Peritoneal dialysis scheme decision-making method and system based on multi-source data and topsis optimization

Abstract

The application provides a peritoneal dialysis scheme decision method and system based on multi-source data and TOPSIS optimization, and the system comprises a feature extraction module, a preprocessing module, a data access and processing module, an intelligent scheme decision module and a result display module; the method comprises the following sub-steps: S1, acquiring patient electronic medical records and extracting medical features and medical feature values therefrom; S2, preprocessing the medical features and medical feature values; S3, generating a multi-source peritoneal dialysis scheme; S4, generating clinical indexes; S5, constructing a decision matrix for TOPSIS multi-objective decision-making; the method generates a dialysis scheme for each patient through the intelligent scheme decision module in combination with multi-source data and a TOPSIS optimization algorithm; the method no longer simply relies on personal experience, and the precision and individualization degree of treatment are improved; the intelligent decision module generates an optimal peritoneal dialysis scheme by acquiring patient electronic medical record data.

Description

Technical Field

[0001] This invention relates to the field of medical and health technology, and in particular to a method and system for peritoneal dialysis protocol decision-making based on multi-source data and TOPSIS optimization. Background Technology

[0002] Peritoneal dialysis utilizes the peritoneum as a semipermeable membrane to regularly infuse prepared dialysis fluid into the patient's peritoneal cavity through a catheter using gravity. The peritoneal dialysis fluid is continuously replaced to remove metabolic products, toxic substances, and correct water and electrolyte imbalances. It is a dialysis treatment different from hemodialysis (HD) and is a common renal replacement therapy. With the increasing number of patients with chronic kidney disease, the clinical demand for peritoneal dialysis continues to rise. However, the traditional peritoneal dialysis treatment model has the following drawbacks: 1. The development of the treatment plan is highly dependent on experience: Traditional peritoneal dialysis treatment requires patients to go to the hospital for regular monitoring. The treatment plan is largely developed and adjusted based on the experience of medical staff. However, the individual conditions of different patients are complex, and the experience and energy of medical staff are limited. 2. Limited functionality of intelligent systems: While existing intelligent monitoring systems can monitor physiological parameters in real time, they lack targeted treatment suggestions and cannot meet the needs of personalized treatment. 3. High pressure on medical resources: Under the traditional model, patients need to come to the hospital frequently for monitoring, and medical staff need to manually adjust the treatment plan, which not only increases the burden of medical treatment for patients, but also exacerbates the operational pressure on medical staff and professional equipment, resulting in low efficiency of medical resource utilization. Summary of the Invention

[0003] The purpose of this invention is to address the shortcomings of existing technologies by proposing a peritoneal dialysis protocol decision-making method and system based on multi-source data and TOPSIS optimization.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: The peritoneal dialysis protocol decision-making system based on multi-source data and TOPSIS optimization includes a feature extraction module, a preprocessing module, a data access and processing module, an intelligent protocol decision-making module, and a results display module. The feature extraction module is used to extract medical features and medical feature values ​​from the patient's electronic medical record; The preprocessing module is used to perform preprocessing on medical feature values, such as data verification, data checking, and data sampling. The data access and processing module is used to acquire the feature data of the user to be decided and generate a multi-source peritoneal dialysis plan and corresponding clinical indicators. The intelligent solution decision module is used to construct a decision matrix to perform TOPSIS multi-objective decision-making and obtain the optimal peritoneal dialysis solution. The results display module is used to display the optimal peritoneal dialysis plan on the user interface of the peritoneal dialysis plan decision system optimized based on multi-source data and TOPSIS.

[0005] This invention also proposes a peritoneal dialysis protocol decision-making method based on multi-source data and TOPSIS optimization, comprising the following sub-steps: S1: Obtain the patient's electronic medical record and extract medical features and medical feature values ​​from it; S11: Obtain the patient's electronic medical record; The electronic medical records of patients awaiting decision-making are collected from the hospital's electronic medical record platform, and electronic medical records of patients who have been continuously undergoing peritoneal dialysis for more than 6 months are selected from them; the electronic medical records of these patients contain the corresponding peritoneal dialysis plans. The peritoneal dialysis protocol includes five protocol indicators: daily exchange frequency, infusion volume, dialysate type, dialysate concentration, and peritoneal retention time. S12: Extract medical features and medical feature values; The feature extraction module extracts medical features and medical feature values ​​from the patient electronic medical records selected in step S11; The medical characteristics include the patient's basic information, medical history, access information, blood tests, urinary tract tests, follow-up records, and other indicators; The medical feature values ​​are the specific numerical values ​​of each medical feature; Other indicators include assessment of dialysis adequacy, assessment of residual renal function (Kt / v, cCr), and incidence of peritonitis.

