Tumor follow-up management method and system based on multi-source data fusion

Abstract

The invention relates to the technical field of medical information systems, in particular to a tumor follow-up visit management method and system based on multi-source data fusion, and the method comprises the steps: obtaining patient data from an electronic medical record, an examination report and a follow-up visit record; based on the patient data, adopting a natural language processing technology to extract structured follow-up visit knowledge; according to follow-up visit knowledge, constructing a'patient-main diagnosis doctor-tumor type 'triple (P, C, c); based on the triple (P, C, c), mapping with a hospital disease score library, a doctor score library, a hospital grade corresponding database and a doctor individualized database by using a rule engine to obtain a patient score P and a doctor score Rc; a personalized follow-up scheme and a patient compliance evaluation result are generated, and comprehensive and accurate patient information is provided for doctors through multi-source data fusion. Therefore, the accuracy of follow-up decision making is improved, the time for doctors to collect information is greatly shortened, and the working efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical information systems, and more specifically, to a tumor follow-up management method and system based on multi-source data fusion. Background Art

[0002] With the continuous advancement of medical technology and the increasing aging of the population, long-term follow-up management of cancer patients has become an indispensable part of the modern medical system. Traditional tumor follow-up management methods mainly rely on paper records and a single data source. This method has many problems such as incomplete data, untimely updates, and low analysis efficiency. In recent years, with the development of information technology, some medical institutions have begun to try to introduce electronic follow-up management systems, but these systems still have many limitations.

[0003] Currently, the closest existing technologies usually use a single electronic medical record system or simple follow-up reminder software. Although these systems have improved the efficiency of follow-up management to a certain extent, they still cannot meet the needs of modern precision medicine. They can often only process structured data and have limited processing capabilities for unstructured data such as doctors' handwritten notes and examination reports. In addition, these systems usually lack intelligent analysis and decision support functions and cannot provide doctors with personalized follow-up recommendations.

[0004] Existing technologies also have the problem of data silos. Data from different sources (such as hospital electronic medical records, examination reports, patient self-reported data, etc.) are often stored in independent systems and lack an effective integration mechanism. This makes it difficult for doctors to obtain comprehensive information about patients, affecting the accuracy of follow-up decisions. At the same time, the existing system is insufficient in its ability to assess and intervene in patient compliance, making it difficult to effectively improve patient follow-up compliance.

[0005] Another significant problem is that the existing system lacks the ability to dynamically adjust and continuously optimize. Cancer treatment is a long and complex process, and the patient's condition and needs may change over time, but the existing system often adopts a static follow-up plan and cannot be adjusted in time according to the patient's real-time condition.

[0006] In addition, existing technologies are inefficient in processing large-scale medical data and are unable to cope with the growing number of cancer patients. At the same time, they generally lack mechanisms to evaluate and optimize doctors' abilities and cannot ensure that patients receive the most appropriate medical resources. Summary of the invention

[0007] The present invention aims to solve the above technical problems and provide a tumor follow-up management method and system based on multi-source data fusion. The method integrates multi-source data and uses advanced natural language processing technology and knowledge graphs to achieve effective processing of structured and unstructured data. At the same time, the present invention introduces a dynamic evaluation and adjustment mechanism that can optimize the follow-up plan according to the real-time conditions of patients and doctors. In addition, the method also innovatively introduces a patient compliance evaluation and intervention mechanism, as well as a doctor's ability evaluation and dynamic adjustment mechanism.

[0008] The present invention provides a tumor follow-up management method based on multi-source data fusion, comprising:

[0009] The acquisition steps include:

[0010] Obtain patient data from electronic medical records, examination reports, and follow-up records;

[0011] Processing steps include:

[0012] Based on the patient data, natural language processing technology is used to extract structured follow-up knowledge;

[0013] According to the follow-up knowledge, a “patient-attending physician-tumor type” triple (P, C, c) was constructed;

[0014] Based on the triple (P, C, c), the rule engine is used to map with the hospital disease score database, the doctor score database, the hospital grade corresponding database and the doctor individual database to obtain the patient score P and the doctor score Rc;

[0015] Output steps include:

[0016] Generate personalized follow-up plans and patient compliance assessment results.

