data acquisition module, a postoperative recovery guidance model establishment module, a patient status monitoring model establishment module, a monitoring optimization module and a comprehensive management module;

[0006] The data collection module collects the age, gender, medical history, surgical information, postoperative physiological data, complication data and postoperative recovery data of gastrointestinal patients undergoing surgery;

[0007] The postoperative recovery guidance model establishment module establishes the postoperative recovery guidance model by defining feature distinguishing factors, defining feature weights, designing local density functions, designing cluster center update formulas, defining composite distribution distance functions, designing data point distribution rules, designing objective functions and clustering;

[0008] The patient status monitoring model establishment module establishes the patient status monitoring model by setting model labels, designing model architecture, multi-layer nonlinear transformation, designing energy functions and designing adaptive weight adjustment mechanisms;

[0009] The monitoring optimization module performs monitoring optimization through initial setting, generating parameter search points, designing adaptive parameters, co-evolutionary search, elite retention and setting search rules;

[0010] The comprehensive management module manages the user's postoperative recovery using a postoperative recovery guidance model and a patient status monitoring model.

[0011] Furthermore, the data acquisition module collects the age, gender, medical history, surgical information, postoperative physiological data, complication data and postoperative recovery data of gastrointestinal patients undergoing surgery; the medical history includes previous medical history, allergy history and family medical history; the surgical information includes the type of surgery, operation time, operation duration and drugs used during the operation; the postoperative physiological data includes heart rate, blood pressure, body temperature, respiratory rate and blood biochemical examination indicators; the complication data is the type of complications that occur after the operation, complication prevention suggestions and complication care suggestions; the postoperative recovery data includes three recovery states: good, normal and abnormal.

[0012] Furthermore, the establishment of the postoperative recovery guidance model module specifically includes the following contents:

[0013] Define the feature distinguishing factor, expressed as follows:

[0014] ;

[0015] Among them, k represents the cluster index, j represents the feature dimension index of the data, represents the feature distinguishing factor of the jth feature of the kth cluster, represents the kth cluster center, represents the number of data points in the kth cluster, i represents the index of the data point used for model training, represents the i-th data point, Indicates that it belongs to the symbol, Indicates that the i-th data point belongs to the k-th cluster, represents the jth feature of the i-th data point, represents the mean of the jth feature in the kth cluster, represents all clusters except the kth cluster, represents the mean of the jth feature in all clusters except the kth cluster;

[0016] Define feature weights as follows:

[0017] ;

[0018] in, represents the feature weight of the jth feature of the kth cluster, represents the feature dimension index of the data, and D represents the total number of feature dimensions of the data. represents the kth cluster The characteristic distinguishing factor of the feature, represents an exponential function with a natural constant as base;

[0019] Design the local density function, expressed as follows:

[0020] ;

[0021] in, represents the local density function, represents the local density value of the i-th data point, l represents the index of the data point, represents the lth data point, represents the set of five data points closest to data point i, represents the density adjustment coefficient, Indicates the modulus length;

[0022] Design the cluster center update formula, which is expressed as follows:

[0023] ;

[0024] in, Represents the jth feature of the cluster center of the kth cluster after updating;

[0025] Define the composite distribution distance function, expressed as follows:

[0026] ;

[0027] in, represents the composite partition distance function, represents the composite distribution distance between the i-th data point and the k-th cluster center, represents the allocation weight;

[0028] Design data point allocation rules, calculate the composite allocation distance between the data point and each cluster center, select the cluster center with the smallest composite allocation distance to the data point, and allocate the data point to the cluster to which this cluster center belongs;

[0029] The objective function is designed as follows:

[0030] ;

[0031] in, represents the objective function value, represents the number of cluster centers, and represents the coordination coefficient;

[0032] Clustering, setting the objective function threshold, first randomly select from all model training data points Data points are used as the initial cluster centers, and then each data point is assigned to a cluster according to the data point assignment rule. The cluster center is then updated using the cluster center update formula, and the data point assignment and cluster center update are repeated. After each data point assignment, the objective function value is calculated. When the objective function value is less than the objective function threshold, the clustering converges.

[0033] Furthermore, the patient status monitoring model module is established, specifically including the following contents:

[0034] Set the model label and set the postoperative recovery data as the label data of the model;

[0035] Design the model architecture. The model is composed of a stack of restricted Boltzmann machines. Each layer is connected through an adaptive weight adjustment mechanism. The hidden units are activated through multiple layers of nonlinear transformations and finally optimized through energy functions.

