Burn patient hemodialysis curative effect prediction system based on big data

Abstract

The invention discloses a burn patient hemodialysis curative effect prediction system based on big data. The burn patient hemodialysis curative effect prediction system comprises a dialysis curative effect prediction data acquisition module, an original data optimization module, a hemodialysis curative effect prediction model construction module, a model effect improvement module and a personalized hemodialysis curative effect prediction module. The invention relates to the technical field of medical data processing, in particular to a burn patient hemodialysis curative effect prediction system based on big data, which innovatively proposes dynamic curative effect prediction weight distribution, and can accurately predict the hemodialysis curative effect according to the current condition of a patient by adjusting the weight of each index in real time; three independent prediction layers are innovatively designed, independent prediction is carried out for different curative effect stages, and the hemodialysis curative effect of a burn patient is comprehensively and accurately evaluated; by improving a particle optimization algorithm for obtaining an optimal hyper-parameter combination of the model, the accuracy of a model output result is improved, and accurate prediction of a hemodialysis curative effect of a burn patient is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data processing, and in particular to a hemodialysis efficacy prediction system for burn patients based on big data. Background Art

[0002] With the continuous progress of burn treatment, burn patients often face a variety of complications such as fluid imbalance, infection, organ failure, etc., especially in the case of impaired renal function or fluid imbalance, hemodialysis has become a common treatment method. The condition of burn patients is complex and changes rapidly. The traditional prediction based on manual prediction by doctors is difficult to adapt to the dynamic changes of patients' conditions. In order to meet this challenge, a hemodialysis efficacy prediction system for burn patients based on big data came into being. By collecting and analyzing a large number of clinical data of burn patients and combining artificial intelligence technology, it can realize real-time monitoring of patients' conditions and predict dialysis efficacy, thereby providing doctors with more scientific decision support; however, traditional hemodialysis In the hemodialysis efficacy prediction system, fixed weights are used to process all data, ignoring the dynamic adjustments of different patients at different treatment stages and when their conditions change, resulting in inaccurate prediction of hemodialysis efficacy when the condition of burn patients changes rapidly. The existing traditional hemodialysis efficacy prediction model cannot flexibly respond to changes in the patient's condition at different treatment stages, and lacks consideration of comprehensive prediction of the hemodialysis efficacy throughout the entire process, resulting in inaccurate and incomplete prediction results. In the existing applicable prediction model, there are inappropriate built-in parameter settings, and the model optimization algorithm has a weak ability to obtain the global optimal solution, resulting in inaccurate prediction results of the final model. Summary of the invention

[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a hemodialysis efficacy prediction system for burn patients based on big data. In view of the technical problem that the traditional hemodialysis efficacy prediction system processes all data with fixed weights, ignores the dynamic adjustment of different patients at different treatment stages and changes in their condition, and leads to inaccurate prediction of hemodialysis efficacy when the condition of burn patients changes rapidly, this solution innovatively proposes a dynamic efficacy prediction weight allocation. By adjusting the weights of various indicators in real time, accurate predictions can be made according to the patient's current condition, avoiding the errors caused by fixed weights. The unique burn condition of each patient is comprehensively considered by dynamically adjusting the weights, which can accurately reflect the physiological and pathological states of burn patients at different stages, thereby improving the accuracy of hemodialysis efficacy prediction. In view of the fact that the existing traditional hemodialysis efficacy prediction model cannot flexibly respond to changes in patients' conditions at different treatment stages, and lacks comprehensive prediction of hemodialysis efficacy throughout the whole process, In order to solve the technical problem that the prediction results are inaccurate and incomplete due to lack of consideration of measurement, this solution innovatively designs three independent prediction layers, each of which makes independent predictions for different efficacy stages, which can comprehensively and accurately evaluate the hemodialysis efficacy of burn patients, avoid the deviation of a single dimension, and help doctors adjust the treatment plan in time to improve the treatment effect, thereby ensuring the comprehensive monitoring of the dialysis efficacy of burn patients and significantly improving the accuracy of the prediction of the hemodialysis efficacy; in view of the technical problem that the existing prediction models have inappropriate built-in parameter settings and the model optimization algorithm has weak ability to obtain the global optimal solution, resulting in inaccurate final model prediction results, this solution enhances the global optimal solution search and local search capabilities by designing particle search optimization factors and inferior particle acceptance probability strategies, thereby optimizing the hyperparameter combination of the hemodialysis efficacy prediction model, significantly improving the accuracy of the model output results, and realizing accurate prediction of the hemodialysis efficacy of burn patients.

[0004] The technical solution adopted by the present invention is as follows: the hemodialysis efficacy prediction system for burn patients based on big data provided by the present invention includes a hemodialysis efficacy prediction data acquisition module, an original data optimization module, a hemodialysis efficacy prediction model construction module, a model effect improvement module and a personalized hemodialysis efficacy prediction module;

[0005] The module for obtaining hemodialysis efficacy prediction data is to obtain the original data of hemodialysis efficacy prediction by collecting data from the hospital information system;

[0006] The raw data optimization module is used to optimize the raw data for hemodialysis efficacy prediction. First, the data quality is improved by raw data preprocessing. Then, the dynamic efficacy prediction weight allocation stage is entered. The feature weight is dynamically adjusted according to the patient's burn severity, complication risk and physiological index changes to obtain the total weight representing the patient's hemodialysis efficacy prediction. Dynamic feature weighting is then performed to obtain hemodialysis efficacy prediction optimization data.