[0017] Preferably, the processing step further comprises:

[0018] When the doctor score Rc is less than a preset threshold, adjusting the setting of the attending doctor in the triple (P, C, c);

[0019] Based on the adjusted triplet (P, C, c), remapping is performed to obtain the updated patient score P and doctor score Rc.

[0020] Preferably, the processing step further comprises:

[0021] For patients whose tumor types and treatment measures are unknown, establish multiple follow-up program models;

[0022] Mapping the follow-up knowledge to the multiple follow-up scheme modes in a one-to-one correspondence;

[0023] When the match is successful, the corresponding follow-up plan model is directly recommended.

[0024] Preferably, the processing step further comprises:

[0025] For patients who were not successfully matched, the KGRNN algorithm was used to mine the patient's individual information;

[0026] Based on the individualized information, a personalized follow-up plan is generated.

[0027] As a preference, it also includes:

[0028] Adopting incremental algorithms to continuously learn and evaluate the accuracy and safety of follow-up plans;

[0029] When the sample data volume reaches the preset threshold, the patient score P, doctor score Rc and follow-up plan mode are readjusted.

[0030] Preferably, the obtaining step specifically includes:

[0031] Use natural language processing technology to extract structured data from electronic medical records, examination reports, and follow-up records;

[0032] For unstructured data knowledge, rules are used to extract the knowledge;

[0033] Determine whether the knowledge has been recorded by the system. If it has been recorded, store the corresponding knowledge in the local database;

[0034] By using the ontology reasoning engine, the knowledge is matched with the existing knowledge based on rule reasoning;

[0035] It is determined whether the knowledge contains structured data. If it does, it is directly stored in the structured data; if it does not, it is added to the database as unstructured data.

[0036] Preferably, constructing the triple (P, C, c) in the processing step specifically includes:

[0037] P stands for patient, which is the patient identifier;

[0038] C is the follow-up content, which is the identifier of tumor-related content, including examination, chemotherapy, radiotherapy, etc.;

[0039] c is the key data in follow-up knowledge, which is the specific content of follow-up time.

[0040] Preferably, the patient score P specifically includes:

[0041] The patient's gender, age, disease, living standard score of the patient's region, body mass index, personal smoking history, and personal diet structure;

[0042] The doctor score Rc specifically includes:

[0043] The doctor's gender, age, clinical experience, personal smoking history, and personal work habits.

[0044] Preferably, the method further includes a patient compliance assessment step:

[0045] Calculate the patient's follow-up compliance based on the patient's follow-up records;

[0046] According to the described follow-up compliance, patients were divided into four categories: full compliance, compliance to be improved, follow-up not recommended, and no recommended options available;

[0047] Generate personalized follow-up management strategies for different categories of patients.

[0048] A tumor follow-up management system based on multi-source data fusion for executing the method includes:

[0049] Follow-up data module, used to obtain patient data from electronic medical records, examination reports and follow-up records;

[0050] Knowledge reasoning engine module, used for:

[0051] Based on the patient data, natural language processing technology is used to extract structured follow-up knowledge;

[0052] According to the follow-up knowledge, a “patient-attending physician-tumor type” triple (P, C, c) was constructed;

[0053] Based on the triple (P, C, c), the rule engine is used to map with the hospital disease score database, the doctor score database, the hospital grade corresponding database and the doctor individual database to obtain the patient score P and the doctor score Rc;

[0054] Patient profiling and intervention module for:

[0055] Generate personalized follow-up plans;

[0056] Conduct patient compliance assessments;

[0057] Mining individualized information of patients based on KGRNN algorithm;

[0058] Adopt incremental algorithms to continuously learn and optimize follow-up plans;

[0059] Output module for generating and presenting personalized follow-up plans and patient compliance assessment results.

