Tumor nursing information collaborative management system based on multi-source data fusion

Abstract

The invention discloses a tumor nursing information collaborative management system based on multi-source data fusion, which adds elaborately designed noise into a query result through a differential privacy technology when patient medication data is processed, and the noise is beneficial to hiding specific data of a single patient, so that the data of the single patient can be more effectively hidden. Therefore, sensitive personal information is effectively prevented from being leaked. The differential privacy technique ensures that when a large amount of patient data is analyzed, the change of any single data record does not significantly affect the final analysis result. By setting a proper privacy budget, the system can maintain the accuracy and reliability of data analysis while protecting data privacy.

Description

Technical Field

[0001] The present invention belongs to the field of tumor nursing, and in particular, relates to a tumor nursing information collaborative management system based on multi-source data fusion. Background Art

[0002] As data in the healthcare field continues to grow and become more complex, the traditional single data source medication plan generation system has been unable to meet the increasingly complex clinical needs. As a complex and highly personalized medical field, cancer care involves a variety of health data of patients, including electronic medical records, medical images, laboratory test results, medication history records, etc. These data often come from different medical systems and devices, and there is a certain gap between them, resulting in inefficient cross-system collaboration and data sharing.

[0003] Existing medication plan generation systems usually rely on a single data source, such as relying solely on the patient's electronic medical record or laboratory test results to formulate a treatment plan. The main problem with this approach is that a single data source cannot fully reflect the patient's health status, especially when facing complex cases, the patient's multi-dimensional data (such as changes in condition, medication history, drug allergic reactions, medical imaging data, etc.) has not been effectively integrated. This lack of data fusion results in the medication recommendation system being unable to fully consider all of the patient's health information, which in turn affects the accuracy and personalization of the treatment plan. In addition, existing systems often lack a dynamic adjustment mechanism and cannot flexibly optimize treatment plans based on patient feedback or changes in condition during treatment, limiting their application in complex medical scenarios. In addition, during the process of data collection and processing, patients' sensitive information, such as disease diagnosis, drug use records, and medical images, is extremely vulnerable to unauthorized access or data leakage risks. Especially when it comes to medication data, this information is directly related to the patient's personal privacy and security. Therefore, how to effectively use data for personalized treatment recommendations while ensuring patient privacy has become a key issue. Existing technologies often have a trade-off between privacy protection and data utilization, and cannot take into account the needs of both. With the increase of medical data and the rising demand for personalized treatment, how to use multi-source data to generate accurate and personalized medication plans while fully protecting patient privacy has become an important issue that needs to be urgently addressed in today's medical technology.

[0004] In view of this, the present invention is proposed. Summary of the invention

[0005] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a tumor nursing information collaborative management system based on multi-source data fusion, which solves the problems raised in the above-mentioned background technology.

[0006] In order to solve the above technical problems, the basic concept of the technical solution adopted by the present invention is:

[0007] A tumor nursing information collaborative management system based on multi-source data fusion, comprising: a data acquisition module: automatically collecting medical data of patients from multiple data sources;

[0008] Data preprocessing module: performs preprocessing such as cleaning, denoising, format standardization, and missing value filling on the collected patient medication data to ensure data consistency and quality;

[0009] Privacy protection module: Through differential privacy technology, noise is added during the processing of medication data to prevent the leakage of patients' sensitive information;

[0010] Data fusion module: integrates medication data and medical data from different data sources to build a complete patient care file;

[0011] Medication management module: Automatically generates personalized medication plans based on the comprehensive data provided by the data fusion module.

[0012] Collaborative management module: provides a collaborative platform for multi-role medical staff, supporting cross-department collaboration and information sharing.

[0013] Optionally, differential privacy technology is used to add noise to the medication data to prevent leakage of sensitive patient information. The steps are:

[0014] Analyze the patient's medication data, calculate the sensitivity of the query function Δf, and select an appropriate privacy budget ε based on the required privacy protection and accuracy requirements, where ε = 0.01;

[0015] According to the obtained sensitivity Δf and privacy budget ε, the noise is generated using the noise generation mechanism of differential privacy, where the distribution of noise is based on the Laplace mechanism, and the expression is: Next, the generated noise is added to the statistical query results of the medication data.

