Gynecological tumor patient intelligent health propaganda method, device and equipment and storage medium

Abstract

This invention discloses an intelligent health education method, device, equipment, and storage medium for gynecological tumor patients. The method acquires the patient's electronic medical record information and patient interaction data. Based on the electronic medical record information, it identifies the gynecological tumor disease type and treatment stage tags. Based on the patient interaction data, it specifically identifies psychological stress signals related to gynecological tumor treatment, generating a profile of the patient's current psychosomatic state. It matches and dynamically generates personalized health education content and psychological counseling plans from a structured knowledge base. Based on the patient's treatment data, it automatically generates reminders for medication and follow-up examinations. It structurally correlates the output records of personalized health education content and the patient's feedback data on the reminders to form a digital pathmap for patient education. This enables dynamic matching and personalized generation of health education content and psychological counseling plans, improving the patient's cognitive adaptability to health information and the effectiveness of psychological support, achieving closed-loop management throughout the entire hospital and inpatient cycle.

Description

Technical Field

[0001] This invention relates to the field of smart medical digital health management technology, and in particular to a method, device, equipment and storage medium for intelligent health education of gynecological tumor patients. Background Technology

[0002] Currently, in gynecologic oncology clinical nursing, health education for patients mainly relies on nurses' verbal explanations, supplemented by some standardized paper manuals or static electronic materials.

[0003] These existing technical solutions are standardized in content and cannot adapt to the personalized needs of patients with different disease types, treatment stages, and cognitive levels; the education process is one-way and lacks effective interaction and feedback mechanisms; moreover, the effectiveness of education is difficult to quantify and evaluate; existing technical problems include: (1) Low efficiency and effectiveness of information transmission: Patients are in a state of stress, which can easily lead to information overload and forgetting, and professional terminology can cause misunderstandings; (2) The content of the education is homogeneous and lacks specificity: it is impossible to dynamically generate exclusive content based on the individual patient's situation (such as after ovarian cancer surgery or during cervical cancer chemotherapy); (3) Lack of immediate interaction and emotional support: Patients' questions cannot be answered accurately and in a simple way at any time, and they lack psychological counseling; (4) The education process is not traceable and there are blind spots in management: the management chain from education to home follow-up has not formed a closed loop, nurses have heavy repetitive work and there are medical safety risks.

[0004] Especially for patients with gynecological tumors (such as ovarian cancer and cervical cancer), health education faces more complex specialist challenges: First, the treatment options (surgery, chemotherapy, targeted therapy, endocrine therapy, etc.) are diverse and have a unique spectrum of side effects (such as ovarian function suppression and neurotoxicity caused by chemotherapy, and radiation enteritis / cystitis caused by radiotherapy). The information that patients need to understand is highly specialized and constantly changing.

[0005] Secondly, patients often experience intense specific psychological stress, such as anxiety about loss of fertility, changes in body shape, sexual dysfunction, and family roles. These psychological states directly affect their treatment decisions and adherence, and general psychological support is often insufficient to address these core pain points.

[0006] Furthermore, the treatment cycle is long and involves multiple stages both inside and outside the hospital, requiring extremely high standards for medication safety and continuous management of follow-up visits across the cycle.

[0007] Existing standardized health education tools or general health robots are completely inadequate to meet the comprehensive needs of specialized, personalized, holistic, and full-cycle management, resulting in a significant information support gap and safety risks for this patient group. Summary of the Invention

[0008] The main objective of this invention is to provide an intelligent health education method, device, equipment, and storage medium for gynecological cancer patients, aiming to solve the technical problems in the existing health education for gynecological cancer patients, such as information homogenization, lack of targeted psychological support, and disconnection from outpatient management leading to untraceable effects.

[0009] In a first aspect, the present invention provides an intelligent health education method for gynecological tumor patients, the intelligent health education method for gynecological tumor patients comprising the following steps: The system acquires the patient's electronic medical record information and patient interaction data, identifies the gynecological tumor disease type and treatment stage label based on the electronic medical record information, and specifically identifies psychological stress signals related to gynecological tumor treatment based on the patient interaction data, thereby generating a current psychosomatic profile of the patient. Based on the current psychosomatic state profile, personalized educational content and psychological counseling plans are dynamically generated from a structured knowledge base and output to the patient through an interactive terminal. Based on the patient's treatment data, medication and follow-up reminders are automatically generated, and the output records of the personalized education content and the patient's feedback data on the reminders are structurally linked to form the patient's digital education path map.

[0010] Optionally, the step of acquiring the patient's electronic medical record information and patient interaction data, identifying the gynecological tumor disease type and treatment stage label based on the electronic medical record information, and specifically identifying psychological stress signals related to gynecological tumor treatment based on the patient interaction data to generate a current psychosomatic profile of the patient includes: The patient's electronic medical record is obtained from the hospital information system through a secure encrypted interface. Natural language processing technology is used to perform structured parsing of the electronic medical record to identify the disease type and TNM stage of the gynecological tumor and determine the current treatment stage label. The patient's voice information is collected in real time through the microphone of the interactive terminal, and the patient's facial expression video data is captured through the camera of the interactive terminal. The facial expression video data is anonymized and desensitized to obtain the target facial expression video data after anonymization and desensitization. The patient's text information is collected in real time through the input module of the interactive terminal. The voice information, the target facial expression video data and the facial expression information are used as patient interaction data. The patient interaction data is semantically compared with a pre-set gynecological tumor psychosomatic keyword feature library, and a multimodal emotion analysis model is used to perform weighted calculations on voice tone and facial micro-expressions to specifically identify psychological stress signals closely related to gynecological tumor treatment. The disease type, treatment stage label, and psychological stress signal are fused and mapped in multiple dimensions to construct a current psychosomatic profile of the patient that includes both physiological treatment progress and psychological stress dimensions.