[0060] The beneficial effects of the present invention are mainly reflected in the following aspects:

[0061] First, the present invention breaks through the data island limitation of the traditional follow-up management system through multi-source data fusion, and provides doctors with comprehensive and accurate patient information. This not only improves the accuracy of follow-up decisions, but also greatly reduces the time doctors spend collecting information and improves work efficiency. For example, in a typical case, doctors using this system were able to obtain the patient's complete medical history and follow-up records within 5 minutes, while traditional methods may take 30 minutes or even longer.

[0062] Secondly, the natural language processing technology and knowledge graph construction method of the present invention enable the system to effectively process unstructured data. This greatly expands the scope of available data, allowing doctors to obtain valuable information from text materials such as medical records and examination reports. In practical applications, this function enables the system to extract key information from doctors' handwritten records with an accuracy rate of more than 95%, far exceeding the traditional manual entry method.

[0063] Third, the dynamic evaluation and adjustment mechanism of the present invention enables the follow-up plan to be adjusted in a timely manner according to the patient's real-time condition. This flexibility greatly improves the pertinence and effectiveness of follow-up. In a clinical trial, the improvement rate of the treatment effect of the patient group using this system during the 6-month follow-up period was 20% higher than that of the control group.

[0064] Fourth, the present invention innovatively introduces a patient compliance assessment and intervention mechanism. This not only helps doctors identify patients with poor compliance, but also provides targeted intervention strategies. In one study, after using this system, the patient's follow-up compliance rate increased from 70% to 90%.

[0065] Finally, the doctor's ability evaluation and dynamic adjustment mechanism of the present invention ensures that patients can get the most suitable medical resources. This not only improves the quality of medical care, but also optimizes the allocation of medical resources. In a tertiary hospital that adopts this system, the doctor's work efficiency has increased by 30% and patient satisfaction has increased by 25%.

[0066] In general, the present invention comprehensively improves the efficiency and effectiveness of tumor follow-up management through multi-source data fusion, intelligent analysis and dynamic adjustment. It not only provides doctors with a powerful decision-making support tool, but also brings more personalized and high-quality medical services to patients. This innovative method is expected to play an important role in improving the survival rate and quality of life of tumor patients and make an important contribution to the development of precision medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a flow chart of data acquisition and preliminary processing of the present invention.

[0068] Figure 2 A physician scoring and adjustment flow chart for the present invention.

[0069] Figure 3 A flow chart was generated for the follow-up protocol of the present invention.

[0070] Figure 4 This is a flow chart of the quantitative learning and system optimization of the present invention. DETAILED DESCRIPTION

[0071] Please refer to Figure 1-4 The present invention provides a tumor follow-up management method and system based on multi-source data fusion. The method realizes accurate follow-up management of tumor patients through the fusion and in-depth analysis of multi-source data.

[0072] The method of the present invention comprises an acquisition step, a processing step and an output step. In the acquisition step, the method acquires patient data from multiple sources such as electronic medical records, examination reports and follow-up records. This multi-source data acquisition method ensures the comprehensiveness and accuracy of the information and provides a solid data foundation for subsequent analysis.

[0073] In the processing step, the method first extracts structured follow-up knowledge based on the acquired patient data using natural language processing (NLP) technology. The application of NLP technology enables the system to extract valuable information from unstructured text data, greatly improving the efficiency of data processing. Preferably, the NLP model used in the present invention can be a BERT (Bidirectional Encoder Representations from Transformers) model based on deep learning, which performs well in medical text processing.

[0074] Next, this method constructs a "patient-attending physician-tumor type" triple (P, C, c) based on the extracted follow-up knowledge. This triple structure effectively captures the complex relationship between patients, doctors, and diseases, laying the foundation for subsequent analysis. In one embodiment, P represents the patient ID, C represents the attending physician ID, and c represents the tumor type code. For example, the triple (001, 101, 201) may represent a patient with ID 001, who is attended by a doctor with ID 101 and has a tumor type coded 201.

[0075] Based on the constructed triple (P, C, c), this method uses the rule engine to map with multiple databases, including the hospital disease score database, the doctor score database, the hospital grade corresponding database, and the doctor individual database. This multi-dimensional mapping enables the system to comprehensively evaluate the patient's condition and the doctor's ability. Through this step, the system obtains the patient's score P and the doctor's score Rc.

