Enteral nutrition intolerance screening evaluation and management system

Abstract

The invention discloses an enteral nutrition intolerance screening evaluation and management system, which belongs to the technical field of medical monitoring and comprises an information input module, a knowledge updating module, an information screening module, a risk scoring module, a scheme recommendation module, a feeding management module, a data analysis module, a user management module, a report generation module and a remote monitoring module. According to the method, the accuracy and reliability of risk assessment can be improved, the screening rate of enteral feeding intolerance high-risk groups is fully improved, the working efficiency of medical staff is improved, standardized and normalized feeding intolerance screening is achieved, the effectiveness and pertinence of treatment can be improved, it is ensured that recommended schemes are comprehensive and balanced, manual intervention is reduced, and the risk assessment efficiency is improved. Powerful decision support is provided for doctors, the actual requirements of patients are better met, the treatment effect and the satisfaction degree of the patients are improved, and the treatment compliance of the patients is enhanced.

Description

Technical Field

[0001] The present invention relates to the field of medical monitoring technology, and in particular to an enteral nutrition intolerance screening, evaluation and management system. Background Art

[0002] Severely ill patients are often hospitalized due to respiratory failure, renal insufficiency, electrolyte metabolism disorder, acid-base imbalance, severe trauma, shock, severe infection, etc. The condition is complicated and the body's stress response is strong, manifested as hypercatabolism and weakened anabolism, resulting in increased energy and protein consumption. In addition, the reduced input or insufficient supply of nutrients leads to gastrointestinal ischemia and hypoxia, and adverse reactions occur, and the patient's nutritional status will decline sharply. Once malnutrition occurs, it will lead to decreased organ function, intestinal structure and barrier function damage, reduced immune function, muscle atrophy, respiratory muscle weakness, and difficulty in weaning from the mechanical ventilation patient. Enteral nutrition is currently the preferred nutritional treatment for critically ill patients, but enteral feeding intolerance is one of its common complications. There is no unified standard for enteral feeding intolerance. Therefore, objective and standardized risk assessment has become the first step in preventing enteral nutrition intolerance in critically ill patients. Risk assessment aims to screen the risk of enteral feeding intolerance in patients with enteral nutrition at an early stage, and provide a basis for identifying high-risk groups of enteral feeding intolerance and taking timely measures to prevent enteral feeding intolerance.

[0003] The accuracy and reliability of risk assessment of the existing enteral nutrition intolerance screening, assessment and management system are low, which reduces the screening rate of high-risk groups for enteral feeding intolerance. In addition, the work efficiency of medical staff is low, and standardized and normalized feeding intolerance screening cannot be achieved. In addition, the effectiveness and pertinence of the treatment of the existing enteral nutrition intolerance screening, assessment and management system are low, which increases the situation of manual intervention, cannot meet the actual needs of patients, and reduces the treatment effect and patient satisfaction. Therefore, we propose an enteral nutrition intolerance screening, assessment and management system. Summary of the invention

[0004] The purpose of the present invention is to solve the defects in the prior art and to propose an enteral nutrition intolerance screening, evaluation and management system.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] An enteral nutrition intolerance screening, evaluation and management system, comprising an information entry module, a knowledge update module, an information screening module, a risk scoring module, a program recommendation module, a feeding management module, a data analysis module, a user management module, a report generation module and a remote monitoring module;

[0007] The information entry module is used to enter the patient's basic information, medical history, nutritional status and laboratory test results;

[0008] The knowledge updating module is used to update the latest medical literature and research results and associate new knowledge with existing data;

[0009] The information screening module is used to screen the input data and identify key factors affecting enteral nutrition tolerance;

[0010] The risk scoring module is used to combine the multi-dimensional data of the patient to perform enteral nutrition intolerance risk scoring on the patient;

[0011] The regimen recommendation module generates a personalized enteral nutrition treatment regimen based on the patient's risk score and the screened patient information;

[0012] The feeding management module is used to monitor the patient's nutritional intake and digestion in real time, and optimize feeding time and nutritional composition ratio;

[0013] The data analysis module is used to conduct comprehensive analysis of all the data collected and generate a detailed data analysis report;

[0014] The user management module is used to manage the permissions and role allocation of system users and record user operation logs;

[0015] The report generation module is used to generate various types of reports and provide electronic and paper report output;

[0016] The remote monitoring module is used to monitor the patient's health status and enteral nutrition treatment process in real time.