[0011] Optionally, the step of matching and dynamically generating personalized educational content and psychological counseling plans from a structured knowledge base based on the current psychosomatic state profile, and outputting them to the patient through an interactive terminal, includes: The current psychosomatic state profile of the patient is deconstructed to extract disease stage labels, treatment plan labels, and psychological stress type labels, and a multidimensional feature vector for retrieval is constructed. In the structured knowledge base, semantic matching and weight calculation are performed using the multidimensional feature vectors to locate the standard nursing education material node corresponding to the disease stage label, the treatment cycle guidance node corresponding to the treatment plan label, and the psychological counseling strategy node matching the psychological stress type label. Using natural language generation and multimedia synthesis technologies, the located standard nursing education material nodes, treatment cycle guidance nodes, and psychological counseling strategy nodes are dynamically assembled and semantically fused to generate personalized education content and psychological counseling plans that are adapted to the patient's current cognitive level and emotional resilience, and then output to the patient through an interactive terminal.

[0012] Optionally, the step of deconstructing the patient's current psychosomatic state profile to extract disease stage labels, treatment plan labels, and psychological stress type labels, and constructing a multidimensional feature vector for retrieval, includes: Access the data structure of the patient's current psychosomatic state profile, parse the physiological treatment dimension data mapped from the electronic medical record information in the data structure, and extract the disease stage label representing the degree of tumor progression and the treatment plan label representing the current medical intervention from the physiological treatment dimension data; The psychological state dimension data formed by patient interaction data analysis in the data structure is parsed, and psychological stress type labels representing the patient's emotional focus are extracted from the psychological state dimension data; The disease stage label, treatment plan label, and psychological stress type label are encoded and combined according to a preset dimensional structure to construct a multidimensional feature vector containing physiological treatment features and psychological state features. The multidimensional feature vector serves as a retrieval index to locate knowledge nodes in the structured knowledge base that match the patient's current state.

[0013] Optionally, the method employs natural language generation and multimedia synthesis technologies to dynamically assemble and semantically fuse the located standard nursing education material nodes, treatment cycle guidance nodes, and psychological counseling strategy nodes to generate personalized education content and psychological counseling plans adapted to the patient's current cognitive level and emotional resilience. These personalized plans are then output to the patient through an interactive terminal, including: The pre-trained natural language generation model is invoked to dynamically adjust the vocabulary complexity and sentence structure of the educational and guidance text to be generated based on the current psychosomatic state profile of the patient. Multimedia synthesis technology is then used to convert the generated text content into speech or combine it with animated videos for presentation. During the dynamic assembly process, the standard nursing education material nodes and the treatment cycle guidance nodes are logically sorted according to the treatment timeline, and the psychological counseling strategy nodes are inserted as an emotional buffer layer between the key information nodes. The semantic fusion algorithm eliminates expression conflicts between different nodes, and dynamically adjusts the tone and information density of the output content according to the psychological stress signal to generate personalized educational content and psychological counseling plans in the form of text, audio and video, which are then output to the patient through the interactive terminal.

[0014] Optionally, the step of automatically generating medication and follow-up reminder tasks based on the patient's treatment data, and structurally associating the output records of the personalized education content and the patient's feedback data on the reminder tasks to form an educational digital path map for the patient, including: By synchronizing patients' in-hospital treatment data through a secure data interface, and using a pre-set treatment plan parser to identify key treatment nodes, reminders for medication and follow-up examinations are automatically generated. The timestamp of the personalized education content push, the patient's viewing time, and interactive click behavior will be captured in real time and recorded as output records; Simultaneously, the patient's confirmation of the reminder task, extension application, or side effect report information is collected as feedback data; using the patient's unique anonymous identifier as an index, the patient interaction data, the key treatment nodes, the output records, and the feedback data are structured, stored, and mapped in a pre-defined time series to construct a visualized digital educational path map for the patient.

[0015] Optionally, the step of synchronizing the patient's in-hospital treatment data through a secure data interface, identifying key treatment nodes using a pre-set treatment plan parser, and automatically generating medication and follow-up reminders includes: The patient's in-hospital treatment data is synchronized through a secure data interface. The in-hospital treatment data includes the start date of the chemotherapy cycle, the duration of targeted drug maintenance therapy, and the postoperative follow-up examination time. Based on the start date of the chemotherapy cycle, the duration of the targeted drug maintenance therapy, and the postoperative follow-up examination time, a medication and follow-up reminder task is generated, which includes the drug name, dosage, side effect warnings, and pre-examination preparations, and is pushed out on time via mobile or robot terminals.

[0016] Secondly, to achieve the above objectives, the present invention also proposes an intelligent health education device for gynecological tumor patients, the intelligent health education device for gynecological tumor patients comprising: The identification and analysis module is used to acquire the patient's electronic medical record information and patient interaction data, identify the gynecological tumor disease type and treatment stage label based on the electronic medical record information, and specifically identify psychological stress signals related to gynecological tumor treatment based on the patient interaction data, and generate a current psychosomatic profile of the patient. The solution generation module is used to match and dynamically generate personalized education content and psychological counseling solutions from a structured knowledge base based on the current psychosomatic state profile, and output them to the patient through an interactive terminal. The reminder association module is used to automatically generate medication and follow-up reminder tasks based on the patient's treatment data, and to structurally associate the output records of the personalized education content and the patient's feedback data on the reminder tasks to form the patient's education digital path map.