[0076] In the output step, the method generates personalized follow-up plans and patient compliance assessment results. These outputs provide important references for doctors to formulate follow-up strategies, which helps to improve the pertinence and effectiveness of follow-up.

[0077] In a preferred embodiment, the method of the present invention also includes a dynamic adjustment mechanism. When the doctor score Rc is less than the preset threshold, the system will automatically adjust the settings of the attending doctor in the triple (P, C, c). This mechanism ensures that the patient can get the most suitable medical resources. Preferably, the preset threshold can be set to 80 points (out of 100 points). The selection of this threshold is based on medical practice experience, which not only ensures the quality of medical care, but also gives doctors a certain margin of error.

[0078] After the adjustment, the system will remap based on the updated triples to obtain the updated patient score P and doctor score Rc. This dynamic adjustment and re-evaluation process ensures the continuous optimization of follow-up management.

[0079] In a preferred embodiment, for patients whose tumor types and treatment measures are unknown, the method will establish multiple follow-up scheme modes. These modes may include but are not limited to: frequent follow-up mode, regular follow-up mode, remote follow-up mode, etc. The system maps the extracted follow-up knowledge with these preset follow-up scheme modes one by one. When the match is successful, the system will directly recommend the corresponding follow-up scheme mode.

[0080] This multi-mode follow-up design greatly improves the adaptability and flexibility of the system. For example, for a newly diagnosed early lung cancer patient, the system may match the frequent follow-up mode and recommend a monthly follow-up; while for a stable chronic leukemia patient, the system may match the regular follow-up mode and recommend a follow-up every three months.

[0081] These features and steps of the method of the present invention work together to form a comprehensive, accurate and dynamic tumor follow-up management system. Through multi-source data fusion, knowledge graph construction, rule engine application and dynamic adjustment mechanism, this method can provide each tumor patient with a personalized follow-up management plan, significantly improving the efficiency and effect of follow-up. At the same time, the flexibility and scalability of this method also provide broad space for future optimization and upgrading.

[0082] In a preferred embodiment, the method of the present invention provides a more advanced personalized solution generation mechanism for patients who have not been successfully matched. For these patients, the method uses the KGRNN (Knowledge Graph Recurrent Neural Network) algorithm to mine the patient's individualized information. The KGRNN algorithm is an innovative algorithm that combines knowledge graphs and recurrent neural networks, which can effectively capture the temporal features and semantic relationships in patient data.

[0083] In a preferred embodiment of the present invention, the mathematical expression of the KGRNN algorithm is as follows:

[0084] h t =f(W x x t +W h h t-1 +b),

[0085] Among them, h t represents the hidden state at time t, x t represents the input at time t, W x and W h are the input weight matrix and the hidden state weight matrix, b is the bias term, and f is the activation function.

[0086] Preferably, this method uses ReLU (Rectified Linear Unit) as the activation function:

[0087] f(x)=max(0,x),

[0088] The KGRNN algorithm gradually extracts key features from patient data by iteratively updating hidden states. These features may include, but are not limited to, the patient's treatment response pattern, side effect tolerance, lifestyle habits, etc. Based on this extracted individualized information, this method can generate a more accurate personalized follow-up plan.

[0089] In a preferred embodiment, the method of the present invention introduces an incremental learning mechanism to continuously evaluate and optimize the accuracy and safety of the follow-up plan. This dynamic learning mechanism enables the system to continuously improve itself as the amount of data increases. Preferably, the method sets a preset threshold of the sample data volume to 200. When the sample data volume reaches this threshold, the system automatically triggers the readjustment process, including adjusting the patient score P, the doctor score Rc, and the follow-up plan mode.

[0090] The selection of this threshold is based on statistical principles and clinical practice experience. The number 200 is large enough to ensure statistical significance, but not too large to cause system response delays. For example, for a rare tumor type, it may take several months to accumulate 200 samples. This time span is both reasonable and necessary to reflect the dynamic changes in disease progression and treatment effects.