[0016] Optionally, the steps of analyzing the patient's medication data and calculating the sensitivity Δf of the query function are:

[0017] Define query function f, which is the statistical query operation to be performed on the patient medication data. Query function f can be used to calculate the frequency, total amount or other relevant statistical data of drug use;

[0018] Calculate the maximum impact of a single patient's medication record change on the query result. Then, calculate the sensitivity Δf by calculating the difference in the query function results between the original data set D and the modified data set D′. The formula is as follows: They are the query results on the original dataset D and the modified dataset D′ respectively;

[0019] Output the calculated sensitivity Δf, which represents the maximum impact of a single data record change on the query result.

[0020] Optionally, medication data from different data sources are fused with medical data to build a complete patient care profile. The steps to fuse medication data with medical data are as follows:

[0021] Use a lookup table to convert features from different data sources into embedding vectors and concatenate them together. For each word in the medical corpus text and the relationship between entities, represent it as a low-dimensional vector;

[0022] The Skip-gram model of the Word2Vec tool is used to train medical corpus text through convolutional neural networks. The model infers the meaning of the current word through the probability of context words, thereby modeling the relationship between words and converting each word in the text into a k-dimensional vector;

[0023] Use the convolution kernel to perform convolution operation on the input medical text to automatically extract features from the text. The convolution kernel extracts features at different positions through a sliding window to generate a feature set. The feature vector sequence is calculated through the convolution layer to generate a new feature set;

[0024] The features output by the convolutional layer are filtered through the pooling layer to extract the most representative features, remove redundant data, and pass them to the next layer. The pooled features are further aggregated to form a global feature vector that represents the entire input data;

[0025] The Softmax function is used to classify the features obtained by the pooling layer to generate the final treatment recommendation or medication plan. The output layer classifies the patient care information through weighted and biased operations and outputs the final decision. Finally, by integrating the patient's medication data, disease information, examination results, treatment plan and other data, a comprehensive patient care file is constructed to provide support for subsequent treatment decisions.

[0026] Optionally, based on the comprehensive data provided by the data fusion module, a personalized medication regimen is automatically generated. This module comprehensively considers the patient's condition, medication history, and drug interactions, and recommends the most appropriate drug combination and dosage in the following steps:

[0027] Based on the fused data, key features are automatically extracted, including the patient's condition, medication history, drug allergy history, complication information, and examination results. The feature extraction process also analyzes the patient's health status, disease type, and historical treatment records.

[0028] Generate personalized medication plans based on the patient's individual characteristics and clinical needs. The medication plan generation process takes into account the patient's condition, drug interactions, drug allergic reactions, treatment history factors, and combines treatment guidelines to automatically recommend the most appropriate drugs, dosages, and routes of administration for patients.

[0029] Optionally, the steps for generating a personalized medication regimen based on the individual characteristics and clinical needs of the patient are:

[0030] Use convolutional neural networks (CNNs) to build a personalized medication regimen generation model. CNN models can automatically learn complex features from patients' health data, especially extracting important information from high-dimensional data;

[0031] Model training is performed based on historical patient data sets, and the patient's health data is used during the training process to optimize the parameters of the CNN model;

[0032] The trained CNN model automatically generates personalized medication plans based on the individual characteristics and clinical needs of the patient. The model will recommend appropriate drugs, dosages, and routes of administration based on the patient's disease type, severity, and history of drug allergies;

[0033] Based on the output of the CNN model, a personalized medication regimen report for the patient is generated, providing detailed drug name, recommended dosage, route of administration, and frequency of use.

[0034] Optionally, the steps to provide a collaboration platform for multi-role medical staff and support cross-department collaboration and information sharing are as follows:

[0035] Create a medical staff information registration system that allows different roles (such as doctors, nurses, pharmacists, laboratory personnel, etc.) to register information. Each medical staff member will obtain a unique role identification according to their function and department, and be authenticated;

[0036] The system assigns corresponding access rights to medical staff of different roles to ensure data security and privacy protection. Different data access and modification permissions are set according to the medical staff's position and department;

[0037] On the platform, medical staff can access the patient's complete health records, including medical history, medication records, treatment plans, examination results, etc. The records are updated in real time through data fusion technology to ensure that medical staff in different departments can obtain the latest information. Through information sharing, doctors, nurses, pharmacists, etc. can collaborate in a timely manner based on the patient's health status and treatment progress to ensure that the patient's treatment plan is consistently implemented.