[0006] S2: Preprocessing of medical features and their values; S21: Perform data validation on each medical characteristic value; The preprocessing module presets the threshold range for each medical feature value, compares each medical feature value with the corresponding threshold range, and if it is within the threshold range, the medical feature value is determined to be normal data; if it is outside the threshold range, the medical feature value is determined to be abnormal data, and proceeds to step S22. S22: Perform data verification on abnormal data; The preprocessing module pushes abnormal data to the corresponding data review specialist via email or other means. The data review specialist receives the abnormal data, performs manual verification, and provides feedback on the verification results, which include verification passed and verification failed. The preprocessing module updates the abnormal data that passes the verification to normal data and removes the data that fails the verification. S23: Conduct regular data spot checks; The preprocessing module periodically samples medical features and their values, randomly selecting a set proportion of medical features and their values. These features and values ​​are then compared with the corresponding original data in the hospital's electronic medical record platform. If they match, the data is deemed correct; otherwise, an anomaly is reported and manual verification is performed, removing data that fails the verification. After preprocessing, a qualified electronic medical record is obtained.

[0007] S3: Generate a multi-source peritoneal dialysis plan; The data access and processing module acquires the feature data of the user to be decided in the hospital system, and generates multi-source peritoneal dialysis plans through three methods: machine prediction learning, similar patient case matching, and large model inference, including peritoneal dialysis plan 1, peritoneal dialysis plan 2, and peritoneal dialysis plan 3. S4: Generate clinical indicators; Based on peritoneal dialysis plans 1, 2 and 3 obtained in step S3, corresponding clinical indicators are calculated and generated respectively. The clinical indicators include Kt / V value, peritonitis risk probability, ultrafiltration volume value and cost per dialysis session. S41: Generate clinical indicators based on peritoneal dialysis protocol 1; The feature data of the user to be decided and peritoneal dialysis scheme 1 are used as joint inputs and respectively input to the pre-trained Kt / V prediction model, peritonitis risk prediction model and ultrafiltration volume prediction model, and respectively output the corresponding Kt / V value, peritonitis risk probability and ultrafiltration volume value. The system contains a corresponding cost price list. Based on the medical supplies (such as dialysis fluid), consumable devices, etc. included in peritoneal dialysis scheme 1, the cost of a single dialysis session is obtained by multiplying each item by the corresponding unit price in the cost price list and then summing the results. Record the Kt / V value for peritoneal dialysis regimen 1. The probability of peritonitis is Ultrafiltration rate value The cost of a single dialysis session is ; S42: Generate clinical indicators based on peritoneal dialysis protocol 2; The feature data of the user to be decided and peritoneal dialysis scheme 2 are used as joint inputs and fed into the pre-built peritonitis risk prediction model to output the peritonitis risk probability. If the patient case with the highest similarity obtained in step S3 has stored Kt / V values ​​and ultrafiltration volume values, then the Kt / V values ​​and ultrafiltration volume values ​​of the similar patients are directly used; if not, then the feature data of the user to be decided and peritoneal dialysis scheme 2 are used as joint inputs, and are respectively input into the pre-trained Kt / V prediction model and ultrafiltration volume prediction model, and output the corresponding Kt / V values ​​and ultrafiltration volume values; and based on the medical supplies (such as dialysis fluid), consumable devices, etc. included in peritoneal dialysis scheme 2, the costs are multiplied by the corresponding unit prices in the cost price list and then summed to obtain the cost of a single dialysis session; Record the Kt / V value for peritoneal dialysis regimen 2. The probability of peritonitis is Ultrafiltration rate value The cost of a single dialysis session is ; S43: Generate clinical indicators based on peritoneal dialysis protocol 3; The feature data of the user to be decided and peritoneal dialysis scheme 3 are used as joint inputs and fed into the pre-trained Kt / V prediction model, peritonitis risk prediction model and ultrafiltration volume prediction model, respectively, and output the corresponding Kt / V value, peritonitis risk probability and ultrafiltration volume value; based on the medical supplies (such as dialysis fluid) and consumable devices included in peritoneal dialysis scheme 3, the cost per dialysis session is obtained by multiplying them by the corresponding unit price in the cost price list and summing them. Record the Kt / V value for peritoneal dialysis regimen 3. The probability of peritonitis is Ultrafiltration rate value The cost of a single dialysis session is .