[0007] The module for constructing a hemodialysis efficacy prediction model is used to obtain a hemodialysis efficacy prediction model, specifically by designing short-term, medium-term and long-term efficacy prediction layers, and training in combination with the total loss function of the model, thereby obtaining a hemodialysis efficacy prediction model;

[0008] The model effect improvement module obtains the optimal hyperparameter combination of the hemodialysis efficacy prediction model by improving the particle swarm optimization algorithm, and adjusts the hyperparameters of the model according to the optimal hyperparameter combination to obtain the hemodialysis efficacy prediction model with improved effect;

[0009] The personalized hemodialysis efficacy prediction module specifically inputs real-time data into the model to obtain hemodialysis efficacy prediction results, and assists doctors in making scientific decisions based on the prediction results.

[0010] Furthermore, the module for obtaining hemodialysis efficacy prediction data specifically obtains original hemodialysis efficacy prediction data from the hospital information system by collecting data; the original hemodialysis efficacy prediction data includes hemodialysis efficacy prediction data of other burn patients and real-time hemodialysis efficacy prediction data of burn patients; the hemodialysis efficacy prediction data of other burn patients and real-time hemodialysis efficacy prediction data of burn patients both include basic information data of patients, physiological data of patients, and hemodialysis data of patients; the hemodialysis efficacy data of other burn patients also includes hemodialysis efficacy results of patients.

[0011] Furthermore, the raw data optimization module is specifically for optimizing the raw data for hemodialysis efficacy prediction, including raw data preprocessing, dynamic efficacy prediction weight allocation and dynamic feature weighting, to obtain hemodialysis efficacy prediction optimized data; specifically including the following steps:

[0012] Raw data preprocessing is used to improve data quality, including data cleaning, data standardization and category data processing;

[0013] Dynamic efficacy prediction weight allocation is used to dynamically adjust the importance of different features in efficacy prediction according to the severity of burns, complications and physiological status of the patient, and obtain the total weight representing the prediction of the patient's hemodialysis efficacy; specifically, it includes the following steps:

[0014] The burn severity weight is calculated by using the area index function and the location correction factor; the formula used is as follows:

[0015] ;

[0016] In the formula, represents the burn severity weight, Indicates the percentage of body surface area burned. Indicates the depth of burns in patients. Indicates the correction factor for burn location;

[0017] The complication weight is calculated based on the complication risk level using the following formula:

[0018] ;

[0019] In the formula, represents the complication weight of burn patients, Represents high-risk complication index, Indicates the intermediate-risk complication index. Indicates low risk complication index;

[0020] The weight of the physiological index is calculated by considering the change of the physiological index; the formula used is as follows:

[0021] ;

[0022] In the formula, represents the weight of the physiological index in the current stage, k represents the stage sensitivity coefficient, and controls the impact intensity of the burn treatment stage. represents the treatment phase adjustment factor, represents the clinical importance coefficient of the static deviation of the i-th physiological index, Represents the sensitivity coefficient of the dynamic change of the i-th physiological index, represents the monitoring value of the i-th physiological index of the burn patient at the last moment, represents the monitoring value of the i-th physiological index of the burn patient at the current moment, represents the normal range of the ith physiological index, represents the rate of change of the ith physiological index;

[0023] Dynamic weight fusion is to integrate various weight factors through hierarchical control and nonlinear fusion, which retains the medical interpretability of the weight system and ensures numerical stability. The formula used is as follows:

[0024] ;

[0025] In the formula, represents the total weight;

[0026] Dynamic feature weighting is used to obtain a weighted feature set as the input of the hemodialysis efficacy prediction model; the formula used is as follows:

[0027] ;

[0028] ;

[0029] In the formula, represents the weighting factor of the jth feature that is dynamically adjusted according to the patient's condition, represents the weighting factor designed based on the clinical importance of the feature itself, represents the weighted feature set, represents the actual value of the jth feature, and n represents the total number of features.

[0030] Furthermore, the hemodialysis efficacy prediction model module is used to obtain the hemodialysis efficacy prediction model, including designing a prediction task sharing layer, designing a short-term efficacy prediction layer, designing a mid-term efficacy prediction layer, designing a long-term efficacy prediction layer and designing a model total loss function; specifically including the following steps:

[0031] A prediction task sharing layer is designed to obtain shared features. Specifically, the long-term and short-term dependencies in the data are captured through the long short-term memory network, and the attention mechanism is introduced to capture important information in the data. The formula used is as follows:

[0032] ;

[0033] ;

[0034] In the formula, Indicates the shared features of the prediction task shared layer output, Indicates the unit runs the function, Indicates the hidden state at the previous moment, Indicates the state of the memory unit at the previous moment, represents the original shared features, represents the weight matrix of the attention mechanism, represents the bias vector of the attention mechanism, Indicates the accumulation of all relevant feature elements;

[0035] A short-term efficacy prediction layer is designed to predict the short-term efficacy of hemodialysis and determine the fluid balance of burn patients after dialysis. The formula used is as follows:

[0036] ;

[0037] In the formula, Represents short-term characteristics, and They represent the weight matrix and bias parameters of the short-term efficacy prediction layer output respectively. It indicates the short-term prediction results of hemodialysis efficacy;

[0038] A mid-term efficacy prediction layer was designed to predict the mid-term efficacy of hemodialysis and determine the risk of complications after dialysis in burn patients. The formula used is as follows:

[0039] ;

[0040] In the formula, Indicates mid-term characteristics, and They represent the weight matrix and bias parameters output by the mid-term efficacy prediction layer, It indicates the mid-term prediction results of hemodialysis efficacy;