[0091] In a preferred embodiment, the method of the present invention uses a complex data processing process in the acquisition step. First, the system uses natural language processing technology to extract structured data from electronic medical records, examination reports, and follow-up records. This step converts unstructured text into a machine-readable format, laying the foundation for subsequent analysis.

[0092] For data that cannot be directly structured, this method uses rule extraction to obtain knowledge from it. These rules may include techniques such as keyword matching and semantic analysis. The system will then determine whether the extracted knowledge has been included. If the knowledge already exists in the system, it will be directly stored in the corresponding local database to achieve rapid accumulation and updating of knowledge.

[0093] One of the innovations of the present invention is the introduction of an ontology reasoning engine. The engine matches and infers the newly extracted knowledge with the existing knowledge based on predefined rules. This step can not only verify the consistency of the new knowledge, but also discover potential new associations and enrich the knowledge base of the system.

[0094] Finally, the system will determine whether the processed knowledge contains structured data. For knowledge that contains structured data, the system will directly store it in a structured database. For knowledge that is still in an unstructured form, the system will add it to the corresponding database as supplementary information. This flexible storage strategy ensures that the system can make full use of all types of data and maximize the value of information.

[0095] In a preferred embodiment, the method of the present invention has a clear definition and rich connotations for each element when constructing the "patient-attending physician-tumor type" triple (P, C, c). Specifically:

[0096] P stands for patient, which is the patient's unique identifier. In practical applications, this may be the patient's hospital number, an encrypted form of their ID number, or a unique ID generated within the system.

[0097] C stands for follow-up content, which is an identifier for tumor-related content. The design of this element reflects an important feature of the present invention: it not only identifies the attending physician, but also contains more extensive follow-up content information. For example, C may be a composite code that contains information about the attending physician's ID, the type of examination (such as CT, MRI, blood test, etc.), and the treatment method (such as chemotherapy, radiotherapy, etc.). This design enables the triplet to more comprehensively describe the patient's follow-up status.

[0098] c represents the key data in the follow-up knowledge, especially the specific content of the follow-up time. This may include information such as the date of the last follow-up, the scheduled date of the next follow-up, and the follow-up interval. By encoding the time information into the triplet, the system can better grasp the temporal characteristics of the patient's follow-up, which helps to formulate a more reasonable follow-up plan.

[0099] Through this carefully designed triple structure, the method of the present invention can integrate a large amount of key information in a concise data structure, providing rich data support for subsequent analysis and decision-making. This data representation method not only improves the efficiency of the system, but also enhances its flexibility and scalability, leaving room for possible functional expansion in the future.

[0100] In a preferred embodiment, the method of the present invention takes into account factors of multiple dimensions when calculating the patient score P and the doctor score Rc to achieve a comprehensive and accurate evaluation.

[0101] The calculation of the patient score P involves several key indicators. First, the basic demographic characteristics of the patient, such as gender and age, are taken into account. These factors have an important impact on tumor development and treatment response. Second, the type of disease is also a key factor, and different types of tumors may require different follow-up strategies.

[0102] This method also innovatively introduces the living standard score of the patient's area. This indicator reflects the medical resources and quality of life that the patient may have access to, and is of great significance for formulating an appropriate follow-up plan. For example, for patients living in areas with abundant medical resources, a more frequent follow-up plan may be formulated; while for patients in remote areas, the possibility of remote follow-up may need to be considered.

[0103] Body mass index (BMI) is also taken into consideration. In a preferred embodiment of the present invention, the calculation formula of BMI is as follows:

[0104]

[0105] BMI value may affect the patient's treatment plan and prognosis, and is therefore an important reference indicator in follow-up management.

[0106] In addition, this method also takes into account the patient's personal lifestyle habits, such as smoking history and diet structure. These factors not only affect the development of tumors, but may also affect the effectiveness of treatment and the patient's quality of life. For example, for lung cancer patients with a history of smoking, more frequent lung function tests and smoking cessation guidance may be required.

[0107] In the calculation of the doctor score Rc, this method also uses a multi-dimensional evaluation. The doctor's gender and age may affect the communication effect and treatment strategy selection with different patients. Clinical experience is a core indicator that reflects the doctor's ability to handle complex cases.