[0008] S5: Construct a decision matrix to perform TOPSIS multi-objective decision-making and obtain the optimal peritoneal dialysis plan; S51: Construct the decision matrix; Based on the specific values ​​of the clinical indicators corresponding to peritoneal dialysis plans 1, 2, and 3 obtained in step S4, a decision matrix is ​​constructed, denoted as decision matrix A; ; S52: Standardize the data in the decision matrix; The decision matrix B is obtained by standardizing the data in the decision matrix using the following standardization formula. ; in, These are the original data values ​​in decision matrix A. The data values ​​are standardized, n=3; ; S53: Obtain the positive ideal solution and the negative ideal solution; The positive ideal solution and the negative ideal solution are obtained by using the cosine method. The positive ideal solution is the maximum value of the data value of each column element in the decision matrix B, and the negative ideal solution is the minimum value of the data value of each column element in the decision matrix B. Let the ideal solution be... The negative ideal solution is ; ; S54: Select the optimal peritoneal dialysis plan; The distance between each data point in decision matrix B and the positive and negative ideal solutions is calculated using the following formula: ; in, For the j-th index, the positive ideal solution is... For the j-th index, the negative ideal solution is... Let the distance between the i-th peritoneal dialysis scheme and the ideal solution be... Let be the distance between the i-th peritoneal dialysis scheme and the negative ideal solution; The following formula is used to calculate the approximation of each peritoneal dialysis regimen to the ideal solution; ; in, To approximate the degree of similarity, ≤1, The closer the value is to 0, the better the corresponding peritoneal dialysis plan. The result Sort the sequence, where the maximum value in the sequence is... The corresponding peritoneal dialysis plan is the optimal peritoneal dialysis plan for the patient to be decided. The optimal peritoneal dialysis plan is rendered through the results display module and displayed on the user interface of the peritoneal dialysis plan decision-making system based on multi-source data and TOPSIS optimization.

[0009] Compared with the prior art, the beneficial effects of the present invention are as follows: This method uses an intelligent treatment plan decision-making module, combined with multi-source data and the TOPSIS optimization algorithm, to generate a dialysis plan for each patient; it no longer relies solely on personal experience, thus improving the accuracy and personalization of treatment. It adopts a multi-objective decision-making method based on TOPSIS, which integrates multiple clinical indicators such as Kt / V value, peritonitis risk, and ultrafiltration volume to generate peritoneal dialysis plans. This not only meets the needs of physiological parameter monitoring but also provides doctors with multi-dimensional decision support, helping them to develop more scientific treatment plans. This method acquires patients' electronic medical records and obtains the optimal peritoneal dialysis plan through an intelligent decision-making module; the system processes the data and provides dialysis plans, reducing the workload of doctors and optimizing the utilization efficiency of medical resources. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the steps of the peritoneal dialysis protocol decision-making method based on multi-source data and TOPSIS optimization according to the present invention. Detailed Implementation

[0011] To provide a further understanding of the purpose, structure, features, and functions of the present invention, detailed descriptions are provided below with reference to specific embodiments.

[0012] The peritoneal dialysis protocol decision-making system based on multi-source data and TOPSIS optimization includes a feature extraction module, a preprocessing module, a data access and processing module, an intelligent protocol decision-making module, and a results display module. The feature extraction module is used to extract medical features and medical feature values ​​from the patient's electronic medical record; The preprocessing module is used to perform preprocessing on medical feature values, such as data verification, data checking, and data sampling. The data access and processing module is used to acquire the feature data of the user to be decided and generate a multi-source peritoneal dialysis plan and corresponding clinical indicators. The intelligent solution decision module is used to construct a decision matrix to perform TOPSIS multi-objective decision-making and obtain the optimal peritoneal dialysis solution. The results display module is used to display the optimal peritoneal dialysis plan on the user interface of the peritoneal dialysis plan decision system optimized based on multi-source data and TOPSIS.

[0013] like Figure 1 As shown, this invention also proposes a peritoneal dialysis protocol decision-making method based on multi-source data and TOPSIS optimization, comprising the following sub-steps: S1: Obtain the patient's electronic medical record and extract medical features and medical feature values ​​from it; S11: Obtain the patient's electronic medical record; The electronic medical records of patients awaiting decision-making are collected from the hospital's electronic medical record platform, and electronic medical records of patients who have been continuously undergoing peritoneal dialysis for more than 6 months are selected from them; the electronic medical records of these patients contain the corresponding peritoneal dialysis plans. The peritoneal dialysis protocol includes five protocol indicators: daily exchange frequency, infusion volume, dialysate type, dialysate concentration, and peritoneal retention time. S12: Extract medical features and medical feature values; The feature extraction module extracts medical features and medical feature values ​​from the patient electronic medical records selected in step S11; The medical characteristics include the patient's basic information, medical history, access information, blood tests, urinary tract tests, follow-up records, and other indicators; The medical feature values ​​are the specific numerical values ​​of each medical feature; Specifically, the basic information includes gender, height, weight, and CKD (chronic kidney disease) stage; The medical history information includes the primary disease, family history, allergy history, comorbidities, complications, mode of initiation of dialysis, initial eGFR (estimated glomerular filtration rate obtained from the first test when the patient starts peritoneal dialysis treatment), initial urine output, age at initiation of dialysis, etc. The access information includes the type of surgery, surgical procedure, and catheter type; The blood tests include complete blood count, blood biochemistry, iron metabolism, blood electrolytes, and infectious disease testing. The urinary tract examination includes routine urinalysis, etc. The follow-up records include weight, blood pressure, routine dialysis fluid analysis (white blood cells, lobulation, bacterial count), general condition (appetite, nutritional status, mental state, sleep), exit infection status, ultrafiltration volume, tunnel infection, etc. Other indicators include assessment of dialysis adequacy, assessment of residual renal function (Kt / v, cCr), and incidence of peritonitis.