[0041] A long-term efficacy prediction layer is designed to predict the long-term efficacy of hemodialysis and judge the recovery of burn patients. Specifically, a gating mechanism is introduced. The formula used is as follows:

[0042] ;

[0043] In the formula, represents the gating signal, Represents long-term characteristics, and They represent the weight matrix and bias parameters of the long-term efficacy prediction layer output respectively. It represents the long-term prediction results of hemodialysis efficacy;

[0044] Obtaining the model prediction results, specifically the hemodialysis efficacy prediction results are composed of the short-term prediction results of the hemodialysis efficacy, the mid-term prediction results of the hemodialysis efficacy and the long-term prediction results of the hemodialysis efficacy;

[0045] The total loss function of the design model is specifically based on the dynamic weight design model. The formula used is as follows:

[0046] ;

[0047] ;

[0048] In the formula, L represents the total loss function value of the model, represents the cross entropy loss function of each prediction layer, represents the mean absolute error of the prediction layer c, represents the dynamic weight of the prediction layer c in the total loss, represents the hyperparameters that control weight adjustment;

[0049] The prediction model training specifically uses the hemodialysis efficacy data of other burn patients as the training data of the model, calculates the loss through the total loss function of the model, and optimizes the model parameters using the gradient descent method to minimize the error, thereby obtaining a hemodialysis efficacy prediction model.

[0050] Furthermore, the model effect improvement module specifically obtains the optimal hyperparameter combination of the hemodialysis efficacy prediction model by improving the particle swarm optimization algorithm to obtain the hemodialysis efficacy prediction model after the effect is improved, including the following steps:

[0051] Initialize the particle swarm, specifically randomly generate the initial position and initial velocity of the particles and calculate the individual fitness value f of the particles i ; The performance of the hemodialysis efficacy prediction model established based on the individual position is used as the individual fitness value;

[0052] Calculate the particle search optimization factor to adjust the particle search parameters. The formula used is as follows:

[0053] ;

[0054] ;

[0055] In the formula, represents the average distance between the i-th particle and all other particles in the k-th iteration, N represents the total number of particles, represents the position of the ith particle in the dth dimension in the kth iteration, represents the position of the jth particle in the dth dimension in the kth iteration, D represents the number of dimensions of the solution space for each particle, represents the weight parameter used to balance the distance, represents the particle search optimization factor in the kth iteration, represents the weight coefficient of the particle in the dth dimension, represents the average distance between the global optimal particle and other particles in the kth iteration, represents the minimum distance between particles in the kth iteration, represents the maximum inter-particle distance in the kth iteration;

[0056] Adaptively adjust particle parameters, specifically adjust inertia weight and individual learning factor according to particle search optimization factor and group learning factor , the formula used is as follows:

[0057] ;

[0058] ;

[0059] ;

[0060] In the formula, represents the inertia weight of the kth iteration, Represents the parameter for adjusting the oscillation amplitude, which is used to control the amplitude of the distance jump. represents the individual learning factor of the kth iteration, represents the population learning factor of the kth iteration, Represents a constant term to prevent the parameter from tending to zero;

[0061] Particle update, specifically through inertia weight, individual learning factor and group learning factor Update particle velocity and particle position ;

[0062] Accepting inferior particles, specifically by calculating the inferior particle acceptance probability, the inferior particles are accepted using the following formula:

[0063] ;

[0064] In the formula, Indicates The acceptance probability of the i-th particle in the iteration is, represents the maximum number of iterations, represents the adjustment parameter of the acceptance probability, represents the particle fitness function, Indicates The position of the i-th particle in the d-th dimension in the iteration;

[0065] Get the optimal hyperparameter combination, specifically when the individual fitness value of the particle f i When the value is higher than the fitness threshold and the maximum number of iterations is reached, the search is terminated and the global optimal position of the individual particle is obtained. The global optimal position of the individual particle specifically refers to the optimal hyperparameter combination of the hemodialysis efficacy prediction model, and the optimal hyperparameter combination of the hemodialysis efficacy prediction model is obtained;

[0066] Obtaining an optimized prediction model, specifically adjusting the hyperparameters of the hemodialysis efficacy prediction model according to the optimal hyperparameter combination of the hemodialysis efficacy prediction model, to obtain a hemodialysis efficacy prediction model with improved effect.

[0067] Furthermore, the personalized hemodialysis efficacy prediction module specifically inputs the real-time burn patient hemodialysis efficacy prediction data into the improved hemodialysis efficacy prediction model to obtain a hemodialysis efficacy prediction result, and assists doctors in making scientific decisions based on the hemodialysis efficacy prediction result.

[0068] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0069] (1) In order to solve the technical problem that the traditional hemodialysis efficacy prediction system processes all data with fixed weights and ignores the dynamic adjustments of different patients at different treatment stages and when their conditions change, resulting in inaccurate prediction of hemodialysis efficacy when the conditions of burn patients change rapidly, this scheme innovatively proposes a dynamic efficacy prediction weight allocation. By adjusting the weights of various indicators in real time, accurate predictions can be made based on the patient's current conditions, avoiding the errors caused by fixed weights. The unique burn conditions of each patient are comprehensively considered by dynamically adjusting the weights, which can accurately reflect the physiological and pathological conditions of burn patients at different stages, thereby improving the accuracy of hemodialysis efficacy prediction.