[0038] After adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art. Of course, any product implementing the present invention does not necessarily need to achieve all the advantages described below at the same time:

[0039] 1. This invention uses differential privacy technology. When processing patient medication data, we add carefully designed noise to the query results. This noise helps hide the specific data of individual patients, thereby effectively preventing the leakage of sensitive personal information. Differential privacy technology ensures that when analyzing a large amount of patient data, changes in any single data record will not significantly affect the final analysis results. By setting an appropriate privacy budget, the system can maintain the accuracy and reliability of data analysis while protecting data privacy.

[0040] 2. Based on the integrated data after fusion, the system of the present invention can automatically generate personalized medication plans for each patient. First, the system forms a complete patient health record by integrating the patient's multidimensional data, including but not limited to the type of disease, the severity of the disease, historical medication records, laboratory test results, etc. Combined with this information, the system can comprehensively assess the patient's treatment needs and risks, and then recommend the most suitable drug combination for them. In order to ensure the safety and effectiveness of medication, the system will also analyze the interaction between the patient's current drugs and historical drugs, avoid potential drug conflicts, and reduce the occurrence of adverse reactions. In addition, the system will automatically exclude drugs that may cause allergic reactions based on the patient's history of drug allergies, thereby protecting the patient's medication safety. According to the patient's specific situation (such as age, weight, renal function, etc.) and the stage of disease development, the system will recommend the most appropriate drug dosage and route of administration (such as oral, intravenous injection, etc.) to ensure the accuracy of treatment.

[0041] The specific implementation modes of the present invention are further described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The drawings described below are only some embodiments. For those skilled in the art, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0043] Figure 1 It is a structural diagram.

[0044] It should be noted that these drawings and textual descriptions are not intended to limit the conceptual scope of the present invention in any way, but are intended to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0045] The present invention will now be described in further detail with reference to the accompanying drawings.

[0046] As shown in FIG. 1 , in this embodiment, a tumor nursing information collaborative management system based on multi-source data fusion is provided, including a data acquisition module: automatically acquiring medical data of patients from multiple data sources, including but not limited to electronic medical records, medical imaging data, laboratory test results, drug use records, etc. of patients, and including medication history information of patients, drug name, dosage, route of administration, duration of medication, etc.;

[0047] Data preprocessing module: performs preprocessing such as cleaning, denoising, format standardization, and missing value filling on the collected patient medication data to ensure data consistency and quality;

[0048] Privacy protection module: Through differential privacy technology, noise is added during the processing of medication data to prevent the leakage of patients' sensitive information;

[0049] Data fusion module: integrates medication data and medical data from different data sources to build a complete patient care file;

[0050] Medication management module: Based on the comprehensive data provided by the data fusion module, a personalized medication regimen is automatically generated. Based on the fused comprehensive data, the system of the present invention can automatically generate a personalized medication regimen for each patient. First, the system forms a complete patient health record by integrating the patient's multidimensional data, including but not limited to the type of disease, the severity of the disease, historical medication records, laboratory test results, etc. Combined with this information, the system can comprehensively assess the patient's treatment needs and risks, and then recommend the most suitable drug combination for them. In order to ensure the safety and effectiveness of medication, the system will also analyze the interaction between the patient's current drugs and historical drugs, avoid potential drug conflicts, and reduce the occurrence of adverse reactions. In addition, the system will automatically exclude drugs that may cause allergic reactions based on the patient's history of drug allergies, thereby protecting the patient's medication safety. According to the patient's specific situation (such as age, weight, renal function, etc.) and the stage of disease development, the system will recommend the most appropriate drug dosage and route of administration (such as oral, intravenous injection, etc.) to ensure the accuracy of treatment.

[0051] Collaborative management module: provides a collaborative platform for multi-role medical staff, supporting cross-department collaboration and information sharing. The collaborative management module provides a centralized collaborative platform for medical staff of different roles, supporting cross-department communication and information sharing. Doctors, nurses, pharmacists and other medical staff can access the patient's complete health records in real time and jointly develop and adjust treatment plans.

[0052] In this embodiment, the steps of adding noise to the medication data processing process to prevent leakage of the patient's sensitive information are as follows:

[0053] Analyze the patient's medication data, calculate the sensitivity of the query function Δf, and select an appropriate privacy budget ε based on the required privacy protection and accuracy requirements, where ε = 0.01;

[0054] According to the obtained sensitivity Δf and privacy budget ε, the noise is generated using the noise generation mechanism of differential privacy, where the distribution of noise is based on the Laplace mechanism, and the expression is: Next, the generated noise is added to the statistical query results of the medication data.