[0017] Thirdly, to achieve the above objectives, the present invention also proposes an intelligent health education device for gynecological tumor patients. The intelligent health education device for gynecological tumor patients includes: a memory, a processor, and an intelligent health education program for gynecological tumor patients stored in the memory and executable on the processor. The intelligent health education program for gynecological tumor patients is configured to implement the steps of the intelligent health education method for gynecological tumor patients as described above.

[0018] Fourthly, to achieve the above objectives, the present invention also proposes a storage medium storing a smart health education program for gynecological tumor patients, wherein the smart health education program for gynecological tumor patients, when executed by a processor, implements the steps of the smart health education method for gynecological tumor patients as described above.

[0019] The intelligent health education method for gynecologic oncology patients proposed in this invention acquires the patient's electronic medical record information and patient interaction data. Based on the electronic medical record information, it identifies the gynecologic oncology disease type and treatment stage tags, and based on the patient interaction data, it specifically identifies psychological stress signals related to gynecologic oncology treatment, generating a current psychosomatic profile of the patient. Based on this current psychosomatic profile, it matches and dynamically generates personalized education content and psychological counseling plans from a structured knowledge base, and outputs them to the patient through an interactive terminal. Based on the patient's treatment data, it automatically generates medication and follow-up reminder tasks, and structurally correlates the output records of the personalized education content and the patient's feedback data on the reminder tasks to form an educational digital path map for the patient. By integrating electronic medical record information and multimodal patient interaction data, it can accurately construct a current psychosomatic profile of the patient, achieving dynamic matching and personalized generation of education content and psychological counseling plans. This significantly improves the patient's cognitive adaptability to health information and the effectiveness of psychological support. Furthermore, through the automatic generation of medication and follow-up tasks and the structured correlation of the entire process of interaction data, a traceable educational digital path map is formed, achieving closed-loop management throughout the entire hospital and outpatient cycle. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the first embodiment of the intelligent health education method for gynecological tumor patients of the present invention; Figure 3 This is a flowchart illustrating the second embodiment of the intelligent health education method for gynecological tumor patients of the present invention; Figure 4 This is a flowchart illustrating the third embodiment of the intelligent health education method for gynecological tumor patients of the present invention. Figure 5 This is a flowchart illustrating the fourth embodiment of the intelligent health education method for gynecological tumor patients of the present invention. Figure 6 This is a schematic diagram of the overall system architecture corresponding to the intelligent health education method for gynecological tumor patients of the present invention; Figure 7 This is a functional block diagram of the first embodiment of the intelligent health education device for gynecological tumor patients of the present invention.

[0021] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0022] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0023] The solution of this invention mainly involves: acquiring the patient's electronic medical record information and patient interaction data; identifying the gynecological tumor disease type and treatment stage label based on the electronic medical record information; specifically identifying psychological stress signals related to gynecological tumor treatment based on the patient interaction data; generating a current psychosomatic state profile of the patient; matching and dynamically generating personalized educational content and psychological counseling plans from a structured knowledge base based on the current psychosomatic state profile; and outputting these to the patient through an interactive terminal; automatically generating medication and follow-up reminder tasks based on the patient's treatment data; and recording the output of the personalized educational content and the patient's feedback on the reminder tasks. The data is structured and correlated to form a digital education path map for the patient. By integrating electronic medical record information and multimodal patient interaction data, a precise profile of the patient's current psychosomatic state can be constructed. This enables dynamic matching and personalized generation of educational content and psychological counseling plans, significantly improving the patient's cognitive adaptability to health information and the effectiveness of psychological support. Furthermore, a traceable digital education path map is formed through the automatic generation of medication and follow-up tasks and the structured correlation of interactive data throughout the entire process. This achieves closed-loop management throughout the entire cycle inside and outside the hospital, solving the technical problems of homogenized information, lack of targeted psychological support, and untraceable effects due to disconnection from outpatient management in the existing technology for health education of gynecologic oncology patients.

[0024] Reference Figure 1 , Figure 1 This is a schematic diagram of the device structure of the hardware operating environment involved in the embodiments of the present invention.

[0025] like Figure 1 As shown, the device may include: a processor 1001, such as a CPU; a communication bus 1002; a user interface 1003; a network interface 1004; and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as a disk drive. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0026] Those skilled in the art will understand that Figure 1 The device structure shown does not constitute a limitation on the device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0027] like Figure 1 As shown, the memory 1005, which serves as a storage medium, may include an operating device, a network communication module, a user interface module, and an intelligent health education program for gynecological tumor patients.

[0028] The device of this invention calls the intelligent health education program for gynecological tumor patients stored in the memory 1005 through the processor 1001 and performs the following operations: The system acquires the patient's electronic medical record information and patient interaction data, identifies the gynecological tumor disease type and treatment stage label based on the electronic medical record information, and specifically identifies psychological stress signals related to gynecological tumor treatment based on the patient interaction data, thereby generating a current psychosomatic profile of the patient. Based on the current psychosomatic state profile, personalized educational content and psychological counseling plans are dynamically generated from a structured knowledge base and output to the patient through an interactive terminal. Based on the patient's treatment data, medication and follow-up reminders are automatically generated, and the output records of the personalized education content and the patient's feedback data on the reminders are structurally linked to form the patient's digital education path map.