[0070] (2) In view of the technical problems that the existing traditional hemodialysis efficacy prediction model cannot flexibly respond to changes in patients' conditions at different stages of treatment and lacks consideration of the comprehensive prediction of the entire hemodialysis efficacy, resulting in inaccurate and incomplete prediction results, this solution innovatively designs three independent prediction layers. Each prediction layer makes independent predictions for different efficacy stages, which can comprehensively and accurately evaluate the hemodialysis efficacy of burn patients, avoid the deviation of a single dimension, help doctors adjust the treatment plan in time, and improve the treatment effect, thereby ensuring the comprehensive monitoring of the hemodialysis efficacy of burn patients and significantly improving the accuracy of hemodialysis efficacy prediction.

[0071] (3) In order to address the technical issues that the existing prediction models have inappropriate built-in parameter settings and the model optimization algorithm has a weak ability to obtain the global optimal solution, which leads to inaccurate final model prediction results, this solution designs particle search optimization factors and inferior particle acceptance probability strategies to enhance the global optimal solution search and local search capabilities, thereby achieving the optimal hyperparameter combination for the hemodialysis efficacy prediction model, significantly improving the accuracy of the model output results and achieving accurate prediction of the hemodialysis efficacy of burn patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A schematic diagram of the modules of the big data-based hemodialysis efficacy prediction system for burn patients provided by the present invention;

[0073] Figure 2 Schematic diagram of the process of the raw data optimization module;

[0074] Figure 3 A flowchart for constructing a hemodialysis efficacy prediction model module;

[0075] Figure 4 This is a flow chart of the model effect improvement module;

[0076] Figure 5 A schematic diagram of the process of dynamic efficacy prediction weight allocation in the raw data optimization module;

[0077] 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

[0078] 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.

[0079] In the description of the present invention, it is necessary to understand 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 referred system or element must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as a limitation on the present invention.

[0080] Example 1, see Figure 1 The big data-based hemodialysis efficacy prediction system for burn patients provided by the present invention includes a hemodialysis efficacy prediction data acquisition module, an original data optimization module, a hemodialysis efficacy prediction model construction module, a model effect improvement module and a personalized hemodialysis efficacy prediction module;

[0081] The module for obtaining hemodialysis efficacy prediction data is to obtain the original data of hemodialysis efficacy prediction from the hospital information system by collecting data, and send the data to the original data optimization module;

[0082] The raw data optimization module receives the data sent by the hemodialysis efficacy prediction data acquisition module, improves the data quality through raw data preprocessing, and then enters the dynamic efficacy prediction weight allocation stage, dynamically adjusts the feature weight according to the patient's burn severity, complication risk and physiological index changes, obtains the total weight representing the patient's hemodialysis efficacy prediction, and performs dynamic feature weighting to obtain hemodialysis efficacy prediction optimization data, and sends the data to the hemodialysis efficacy prediction model construction module;

[0083] The hemodialysis efficacy prediction model building module receives the data sent by the original data optimization module, and is used to obtain the hemodialysis efficacy prediction model, specifically by designing short-term, medium-term and long-term efficacy prediction layers, and training in combination with the model total loss function, so as to obtain the hemodialysis efficacy prediction model, and send the data to the model effect improvement module;

[0084] The model effect improvement module receives the data sent by the hemodialysis efficacy prediction model construction module, obtains the optimal hyperparameter combination of the hemodialysis efficacy prediction model by improving the particle swarm optimization algorithm, adjusts the hyperparameters of the model according to the optimal hyperparameter combination, obtains the hemodialysis efficacy prediction model after the effect is improved, and sends the data to the personalized hemodialysis efficacy prediction module;

[0085] The personalized hemodialysis efficacy prediction module receives data sent by the model effect improvement module, inputs real-time data into the model, obtains hemodialysis efficacy prediction results, and assists doctors in making scientific decisions based on the prediction results.

[0086] Example 2, see Figure 1 This embodiment is based on the above embodiment, and the module for obtaining hemodialysis efficacy prediction data is specifically to obtain the original data of hemodialysis efficacy prediction from the hospital information system by collecting; the original data of hemodialysis efficacy prediction includes hemodialysis efficacy prediction data of other burn patients and real-time hemodialysis efficacy prediction data of burn patients; the hemodialysis efficacy prediction data of other burn patients and real-time hemodialysis efficacy prediction data of burn patients both include basic information data of patients, physiological data of patients, and hemodialysis data of patients; the hemodialysis efficacy data of other burn patients also includes hemodialysis efficacy results of patients; the hemodialysis efficacy results of patients are specifically short-term efficacy effects, medium-term efficacy effects, and long-term efficacy effects; the efficacy effects are specifically good, average, and poor; the basic information data of patients include age, gender, weight, underlying diseases, patient burn depth level, burn area percentage, and burn site; the physiological data of patients include heart rate, blood pressure, urine volume, electrolyte level, and body temperature; the hemodialysis data of patients include blood biochemical indexes, renal function indexes, blood routine indexes, coagulation function indexes, dialysis times, complications, and treatment records.