[0014] S2: Preprocessing of medical features and their values; S21: Perform data validation on each medical characteristic value; The preprocessing module presets the threshold range for each medical feature value, compares each medical feature value with the corresponding threshold range, and if it is within the threshold range, the medical feature value is determined to be normal data; if it is outside the threshold range, the medical feature value is determined to be abnormal data, and proceeds to step S22. S22: Perform data verification on abnormal data; The preprocessing module pushes abnormal data to the corresponding data review specialist via email or other means. The data review specialist receives the abnormal data, performs manual verification, and provides feedback on the verification results, which include verification passed and verification failed. Specifically, based on the patient's relevant medical records, test reports, and other information, the service provider verifies whether the abnormal data is due to an entry error or a special condition of the patient. If it is an entry error, the service provider will report "verification failed"; if it is a special condition of the patient (such as the patient having a rare blood pressure abnormality), the service provider will report "verification passed". The preprocessing module updates the abnormal data that passes the verification to normal data and removes the data that fails the verification. S23: Conduct regular data spot checks; The preprocessing module periodically samples medical features and their values, randomly selecting a set proportion of medical features and their values. It then compares these medical features and their values ​​with the corresponding original data in the hospital's electronic medical record platform. If they match, the data is deemed correct; otherwise, an anomaly is reported and manual verification is performed, removing data that fails the verification. After preprocessing, a qualified electronic medical record is obtained.

[0015] This invention incorporates data verification, inspection, and sampling mechanisms during the data preprocessing stage to ensure the accuracy of medical feature data extracted from patients' electronic medical records. In case of data anomalies, the data is pushed to data reviewers for manual verification, thereby improving data accuracy and reducing errors in decision-making caused by data issues.

[0016] S3: Generate a multi-source peritoneal dialysis plan; The data access and processing module acquires the feature data of the user to be decided in the hospital system, and generates multi-source peritoneal dialysis plans through three methods: machine prediction learning, similar patient case matching, and large model inference, including peritoneal dialysis plan 1, peritoneal dialysis plan 2, and peritoneal dialysis plan 3. Specifically, Scheme 1 is generated as follows: Extract medical features and their values ​​related to the five protocol indicators from qualified electronic medical records and construct a dataset; The medical characteristics include CKD stage, initial eGFR, weight, comorbidities, etc. Using machine learning algorithms such as SVM, XGBooost, and Random Forest in the Python sklearn tool, classification and regression models were built and trained for the five scheme indicators respectively. A classification model was constructed for the classification indicators (dialysis fluid type, daily exchange frequency, and dialys fluid concentration), and a regression model was constructed for the numerical indicators (infusion volume and retention time). The classification or regression model is trained using the dataset to obtain the trained classification or regression model; Based on the feature data of the users to be decided, the classification indicators in the treatment plan indicators of the patients to be decided are predicted by the trained classification model, and the final prediction results of the classification indicators are determined by voting method. Specifically, each classification model is used to predict the classification indicators (dialysis fluid type, daily exchange frequency, and dialysate concentration), and each classification model outputs a prediction result. The prediction results of all classification models for the same classification indicator are counted by voting, and the prediction result with the most votes is the final prediction result. If the votes are tied, the prediction result of the classification model with the highest pre-set priority will be used as the final prediction result for the corresponding classification indicator. Based on the feature data of the users to be decided, the numerical indicators in the treatment plan indicators of the patients to be decided are predicted by the trained regression model. The weighted average method is used to determine the final prediction results of the numerical indicators. The final prediction results of the five indicators are integrated to obtain peritoneal dialysis plan 1. The generation method for Scheme 2 is as follows: Each patient's medical characteristics, medical characteristic values, and corresponding peritoneal dialysis plan in the qualified electronic medical records are treated as a patient case, and a treatment recommendation database is constructed. The treatment recommendation database contains all patient cases in the qualified electronic medical records. The feature data of the actual users to be decided are converted into a format consistent with the database. The similarity between the users to be decided and all patient cases in the database is calculated one by one using a similarity algorithm (such as cosine similarity). The patient case with the highest similarity is selected, and the peritoneal dialysis plan in the patient case is taken as peritoneal dialysis plan 2. The generation method for Scheme 3 is as follows: The feature data of the user to be decided is input into a large model such as Deepseek. Based on instructions such as prompt word engineering, the large model outputs relevant content of peritoneal dialysis plan. The rule engine is called to verify the relevant content of the generated peritoneal dialysis plan based on constraints such as peritoneal dialysis clinical diagnosis and treatment guidelines. After filtering out illegal plans, peritoneal dialysis plan 3 is obtained.