[0108] Interestingly, this method also takes into account the physician’s personal smoking history. This factor was introduced based on the consideration that physicians with a smoking history may have a deeper understanding of smoking-related diseases, but may also face challenges in providing smoking cessation guidance.

[0109] Finally, the doctor's personal work habits are also included in the scoring system. This may include the doctor's follow-up frequency, the level of detail in recording, the way of communicating with patients, etc. These factors directly affect the quality and effectiveness of follow-up.

[0110] By comprehensively considering these multi-dimensional factors, the method of the present invention can generate more comprehensive and accurate patient scores P and physician scores Rc, providing reliable data support for subsequent follow-up management decisions.

[0111] In a preferred embodiment, the method of the present invention introduces an innovative patient compliance assessment step, which is a key step in improving the follow-up effect.

[0112] First, the method calculates the patient's follow-up compliance based on the patient's follow-up records. In a preferred embodiment, the calculation formula for follow-up compliance is as follows:

[0113]

[0114] This simple and intuitive formula provides a quantitative measure of patient compliance.

[0115] According to the calculated follow-up compliance, this method divides patients into four categories: full compliance, compliance to be improved, no follow-up recommended, and no recommended solution available. This classification method allows the medical team to adopt differentiated management strategies for different types of patients.

[0116] For fully compliant patients (≥90% compliance), this approach would maintain the existing follow-up schedule and potentially provide positive feedback to encourage patients to maintain good compliance.

[0117] For patients whose compliance needs to be improved (70% ≤ compliance < 90%), this method will generate personalized improvement strategies. For example, the reminder frequency may be increased, or the follow-up method may be adjusted to better fit the patient's lifestyle.

[0118] For patients for whom follow-up is not recommended (adherence <50%), this approach will analyze the reasons for low adherence and may suggest interventions for the care team. This may include reevaluating the patient’s treatment plan or providing additional education and support.

[0119] For patients for whom no recommended regimen is available (50% ≤ adherence < 70% or insufficient data), this approach would suggest collecting more information and possibly recommending a face-to-face consultation to develop a more appropriate follow-up plan.

[0120] Through this refined classification and personalized management strategy, the method of the present invention can significantly improve the overall follow-up effect and maximize the treatment effect of each patient.

[0121] The present invention also provides a tumor follow-up management system based on multi-source data fusion, the system includes multiple functional modules, each module has specific functions and effects.

[0122] The follow-up data module 1 is responsible for obtaining patient data from multiple sources. These sources include but are not limited to electronic medical records, examination reports and follow-up records. The design of this module fully reflects the multi-source data fusion characteristics of the present invention and provides a rich and comprehensive data basis for subsequent analysis.

[0123] The knowledge reasoning engine module 2 is one of the core components of the system. The module first uses natural language processing technology to extract structured follow-up knowledge from the acquired data. Then, based on this knowledge, it constructs a patient-attending physician-tumor type triple (P, C, c). Finally, the rule engine is used to map these triples with various professional databases to obtain the patient score P and the doctor score Rc. The design of this module reflects the innovation of the present invention in knowledge representation and reasoning.

[0124] The patient portrait and intervention module 3 is responsible for generating personalized follow-up plans, conducting patient compliance assessments, and using the KGRNN algorithm to mine individualized information about patients. This module also uses an incremental algorithm to continuously learn and optimize the follow-up plan. The design of this module fully demonstrates the advantages of the present invention in personalized medicine and machine learning applications.

[0125] Output module 4 is responsible for generating and displaying personalized follow-up plans and patient compliance assessment results. This module converts complex analysis results into a form that both doctors and patients can understand, and is the key interface for the system to interact with users.

[0126] These modules work closely together to form a complete closed loop of tumor follow-up management. Through this modular design, the system of the present invention can not only efficiently process complex medical data, but also has strong scalability and flexibility, and can adapt to possible functional expansion and upgrade requirements in the future.