[0087] Example 3, see Figure 1 , Figure 2 and Figure 5 This embodiment is based on the above embodiment, and the raw data optimization module is specifically used to optimize the raw data for hemodialysis efficacy prediction, including raw data preprocessing, dynamic efficacy prediction weight allocation and dynamic feature weighting, to obtain hemodialysis efficacy prediction optimization data; specifically includes the following steps:

[0088] Raw data preprocessing is used to improve data quality, including data cleaning, data standardization and category data processing; specifically, it includes the following steps:

[0089] Data cleaning is to deal with missing values, outliers and duplicate values;

[0090] Data standardization is to use Z-score standardization method to standardize numerical data;

[0091] Categorical data processing is to encode categorical data using one-hot encoding and convert it into numerical features;

[0092] Dynamic efficacy prediction weight allocation is used to dynamically adjust the importance of different features in efficacy prediction according to the severity of burns, complications and physiological status of the patient, and obtain the total weight representing the prediction of the patient's hemodialysis efficacy; specifically, it includes the following steps:

[0093] The burn severity weight is calculated by using the area index function and the location correction factor; the formula used is as follows:

[0094] ;

[0095] In the formula, represents the burn severity weight, Indicates the percentage of body surface area burned. Indicates the depth of burns in patients. Indicates the burn site correction factor; the burn site correction factor is divided according to the burn site, the facial burn site correction factor is 1.8, the perineum burn site correction factor is 1.5, the back burn site correction factor is 1.2, the abdomen burn site correction factor is 1.3, and the limbs burn site correction factor is 1.0;

[0096] The complication weight is calculated based on the complication risk level using the following formula:

[0097] ;

[0098] In the formula, represents the complication weight of burn patients, It indicates a high-risk complication index, specifically 0.3. It represents the medium-risk complication index, specifically 0.2. It indicates a low-risk complication index, specifically 0.1;

[0099] The weight of the physiological index is calculated by considering the change of the physiological index; the formula used is as follows:

[0100] ;

[0101] In the formula, Indicates the weight of the physiological index at the current stage, k indicates the stage sensitivity coefficient, which controls the impact intensity of the burn treatment stage, and the value range is , Indicates the treatment stage adjustment factor, shock period: 1.2, infection period: 1.5, recovery period: 0.8, represents the clinical importance coefficient of the static deviation of the i-th physiological index, Represents the sensitivity coefficient of the dynamic change of the i-th physiological index, represents the monitoring value of the i-th physiological index of the burn patient at the last moment, represents the monitoring value of the i-th physiological index of the burn patient at the current moment, represents the normal range of the ith physiological index, represents the rate of change of the ith physiological index;

[0102] Dynamic weight fusion is to integrate various weight factors through hierarchical control and nonlinear fusion, which retains the medical interpretability of the weight system and ensures numerical stability. The formula used is as follows:

[0103] ;

[0104] In the formula, represents the total weight;

[0105] Dynamic feature weighting is used to obtain a weighted feature set as the input of the hemodialysis efficacy prediction model; the formula used is as follows:

[0106] ;

[0107] ;

[0108] In the formula, represents the weighting factor of the jth feature that is dynamically adjusted according to the patient's condition, represents the weighting factor designed based on the clinical importance of the feature itself, represents the weighted feature set, represents the actual value of the jth feature, and n represents the total number of features.

[0109] By performing the above operations, this solution innovatively proposes a dynamic weight allocation for prediction of therapeutic effects. By adjusting the weights of various indicators in real time, accurate predictions can be made based on the patient's current condition, avoiding the errors caused by fixed weights. The unique burn condition of each patient is comprehensively considered by dynamically adjusting the weights, which can accurately reflect the physiological and pathological status of burn patients at different stages, thereby improving the accuracy of hemodialysis efficacy prediction.

[0110] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. The hemodialysis efficacy prediction model module is used to obtain the hemodialysis efficacy prediction model, including designing a prediction task sharing layer, designing a short-term efficacy prediction layer, designing a mid-term efficacy prediction layer, designing a long-term efficacy prediction layer and designing a model total loss function; specifically including the following steps:

[0111] A prediction task sharing layer is designed to obtain shared features. Specifically, the long-term and short-term dependencies in the data are captured through the long short-term memory network, and the attention mechanism is introduced to capture important information in the data. The formula used is as follows:

[0112] ;

[0113] ;

[0114] In the formula, Indicates the shared features of the prediction task shared layer output, Indicates the unit runs the function, Indicates the hidden state at the previous moment, Indicates the state of the memory unit at the previous moment, represents the original shared features, represents the weight matrix of the attention mechanism, represents the bias vector of the attention mechanism, Indicates the accumulation of all relevant feature elements;

[0115] A short-term efficacy prediction layer is designed to predict the short-term efficacy of hemodialysis and determine the fluid balance of burn patients after dialysis. The formula used is as follows:

[0116] ;

[0117] ;

[0118] In the formula, Represents short-term characteristics, represents the batch normalization function, and Respectively represent the weight matrix and bias parameters of the short-term efficacy prediction layer, and They represent the weight matrix and bias parameters of the short-term efficacy prediction layer output respectively. It indicates the short-term prediction results of hemodialysis efficacy;

[0119] A mid-term efficacy prediction layer was designed to predict the mid-term efficacy of hemodialysis and determine the risk of complications after dialysis in burn patients. The formula used is as follows:

[0120] ;

[0121] ;

[0122] In the formula, Indicates mid-term characteristics, and Respectively represent the weight matrix and bias parameter of the mid-term efficacy prediction layer, and They represent the weight matrix and bias parameters output by the mid-term efficacy prediction layer, It indicates the mid-term prediction results of hemodialysis efficacy;

[0123] A long-term efficacy prediction layer is designed to predict the long-term efficacy of hemodialysis and judge the recovery of burn patients. Specifically, a gating mechanism is introduced. The formula used is as follows:

[0124] ;

[0125] ;

[0126] ;