[0017] S4: Generate clinical indicators; Based on peritoneal dialysis plans 1, 2 and 3 obtained in step S3, corresponding clinical indicators are calculated and generated respectively. The clinical indicators include Kt / V value, peritonitis risk probability, ultrafiltration volume value and cost per dialysis session. S41: Generate clinical indicators based on peritoneal dialysis protocol 1; The feature data of the user to be decided and peritoneal dialysis scheme 1 are used as joint inputs and respectively input to the pre-trained Kt / V prediction model, peritonitis risk prediction model and ultrafiltration volume prediction model, and respectively output the corresponding Kt / V value, peritonitis risk probability and ultrafiltration volume value. The system contains a corresponding cost price list. Based on the medical supplies (such as dialysis fluid), consumable devices, etc. included in peritoneal dialysis scheme 1, the cost of a single dialysis session is obtained by multiplying each item by the corresponding unit price in the cost price list and then summing the results. Record the Kt / V value for peritoneal dialysis regimen 1. The probability of peritonitis is Ultrafiltration rate value The cost of a single dialysis session is ; S42: Generate clinical indicators based on peritoneal dialysis protocol 2; The feature data of the user to be decided and peritoneal dialysis scheme 2 are used as joint inputs and fed into the pre-built peritonitis risk prediction model to output the peritonitis risk probability. If the patient case with the highest similarity obtained in step S3 has stored Kt / V values ​​and ultrafiltration volume values, then the Kt / V values ​​and ultrafiltration volume values ​​of the similar patients are directly used; if not, then the feature data of the user to be decided and peritoneal dialysis scheme 2 are used as joint inputs, and are respectively input into the pre-trained Kt / V prediction model and ultrafiltration volume prediction model, and output the corresponding Kt / V values ​​and ultrafiltration volume values; and based on the medical supplies (such as dialysis fluid), consumable devices, etc. included in peritoneal dialysis scheme 2, the costs are multiplied by the corresponding unit prices in the cost price list and then summed to obtain the cost of a single dialysis session; Record the Kt / V value for peritoneal dialysis regimen 2. The probability of peritonitis is Ultrafiltration rate value The cost of a single dialysis session is ; S43: Generate clinical indicators based on peritoneal dialysis protocol 3; The feature data of the user to be decided and peritoneal dialysis scheme 3 are used as joint inputs and fed into the pre-trained Kt / V prediction model, peritonitis risk prediction model and ultrafiltration volume prediction model, respectively, and output the corresponding Kt / V value, peritonitis risk probability and ultrafiltration volume value; based on the medical supplies (such as dialysis fluid) and consumable devices included in peritoneal dialysis scheme 3, the cost per dialysis session is obtained by multiplying them by the corresponding unit price in the cost price list and summing them. Record the Kt / V value for peritoneal dialysis regimen 3. The probability of peritonitis is Ultrafiltration rate value The cost of a single dialysis session is .

[0018] By combining multiple peritoneal dialysis plan generation methods and relying on machine learning algorithms (such as SVM, XGBoost, random forest, etc.) for personalized recommendations, and utilizing similar patient case matching and large model inference, multiple peritoneal dialysis plans are generated for patients from different perspectives. By comprehensively utilizing the advantages of each method, avoiding the bias and limitations of a single method, and calculating the clinical indicators of each plan, we can ensure that individual differences are maximized and treatment outcomes are improved.

[0019] This invention generates multiple dialysis plans by combining clinical indicators such as Kt / V value, peritonitis risk, and ultrafiltration volume, and calculates the cost per dialysis session for each plan. This further helps doctors optimize the use of medical resources while ensuring treatment effectiveness and reducing the economic burden on patients.