[0017] As a further solution of the present invention, the specific steps of associating the new knowledge with the existing data in the knowledge updating module are as follows:

[0018] Step 1: Regularly crawl the latest research literature from various medical databases and medical journals through crawler technology, clean the captured literature data, remove duplicate literature through text similarity algorithm, and then use natural language processing to pre-process the cleaned literature data and preliminarily classify and label the literature content;

[0019] Step 2: Use the pre-trained BERT model to extract key entities from the literature, extract entity relationships through rule matching methods, and generate "entity 1-relationship-entity 2" triples. Add each group of extracted entities as a node to the Neo4j graph database, and establish edges between each group of nodes based on the extracted relationships;

[0020] Step 3: Check whether the new nodes and edges are repeated with the existing nodes and edges. If so, delete the repeated nodes and edges, merge the new knowledge with the existing knowledge, update the knowledge graph, and then record the content and time of each update, calculate the cosine similarity of each group of nodes, and then use the Apriori algorithm to mine the association rules between the new knowledge and the existing data.

[0021] As a further embodiment of the present invention, the medical database in step 1 specifically includes PubMed, Cochrane Library, Embase, Web of Science, Scopus, ClinicalTrials.gov, MEDLINE, Google Scholar, Ovid, CINAHL, LILACS, PsycINFO and IEEE Xplore.

[0022] As a further solution of the present invention, the specific steps of identifying key factors of the information screening module are as follows:

[0023] Step 1: Remove character and format errors in the entered patient data through regular expressions and rule matching, then fill in the missing values ​​in the data through the mean filling method, and then normalize the processed patient data;

[0024] Step 2: Retrieve corresponding entities and relationships in the knowledge graph based on the processed patient data, match the identified entities with the nodes in the knowledge graph, and extract feature information corresponding to enteral nutrition tolerance;

[0025] Step 3: Preprocess the corresponding historical patient data and divide it into a training set and a test set. Then, build and initialize a set of screening and recognition models through the TensorFlow deep learning framework, and then input the training set data into the screening and recognition model;

[0026] Step 4: Divide the training set into small batches of data and input them into the model for forward propagation. The screening recognition model calculates the predicted value based on the current parameters and the loss value of the loss function. Based on the chain rule, the screening recognition model transfers the loss value backward from the output layer and updates the model parameters according to the gradient of the loss function.

[0027] Step 5: After each training cycle, use the test set to verify the screening recognition model and evaluate its performance on unseen data. Adjust the hyperparameters of the screening recognition model based on the evaluation results. If the model performance no longer improves or begins to decline, terminate the training early. Otherwise, continue to train the screening recognition model until the preset number of training times is reached.

[0028] Step 6: Input the latest patient feature information into the screening and recognition model for forward propagation, and output the prediction results. Analyze the output results of the model, identify key factors, important clinical features and potential risk points, and then generate visual charts to display the prediction results and key factors. At the same time, regularly collect expert opinions, adjust feature selection and model parameters, and retrain and optimize the model.

[0029] As a further solution of the present invention, the specific steps of the risk scoring module for enteral nutrition intolerance risk scoring are as follows:

[0030] Step ①: Represent the patient's health status as a tree node. Each node contains the patient's current clinical characteristics and potential risk points to build a set of search trees, determine the actions that can be taken in different states, and use the possible actions from one state to another as the edges of the tree. Then, use the patient's current state as the root node of the search tree.