[0127] In the formula, represents the gating signal, Represents long-term characteristics, and Respectively represent the weight matrix and bias parameter of the long-term efficacy prediction layer, and They represent the weight matrix and bias parameters of the long-term efficacy prediction layer output respectively. and Respectively represent the weight matrix and bias term parameters of the gating mechanism, It represents the long-term prediction results of hemodialysis efficacy;

[0128] Obtaining the model prediction results, specifically the hemodialysis efficacy prediction results are composed of the short-term prediction results of the hemodialysis efficacy, the mid-term prediction results of the hemodialysis efficacy and the long-term prediction results of the hemodialysis efficacy;

[0129] The total loss function of the design model is specifically based on the dynamic weight design model. The formula used is as follows:

[0130] ;

[0131] ;

[0132] In the formula, L represents the total loss function value of the model, represents the cross entropy loss function of each prediction layer, represents the mean absolute error of the prediction layer c, represents the dynamic weight of the prediction layer c in the total loss, represents the hyperparameters that control weight adjustment;

[0133] The prediction model training specifically uses the hemodialysis efficacy data of other burn patients as the training data of the model, calculates the loss through the total loss function of the model, and optimizes the model parameters using the gradient descent method to minimize the error, thereby obtaining a hemodialysis efficacy prediction model.

[0134] By performing the above operations, in order to address the technical problems that the existing traditional hemodialysis efficacy prediction model cannot flexibly respond to changes in patients' conditions at different treatment stages and lacks consideration of the comprehensive prediction of the entire hemodialysis efficacy, resulting in inaccurate and incomplete prediction results, this solution innovatively designs three independent prediction layers. Each prediction layer makes independent predictions for different efficacy stages, which can comprehensively and accurately evaluate the hemodialysis efficacy of burn patients, avoid the deviation of a single dimension, help doctors adjust treatment plans in a timely manner, and improve treatment effects, thereby ensuring comprehensive monitoring of the dialysis efficacy of burn patients and significantly improving the accuracy of hemodialysis efficacy prediction.

[0135] Example 5, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. The model effect improvement module specifically obtains the optimal hyperparameter combination of the hemodialysis efficacy prediction model by improving the particle swarm optimization algorithm to obtain the hemodialysis efficacy prediction model after the effect is improved, and includes the following steps:

[0136] Initialize the particle swarm, specifically randomly generate the initial position and initial velocity of the particles and calculate the individual fitness value f of the particles i ; The performance of the hemodialysis efficacy prediction model established based on the individual position is used as the individual fitness value;

[0137] Calculate the particle search optimization factor to adjust the particle search parameters. The formula used is as follows:

[0138] ;

[0139] ;

[0140] In the formula, represents the average distance between the i-th particle and all other particles in the k-th iteration, N represents the total number of particles, represents the position of the ith particle in the dth dimension in the kth iteration, represents the position of the jth particle in the dth dimension in the kth iteration, D represents the number of dimensions of the solution space for each particle, represents the weight parameter used to balance the distance, represents the particle search optimization factor in the kth iteration, represents the weight coefficient of the particle in the dth dimension, represents the average distance between the global optimal particle and other particles in the kth iteration, represents the minimum distance between particles in the kth iteration, represents the maximum inter-particle distance in the kth iteration;

[0141] Adaptively adjust particle parameters, specifically adjust inertia weight and individual learning factor according to particle search optimization factor and group learning factor , the formula used is as follows:

[0142] ;

[0143] ;

[0144] ;

[0145] In the formula, represents the inertia weight of the kth iteration, Represents the parameter for adjusting the oscillation amplitude, with a value range of [0,1], used to control the amplitude of the distance jump. represents the individual learning factor of the kth iteration, represents the population learning factor of the kth iteration, Represents a constant term to prevent the parameter from tending to zero;

[0146] Particle update, the formula used is as follows:

[0147] ;

[0148] In the formula, Indicates The velocity of the ith particle in the dth dimension in the iteration, Indicates The velocity of the ith particle in the dth dimension in the iteration, Indicates The local optimal position of individual particles in the iteration, represents the global optimal position of the particle, and represents a random number in the range [0,1], Indicates The position of the i-th particle in the d-th dimension in the iteration;

[0149] Accepting inferior particles, specifically by calculating the inferior particle acceptance probability, the inferior particles are accepted using the following formula:

[0150] ;

[0151] In the formula, Indicates The acceptance probability of the i-th particle in the iteration is, represents the maximum number of iterations, represents the adjustment parameter of the acceptance probability, represents the particle fitness function;

[0152] Get the optimal hyperparameter combination, specifically when the individual fitness value of the particle f i When the value is higher than the fitness threshold and the maximum number of iterations is reached, the search is terminated and the global optimal position of the individual particle is obtained. The global optimal position of the individual particle specifically refers to the optimal hyperparameter combination of the hemodialysis efficacy prediction model, and the optimal hyperparameter combination of the hemodialysis efficacy prediction model is obtained;

[0153] Obtaining an optimized prediction model, specifically adjusting the hyperparameters of the hemodialysis efficacy prediction model according to the optimal hyperparameter combination of the hemodialysis efficacy prediction model, to obtain a hemodialysis efficacy prediction model with improved effect.

[0154] By performing the above operations, in view of the technical problems that the existing prediction models have inappropriate built-in parameter settings and the model optimization algorithm has weak ability to obtain the global optimal solution, which leads to inaccurate final model prediction results, this solution designs particle search optimization factors and inferior particle acceptance probability strategies to enhance the global optimal solution search and local search capabilities, thereby achieving the optimal hyperparameter combination of the hemodialysis efficacy prediction model, significantly improving the accuracy of the model output results, and realizing accurate prediction of the hemodialysis efficacy of burn patients.