[0036] Multi-layer nonlinear transformation, expressed as follows:

[0037] ;

[0038] Among them, u represents the layer index of the model, p and q represent the indexes of hidden units and visible units respectively. represents the activation value of the pth hidden unit in the uth layer of the model, represents a multi-layer nonlinear transformation function, represents the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model, represents the activation value of the qth visible unit in the u-1th layer of the model, represents the bias of the p-th hidden unit in the u-th layer of the model, Represents the input value of the multi-layer nonlinear transformation function, represents the hyperbolic tangent function, and represents the transformation coefficient;

[0039] Design the energy function, which is expressed as follows:

[0040] ;

[0041] in, represents the activation value of the qth visible unit, represents the activation value of the pth hidden unit, represents the energy function value of the qth visible unit and the pth hidden unit, represents the weight between the pth hidden unit and the qth visible unit;

[0042] Design an adaptive weight adjustment mechanism, which is expressed as follows:

[0043] ;

[0044] Among them, t represents the index of the time point, represents the value of the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model at time t, represents the value of the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model at time t-1, represents the regulation rate, represents the adjustment of the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model at time t, and represents the balance coefficient, E represents the energy function, S represents the activation state of the hidden layer, Represents the energy function for weight Find the partial derivative, Represents the activation state of the hidden layer to the weight Find the partial derivative.

[0045] Furthermore, the monitoring optimization module specifically includes the following contents:

[0046] Initial settings: set the search parameters, including the number of layers of the model and the number of hidden units in each layer, set the accuracy of the model as the parameter performance value, set the maximum number of parameter searches and the parameter performance qualification value;

[0047] Generate parameter search points, expressed as follows:

[0048] ;

[0049] in, Indicates the position of the generated initial parameter search point, PB indicates the maximum value of the search range, and LB indicates the minimum value of the search range. represents a randomly selected position within the search range. Represents a number ranging from 0 to 1;

[0050] Design the adaptive parameters, expressed as follows:

[0051] ;

[0052] in, Indicates the number of parameter searches. and Indicated in Adaptive parameters during secondary parameter search, and Indicated in Adaptive parameters during secondary parameter search, Indicates the maximum number of parameter searches, represents the maximum parameter performance value among all searched parameter positions, represents the minimum parameter performance value among all searched parameter positions, represents the average parameter performance of all searched parameter positions;

[0053] Coevolutionary search, expressed as follows:

[0054] ;

[0055] in, Indicates that the parameter search point is The parameter positions obtained by the co-evolutionary search are Indicates that the parameter search point is The parameter positions obtained by the co-evolutionary search are Indicates that all parameter search points are The average value of the parameter positions obtained by the coevolutionary search, Indicates that the parameter search point is The optimal parameter position is obtained by the second co-evolutionary search;

[0056] Elite retention: After each co-evolutionary search, the worst parameter position is first updated to the optimal parameter position obtained in the previous parameter search, and then the average position of the searched parameter points is calculated;

[0057] ;

[0058] in, Indicates that the parameter search point is The worst parameter position is obtained by the co-evolutionary search. Indicates that all parameter search points are The average value of the parameter positions obtained by the coevolutionary search, H represents the total number of parameter search points, Indicates that all parameter search points except the parameter search point that searches for the worst parameter position are performed. The sum of the parameter positions obtained by the coevolutionary search;

[0059] Set the search rules, first generate the initial parameter search points, use the parameter search points to perform co-evolutionary search, perform an elite retention operation after each search, and when the parameter performance value of the searched parameter point is greater than the parameter performance qualified value, the search ends and the parameter with the largest parameter performance value among all searched parameter positions is output; when the search times are greater than the maximum search times, search again; otherwise, the search times are increased by one and the next search is performed.

[0060] Furthermore, the comprehensive management module collects the user's age, gender, medical history, surgical information, and postoperative physiological data, and inputs the data into a postoperative recovery guidance model to obtain data belonging to the same cluster as the user data, and provides the user with complication prevention suggestions and complication care suggestions for the data in the same cluster; the user data is then input into a patient status monitoring model, and the model outputs the user's postoperative recovery data. When the status of the postoperative recovery data is abnormal, an early warning signal is issued.