[0055] In this embodiment, the steps of analyzing the patient's medication data and calculating the sensitivity Δf of the query function are as follows:

[0056] Define query function f, which is the statistical query operation to be performed on the patient medication data. Query function f can be used to calculate the frequency, total amount or other relevant statistical data of drug use;

[0057] Calculate the maximum impact of changes in a single patient's medication record on the query results. For example, if the query function is to calculate the frequency of use of drug A, consider the impact of changes in the patient's medication record (such as changing to not using drug A) on the frequency of use of drug A;

[0058] The sensitivity Δf is calculated by calculating the difference in the query function results between the original dataset D and the modified dataset D′, and the formula is as follows: They are the query results on the original dataset D and the modified dataset D′ respectively;

[0059] Output the calculated sensitivity Δf, which represents the maximum impact of a single data record change on the query result.

[0060] In this embodiment, when the medication data and medical data from different data sources are integrated to construct a complete patient care file, the medication data and medical data are integrated by using a word embedding layer, a convolutional neural network and a pooling technology. The steps are as follows:

[0061] Use a lookup table to convert features from different data sources (such as medication data, disease information, treatment plans, etc.) into embedding vectors and concatenate them together. For each word in the medical corpus text and the relationship between entities (such as the association between drugs and diseases), represent it as a low-dimensional vector (such as p1 and p2 dimensional vectors);

[0062] The Skip-gram model of the Word2Vec tool is used to train medical corpus text through convolutional neural networks. The model infers the meaning of the current word through the probability of context words, thereby modeling the relationship between words and converting each word in the text into a k-dimensional vector;

[0063] Use the convolution kernel to perform convolution operation on the input medical text to automatically extract features from the text. The convolution kernel extracts features at different positions through a sliding window to generate a feature set. The feature vector sequence (such as x1, x2, x3, ..., x m ) Generate a new feature set through convolutional layer calculation;

[0064] The features output by the convolutional layer are filtered through the pooling layer (using the maximum pooling method), the most representative features are extracted, redundant data is removed, and it is passed to the next layer. The pooled features are further aggregated to form a global feature vector that represents the entire input data;

[0065] The Softmax function is used to classify the features obtained by the pooling layer to generate the final treatment recommendation or medication plan. The output layer classifies the patient care information through weighted and biased operations and outputs the final decision. Finally, by integrating the patient's medication data, disease information, examination results, treatment plan and other data, a comprehensive patient care file is constructed to provide support for subsequent treatment decisions.

[0066] In this embodiment, based on the comprehensive data provided by the data fusion module, a personalized medication regimen is automatically generated. The module comprehensively considers the patient's condition, medication history, and drug interactions, and recommends the most appropriate drug combination and dosage in the following steps:

[0067] Based on the fused data, key features are automatically extracted, including the patient's condition, medication history, drug allergy history, complication information, and examination results. The feature extraction process also analyzes the patient's health status, disease type, and historical treatment records.

[0068] Generate personalized medication plans based on the patient's individual characteristics and clinical needs. The medication plan generation process takes into account the patient's condition, drug interactions, drug allergic reactions, treatment history factors, and combines treatment guidelines to automatically recommend the most appropriate drugs, dosages, and routes of administration for patients.

[0069] In this embodiment, the steps of generating a personalized medication regimen according to the individual characteristics and clinical needs of the patient are as follows:

[0070] Use convolutional neural networks (CNNs) to build a personalized medication regimen generation model. CNN models can automatically learn complex features from patients' health data, especially extracting important information from high-dimensional data (such as diseases, treatment history, drug reactions, etc.);

[0071] Model training is based on historical patient data sets. During the training process, the patient's health data (such as drug use history, treatment effects, adverse drug reactions, etc.) is used to optimize the parameters of the CNN model.

[0072] The trained CNN model automatically generates personalized medication plans based on the individual characteristics and clinical needs of the patient. The model will recommend appropriate drugs, dosages, and routes of administration based on the patient's disease type, severity, and history of drug allergies;

[0073] Based on the output of the CNN model, a personalized medication regimen report for the patient is generated, providing detailed drug name, recommended dosage, route of administration, and frequency of use.