[0031] Step ②: Use the upper confidence interval algorithm to select the child node with the highest UCB value, expand the selected node, generate all possible child nodes, start from the newly expanded node, perform random action simulation, and evaluate the effects of different nutrition plans on patients, and record the simulation results;

[0032] Step 3: Backtrack the simulation results and update the number of visits and average score of each node on the path. Check the number of visits and score changes of the nodes in the tree to determine whether to continue the simulation. If the score change threshold does not exceed the preset convergence threshold, stop the search. Otherwise, continue the search.

[0033] Step ④: Based on the simulation results of the terminal nodes, calculate the patient's overall risk score, and then generate a visual chart through Matplotlib to display the patient's risk score and influencing factors. Based on the paths and nodes in the search tree, output the clinical characteristics and actions that have the greatest impact on the final risk score.

[0034] As a further solution of the present invention, the specific steps of generating the enteral nutrition treatment plan in the plan recommendation module are as follows:

[0035] Step I: The screened patient information and risk scores are used as antigens, and then the initial antibody group is generated by random initialization method, each antibody represents a possible enteral nutrition treatment plan;

[0036] Step II: Use the affinity function to calculate the affinity value of each antibody, evaluate its matching degree with the patient data, select antibodies with affinity higher than the preset selection threshold for clonal expansion, generate a clonal antibody population, and then perform random perturbations and cross-recombination operations on the cloned antibodies to generate a variety of treatment options;

[0037] Step III: Calculate the affinity value of the antibody after clonal mutation, select antibodies with affinity greater than the preset affinity threshold to enter the mature antibody group, optimize the mature antibody, and repeat the antibody iteration until the maximum number of iterations is reached;

[0038] Step IV: Output the antibody with the highest affinity as the final recommended personalized enteral nutrition treatment plan, and record and save the detailed information of the plan. Clinical experts evaluate the recommended enteral nutrition treatment plan, collect expert feedback, analyze the advantages and disadvantages of the recommended plan and adjust the plan.

[0039] As a further solution of the present invention, the specific steps of optimizing the feeding time and nutritional composition ratio of the feeding management module are as follows:

[0040] Step 1: According to the recommended enteral nutrition treatment plan, set the initial feeding schedule and nutrient composition ratio, set the initial temperature, annealing rate and termination temperature, and calculate the objective function value of the current feeding plan;

[0041] Step 2: Fine-tune the current feeding schedule and nutrient ratio, calculate the objective function value of the adjusted feeding plan, and compare it with the objective function value of the current solution. If the objective function value of the latest feeding plan is less than the minimum objective function value of the original plan, it is accepted unconditionally. Otherwise, the acceptance probability is calculated according to the Metropolis criterion, and based on the acceptance probability, a random number is used to decide whether to accept the latest feeding plan.

[0042] Step 3: After each iteration, based on the preset annealing rate, the temperature is gradually reduced using an exponential descent strategy. If the current temperature is lower than the preset termination temperature, the iteration is stopped. Otherwise, the scheme is iterated repeatedly. After all iterations, the feeding scheme with the smallest objective function value is selected as the final optimal solution, and the optimal feeding scheme is recorded and saved.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. The present invention represents the patient's health status as a tree node, each node contains the patient's current clinical characteristics and potential risk points, so as to construct a set of search trees, determine the actions that can be taken in different states, and use the possible actions from one state to another as the edges of the tree. Then, the patient's current state is used as the root node of the search tree, and the child node with the highest UCB value is selected. The selected node is expanded to generate all possible child nodes. Starting from the newly expanded node, random action simulation is performed, and the effects of different nutritional programs on patients are evaluated. The simulation results are recorded, the simulation results are backtracked, and the number of visits and the average score of each node on the path are updated. Check the number of visits and score changes of the nodes in the tree to determine whether the simulation needs to continue. If the score change threshold does not exceed the preset convergence threshold, stop the search, otherwise continue the search. Based on the simulation results of the terminal nodes, calculate the patient's overall risk score, and then generate a visual chart through Matplotlib to display the patient's risk score and influencing factors. Based on the paths and nodes in the search tree, output the clinical characteristics and actions that have the greatest impact on the final risk score. This can improve the accuracy and reliability of risk assessment, fully increase the screening rate of high-risk groups for enteral feeding intolerance, improve the work efficiency of medical staff, and achieve standardized and normalized feeding intolerance screening.