[0061] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0062] (1) To address the problems of uneven feature weight distribution, easy neglect of local data, and inaccurate cluster center update in the traditional clustering method in the postoperative recovery guidance model, this scheme defines feature discrimination factors and feature weights, designs local density functions, designs cluster center update formulas, defines composite allocation distance functions, designs data point allocation rules and objective functions, improves the model's ability to capture local features and balance global and local features, and improves the convergence of the model and the stability of the clustering results.

[0063] (2) To address the problems of single model architecture, weak activation ability of hidden units, insufficient feature extraction ability and low monitoring accuracy in traditional patient status monitoring models, this scheme enhances the activation ability of hidden units and the model's feature extraction ability for data by designing model architecture, multi-layer nonlinear transformation, energy function and adaptive weight adjustment mechanism, thereby improving the model's adaptability to patients' postoperative recovery status and the accuracy of real-time monitoring.

[0064] (3) To address the problems of low search efficiency, poor stability and low quality of search results in the monitoring and optimization module of traditional parameter search methods, this scheme generates parameter search points, designs adaptive parameters, performs co-evolutionary search, retains elites and sets search rules, combines randomness and goal orientation to generate initial parameter points, improves the starting point quality and convergence speed of parameter search, retains the optimal parameter position in each search, improves the stability of parameter search and the quality of results, and improves the flexibility and adaptability of parameter search. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A schematic diagram of a gastrointestinal patient postoperative recovery management system based on big data provided by the present invention;

[0066] Figure 2 A schematic diagram of the postoperative recovery guidance model module;

[0067] Figure 3 A schematic diagram of a patient status monitoring model module is provided;

[0068] Figure 4 Schematic diagram of the monitoring optimization module;

[0069] Figure 5 A schematic diagram of the search rules.

[0070] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0072] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0073] Example 1, see Figure 1 The gastrointestinal patient postoperative recovery management system based on big data provided by the present invention includes a data acquisition module, a postoperative recovery guidance model establishment module, a patient status monitoring model establishment module, a monitoring optimization module and a comprehensive management module;

[0074] The data acquisition module collects the age, gender, medical history, surgical information, postoperative physiological data, complication data and postoperative recovery data of gastrointestinal patients after surgery; the medical history includes past medical history, allergy history and family medical history; the surgical information includes the type of surgery, surgery time, surgery duration and drugs used during surgery; the postoperative physiological data includes heart rate, blood pressure, body temperature, respiratory rate and blood biochemical examination indicators; the complication data is the type of complication that occurs after surgery, complication prevention suggestions and complication nursing suggestions; the postoperative recovery data includes three recovery states: good, normal and abnormal; and the data is sent to the module for establishing a postoperative recovery guidance model and the module for establishing a patient status monitoring model;

[0075] The postoperative recovery guidance model establishment module receives data sent by the data acquisition module, establishes a postoperative recovery guidance model by defining feature differentiation factors, defining feature weights, designing local density functions, designing cluster center update formulas, defining composite distribution distance functions, designing data point distribution rules, designing objective functions and clustering, and sends the data to the comprehensive management module;

[0076] The patient status monitoring model establishment module receives data sent by the data acquisition module, establishes the patient status monitoring model by setting model labels, designing model architecture, multi-layer nonlinear transformation, designing energy functions and designing adaptive weight adjustment mechanisms, and sends the data to the monitoring optimization module;

[0077] The monitoring optimization module receives data sent by the patient status monitoring model establishment module, performs monitoring optimization through initial setting, generating parameter search points, designing adaptive parameters, co-evolutionary search, elite retention and setting search rules, and sends the data to the comprehensive management module;

[0078] The comprehensive management module receives data sent by the postoperative recovery guidance model establishment module and the monitoring optimization module, and manages the postoperative recovery of the user using the postoperative recovery guidance model and the patient status monitoring model.