[0074] In this embodiment, the steps of providing a collaboration platform for multi-role medical staff and supporting cross-department collaboration and information sharing are as follows:

[0075] Create a medical staff information registration system that allows different roles (such as doctors, nurses, pharmacists, laboratory personnel, etc.) to register information. Each medical staff member will obtain a unique role identification according to their function and department, and be authenticated;

[0076] The system assigns corresponding access rights to medical staff with different roles to ensure data security and privacy protection. Different data access and modification permissions are set according to the positions and departments of medical staff. For example, doctors can access patients' medical records and treatment plans, nurses can view nursing records, and pharmacists can access drug prescriptions and medication history.

[0077] On the platform, medical staff can access the patient's complete health records, including medical history, medication records, treatment plans, examination results, etc. The records are updated in real time through data fusion technology to ensure that medical staff in different departments can obtain the latest information. Through information sharing, doctors, nurses, pharmacists, etc. can collaborate in a timely manner based on the patient's health status and treatment progress to ensure that the patient's treatment plan is consistently implemented.

[0078] The platform provides real-time collaboration and communication functions across departments. Medical staff can communicate with experts or medical staff from other departments as needed to discuss patients' treatment options and care plans.

[0079] The platform supports multi-dimensional data sharing and analysis, and medical staff can conduct comprehensive analysis based on the patient's health records. Through the data visualization function, doctors can view the patient's treatment progress, drug use effects, symptom changes and other information, and share it with medical staff in other departments.

[0080] On the platform, medical staff can collaborate and discuss based on the patient's feedback during treatment and adjust the treatment plan. For example, nurses can provide feedback on changes in patients' symptoms, pharmacists can provide advice on drug use, and doctors can adjust treatment plans based on patients' feedback.

[0081] Description of prior art solutions

[0082] Comparative Example 1: Medication Plan Generation System Based on a Single Data Source

[0083] In the prior art, some medication regimen generation systems are based on a single data source, such as formulating medication regimens through electronic medical records (EMR) or laboratory test data. These systems usually generate recommended drugs and dosages through simple rules or algorithms based on fixed standards. For example:

[0084] Data source: Only rely on disease information and patient medication history in electronic medical records.

[0085] Recommendation method: Based on the patient's disease type and historical medication records, drug recommendations are made using rules or linear regression models.

[0086] Limitations: Lack of integration of multi-source data (such as medical images, laboratory test results, etc.), so it cannot fully reflect the patient's health status and individual needs. In addition, these systems do not have the ability to consider complex factors such as drug interactions and drug allergic reactions.

[0087] Comparative Example 2: Medication recommendations based on traditional clinical treatment guidelines

[0088] Another common technology solution is a medication recommendation system based on traditional treatment guidelines. These systems recommend drugs and dosages by matching the patient's disease type with standard treatment guidelines. For example:

[0089] Data source: Mainly based on the patient's disease type and clinical treatment guidelines.

[0090] Recommendation method: Generate drug recommendations by comparing with standard treatment guidelines (such as cancer treatment guidelines).

[0091] Limitations: Such systems ignore individual differences among patients (such as drug allergies, historical medication reactions, etc.) and cannot be dynamically adjusted based on real-time treatment effects.

[0092] 2. Comparative experiment with the same conditions

[0093] In order to fairly compare the effects of this solution with the existing solution, the following equal conditions are set:

[0094] Experimental design conditions:

[0095] Patient data source: Use the same data set, including the patient's medical history, medication history, laboratory test results, medical imaging data, etc. Treatment goal: Solve the problem of generating personalized medication plans for cancer patients, including drug selection, dosage adjustment, and administration route. Privacy protection: All experiments apply differential privacy technology, and the noise intensity is set to ε = 0.1 to ensure the consistency of privacy protection in all schemes.

[0096] A total of 294,149 data items were collected in this experiment, including disease diagnosis items, medical subjects, disease names, and drug names. The word frequency statistics strategy was applied to analyze the acquired data, understand the collected medical corpus knowledge, and summarize it based on the corpus text. A total of 10 entity relationship types were defined. The data annotation part mainly relied on the medical corpus annotation system developed by ourselves. After confirming the format of the corpus, the annotation work was directly carried out. In the expected annotation system, only the entities to be annotated and the various relationships mentioned above can be selected to complete the annotation of entities and their relationships. Using this system can reduce tedious manual operations and improve the accuracy. Finally, semi-automatic annotation based on this system is completed

[11] . The results of text annotation are shown in Example 1. The annotated part of the results was manually proofread to obtain the final experimental data.