[0045] 2. The present invention uses the screened patient information and risk scores as antigens, and then generates an initial antibody group through a random initialization method, each antibody represents a possible enteral nutrition treatment plan, uses an affinity function to calculate the affinity value of each antibody, evaluates its matching degree with the patient data, selects antibodies with affinity higher than a preset selection threshold for cloning and expansion, generates a clonal antibody group, and then performs random perturbations and cross-recombination operations on the cloned antibodies to generate diversified treatment plans, calculates the affinity value of the antibody after clonal mutation, selects antibodies with affinity greater than the preset affinity threshold to enter the mature antibody group, and then performs cross-recombination operations on the mature antibodies. Optimize and repeat antibody iteration until the maximum number of iterations is reached, and output the antibody with the highest affinity as the final recommended personalized enteral nutrition treatment plan. At the same time, record and save the detailed information of the plan. Clinical experts evaluate the recommended enteral nutrition treatment plan, collect expert feedback, analyze the advantages and disadvantages of the recommended plan and adjust the plan, which can improve the effectiveness and pertinence of the treatment, ensure that the recommended plan is comprehensive and balanced, reduce manual intervention, provide strong decision-making support for doctors, better meet the actual needs of patients, improve treatment effects and patient satisfaction, and enhance patient treatment compliance. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] 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.

[0047] Figure 1 This is a system block diagram of an enteral nutrition intolerance screening, evaluation and management system proposed by the present invention. DETAILED DESCRIPTION

[0048] The technical solutions in the embodiments of the present invention will be described clearly and completely 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.

[0049] Example 1

[0050] Reference Figure 1 , an enteral nutrition intolerance screening, assessment and management system, including an information entry module, a knowledge update module, an information screening module, a risk scoring module, a plan recommendation module, a feeding management module, a data analysis module, a user management module, a report generation module and a remote monitoring module.

[0051] The information entry module is used to enter the patient's basic information, medical history, nutritional status and laboratory test results; the knowledge update module is used to update the latest medical literature and research results and associate new knowledge with existing data.

[0052] Specifically, the latest research literature is regularly captured from various medical databases and medical journals through crawler technology, the captured literature data is cleaned, and duplicate literature is removed through text similarity algorithm. Then, natural language processing is used to preprocess the cleaned literature data, and the literature content is preliminarily classified and labeled. The pre-trained BERT model is used to extract key entities from the literature, and the entity relationship is extracted through rule matching method, and "entity 1-relationship-entity 2" triples are generated. Each group of extracted entities is added as a node to the Neo4j graph database, and edges are established between each group of nodes according to the extracted relationship. It is checked whether the new nodes and edges are repeated with the existing nodes and edges. If so, the duplicate nodes and edges are deleted, the new knowledge is merged with the existing knowledge, and the knowledge graph is updated. Then, the content and time of each update are recorded, the cosine similarity of each group of nodes is calculated, and the association rules between the new knowledge and the existing data are mined through the Apriori algorithm.

[0053] It should be further explained that the medical databases specifically include PubMed, Cochrane Library, Embase, Web of Science, Scopus, ClinicalTrials.gov, MEDLINE, Google Scholar, Ovid, CINAHL, LILACS, PsycINFO and IEEE Xplore.

[0054] The information screening module is used to screen the input data and identify the key factors that affect enteral nutrition tolerance.