[0029] The device of this invention calls the intelligent health education program for gynecological tumor patients stored in the memory 1005 through the processor 1001, and also performs the following operations: The patient's electronic medical record is obtained from the hospital information system through a secure encrypted interface. Natural language processing technology is used to perform structured parsing of the electronic medical record to identify the disease type and TNM stage of the gynecological tumor and determine the current treatment stage label. The patient's voice information is collected in real time through the microphone of the interactive terminal, and the patient's facial expression video data is captured through the camera of the interactive terminal. The facial expression video data is anonymized and desensitized to obtain the target facial expression video data after anonymization and desensitization. The patient's text information is collected in real time through the input module of the interactive terminal. The voice information, the target facial expression video data and the facial expression information are used as patient interaction data. The patient interaction data is semantically compared with a pre-set gynecological tumor psychosomatic keyword feature library, and a multimodal emotion analysis model is used to perform weighted calculations on voice tone and facial micro-expressions to specifically identify psychological stress signals closely related to gynecological tumor treatment. The disease type, treatment stage label, and psychological stress signal are fused and mapped in multiple dimensions to construct a current psychosomatic profile of the patient that includes both physiological treatment progress and psychological stress dimensions.

[0030] The device of this invention calls the intelligent health education program for gynecological tumor patients stored in the memory 1005 through the processor 1001, and also performs the following operations: The current psychosomatic state profile of the patient is deconstructed to extract disease stage labels, treatment plan labels, and psychological stress type labels, and a multidimensional feature vector for retrieval is constructed. In the structured knowledge base, semantic matching and weight calculation are performed using the multidimensional feature vectors to locate the standard nursing education material node corresponding to the disease stage label, the treatment cycle guidance node corresponding to the treatment plan label, and the psychological counseling strategy node matching the psychological stress type label. Using natural language generation and multimedia synthesis technologies, the located standard nursing education material nodes, treatment cycle guidance nodes, and psychological counseling strategy nodes are dynamically assembled and semantically fused to generate personalized education content and psychological counseling plans that are adapted to the patient's current cognitive level and emotional resilience, and then output to the patient through an interactive terminal.

[0031] The device of this invention calls the intelligent health education program for gynecological tumor patients stored in the memory 1005 through the processor 1001, and also performs the following operations: Access the data structure of the patient's current psychosomatic state profile, parse the physiological treatment dimension data mapped from the electronic medical record information in the data structure, and extract the disease stage label representing the degree of tumor progression and the treatment plan label representing the current medical intervention from the physiological treatment dimension data; The psychological state dimension data formed by patient interaction data analysis in the data structure is parsed, and psychological stress type labels representing the patient's emotional focus are extracted from the psychological state dimension data; The disease stage label, treatment plan label, and psychological stress type label are encoded and combined according to a preset dimensional structure to construct a multidimensional feature vector containing physiological treatment features and psychological state features. The multidimensional feature vector serves as a retrieval index to locate knowledge nodes in the structured knowledge base that match the patient's current state.

[0032] The device of this invention calls the intelligent health education program for gynecological tumor patients stored in the memory 1005 through the processor 1001, and also performs the following operations: The pre-trained natural language generation model is invoked to dynamically adjust the vocabulary complexity and sentence structure of the educational and guidance text to be generated based on the current psychosomatic state profile of the patient. Multimedia synthesis technology is then used to convert the generated text content into speech or combine it with animated videos for presentation. During the dynamic assembly process, the standard nursing education material nodes and the treatment cycle guidance nodes are logically sorted according to the treatment timeline, and the psychological counseling strategy nodes are inserted as an emotional buffer layer between the key information nodes. The semantic fusion algorithm eliminates expression conflicts between different nodes, and dynamically adjusts the tone and information density of the output content according to the psychological stress signal to generate personalized educational content and psychological counseling plans in the form of text, audio and video, which are then output to the patient through the interactive terminal.

[0033] The device of this invention calls the intelligent health education program for gynecological tumor patients stored in the memory 1005 through the processor 1001, and also performs the following operations: By synchronizing patients' in-hospital treatment data through a secure data interface, and using a pre-set treatment plan parser to identify key treatment nodes, reminders for medication and follow-up examinations are automatically generated. The timestamp of the personalized education content push, the patient's viewing time, and interactive click behavior will be captured in real time and recorded as output records; Simultaneously, the patient's confirmation of the reminder task, extension application, or side effect report information is collected as feedback data; using the patient's unique anonymous identifier as an index, the patient interaction data, the key treatment nodes, the output records, and the feedback data are structured, stored, and mapped in a pre-defined time series to construct a visualized digital educational path map for the patient.

[0034] The device of this invention calls the intelligent health education program for gynecological tumor patients stored in the memory 1005 through the processor 1001, and also performs the following operations: The patient's in-hospital treatment data is synchronized through a secure data interface. The in-hospital treatment data includes the start date of the chemotherapy cycle, the duration of targeted drug maintenance therapy, and the postoperative follow-up examination time. Based on the start date of the chemotherapy cycle, the duration of the targeted drug maintenance therapy, and the postoperative follow-up examination time, a medication and follow-up reminder task is generated, which includes the drug name, dosage, side effect warnings, and pre-examination preparations, and is pushed out on time via mobile or robot terminals.