[0097] Evaluation criteria:

[0098] Accuracy: The degree of match between the generated medication regimen and the standard treatment guidelines (i.e., the correctness of the recommendations such as drug name and dosage). Personalization: Whether the regimen can be automatically adjusted according to the individual characteristics of the patient (such as condition, medication history, drug allergies, etc.). Drug safety: The accuracy of predictions of drug interactions and drug allergic reactions. Treatment effect: Clinical feedback from patients after treatment, including changes in efficacy and side effects. Processing time: The time required from data input to the generation of a personalized medication regimen.

[0099] Experimental steps:

[0100] Data preprocessing: Clean, denoise, and standardize the patient's health data.

[0101] Medication plan generation: Use this plan and existing plans to generate personalized medication plans.

[0102] Comparison of evaluation indicators: Evaluate the accuracy and personalization of each regimen by comparing it with standard treatment guidelines.

[0103] Safety assessment: Check the accuracy of identification of drug interactions and allergic reactions in the generated medication regimen.

[0104] Clinical effect feedback: Analyze the patient's feedback data after treatment and evaluate the actual effect of the medication plan.

[0105]

[0106] Accuracy: This solution is significantly better than existing solutions in terms of the accuracy of drug recommendations, especially through the fusion of multi-source data, which enables the matching degree between recommendations and treatment guidelines to reach 95%, higher than the 80% and 85% of existing solutions.

[0107] Degree of personalization: This program is more outstanding in personalized recommendations, taking into account the individual characteristics of patients (such as medical conditions, drug allergies, etc.), and the degree of personalization reaches 92%, far exceeding the 70% and 75% of existing programs.

[0108] Drug safety: The prediction accuracy of drug interactions and allergic reactions in this solution reaches 98%, which is significantly higher than the existing technology solutions (85% and 90%).

[0109] Improved efficacy and reduced side effects: This regimen also performed better in terms of patient treatment effect and reduced side effects, with an improvement in efficacy of 90% and a reduction in side effects of 85%, which is a significant improvement over existing regimens (70% / 80% efficacy and 50% / 60% reduction in side effects).

[0110] Processing time: This solution takes a little longer to process, mainly because it requires the fusion of multi-source data and more complex personalized analysis, but this increased time is acceptable compared to the higher accuracy and personalization it provides.

[0111] The present invention is not limited to the above-mentioned embodiments. Anyone should be aware that any structural changes made under the enlightenment of the present invention, and any technical solutions that are the same or similar to the present invention, fall within the protection scope of the present invention. The technology, shape, and structural parts not described in detail in the present invention are all well-known technologies.

Claims

1. A tumor nursing information collaborative management system based on multi-source data fusion, characterized in that: include: Data collection module: automatically collects patients' medical data from multiple data sources; Data preprocessing module: performs preprocessing such as cleaning, denoising, format standardization, and missing value filling on the collected patient medication data to ensure data consistency and quality; Privacy protection module: Through differential privacy technology, noise is added during the processing of medication data to prevent the leakage of patients' sensitive information; Data fusion module: integrates medication data and medical data from different data sources to build a complete patient care file; Medication management module: Automatically generates personalized medication plans based on the comprehensive data provided by the data fusion module. Collaborative management module: provides a collaborative platform for multi-role medical staff, supporting cross-department collaboration and information sharing.

2. According to claim 1, a tumor nursing information collaborative management system based on multi-source data fusion is characterized in that: Using differential privacy technology, noise is added to the medication data processing to prevent the leakage of sensitive patient information. The steps are as follows: Analyze the patient's medication data, calculate the sensitivity of the query function Δf, and select an appropriate privacy budget ε based on the required privacy protection and accuracy requirements, where ε = 0.01; According to the obtained sensitivity Δf and privacy budget ε, the noise is generated using the noise generation mechanism of differential privacy, where the distribution of noise is based on the Laplace mechanism, and the expression is: Next, the generated noise is added to the statistical query results of the medication data.