[0055] Specifically, regular expressions and rule matching are used to remove characters and format errors in the entered patient data, and then the missing values ​​in the data are filled by the mean filling method, and then the processed patient data is normalized, and the corresponding entities and relationships are retrieved in the knowledge graph according to the processed patient data, and the identified entities are matched with the nodes in the knowledge graph, and the feature information corresponding to enteral nutrition tolerance is extracted, and the corresponding historical patient data is preprocessed and divided into training sets and test sets. Then, it is constructed and initialized through the TensorFlow deep learning framework, and then the training set data is input into the screening and recognition model, and the training set is divided into small batches of data and input into the model for forward propagation. The screening and recognition model calculates the predicted value according to the current parameters, and calculates the loss value of the loss function. Based on the chain rule, the screening and recognition model transfers the loss value backward from the output layer, and updates the model parameters according to the gradient of the loss function. After each training cycle, the test set is used to verify the screening and recognition model to evaluate its performance on unseen data. The hyperparameters of the screening and recognition model are adjusted according to the evaluation results. If the model performance no longer improves or begins to decline, the training is terminated early. Otherwise, the screening and recognition model is continued to be trained until the preset number of training times is reached. The latest patient feature information is input into the screening and recognition model for forward propagation, and the prediction results are output. The output results of the model are analyzed to identify key factors, important clinical features and potential risk points. Then, a visual chart is generated to display the prediction results and key factors. At the same time, expert opinions are collected regularly, feature selection and model parameters are adjusted, and the model is retrained and optimized.

[0056] The risk scoring module is used to combine the patient's multidimensional data to perform enteral nutrition intolerance risk scoring on the patient.

[0057] Specifically, the patient's health status is represented as a tree node, each node contains the patient's current clinical characteristics and potential risk points, so as to construct a set of search trees, determine the actions that can be taken in different states, and use the possible actions from one state to another as the edges of the tree. Then, the patient's current state is used as the root node of the search tree, and the upper confidence interval algorithm is used to select the child node with the highest UCB value. The selected node is expanded to generate all possible child nodes. Starting from the newly expanded node, random action simulation is performed, and the impact of different nutritional plans on patients is evaluated. The simulation results are recorded, the simulation results are backtracked, and the number of visits and average scores of each node on the path are updated. The number of visits and score changes of the nodes in the tree are checked to determine whether the simulation needs to be continued. If the score change threshold does not exceed the preset convergence threshold, the search is stopped, otherwise the search is continued. Based on the simulation results of the terminal node, the patient's overall risk score is calculated, and then a visual chart is generated through Matplotlib to display the patient's risk score and influencing factors, and based on the path and nodes in the search tree, the clinical characteristics and actions that have the greatest impact on the final risk score are output.

[0058] It should be further explained that the risk scoring criteria specifically include abdominal distension, abdominal pain, diarrhea, nausea and vomiting.

[0059] Example 2

[0060] Reference Figure 1 , an enteral nutrition intolerance screening, assessment and management system, including an information entry module, a knowledge update module, an information screening module, a risk scoring module, a plan recommendation module, a feeding management module, a data analysis module, a user management module, a report generation module and a remote monitoring module.

[0061] The regimen recommendation module generates a personalized enteral nutrition treatment plan based on the patient's risk score and screened patient information.

[0062] Specifically, the screened patient information and risk scores are used as antigens, and then an initial antibody group is generated through a random initialization method. Each antibody represents a possible enteral nutrition treatment plan. The affinity function is used to calculate the affinity value of each antibody, and the degree of matching with the patient data is evaluated. Antibodies with an affinity higher than a preset selection threshold are selected for cloning and expansion to generate a clonal antibody group. The cloned antibodies are then subjected to random perturbation and cross-recombination operations to produce diversified treatment plans. The affinity values ​​of the antibodies after clonal mutation are calculated, and antibodies with an affinity greater than a preset affinity threshold are selected to enter the mature antibody group. The mature antibodies are then optimized, and the antibody iteration is repeated until the maximum number of iterations is reached. The antibody with the highest affinity is output as the final recommended personalized enteral nutrition treatment plan. At the same time, the detailed information of the plan is recorded and saved. Clinical experts evaluate the recommended enteral nutrition treatment plan, collect expert feedback, analyze the advantages and disadvantages of the recommended plan, and adjust the plan.

[0063] The feeding management module is used to monitor the patient's nutritional intake and digestion in real time, and optimize feeding time and nutritional composition ratio.