[0155] Example 6, see Figure 1 This embodiment is based on the above embodiment. The personalized hemodialysis efficacy prediction module specifically inputs the real-time burn patient hemodialysis efficacy prediction data into the hemodialysis efficacy prediction model after the effect is improved to obtain the hemodialysis efficacy prediction result, and assists doctors in making scientific decisions based on the hemodialysis efficacy prediction result.

[0156] 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.

[0157] 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.

[0158] 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 big data-based hemodialysis efficacy prediction system for burn patients, characterized by: It includes a module for obtaining hemodialysis efficacy prediction data, a module for optimizing raw data, a module for building a hemodialysis efficacy prediction model, a module for improving model effects, and a module for personalized hemodialysis efficacy prediction; The module for obtaining hemodialysis efficacy prediction data is to obtain the original data of hemodialysis efficacy prediction by collecting data from the hospital information system; The raw data optimization module is used to optimize the raw data for hemodialysis efficacy prediction, firstly improve the data quality by preprocessing the raw data, then enter the dynamic efficacy prediction weight allocation stage, dynamically adjust the feature weight according to the patient's burn severity, complication risk and physiological index changes, obtain the total weight representing the patient's hemodialysis efficacy prediction, and perform dynamic feature weighting to obtain the hemodialysis efficacy prediction optimization data; The module for constructing a hemodialysis efficacy prediction model is used to obtain a hemodialysis efficacy prediction model, specifically by designing short-term, medium-term and long-term efficacy prediction layers, each prediction layer performs independent predictions for different efficacy stages, and combines the total loss function of the model for training, thereby obtaining a hemodialysis efficacy prediction model; The model effect improvement module specifically improves the particle swarm optimization algorithm by designing a particle search optimization factor and a low-quality particle acceptance probability strategy, obtains the optimal hyperparameter combination of the hemodialysis efficacy prediction model, and adjusts the model hyperparameters according to the optimal hyperparameter combination to obtain the hemodialysis efficacy prediction model with improved effect; The personalized hemodialysis efficacy prediction module specifically inputs real-time data into the model to obtain hemodialysis efficacy prediction results, and assists doctors in making scientific decisions based on the prediction results.

2. The big data-based hemodialysis efficacy prediction system for burn patients according to claim 1 is characterized by: The original data optimization module specifically includes the following steps: Raw data preprocessing is used to improve data quality, including data cleaning, data standardization and category data processing; Dynamic efficacy prediction weight allocation is used to dynamically adjust the importance of different features in efficacy prediction according to the severity of burns, complications and physiological status of the patient, and obtain the total weight representing the patient's hemodialysis efficacy prediction ; Dynamic feature weighting is used to obtain a weighted feature set as the input of the hemodialysis efficacy prediction model; the formula used is as follows: ； ； In the formula, represents the weighting factor of the jth feature that is dynamically adjusted according to the patient's condition, represents the weighting factor designed based on the clinical importance of the feature itself, represents the weighted feature set, represents the actual value of the jth feature, and n represents the total number of features.

3. The big data-based hemodialysis efficacy prediction system for burn patients according to claim 1 is characterized by: The dynamic efficacy prediction weight allocation specifically includes the following steps: The burn severity weight is calculated by using the area index function and the location correction factor; the formula used is as follows: ； In the formula, represents the burn severity weight, Indicates the percentage of body surface area burned. Indicates the depth of burns in patients. Indicates the correction factor for burn location; The complication weight is calculated based on the complication risk level using the following formula: ； In the formula, represents the complication weight of burn patients, Represents high-risk complication index, Indicates the intermediate-risk complication index. Indicates low risk complication index; The weight of the physiological index is calculated by considering the change of the physiological index; the formula used is as follows: ； In the formula, represents the weight of the physiological index in the current stage, k represents the stage sensitivity coefficient, and controls the impact intensity of the burn treatment stage. represents the treatment phase adjustment factor, represents the clinical importance coefficient of the static deviation of the i-th physiological index, Represents the sensitivity coefficient of the dynamic change of the i-th physiological index, represents the monitoring value of the i-th physiological index of the burn patient at the last moment, represents the monitoring value of the i-th physiological index of the burn patient at the current moment, represents the normal range of the ith physiological index, represents the rate of change of the ith physiological index; Dynamic weight fusion is to integrate various weight factors through hierarchical control and nonlinear fusion, which retains the medical interpretability of the weight system and ensures numerical stability. The formula used is as follows: ； In the formula, Represents the total weight.