3. The tumor nursing information collaborative management system based on multi-source data fusion according to claim 1 is characterized in that: The steps for analyzing the patient's medication data and calculating the sensitivity Δf of the query function are as follows: Define query function f, which is the statistical query operation to be performed on the patient medication data. Query function f can be used to calculate the frequency, total amount or other relevant statistical data of drug use; Calculate the maximum impact of a single patient's medication record change on the query result. Then, calculate the sensitivity Δf by calculating the difference in the query function results between the original dataset D and the modified dataset D′. The formula is as follows: They are the query results on the original dataset D and the modified dataset D′ respectively; Output the calculated sensitivity Δf, which represents the maximum impact of a single data record change on the query result.

4. The tumor nursing information collaborative management system based on multi-source data fusion according to claim 1, characterized in that: When integrating medication data from different data sources with medical data to build a complete patient care profile, the steps to integrate medication data with medical data are as follows: Use a lookup table to convert features from different data sources into embedding vectors and concatenate them together. For each word in the medical corpus text and the relationship between entities, represent it as a low-dimensional vector; The Skip-gram model of the Word2Vec tool is used to train medical corpus text through convolutional neural networks. The model infers the meaning of the current word through the probability of context words, thereby modeling the relationship between words and converting each word in the text into a k-dimensional vector; Use the convolution kernel to perform convolution operation on the input medical text to automatically extract features from the text. The convolution kernel extracts features at different positions through a sliding window to generate a feature set. The feature vector sequence is calculated through the convolution layer to generate a new feature set; The features output by the convolutional layer are filtered through the pooling layer to extract the most representative features, remove redundant data, and pass them to the next layer. The pooled features are further aggregated to form a global feature vector that represents the entire input data; The Softmax function is used to classify the features obtained by the pooling layer to generate the final treatment recommendation or medication plan. The output layer classifies the patient care information through weighted and biased operations and outputs the final decision. Finally, by integrating the patient's medication data, disease information, examination results, treatment plan and other data, a comprehensive patient care file is constructed to provide support for subsequent treatment decisions.

5. The tumor nursing information collaborative management system based on multi-source data fusion according to claim 1, characterized in that: Based on the comprehensive data provided by the data fusion module, a personalized medication plan is automatically generated. This module comprehensively considers the patient's condition, medication history, and drug interactions, and recommends the most appropriate drug combination and dosage steps as follows: Based on the fused data, key features are automatically extracted, including the patient's condition, medication history, drug allergy history, complication information, and examination results. The feature extraction process also analyzes the patient's health status, disease type, and historical treatment records. Generate personalized medication plans based on the patient's individual characteristics and clinical needs. The medication plan generation process takes into account the patient's condition, drug interactions, drug allergic reactions, treatment history factors, and combines treatment guidelines to automatically recommend the most appropriate drugs, dosages, and routes of administration for patients.

6. The tumor nursing information collaborative management system based on multi-source data fusion according to claim 1, characterized in that: The steps to generate a personalized medication regimen based on the patient's individual characteristics and clinical needs are: Use convolutional neural networks (CNNs) to build a personalized medication regimen generation model. CNN models can automatically learn complex features from patients' health data, especially extracting important information from high-dimensional data; Model training is performed based on historical patient data sets, and the patient's health data is used during the training process to optimize the parameters of the CNN model; The trained CNN model automatically generates personalized medication plans based on the individual characteristics and clinical needs of the patient. The model will recommend appropriate drugs, dosages, and routes of administration based on the patient's disease type, severity, and history of drug allergies; Based on the output of the CNN model, a personalized medication regimen report for the patient is generated, providing detailed drug name, recommended dosage, route of administration, and frequency of use.

7. The tumor nursing information collaborative management system based on multi-source data fusion according to claim 1 is characterized in that: The steps to provide a collaborative platform for multi-role medical staff and support cross-department collaboration and information sharing are as follows: Create a medical staff information registration system that allows different roles (such as doctors, nurses, pharmacists, laboratory personnel, etc.) to register information. Each medical staff member will obtain a unique role identification according to their function and department, and be authenticated; The system assigns corresponding access rights to medical staff in different roles to ensure data security and privacy protection. Set different data access and modification permissions based on the medical staff’s position and department; On the platform, medical staff can access the patient's complete health records, including medical history, medication records, treatment plans, examination results, etc. The records are updated in real time through data fusion technology to ensure that medical staff in different departments can obtain the latest information. Through information sharing, doctors, nurses, pharmacists, etc. can collaborate in a timely manner based on the patient's health status and treatment progress to ensure that the patient's treatment plan is consistently implemented.

Images