[0020] S5: Construct a decision matrix for TOPSIS multi-objective decision-making; S51: Construct the decision matrix; Based on the specific values ​​of the clinical indicators corresponding to peritoneal dialysis plans 1, 2, and 3 obtained in step S4, a decision matrix is ​​constructed, denoted as decision matrix A; ; S52: Standardize the data in the decision matrix; The decision matrix B is obtained by standardizing the data in the decision matrix using the following standardization formula. ; in, These are the original data values ​​in decision matrix A. The data values ​​are standardized, n=3; ; S53: Obtain the positive ideal solution and the negative ideal solution; The positive ideal solution and the negative ideal solution are obtained by using the cosine method. The positive ideal solution is the maximum value of the data value of each column element in the decision matrix B, and the negative ideal solution is the minimum value of the data value of each column element in the decision matrix B. Let the ideal solution be... The negative ideal solution is ; ; S54: Select the optimal peritoneal dialysis plan; The distance between each data point in decision matrix B and the positive and negative ideal solutions is calculated using the following formula: ; Where i corresponds to the peritoneal dialysis plan (1=peritoneal dialysis plan 1, 2=peritoneal dialysis plan 2, 3=peritoneal dialysis plan 3), and j corresponds to the clinical indicators (1=Kt / V value, 2=peritoneal inflammation risk probability, 3=ultrafiltration volume value, 4=cost per dialysis session). For the i-th peritoneal dialysis regimen, the standardized data value under the j-th clinical indicator is... For the j-th index, For the j-th index, the negative ideal solution is... Let the distance between the i-th peritoneal dialysis scheme and the ideal solution be... Let be the distance between the i-th peritoneal dialysis scheme and the negative ideal solution; The following formula is used to calculate the approximation of each peritoneal dialysis regimen to the ideal solution; ; in, To approximate the degree of similarity, ≤1, The closer the value is to 0, the better the corresponding peritoneal dialysis plan. The result Sort the sequence, where the maximum value in the sequence is... The corresponding peritoneal dialysis plan is the optimal peritoneal dialysis plan for the patient to be decided. The optimal peritoneal dialysis plan is rendered and displayed in the results display module on the user interface of the peritoneal dialysis plan decision-making system based on multi-source data and TOPSIS optimization.

[0021] This invention employs the TOPSIS multi-objective decision-making method, constructing a decision matrix and combining the distance between the positive and negative ideal solutions to select the optimal treatment plan, ensuring that patients receive the best possible care. The invention also features a results display module that visually presents the optimized peritoneal dialysis plan to the user, allowing doctors to directly see the decision results on the user interface. This facilitates timely adjustments to the treatment plan and allows for thorough communication with patients. Doctors no longer need to manually calculate and compare various indicators; they can directly and quickly compare the advantages and disadvantages of the three plans based on indicator values, and then make a decision based on clinical experience. This significantly reduces the difficulty of decision-making, while also reducing the workload of manual calculations and data processing, thus optimizing the efficiency of medical resource utilization.

[0022] Furthermore, the peritoneal dialysis protocols that are ultimately adopted and validated in clinical practice, along with their results, are fed back into the peritoneal dialysis protocol decision-making system optimized based on multi-source data and TOPSIS. This feedback is used to optimize the machine learning model, update the protocol database, and improve the system's intelligence.

[0023] The present invention has been described in the above-described embodiments; however, these embodiments are merely examples for implementing the present invention. It must be noted that the disclosed embodiments do not limit the scope of the present invention. Conversely, any modifications and refinements made without departing from the spirit and scope of the present invention are within the scope of patent protection of the present invention.

Claims

1. A peritoneal dialysis protocol decision-making method based on multi-source data and TOPSIS optimization, characterized in that: Includes the following steps: S1: Obtain the patient's electronic medical record and extract medical features and medical feature values ​​from it; S11: Obtain the patient's electronic medical record; S12: Extract medical features and medical feature values; S2: Preprocessing of medical features and their values; S21: Perform data validation on each medical characteristic value; S22: Perform data verification on abnormal data; S23: Conduct regular data spot checks; S3: Generate a multi-source peritoneal dialysis plan; The data access and processing module acquires the feature data of the user to be decided in the hospital system, and generates multi-source peritoneal dialysis plans through three methods: machine prediction learning, similar patient case matching, and large model inference, including peritoneal dialysis plan 1, peritoneal dialysis plan 2, and peritoneal dialysis plan 3. S4: Generate clinical indicators; Based on peritoneal dialysis plans 1, 2 and 3 obtained in step S3, corresponding clinical indicators are calculated and generated respectively. The clinical indicators include Kt / V value, peritonitis risk probability, ultrafiltration volume value and cost per dialysis session. S5: Construct a decision matrix for TOPSIS multi-objective decision-making; S51: Construct the decision matrix; S52: Standardize the data in the decision matrix; S53: Obtain the positive ideal solution and the negative ideal solution; S54: Select the optimal peritoneal dialysis plan.