[0064] Specifically, according to the recommended enteral nutrition treatment plan, the initial feeding schedule and the ratio of nutrient components are set, and the initial temperature, annealing rate and termination temperature are set, the objective function value of the current feeding plan is calculated, and fine-tuning is performed on the current feeding schedule and the ratio of nutrient components, and the objective function value of the adjusted feeding plan is calculated and compared with the objective function value of the current solution. If the objective function value of the latest feeding plan is less than the minimum value of the objective function of the original plan, it is unconditionally accepted. Otherwise, the acceptance probability is calculated according to the Metropolis criterion, and based on the acceptance probability, a random number is used to decide whether to accept the latest feeding plan. After each iteration, based on the preset annealing rate, an exponential decline strategy is used to gradually reduce the temperature. If the current temperature is lower than the preset termination temperature, the iteration is stopped. Otherwise, the plan iteration is repeated. After all iterations, the feeding plan with the smallest objective function value is selected as the final optimal solution, and the optimal feeding plan is recorded and saved.

[0065] The data analysis module is used to conduct comprehensive analysis of all data collected and generate detailed data analysis reports; the user management module is used to manage the permissions and role assignments of system users and record user operation logs; the report generation module is used to generate various types of reports and provide electronic and paper report outputs; the remote monitoring module is used to monitor the patient's health status and enteral nutrition treatment process in real time.

Claims

1. An enteral nutrition intolerance screening, assessment and management system, characterized in that: It includes information entry module, knowledge update module, information screening module, risk scoring module, plan recommendation module, feeding management module, data analysis module, user management module, report generation module and remote monitoring module; The information entry module is used to enter the patient's basic information, medical history, nutritional status and laboratory test results; The knowledge updating module is used to update the latest medical literature and research results and associate new knowledge with existing data; The information screening module is used to screen the input data and identify key factors affecting enteral nutrition tolerance; The risk scoring module is used to combine the multi-dimensional data of the patient to perform enteral nutrition intolerance risk scoring on the patient; The regimen recommendation module generates a personalized enteral nutrition treatment regimen based on the patient's risk score and the screened patient information; The feeding management module is used to monitor the patient's nutritional intake and digestion in real time, and optimize feeding time and nutritional composition ratio; The data analysis module is used to conduct comprehensive analysis of all the data collected and generate a detailed data analysis report; The user management module is used to manage the permissions and role allocation of system users and record user operation logs; The report generation module is used to generate various types of reports and provide electronic and paper report output; The remote monitoring module is used to monitor the patient's health status and enteral nutrition treatment process in real time.

2. The enteral nutrition intolerance screening, assessment and management system according to claim 1, characterized in that: The specific steps of associating the new knowledge of the knowledge updating module with the existing data are as follows: Step 1: Regularly crawl the latest research literature from various medical databases and medical journals through crawler technology, clean the captured literature data, remove duplicate literature through text similarity algorithm, and then use natural language processing to pre-process the cleaned literature data and preliminarily classify and label the literature content; Step 2: Use the pre-trained BERT model to extract key entities from the literature, extract entity relationships through rule matching methods, and generate "entity 1-relationship-entity 2" triples. Add each group of extracted entities as a node to the Neo4j graph database, and establish edges between each group of nodes based on the extracted relationships. Step 3: Check whether the new nodes and edges are repeated with the existing nodes and edges. If so, delete the repeated nodes and edges, merge the new knowledge with the existing knowledge, update the knowledge graph, and then record the content and time of each update, calculate the cosine similarity of each group of nodes, and then use the Apriori algorithm to mine the association rules between the new knowledge and the existing data.

3. The enteral nutrition intolerance screening, assessment and management system according to claim 2, characterized in that: The medical databases described in step 1 specifically include PubMed, Cochrane Library, Embase, Web of Science, Scopus, ClinicalTrials.gov, MEDLINE, Google Scholar, Ovid, CINAHL, LILACS, PsycINFO and IEEEXplore.