[0079] Example 2, see Figure 1 and Figure 2 This embodiment is based on the above embodiment, and the postoperative recovery guidance model module is established, specifically including the following contents:

[0080] Define the feature discrimination factor to quantify the difference between features within and between clusters, expressed as follows:

[0081] ;

[0082] Among them, k represents the cluster index, j represents the feature dimension index of the data, represents the feature distinguishing factor of the jth feature of the kth cluster, represents the kth cluster center, represents the number of data points in the kth cluster, i represents the index of the data point used for model training, represents the i-th data point, Indicates that it belongs to the symbol, Indicates that the i-th data point belongs to the k-th cluster, represents the jth feature of the i-th data point, represents the mean of the jth feature in the kth cluster, represents all clusters except the kth cluster, represents the mean of the jth feature in all clusters except the kth cluster;

[0083] Define feature weights and dynamically assign the importance of each feature in clustering based on the feature distinguishing factor, as shown below:

[0084] ;

[0085] in, represents the feature weight of the jth feature of the kth cluster, represents the feature dimension index of the data, and D represents the total number of feature dimensions of the data. represents the kth cluster The characteristic distinguishing factor of the feature, represents an exponential function with a natural constant as base;

[0086] Design a local density function to quantify the distribution density of data points in the neighborhood and capture the local structural characteristics of the data, which is expressed as follows:

[0087] ;

[0088] in, represents the local density function, represents the local density value of the i-th data point, l represents the index of the data point, represents the lth data point, represents the set of five data points closest to data point i, represents the density adjustment coefficient, Indicates the modulus length;

[0089] The cluster center update formula is designed to dynamically update the location of the cluster center through weighted averaging and local density adjustment, which is expressed as follows:

[0090] ;

[0091] in, Represents the jth feature of the cluster center of the kth cluster after updating;

[0092] Define the composite distribution distance function, combining feature differences and local density to quantify the distance between data points and cluster centers, as follows:

[0093] ;

[0094] in, represents the composite partition distance function, represents the composite distribution distance between the i-th data point and the k-th cluster center, represents the allocation weight;

[0095] Design data point allocation rules, calculate the composite allocation distance between the data point and each cluster center, select the cluster center with the smallest composite allocation distance to the data point, and allocate the data point to the cluster to which this cluster center belongs;

[0096] The objective function is designed by comprehensively considering the compactness within the cluster, the local density and the sparsity of the feature weights, which is expressed as follows:

[0097] ;

[0098] in, represents the objective function value, represents the number of cluster centers, and represents the coordination coefficient;

[0099] Clustering, setting the objective function threshold, first randomly select from all model training data points Data points are used as the initial cluster centers, and then each data point is assigned to a cluster according to the data point assignment rule. The cluster center is then updated using the cluster center update formula, and the data point assignment and cluster center update are repeated. After each data point assignment, the objective function value is calculated. When the objective function value is less than the objective function threshold, the clustering converges.

[0100] By performing the above operations, in order to address the problems of uneven feature weight distribution, easy neglect of local data, and inaccurate cluster center update in the traditional clustering method in the postoperative recovery guidance model, this scheme defines feature discrimination factors and feature weights, designs local density functions, designs cluster center update formulas, defines composite allocation distance functions, designs data point allocation rules and objective functions, so as to improve the model's ability to capture local features and its ability to balance global and local features, thereby improving the convergence of the model and the stability of the clustering results.

[0101] Example 3, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the patient status monitoring model module is established, specifically including the following contents:

[0102] Set the model label and set the postoperative recovery data as the label data of the model;

[0103] Design the model architecture. The model is composed of a stack of restricted Boltzmann machines. Each layer is connected through an adaptive weight adjustment mechanism. The hidden units are activated through multiple layers of nonlinear transformations and finally optimized through energy functions.

[0104] Multi-layer nonlinear transformation, expressed as follows:

[0105] ;

[0106] Among them, u represents the layer index of the model, p and q represent the indexes of hidden units and visible units respectively. represents the activation value of the pth hidden unit in the uth layer of the model, represents a multi-layer nonlinear transformation function, represents the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model, represents the activation value of the qth visible unit in the u-1th layer of the model, represents the bias of the p-th hidden unit in the u-th layer of the model, Represents the input value of the multi-layer nonlinear transformation function, represents the hyperbolic tangent function, and represents the transformation coefficient;

[0107] The energy function is designed to quantify the interaction between visible units and hidden units and measure the stability of the model, which is expressed as follows:

[0108] ;

[0109] in, represents the activation value of the qth visible unit, represents the activation value of the pth hidden unit, represents the energy function value of the qth visible unit and the pth hidden unit, represents the weight between the pth hidden unit and the qth visible unit;

[0110] Design an adaptive weight adjustment mechanism, which is expressed as follows:

[0111] ;

[0112] Among them, t represents the index of the time point, represents the value of the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model at time t, represents the value of the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model at time t-1, represents the regulation rate, represents the adjustment of the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model at time t, and represents the balance coefficient, E represents the energy function, S represents the activation state of the hidden layer, Represents the energy function for weight Find the partial derivative, Represents the activation state of the hidden layer to the weight Find the partial derivative.