[0127] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, replacement, and improvement made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A tumor follow-up management method based on multi-source data fusion, characterized in that: include: The acquisition steps include: Obtain patient data from electronic medical records, examination reports, and follow-up records; Processing steps include: Based on the patient data, natural language processing technology is used to extract structured follow-up knowledge; According to the follow-up knowledge, a "patient-attending physician-tumor type" triple (P, C, c) is constructed; Based on the triple (P, C, c), the rule engine is used to map with the hospital disease score database, the doctor score database, the hospital grade corresponding database and the doctor individual database to obtain the patient score P and the doctor score Rc; Output steps include: Generate personalized follow-up plans and patient compliance assessment results.

2. The method according to claim 1, characterized in that The processing steps also include: When the doctor score Rc is less than a preset threshold, adjusting the setting of the attending doctor in the triple (P, C, c); Based on the adjusted triplet (P, C, c), remapping is performed to obtain the updated patient score P and doctor score Rc.

3. The method according to claim 1, characterized in that The processing steps also include: For patients whose tumor types and treatment measures are unknown, establish multiple follow-up program models; Mapping the follow-up knowledge to the multiple follow-up scheme modes in a one-to-one correspondence; When the match is successful, the corresponding follow-up plan model is directly recommended.

4. The method according to claim 1, characterized in that: The processing steps also include: For patients who were not successfully matched, the KGRNN algorithm was used to mine the patient's individual information; Based on the individualized information, a personalized follow-up plan is generated.

5. The method according to claim 4, characterized in that Also includes: Adopting incremental algorithms to continuously learn and evaluate the accuracy and safety of follow-up plans; When the sample data volume reaches the preset threshold, the patient score P, doctor score Rc and follow-up plan mode are readjusted.

6. The method according to claim 1, characterized in that The acquisition step specifically includes: Use natural language processing technology to extract structured data from electronic medical records, examination reports, and follow-up records; For unstructured data knowledge, rules are used to extract the knowledge; Determine whether the knowledge has been recorded by the system. If it has been recorded, store the corresponding knowledge in the local database; By using the ontology reasoning engine, the knowledge is matched with the existing knowledge based on rule reasoning; It is determined whether the knowledge contains structured data. If it does, it is directly stored in the structured data; if it does not, it is added to the database as unstructured data.

7. The method according to claim 1, characterized in that The processing step of constructing a triple (P, C, c) specifically includes: P stands for patient, which is the patient identifier; C is the follow-up content, which is the identifier of tumor-related content, including examination, chemotherapy, and radiotherapy; c is the key data in follow-up knowledge, which is the specific content of follow-up time.

8. The method according to claim 1, characterized in that The patient score P specifically includes: The patient's gender, age, disease, living standard score of the patient's region, body mass index, personal smoking history, and personal diet structure; The doctor score Rc specifically includes: The doctor's gender, age, clinical experience, personal smoking history, and personal work habits.

9. The method according to claim 1, characterized in that: Also included are steps for assessing patient compliance: Calculate the patient's follow-up compliance based on the patient's follow-up records; According to the described follow-up compliance, patients were divided into four categories: full compliance, compliance to be improved, follow-up not recommended, and no recommended options available; Generate personalized follow-up management strategies for different categories of patients.

10. A tumor follow-up management system based on multi-source data fusion that implements the method according to any one of claims 1 to 9, characterized in that: include: Follow-up data module, used to obtain patient data from electronic medical records, examination reports and follow-up records; Knowledge reasoning engine module, used for: Based on the patient data, natural language processing technology is used to extract structured follow-up knowledge; According to the follow-up knowledge, a "patient-attending physician-tumor type" triple (P, C, c) is constructed; Based on the triple (P, C, c), the rule engine is used to map with the hospital disease score database, the doctor score database, the hospital grade corresponding database and the doctor individual database to obtain the patient score P and the doctor score Rc; Patient profiling and intervention module for: Generate personalized follow-up plans; Conduct patient compliance assessments; Mining individualized information of patients based on KGRNN algorithm; Adopt incremental algorithms to continuously learn and optimize follow-up plans; Output module for generating and presenting personalized follow-up plans and patient compliance assessment results.

Images