[0035] This embodiment, through the above-described scheme, acquires the patient's electronic medical record information and patient interaction data. Based on the electronic medical record information, it identifies the type of gynecological tumor and its treatment stage. Based on the patient interaction data, it specifically identifies psychological stress signals related to gynecological tumor treatment, generating a current psychosomatic profile of the patient. Based on this profile, it matches and dynamically generates personalized educational content and psychological counseling plans from a structured knowledge base, and outputs them to the patient through an interactive terminal. Based on the patient's treatment data, it automatically generates medication and follow-up reminders, and structurally correlates the output records of the personalized educational content with the patient's feedback data on these reminders, forming a digital educational path map for the patient. By integrating electronic medical record information and multimodal patient interaction data, it accurately constructs a current psychosomatic profile of the patient, achieving dynamic matching and personalized generation of educational content and psychological counseling plans. This significantly improves the patient's cognitive adaptability to health information and the effectiveness of psychological support. Furthermore, through the automatic generation of medication and follow-up tasks and the structured correlation of the entire process of interactive data, it forms a traceable digital educational path map, achieving closed-loop management throughout the entire hospital and outpatient cycle.

[0036] Based on the above hardware structure, an embodiment of the intelligent health education method for gynecological tumor patients of the present invention is proposed.

[0037] Reference Figure 2 , Figure 2 This is a flowchart illustrating the first embodiment of the intelligent health education method for gynecological tumor patients of the present invention.

[0038] In the first embodiment, the intelligent health education method for gynecological tumor patients includes the following steps: Step S10: Obtain the patient's electronic medical record information and patient interaction data; identify the gynecological tumor disease type and treatment stage label based on the electronic medical record information; and specifically identify psychological stress signals related to gynecological tumor treatment based on the patient interaction data to generate a current psychosomatic profile of the patient.

[0039] It should be noted that the process involves acquiring the patient's electronic medical record information and patient interaction data, identifying the gynecological tumor disease type and treatment stage label based on the electronic medical record information, and specifically identifying psychological stress signals related to gynecological tumor treatment based on the patient interaction data. By combining the gynecological tumor disease type and treatment stage label with the psychological stress signals, a current psychosomatic profile that can simultaneously reflect the patient's physiological treatment progress and psychological stress state is generated, providing an accurate data foundation for the subsequent generation of personalized treatment plans.

[0040] Step S20: Based on the current psychosomatic state profile, match and dynamically generate personalized educational content and psychological counseling plans from the structured knowledge base, and output them to the patient through the interactive terminal.

[0041] It should be understood that, based on the current psychosomatic state profile, the system retrieves knowledge nodes in the structured knowledge base that match the patient's disease stage and psychological state, and uses the search results to dynamically combine and generate personalized educational content and guidance plans that combine medical accuracy and psychological adaptability. The generated plans are then presented to the patient through an interactive terminal, thereby achieving precise intervention and information delivery based on the patient's real-time psychosomatic state.

[0042] Step S30: Automatically generate medication and follow-up reminder tasks based on the patient's treatment data, and structurally associate the output records of the personalized education content and the patient's feedback data on the reminder tasks to form the patient's education digital path map.

[0043] Understandably, medication and follow-up reminders are automatically triggered based on the patient's treatment data to ensure the continuity of the treatment plan. At the same time, the push records of personalized education content are structurally linked with the patient's feedback on the execution of the reminder tasks, forming an educational digital path map that runs through the treatment cycle, thereby enabling full-process quantitative tracking and management of patient compliance and education effectiveness.

[0044] This embodiment, through the above-described scheme, acquires the patient's electronic medical record information and patient interaction data. Based on the electronic medical record information, it identifies the type of gynecological tumor and its treatment stage. Based on the patient interaction data, it specifically identifies psychological stress signals related to gynecological tumor treatment, generating a current psychosomatic profile of the patient. Based on this profile, it matches and dynamically generates personalized educational content and psychological counseling plans from a structured knowledge base, and outputs them to the patient through an interactive terminal. Based on the patient's treatment data, it automatically generates medication and follow-up reminders, and structurally correlates the output records of the personalized educational content with the patient's feedback data on these reminders, forming a digital educational path map for the patient. By integrating electronic medical record information and multimodal patient interaction data, it accurately constructs a current psychosomatic profile of the patient, achieving dynamic matching and personalized generation of educational content and psychological counseling plans. This significantly improves the patient's cognitive adaptability to health information and the effectiveness of psychological support. Furthermore, through the automatic generation of medication and follow-up tasks and the structured correlation of the entire process of interactive data, it forms a traceable digital educational path map, achieving closed-loop management throughout the entire hospital and outpatient cycle.

[0045] Furthermore, Figure 3 This is a flowchart illustrating the second embodiment of the intelligent health education method for gynecological tumor patients of the present invention, as shown below. Figure 3 As shown, based on the first embodiment, a second embodiment of the intelligent health education method for gynecological tumor patients of the present invention is proposed. In this embodiment, step S10 specifically includes the following steps: Step S11: Obtain the patient's electronic medical record from the hospital information system through a secure encrypted interface, perform structured parsing of the electronic medical record using natural language processing technology, identify the disease type and TNM stage of the gynecological tumor, and determine the current treatment stage label.