[0113] By performing the above operations, this solution aims to address the problems of single model architecture, weak activation ability of hidden units, insufficient feature extraction ability and low monitoring accuracy in traditional patient status monitoring models. By designing model architecture, multi-layer nonlinear transformation, energy function and adaptive weight adjustment mechanism, this solution enhances the activation ability of hidden units, enhances the model's feature extraction ability for data, and improves the model's adaptability to patients' postoperative recovery status and the accuracy of real-time monitoring.

[0114] Example 4, see Figure 1 , Figure 4 and Figure 5 This embodiment is based on the above embodiment, and the monitoring optimization module specifically includes the following contents:

[0115] Initial settings: set the search parameters, including the number of layers of the model and the number of hidden units in each layer, set the accuracy of the model as the parameter performance value, set the maximum number of parameter searches and the parameter performance qualification value;

[0116] Generate parameter search points. Generate initial parameter search points by randomizing and adjusting coefficients, as shown below:

[0117] ;

[0118] in, Indicates the position of the generated initial parameter search point, PB indicates the maximum value of the search range, and LB indicates the minimum value of the search range. represents a randomly selected position within the search range. Represents a number ranging from 0 to 1;

[0119] Design adaptive parameters, dynamically adjust the search step size, and balance global search and local search, as shown below:

[0120] ;

[0121] in, Indicates the number of parameter searches. and Indicated in Adaptive parameters during secondary parameter search, and Indicated in Adaptive parameters during secondary parameter search, Indicates the maximum number of parameter searches, represents the maximum parameter performance value among all searched parameter positions, represents the minimum parameter performance value among all searched parameter positions, represents the average parameter performance of all searched parameter positions;

[0122] Coevolutionary search, expressed as follows:

[0123] ;

[0124] in, Indicates that the parameter search point is The parameter positions obtained by the co-evolutionary search are Indicates that the parameter search point is The parameter positions obtained by the co-evolutionary search are Indicates that all parameter search points are The average value of the parameter positions obtained by the coevolutionary search, Indicates that the parameter search point is The optimal parameter position is obtained by the second co-evolutionary search;

[0125] Elite retention: After each co-evolutionary search, the worst parameter position is first updated to the optimal parameter position obtained in the previous parameter search, and then the average position of the searched parameter points is calculated;

[0126] ;

[0127] in, Indicates that the parameter search point is The worst parameter position is obtained by the co-evolutionary search. Indicates that all parameter search points are The average value of the parameter positions obtained by the coevolutionary search, H represents the total number of parameter search points, Indicates that all parameter search points except the parameter search point that searches for the worst parameter position are performed. The sum of the parameter positions obtained by the coevolutionary search;

[0128] Set the search rules, first generate the initial parameter search points, use the parameter search points to perform co-evolutionary search, perform an elite retention operation after each search, and when the parameter performance value of the searched parameter point is greater than the parameter performance qualified value, the search ends and the parameter with the largest parameter performance value among all searched parameter positions is output; when the search times are greater than the maximum search times, search again; otherwise, the search times are increased by one and the next search is performed.

[0129] By performing the above operations, in order to address the problems of low search efficiency, poor stability and low quality of search results in the monitoring and optimization module of traditional parameter search methods, this solution generates parameter search points, designs adaptive parameters, performs co-evolutionary search, retains elites and sets search rules, combines randomness and goal orientation to generate initial parameter points, improves the starting point quality and convergence speed of parameter search, retains the optimal parameter position in each search, improves the stability of parameter search and the quality of results, and improves the flexibility and adaptability of parameter search.

[0130] Example 5, see Figure 1 This embodiment is based on the above embodiment. The comprehensive management module collects the user's age, gender, medical history, surgical information, and postoperative physiological data, inputs the data into the postoperative recovery guidance model, obtains data belonging to the same cluster as the user data, and provides the user with complication prevention suggestions and complication care suggestions for the data in the same cluster; then inputs the user data into the patient status monitoring model, and the model outputs the user's postoperative recovery data. When the status of the postoperative recovery data is abnormal, an early warning signal is issued.