2. The peritoneal dialysis protocol decision-making method based on multi-source data and TOPSIS optimization as described in claim 1, characterized in that: The specific details of step S1 are as follows: S11: Obtain the patient's electronic medical record; The hospital's electronic medical record platform collects the electronic medical records of patients awaiting decision-making, and filters out the electronic medical records of patients who have been undergoing peritoneal dialysis for more than a set time; the electronic medical records of these patients contain the corresponding peritoneal dialysis plans. The peritoneal dialysis protocol includes five protocol indicators: daily exchange frequency, infusion volume, dialysate type, dialysate concentration, and peritoneal retention time. S12: Extract medical features and medical feature values; The feature extraction module extracts medical features and medical feature values ​​from the patient electronic medical records selected in step S11; The medical characteristics include the patient's basic information, medical history, access information, blood tests, urinary tract tests, and follow-up records.

3. The peritoneal dialysis protocol decision-making method based on multi-source data and TOPSIS optimization as described in claim 1, characterized in that: The specific details of step S2 are as follows: S21: Perform data validation on each medical characteristic value; The preprocessing module presets the threshold range for each medical feature value, compares each medical feature value with the corresponding threshold range, and if it is within the threshold range, the medical feature value is determined to be normal data; if it is outside the threshold range, the medical feature value is determined to be abnormal data, and proceeds to step S22. S22: Perform data verification on abnormal data; The preprocessing module pushes abnormal data to the corresponding data review specialist via email. The data review specialist receives the abnormal data, performs manual verification, and provides feedback on the verification results, which include verification passed or verification failed. The preprocessing module updates the abnormal data that passes the verification to normal data and removes the data that fails the verification. S23: Conduct regular data spot checks; The preprocessing module periodically samples medical features and medical feature values, randomly selecting a set proportion of medical features and medical feature values, and comparing the medical features and medical feature values ​​with the corresponding original data in the hospital's electronic medical record platform. If they match, the data is determined to be correct. If there is a discrepancy, report the anomaly and conduct a manual review, removing the data that fails the review. After preprocessing, a qualified electronic medical record is obtained.

4. The peritoneal dialysis protocol decision-making method based on multi-source data and TOPSIS optimization as described in claim 1, characterized in that: In step S3, Scheme 1 is generated as follows: Extract medical features and their values ​​related to the five protocol indicators from qualified electronic medical records and construct a dataset; We used machine learning algorithms to build and train classification and regression models for the five scheme indicators respectively; For the classification indicators, including dialysate type, daily exchange frequency, and dialysate concentration, a classification model was constructed; for the numerical indicators, including infusion volume and retention time, a regression model was constructed. The classification or regression model is trained using the dataset to obtain the trained classification or regression model; Based on the feature data of the users to be decided, the classification indicators in the treatment plan indicators of the patients to be decided are predicted by the trained classification model, and the final prediction results of the classification indicators are determined by voting method. Based on the feature data of the users to be decided, the numerical indicators in the treatment plan indicators of the patients to be decided are predicted by the trained regression model. The weighted average method is used to determine the final prediction results of the numerical indicators. The final prediction results of the five indicators are integrated to obtain peritoneal dialysis plan 1. The generation method for Scheme 2 is as follows: Each patient's medical characteristics, medical characteristic values, and corresponding peritoneal dialysis plan in the qualified electronic medical records are treated as a patient case, and a treatment recommendation database is constructed. The treatment recommendation database contains all patient cases in the qualified electronic medical records. The feature data of the actual users to be decided are converted into a format consistent with the database. The similarity between the users to be decided and all patient cases in the database is calculated one by one using a similarity algorithm. The patient case with the highest similarity is selected, and the peritoneal dialysis plan in the patient case is taken as peritoneal dialysis plan 2. The generation method for Scheme 3 is as follows: The feature data of the user to be decided is input into the large model, and the relevant content of the peritoneal dialysis plan is output. The rule engine is called to verify the generated peritoneal dialysis plan based on the constraints of the peritoneal dialysis clinical diagnosis and treatment guidelines. After filtering out the illegal plans, peritoneal dialysis plan 3 is obtained.