[0046] It should be noted that the privacy and compliance of medical data transmission are ensured through secure encrypted interfaces. Natural language processing technology is used to parse unstructured electronic medical record text into structured data, thereby accurately identifying the disease type of gynecological tumors (primary tumor, regional lymph nodes, distant metastasis, TNM staging, and current treatment stage labels), providing a reliable medical basis for the subsequent construction of a patient's psychosomatic profile.

[0047] Step S12: Collect the patient's voice information in real time through the microphone of the interactive terminal, capture the patient's facial expression video data through the camera of the interactive terminal, anonymize and desensitize the facial expression video data to obtain the anonymized and desensitized target facial expression video data, collect the patient's text information in real time through the input module of the interactive terminal, and use the voice information, the target facial expression video data and the facial expression information as patient interaction data.

[0048] Understandably, the interactive terminal acquires the patient's voice, facial expression video, and text information through its microphone, camera, and input module, respectively. Sensitive facial expression video data is anonymized and desensitized to protect the patient's privacy. Finally, the processed multi-source data is integrated into patient interaction data, providing comprehensive and compliant data input for subsequent specific identification of psychological stress signals.

[0049] Step S13: The patient interaction data is semantically compared with the pre-set gynecological tumor psychosomatic keyword feature library, and the voice tone and facial micro-expressions are weighted and calculated in combination with a multimodal emotion analysis model to specifically identify psychological stress signals closely related to gynecological tumor treatment.

[0050] It should be understood that by semantically comparing patient interaction data with a pre-set specialty keyword feature library, and combining multimodal sentiment analysis models to perform weighted calculations on voice tone and facial micro-expressions, it is possible to comprehensively integrate textual semantics and non-verbal emotional features, accurately capture the specific psychological stress responses generated by patients during treatment, thereby ensuring that the identified psychological stress signals are specifically targeted at gynecologic oncology.

[0051] In a specific implementation, the pre-set gynecological tumor psychosomatic keyword feature library contains words and their emotional expression patterns related to the following categories: ovarian function preservation and loss, menopausal syndrome symptoms, lymphedema risk, chemotherapy and radiotherapy side effects, fertility anxiety, body shape change anxiety, and sexual dysfunction; the specific identification of psychological stress signals closely related to gynecological tumor treatment includes identifying emotional expression patterns corresponding to at least one of the above categories.

[0052] Step S14: Perform multi-dimensional data fusion and mapping of the disease type, the treatment stage label, and the psychological stress signal to construct a current psychosomatic profile of the patient that includes the dimensions of physiological treatment progress and psychological stress.

[0053] Understandably, by fusing and mapping multi-dimensional data such as disease type, treatment stage labels, and psychological stress signals, the constructed profile can simultaneously cover both physiological treatment progress and psychological stress dimensions, thereby comprehensively reflecting the patient's overall physical and mental condition during the treatment process. This provides a complete data model foundation for subsequently matching personalized education content and psychological counseling plans.

[0054] This embodiment, through the above-described scheme, obtains the patient's electronic medical record from the hospital information system via a secure encrypted interface. Natural language processing technology is used to perform structured parsing of the electronic medical record, identifying the disease type and TNM stage of the gynecological tumor, and determining the current treatment stage label. The patient's voice information is collected in real-time via the microphone of the interactive terminal, and facial expression video data is captured via the camera of the interactive terminal. This facial expression video data is anonymized and desensitized to obtain anonymized and desensitized target facial expression video data. The patient's text information is collected in real-time via the input module of the interactive terminal. The voice information, the target facial expression video data, and the facial expression information are used as patient interaction data. The patient interaction data is semantically compared with a pre-set gynecological tumor psychosomatic keyword feature database, and the results are then analyzed. The multimodal sentiment analysis model uses weighted calculations of speech tone and facial micro-expressions to specifically identify psychological stress signals closely related to gynecological tumor treatment. It integrates and maps the disease type, treatment stage, and psychological stress signals across multiple dimensions to construct a current psychosomatic profile of the patient, encompassing both physiological treatment progress and psychological stress. Secure encryption and anonymization ensure the safety of patient medical data and privacy information. Utilizing natural language processing and the multimodal sentiment analysis model, it achieves accurate and specific identification of gynecological tumor disease types, treatment stages, and psychological stress signals. The multi-dimensional data fusion constructs a current psychosomatic profile including both physiological treatment progress and psychological stress, providing a comprehensive, accurate, and compliant data model foundation for generating personalized educational content and psychological counseling plans.

[0055] Furthermore, Figure 4 This is a flowchart illustrating the third embodiment of the intelligent health education method for gynecological tumor patients of the present invention, as shown below. Figure 4 As shown, based on the first embodiment, a third embodiment of the intelligent health education method for gynecological tumor patients of the present invention is proposed. In this embodiment, step S20 specifically includes the following steps: Step S21: Deconstruct the current psychosomatic state profile of the patient, extract disease stage labels, treatment plan labels, and psychological stress type labels, and construct a multidimensional feature vector for retrieval.

[0056] It should be noted that by deconstructing the features of the portrait, disease stage tags and treatment plan tags representing the progress of physiological treatment, as well as psychological stress type tags representing psychological state, are extracted. These tags are then combined to construct a multi-dimensional feature vector, thereby providing a standardized retrieval basis for accurate semantic matching and content positioning in the subsequent structured knowledge base.