4. The system for predicting the therapeutic effect of hemodialysis on burn patients based on big data according to claim 1 is characterized by: The hemodialysis efficacy prediction model building module is used to obtain the hemodialysis efficacy prediction model, and specifically includes the following steps: A prediction task sharing layer is designed to obtain shared features. Specifically, the long-term and short-term dependencies in the data are captured through the long short-term memory network, and the attention mechanism is introduced to capture important information in the data. The formula used is as follows: ； ； In the formula, Indicates the shared features of the prediction task shared layer output, Indicates the unit runs the function, Indicates the hidden state at the previous moment, Indicates the state of the memory unit at the previous moment, represents the original shared features, represents the weight matrix of the attention mechanism, represents the bias vector of the attention mechanism, Indicates the accumulation of all relevant feature elements; A short-term efficacy prediction layer is designed to predict the short-term efficacy of hemodialysis and determine the fluid balance of burn patients after dialysis. The formula used is as follows: ； In the formula, Represents short-term characteristics, and They represent the weight matrix and bias parameters of the short-term efficacy prediction layer output respectively. It indicates the short-term prediction results of hemodialysis efficacy; A mid-term efficacy prediction layer was designed to predict the mid-term efficacy of hemodialysis and determine the risk of complications after dialysis in burn patients. The formula used is as follows: ； In the formula, Indicates mid-term characteristics, and They represent the weight matrix and bias parameters output by the mid-term efficacy prediction layer, It indicates the mid-term prediction results of hemodialysis efficacy; A long-term efficacy prediction layer is designed to predict the long-term efficacy of hemodialysis and judge the recovery of burn patients. Specifically, a gating mechanism is introduced. The formula used is as follows: ； In the formula, represents the gating signal, Represents long-term characteristics, and They represent the weight matrix and bias parameters of the long-term efficacy prediction layer output respectively. It represents the long-term prediction results of hemodialysis efficacy; Obtaining the model prediction results, specifically the hemodialysis efficacy prediction results are composed of the short-term prediction results of the hemodialysis efficacy, the mid-term prediction results of the hemodialysis efficacy and the long-term prediction results of the hemodialysis efficacy; The total loss function of the design model is specifically based on the dynamic weight design model. The formula used is as follows: ； ； In the formula, L represents the total loss function value of the model, represents the cross entropy loss function of each prediction layer, represents the mean absolute error of the prediction layer c, represents the dynamic weight of the prediction layer c in the total loss, represents the hyperparameters that control weight adjustment; The prediction model training specifically uses the hemodialysis efficacy data of other burn patients as the training data of the model, calculates the loss through the total loss function of the model, and optimizes the model parameters using the gradient descent method to minimize the error, thereby obtaining a hemodialysis efficacy prediction model.

5. The system for predicting the therapeutic effect of hemodialysis on burn patients based on big data according to claim 1 is characterized in that: The model effect improvement module specifically includes the following steps: Initialize the particle swarm, specifically randomly generate the initial position and initial velocity of the particles and calculate the individual fitness value f of the particles i ; The performance of the hemodialysis efficacy prediction model established based on the individual position is used as the individual fitness value; Calculate the particle search optimization factor to adjust the particle search parameters. The formula used is as follows: ； ； In the formula, represents the average distance between the i-th particle and all other particles in the k-th iteration, N represents the total number of particles, represents the position of the ith particle in the dth dimension in the kth iteration, represents the position of the jth particle in the dth dimension in the kth iteration, D represents the number of dimensions of the solution space for each particle, represents the weight parameter used to balance the distance, represents the particle search optimization factor in the kth iteration, represents the weight coefficient of the particle in the dth dimension, represents the average distance between the global optimal particle and other particles in the kth iteration, represents the minimum distance between particles in the kth iteration, represents the maximum inter-particle distance in the kth iteration; Adaptively adjust particle parameters, specifically adjust inertia weight and individual learning factor according to particle search optimization factor and group learning factor , the formula used is as follows: ； ； ； In the formula, represents the inertia weight of the kth iteration, Represents the parameter for adjusting the oscillation amplitude, which is used to control the amplitude of the distance jump. represents the individual learning factor of the kth iteration, represents the population learning factor of the kth iteration, Represents a constant term to prevent the parameter from tending to zero; Particle update, specifically through inertia weight, individual learning factor and group learning factor Update particle velocity and particle position ; Accepting inferior particles, specifically by calculating the inferior particle acceptance probability, the inferior particles are accepted using the following formula: ； In the formula, Indicates The acceptance probability of the i-th particle in the iteration is, represents the maximum number of iterations, represents the adjustment parameter of the acceptance probability, represents the particle fitness function, Indicates The position of the i-th particle in the d-th dimension in the iteration; Get the optimal hyperparameter combination, specifically when the individual fitness value of the particle f i When the value is higher than the fitness threshold and the maximum number of iterations is reached, the search is terminated and the global optimal position of the individual particle is obtained. The global optimal position of the individual particle specifically refers to the optimal hyperparameter combination of the hemodialysis efficacy prediction model, and the optimal hyperparameter combination of the hemodialysis efficacy prediction model is obtained; Obtaining an optimized prediction model, specifically adjusting the hyperparameters of the hemodialysis efficacy prediction model according to the optimal hyperparameter combination of the hemodialysis efficacy prediction model, to obtain a hemodialysis efficacy prediction model with improved effect.

6. The big data-based hemodialysis efficacy prediction system for burn patients according to claim 1, characterized in that: The personalized hemodialysis efficacy prediction module specifically inputs the real-time burn patient hemodialysis efficacy prediction data into the improved hemodialysis efficacy prediction model to obtain the hemodialysis efficacy prediction result, and assists doctors in making scientific decisions based on the hemodialysis efficacy prediction result.

7. The big data-based hemodialysis efficacy prediction system for burn patients according to claim 1, characterized in that: The module for obtaining hemodialysis efficacy prediction data specifically obtains hemodialysis efficacy prediction raw data from the hospital information system by collecting data; the hemodialysis efficacy prediction raw data includes hemodialysis efficacy prediction data of other burn patients and real-time hemodialysis efficacy prediction data of burn patients; The other burn patients' hemodialysis efficacy prediction data and the real-time burn patients' hemodialysis efficacy prediction data both include patients' basic information data, patients' physiological data, and patients' hemodialysis data; the other burn patients' hemodialysis efficacy data also include patients' hemodialysis efficacy results.

Images