5. The peritoneal dialysis protocol decision-making method based on multi-source data and TOPSIS optimization as described in claim 1, characterized in that: The specific details of step S4 are as follows: S41: Generate clinical indicators based on peritoneal dialysis protocol 1; The feature data of the user to be decided and peritoneal dialysis scheme 1 are used as joint inputs and respectively input to the pre-trained Kt / V prediction model, peritonitis risk prediction model and ultrafiltration volume prediction model, and respectively output the corresponding Kt / V value, peritonitis risk probability and ultrafiltration volume value. Based on the cost price details table included in the system, the cost of a single dialysis session is calculated for peritoneal dialysis scheme 1 to obtain the cost of a single dialysis session. Record the Kt / V value for peritoneal dialysis regimen 1. The probability of peritonitis is Ultrafiltration rate value The cost of a single dialysis session is ; S42: Generate clinical indicators based on peritoneal dialysis protocol 2; The feature data of the user to be decided and peritoneal dialysis scheme 2 are used as joint inputs and fed into the pre-built peritonitis risk prediction model to output the peritonitis risk probability. If the patient case with the highest similarity obtained in step S3 has stored Kt / V values ​​and ultrafiltration volume values, then the Kt / V values ​​and ultrafiltration volume values ​​of the similar patients are directly used; if not, then the feature data of the user to be decided and peritoneal dialysis scheme 2 are used as joint inputs and input into the pre-trained Kt / V prediction model and ultrafiltration volume prediction model, respectively, and the corresponding Kt / V values ​​and ultrafiltration volume values ​​are output; based on the cost price details table contained in the system, the cost of a single dialysis session for peritoneal dialysis scheme 2 is calculated to obtain the cost of a single dialysis session; Record the Kt / V value for peritoneal dialysis regimen 2. The probability of peritonitis is Ultrafiltration rate value The cost of a single dialysis session is ; S43: Generate clinical indicators based on peritoneal dialysis protocol 3; The feature data of the user to be decided and peritoneal dialysis scheme 3 are used as joint inputs and fed into the pre-trained Kt / V prediction model, peritonitis risk prediction model and ultrafiltration volume prediction model, respectively, and output the corresponding Kt / V value, peritonitis risk probability and ultrafiltration volume value; based on the cost price list included in the system, the cost of a single dialysis session of peritoneal dialysis scheme 3 is calculated to obtain the cost of a single dialysis session. Record the Kt / V value for peritoneal dialysis regimen 3. The probability of peritonitis is Ultrafiltration rate value The cost of a single dialysis session is .

6. The peritoneal dialysis protocol decision-making method based on multi-source data and TOPSIS optimization as described in claim 1, characterized in that: The specific details of step S5 are as follows: S51: Construct the decision matrix; Based on the specific values ​​of the clinical indicators corresponding to peritoneal dialysis plans 1, 2, and 3 obtained in step S4, a decision matrix is ​​constructed, denoted as decision matrix A; ； S52: Standardize the data in the decision matrix; The decision matrix B is obtained by standardizing the data in the decision matrix using the following standardization formula. ； in, These are the original data values ​​in decision matrix A. The data values ​​are standardized, n=3; ； S53: Obtain the positive ideal solution and the negative ideal solution; The positive ideal solution and the negative ideal solution are obtained by using the cosine method. The positive ideal solution is the maximum value of the data value of each column element in the decision matrix B, and the negative ideal solution is the minimum value of the data value of each column element in the decision matrix B. Let the ideal solution be... The negative ideal solution is ; S54: Select the optimal peritoneal dialysis plan; The distance between each data point in decision matrix B and the positive and negative ideal solutions is calculated using the following formula: ； Where i corresponds to the peritoneal dialysis plan, j corresponds to the clinical indicators, For the i-th peritoneal dialysis regimen, the standardized data value under the j-th clinical indicator is... For the j-th index, the positive ideal solution is... For the j-th index, the negative ideal solution is... Let the distance between the i-th peritoneal dialysis scheme and the ideal solution be... Let be the distance between the i-th peritoneal dialysis scheme and the negative ideal solution; The following formula is used to calculate the approximation of each peritoneal dialysis regimen to the ideal solution; ； in, To approximate the degree of similarity, ≤1; The result Sort the sequence, where the maximum value in the sequence is... The corresponding peritoneal dialysis plan is the optimal peritoneal dialysis plan for the patient to be decided. The optimal peritoneal dialysis plan is rendered and displayed in the results display module on the user interface of the peritoneal dialysis plan decision-making system based on multi-source data and TOPSIS optimization.

7. A peritoneal dialysis protocol decision-making system based on multi-source data and TOPSIS optimization for implementing the method of any one of claims 1-6. Its features are: It includes a feature extraction module, a preprocessing module, a data access and processing module, an intelligent solution decision-making module, and a results display module; The feature extraction module is used to extract medical features and medical feature values ​​from the patient's electronic medical record; The preprocessing module is used to perform data verification, data checking, and data sampling of medical feature values. The data access and processing module is used to acquire the feature data of the user to be decided and generate a multi-source peritoneal dialysis plan and corresponding clinical indicators. The intelligent solution decision module is used to construct a decision matrix to perform TOPSIS multi-objective decision-making and obtain the optimal peritoneal dialysis solution. The results display module is used to display the optimal peritoneal dialysis plan on the user interface of the peritoneal dialysis plan decision system optimized based on multi-source data and TOPSIS.

Images