[0057] Furthermore, step S21 specifically includes the following steps: Access the data structure of the patient's current psychosomatic state profile, parse the physiological treatment dimension data mapped from the electronic medical record information in the data structure, and extract the disease stage label representing the degree of tumor progression and the treatment plan label representing the current medical intervention from the physiological treatment dimension data; The psychological state dimension data formed by patient interaction data analysis in the data structure is parsed, and psychological stress type labels representing the patient's emotional focus are extracted from the psychological state dimension data; The disease stage label, treatment plan label, and psychological stress type label are encoded and combined according to a preset dimensional structure to construct a multidimensional feature vector containing physiological treatment features and psychological state features. The multidimensional feature vector serves as a retrieval index to locate knowledge nodes in the structured knowledge base that match the patient's current state.

[0058] It should be understood that by accessing the underlying data structure of the patient profile, the physiological treatment dimension data derived from electronic medical records and the psychological state dimension data derived from patient interaction data analysis are analyzed separately. From these, disease stage labels representing the degree of tumor progression, treatment plan labels representing current medical interventions, and psychological stress type labels representing the patient's emotional focus are accurately extracted, achieving multi-dimensional quantification of the patient's physical and mental state. Subsequently, these multi-source labels are encoded and combined according to a preset dimensional structure to construct a multi-dimensional feature vector that simultaneously contains physiological treatment features and psychological state features. This vector serves as a standardized retrieval index, ensuring that knowledge nodes matching the patient's current physiological and psychological state are accurately located in the structured knowledge base. This establishes a data link from the patient state profile to the matching of knowledge base content, providing accurate data guidance for the dynamic assembly of subsequent content.

[0059] Step S22: In the structured knowledge base, semantic matching and weight calculation are performed using the multidimensional feature vector to locate the standard nursing education material node corresponding to the disease stage label, the treatment cycle guidance node corresponding to the treatment plan label, and the psychological counseling strategy node matching the psychological stress type label.

[0060] Understandably, through semantic matching and weight calculation, disease stage labels, treatment plan labels, and psychological stress type labels are mapped to the corresponding standard nursing education material nodes, treatment cycle guidance nodes, and psychological counseling strategy nodes, respectively.

[0061] Step S23: Using natural language generation technology and multimedia synthesis technology, the located standard nursing education material nodes, treatment cycle guidance nodes and psychological counseling strategy nodes are dynamically assembled and semantically fused to generate personalized education content and psychological counseling plans that are adapted to the patient's current cognitive level and emotional tolerance, and output to the patient through the interactive terminal.

[0062] It should be understood that by using natural language generation technology and multimedia synthesis technology, the standard nursing education material nodes, treatment cycle guidance nodes and psychological counseling strategy nodes located are dynamically assembled and semantically fused to ensure that the generated content is adapted to the patient's current cognitive level and emotional tolerance, and is output to the patient through the interactive terminal, thereby achieving the effective transmission and presentation of precise intervention information.

[0063] Furthermore, step S23 specifically includes the following steps: The pre-trained natural language generation model is invoked to dynamically adjust the vocabulary complexity and sentence structure of the educational and guidance text to be generated based on the current psychosomatic state profile of the patient. Multimedia synthesis technology is then used to convert the generated text content into speech or combine it with animated videos for presentation. During the dynamic assembly process, the standard nursing education material nodes and the treatment cycle guidance nodes are logically sorted according to the treatment timeline, and the psychological counseling strategy nodes are inserted as an emotional buffer layer between the key information nodes. The semantic fusion algorithm eliminates expression conflicts between different nodes, and dynamically adjusts the tone and information density of the output content according to the psychological stress signal to generate personalized educational content and psychological counseling plans in the form of text, audio and video, which are then output to the patient through the interactive terminal.

[0064] It should be noted that by calling a pre-trained natural language generation model, the vocabulary complexity and sentence structure of the text to be generated are dynamically adjusted according to the patient's current psychosomatic state profile to ensure that the content is suitable for the patient's cognitive level. Multimedia synthesis technology is used to achieve multimodal conversion of text into speech or animated video. During the dynamic assembly process, standard nursing education materials and treatment cycle guidance nodes are logically sorted according to the treatment timeline, and psychological counseling strategy nodes are inserted as an emotional buffer layer between key information nodes to balance the rigor of medical information with the gentleness of psychological comfort. Semantic fusion algorithms are used to eliminate expression conflicts between different nodes, and the tone and information density of the output content are dynamically adjusted according to psychological stress signals to generate personalized solutions in the form of text, speech and video, which are then output through an interactive terminal, thereby achieving precise intervention and effective information transmission based on the patient's real-time psychosomatic state.

[0065] This embodiment, through the above-described scheme, deconstructs the patient's current psychosomatic state profile to extract disease stage labels, treatment plan labels, and psychological stress type labels, and constructs a multidimensional feature vector for retrieval. In the structured knowledge base, the multidimensional feature vector is used for semantic matching and weight calculation to locate standard nursing education material nodes corresponding to the disease stage labels, treatment cycle guidance nodes corresponding to the treatment plan labels, and psychological counseling strategy nodes matching the psychological stress type labels. Natural language generation technology and multimedia synthesis technology are used to integrate the located standard nursing education material nodes, treatment cycle guidance nodes, and psychological counseling strategy nodes. Dynamic assembly and semantic fusion are performed to generate personalized health education content and psychological counseling plans that are adapted to the patient's current cognitive level and emotional tolerance. These plans are then output to the patient through an interactive terminal. By using feature deconstruction and multi-dimensional feature vector construction, the system can achieve accurate semantic matching between the patient's mental and physical state and a structured knowledge base. This ensures that the located nursing education, treatment guidance, and psychological counseling nodes are highly consistent with the patient's actual needs. Furthermore, by using natural language generation and multimedia synthesis technologies, the system dynamically assembles and generates personalized plans that are adapted to the patient's current cognitive level and emotional tolerance. This significantly improves the relevance, readability, and psychological acceptance of health education content, enabling precise intervention and effective information delivery based on the patient's real-time status.