[0131] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0132] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

[0133] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A gastrointestinal patient postoperative recovery management system based on big data, characterized by: It includes data collection module, postoperative recovery guidance model establishment module, patient status monitoring model establishment module, monitoring optimization module and comprehensive management module; The data collection module collects the age, gender, medical history, surgical information, postoperative physiological data, complication data and postoperative recovery data of gastrointestinal patients undergoing surgery; The postoperative recovery guidance model establishment module establishes the postoperative recovery guidance model by defining feature distinguishing factors, defining feature weights, designing local density functions, designing cluster center update formulas, defining composite distribution distance functions, designing data point distribution rules, designing objective functions and clustering; Among them, the feature distinguishing factor is defined as follows: ； Among them, k represents the cluster index, j represents the feature dimension index of the data, represents the feature distinguishing factor of the jth feature of the kth cluster, represents the kth cluster center, represents the number of data points in the kth cluster, i represents the index of the data point used for model training, represents the i-th data point, Indicates that it belongs to the symbol, Indicates that the i-th data point belongs to the k-th cluster, represents the jth feature of the i-th data point, represents the mean of the jth feature in the kth cluster, represents all clusters except the kth cluster, represents the mean of the jth feature in all clusters except the kth cluster; The patient status monitoring model establishment module establishes the patient status monitoring model by setting model labels, designing model architecture, multi-layer nonlinear transformation, designing energy functions and designing adaptive weight adjustment mechanisms; The monitoring optimization module performs monitoring optimization through initial setting, generating parameter search points, designing adaptive parameters, co-evolutionary search, elite retention and setting search rules; The comprehensive management module manages the user's postoperative recovery using a postoperative recovery guidance model and a patient status monitoring model.

2. The gastrointestinal patient postoperative recovery management system based on big data according to claim 1 is characterized by: The postoperative recovery guidance model module is established, specifically including the following contents: Define the characteristic distinguishing factors; Define feature weights as follows: ； in, represents the feature weight of the jth feature of the kth cluster, represents the feature dimension index of the data, and D represents the total number of feature dimensions of the data. represents the kth cluster The characteristic distinguishing factor of the feature, represents an exponential function with a natural constant as base; Design the local density function, expressed as follows: ； in, represents the local density function, represents the i-th data point, represents the local density value of the i-th data point, l represents the index of the data point, represents the lth data point, represents the set of five data points closest to data point i, represents the density adjustment coefficient, Indicates the modulus length; Design the cluster center update formula, which is expressed as follows: ； in, Represents the jth feature of the cluster center of the kth cluster after updating; Define the composite distribution distance function, expressed as follows: ； in, represents the composite partition distance function, represents the composite distribution distance between the i-th data point and the k-th cluster center, represents the allocation weight; Design data point allocation rules, calculate the composite allocation distance between the data point and each cluster center, select the cluster center with the smallest composite allocation distance to the data point, and allocate the data point to the cluster to which this cluster center belongs; The objective function is designed as follows: ； in, represents the objective function value, represents the number of cluster centers, and represents the coordination coefficient; Clustering, setting the objective function threshold, first randomly select from all model training data points Data points are used as the initial cluster centers, and then each data point is assigned to a cluster according to the data point assignment rule. The cluster center is then updated using the cluster center update formula, and the data point assignment and cluster center update are repeated. After each data point assignment, the objective function value is calculated. When the objective function value is less than the objective function threshold, the clustering converges.