4. The enteral nutrition intolerance screening, assessment and management system according to claim 2, characterized in that: The specific steps of identifying key factors of the information screening module are as follows: Step 1: Remove character and format errors in the entered patient data through regular expressions and rule matching, then fill in the missing values ​​in the data through the mean filling method, and then normalize the processed patient data; Step 2: Retrieve corresponding entities and relationships in the knowledge graph based on the processed patient data, match the identified entities with the nodes in the knowledge graph, and extract feature information corresponding to enteral nutrition tolerance; Step 3: Preprocess the corresponding historical patient data and divide it into a training set and a test set. Then, build and initialize a set of screening and recognition models through the TensorFlow deep learning framework, and then input the training set data into the screening and recognition model; Step 4: Divide the training set into small batches of data and input them into the model for forward propagation. The screening recognition model calculates the predicted value based on the current parameters and the loss value of the loss function. Based on the chain rule, the screening recognition model transfers the loss value backward from the output layer and updates the model parameters according to the gradient of the loss function. Step 5: After each training cycle, use the test set to verify the screening recognition model and evaluate its performance on unseen data. Adjust the hyperparameters of the screening recognition model based on the evaluation results. If the model performance no longer improves or begins to decline, terminate the training early. Otherwise, continue to train the screening recognition model until the preset number of training times is reached. Step 6: Input the latest patient feature information into the screening and recognition model for forward propagation, and output the prediction results. Analyze the output results of the model, identify key factors, important clinical features and potential risk points, and then generate visual charts to display the prediction results and key factors. At the same time, regularly collect expert opinions, adjust feature selection and model parameters, and retrain and optimize the model.

5. The enteral nutrition intolerance screening, assessment and management system according to claim 4, characterized in that: The specific steps of the risk scoring module for enteral nutrition intolerance risk scoring are as follows: Step ①: Represent the patient's health status as a tree node. Each node contains the patient's current clinical characteristics and potential risk points to build a set of search trees, determine the actions that can be taken in different states, and use the possible actions from one state to another as the edges of the tree. Then, use the patient's current state as the root node of the search tree. Step ②: Use the upper confidence interval algorithm to select the child node with the highest UCB value, expand the selected node, generate all possible child nodes, start from the newly expanded node, perform random action simulation, and evaluate the effects of different nutrition plans on patients, and record the simulation results; Step 3: Backtrack the simulation results and update the number of visits and average score of each node on the path. Check the number of visits and score changes of the nodes in the tree to determine whether to continue the simulation. If the score change threshold does not exceed the preset convergence threshold, stop the search. Otherwise, continue the search. Step ④: Based on the simulation results of the terminal nodes, calculate the patient's overall risk score, and then generate a visual chart through Matplotlib to display the patient's risk score and influencing factors. Based on the paths and nodes in the search tree, output the clinical characteristics and actions that have the greatest impact on the final risk score.

6. The enteral nutrition intolerance screening, assessment and management system according to claim 1, characterized in that: The specific steps of generating the enteral nutrition treatment plan in the plan recommendation module are as follows: Step I: The screened patient information and risk scores are used as antigens, and then the initial antibody group is generated by random initialization method, each antibody represents a possible enteral nutrition treatment plan; Step II: Use the affinity function to calculate the affinity value of each antibody, evaluate its matching degree with the patient data, select antibodies with affinity higher than the preset selection threshold for clonal expansion, generate a clonal antibody population, and then perform random perturbations and cross-recombination operations on the cloned antibodies to generate a variety of treatment options; Step III: Calculate the affinity value of the antibody after clonal mutation, select antibodies with affinity greater than the preset affinity threshold to enter the mature antibody group, optimize the mature antibody, and repeat the antibody iteration until the maximum number of iterations is reached; Step IV: Output the antibody with the highest affinity as the final recommended personalized enteral nutrition treatment plan, and record and save the detailed information of the plan. Clinical experts evaluate the recommended enteral nutrition treatment plan, collect expert feedback, analyze the advantages and disadvantages of the recommended plan and adjust the plan.

Images