3. The gastrointestinal patient postoperative recovery management system based on big data according to claim 1 is characterized by: The patient status monitoring model module is established, specifically including the following contents: Set the model label and set the postoperative recovery data as the label data of the model; Design the model architecture. The model is composed of a stack of restricted Boltzmann machines. Each layer is connected through an adaptive weight adjustment mechanism. The hidden units are activated through multiple layers of nonlinear transformations and finally optimized through energy functions. Multi-layer nonlinear transformation, expressed as follows: ； Among them, u represents the layer index of the model, p and q represent the indexes of hidden units and visible units respectively. represents the activation value of the pth hidden unit in the uth layer of the model, represents a multi-layer nonlinear transformation function, represents the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model, represents the activation value of the qth visible unit in the u-1th layer of the model, represents the bias of the p-th hidden unit in the u-th layer of the model, Represents the input value of the multi-layer nonlinear transformation function, represents the hyperbolic tangent function, and represents the transformation coefficient; Design the energy function, which is expressed as follows: ； in, represents the activation value of the qth visible unit, represents the activation value of the pth hidden unit, represents the energy function value of the qth visible unit and the pth hidden unit, represents the weight between the pth hidden unit and the qth visible unit; Design an adaptive weight adjustment mechanism, which is expressed as follows: ； Among them, t represents the index of the time point, represents the value of the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model at time t, represents the value of the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model at time t-1, represents the regulation rate, represents the adjustment of the weight between the p-th hidden unit and the q-th visible unit of the u-th layer of the model at time t, and represents the balance coefficient, E represents the energy function, S represents the activation state of the hidden layer, Represents the energy function for weight Find the partial derivative, Represents the activation state of the hidden layer to the weight Find the partial derivative.

4. The gastrointestinal patient postoperative recovery management system based on big data according to claim 1 is characterized by: The monitoring optimization module specifically Includes the following: Initial settings: set the search parameters, including the number of layers of the model and the number of hidden units in each layer, set the accuracy of the model as the parameter performance value, set the maximum number of parameter searches and the parameter performance qualification value; Generate parameter search points, expressed as follows: ； in, Indicates the position of the generated initial parameter search point, PB indicates the maximum value of the search range, and LB indicates the minimum value of the search range. represents a randomly selected position within the search range. Represents a number ranging from 0 to 1; Design the adaptive parameters, expressed as follows: ； in, Indicates the number of parameter searches. and Indicated in Adaptive parameters during secondary parameter search, and Indicated in Adaptive parameters during secondary parameter search, Indicates the maximum number of parameter searches, represents the maximum parameter performance value among all searched parameter positions, represents the minimum parameter performance value among all searched parameter positions, represents the average parameter performance of all searched parameter positions; Coevolutionary search, expressed as follows: ； in, Indicates that the parameter search point is The parameter positions obtained by the co-evolutionary search are Indicates that the parameter search point is The parameter positions obtained by the co-evolutionary search are Indicates that all parameter search points are The average value of the parameter positions obtained by the coevolutionary search, Indicates that the parameter search point is The optimal parameter position is obtained by the second co-evolutionary search; Elite retention: After each co-evolutionary search, the worst parameter position is first updated to the optimal parameter position obtained in the previous parameter search, and then the average position of the searched parameter points is calculated; ； in, Indicates that the parameter search point is The worst parameter position is obtained by the co-evolutionary search. Indicates that all parameter search points are The average value of the parameter positions obtained by the coevolutionary search, H represents the total number of parameter search points, Indicates that all parameter search points except the parameter search point that searches for the worst parameter position are performed. The sum of the parameter positions obtained by the coevolutionary search; Set the search rules, first generate the initial parameter search points, use the parameter search points to perform co-evolutionary search, perform an elite retention operation after each search, and when the parameter performance value of the searched parameter point is greater than the parameter performance qualified value, the search ends and the parameter with the largest parameter performance value among all searched parameter positions is output; when the search times are greater than the maximum search times, search again; otherwise, the search times are increased by one and the next search is performed.

5. The gastrointestinal patient postoperative recovery management system based on big data according to claim 1 is characterized by: The data acquisition module collects the age, gender, medical history, surgical information, postoperative physiological data, complication data and postoperative recovery data of gastrointestinal patients who have undergone surgery; the medical history includes past medical history, allergy history and family medical history; the surgical information includes the type of surgery, surgery time, surgery duration and drugs used during the surgery; the postoperative physiological data includes heart rate, blood pressure, body temperature, respiratory rate and blood biochemical examination indicators; the complication data includes the type of complication that occurs after surgery, complication prevention suggestions and complication nursing suggestions; the postoperative recovery data includes three recovery states: good, normal and abnormal; The comprehensive management module collects the user's age, gender, medical history, surgical information, and postoperative physiological data, and inputs the data into a postoperative recovery guidance model to obtain data belonging to the same cluster as the user data, and provides the user with complication prevention suggestions and complication care suggestions for the data in the same cluster; the user data is then input into a patient status monitoring model, and the model outputs the user's postoperative recovery data. When the status of the postoperative recovery data is abnormal, an early warning signal is issued.

Images