[0066] Furthermore, Figure 5 This is a flowchart illustrating the fourth embodiment of the intelligent health education method for gynecological tumor patients of the present invention, as shown below. Figure 5 As shown, based on the first embodiment, a fourth embodiment of the intelligent health education method for gynecological tumor patients of the present invention is proposed. In this embodiment, step S30 specifically includes the following steps: Step S31: Synchronize the patient's in-hospital treatment data through a secure data interface, identify key treatment nodes using a pre-set treatment plan parser, and automatically generate reminders for medication and follow-up examinations.

[0067] It should be noted that by synchronizing patients' in-hospital treatment data through a secure data interface to ensure the authority of the data source and the security of transmission, a pre-built treatment plan parser is used to analyze the treatment data to identify key treatment nodes, and reminders for medication and follow-up examinations are automatically generated based on the identification results. This enables automated tracking and management of patients' out-of-hospital treatment plans, ensuring the continuity and standardization of the treatment process.

[0068] Furthermore, step S31 specifically includes the following steps: The patient's in-hospital treatment data is synchronized through a secure data interface. The in-hospital treatment data includes the start date of the chemotherapy cycle, the duration of targeted drug maintenance therapy, and the postoperative follow-up examination time. Based on the start date of the chemotherapy cycle, the duration of the targeted drug maintenance therapy, and the postoperative follow-up examination time, a medication and follow-up reminder task is generated, which includes the drug name, dosage, side effect warnings, and pre-examination preparations, and is pushed out on time via mobile or robot terminals.

[0069] It should be understood that by synchronizing in-hospital treatment data, including the start date of chemotherapy cycles, the duration of targeted drug maintenance therapy, and postoperative follow-up examination times, through a secure data interface, reminders are generated that include drug names, dosages, side effect warnings, and pre-examination preparations. These reminders are then pushed out on time via mobile devices or robotic terminals, thereby achieving automated tracking and precise prompts for key treatment milestones. This ensures that patients can receive timely medication guidance and follow-up examination preparation information that matches their treatment progress.

[0070] Step S32: The push timestamp of the personalized education content, the patient's viewing time, and the interactive click behavior are captured in real time and recorded as output records.

[0071] Understandably, by capturing push timestamps in real time to determine the specific moment of content delivery, recording the duration of patient viewing to assess the completeness of information reception, and collecting interactive click behavior to reflect the patient's active participation, these key indicators are ultimately recorded as outputs. This provides objective behavioral data support for the subsequent construction of an educational digital roadmap, enabling quantifiable recording and traceability of the effectiveness of educational content delivery and patient compliance.

[0072] Step S33: Simultaneously collect the patient's confirmation of execution of the reminder task, extension application, or side effect reporting information as feedback data; using the patient's unique anonymous identifier as an index, the patient interaction data, the key treatment nodes, the output records, and the feedback data are structured, stored, and mapped in a preset time series to construct a visualized digital educational path map for the patient.

[0073] It should be understood that by collecting patients' confirmation of reminder tasks, extension requests, or side effect reports as feedback data, and using the patient's unique anonymous identifier as an index, patient interaction data, key treatment nodes, output records, and feedback data are structured, stored, and mapped in a pre-defined time series. This creates a visualized digital educational path map, enabling unified management and traceability of patient treatment and educational data throughout the entire treatment cycle, and ensuring full-process traceability of data while protecting privacy.

[0074] In a specific implementation, based on the patient's digital education pathmap, the effectiveness of education and factors influencing adherence can be evaluated. The digital education pathmap contains a traceable longitudinal dataset, which is associated with the patient's anonymized ID, treatment stage tags, educational content push records, patient Q&A interaction records, and time-series data of psychosomatic state assessment results.

[0075] In the specific implementation, see Figure 6 , Figure 6 This is a schematic diagram of the overall system architecture corresponding to the intelligent health education method for gynecological tumor patients of the present invention, as shown below. Figure 6 As shown, the system includes a user interaction layer, an AI processing core layer, and a data and service layer, which are connected by data flow. The AI ​​processing core layer specifically includes an intent recognition module, a knowledge retrieval and generation module, and a personalized content generation engine. The data and service layer specifically includes a structured knowledge base, a patient file and interaction log library, and intelligent reminder and tracking services. Its core is to build a collaborative system of "hardware carrier + AI brain + cloud knowledge base".

[0076] (1) Personalized education content generation engine: By integrating the patient's electronic medical record information, the system automatically identifies key tags such as specific diseases (e.g., ovarian cancer, cervical cancer), treatment stages (preoperative, chemotherapy, rehabilitation), and current complications, and dynamically combines and generates personalized education plans in the form of text, images, audio and video from the structured knowledge base to ensure "a thousand people, a thousand faces".

[0077] (2) Gynecologic Oncology Specialty Fine-tuned Large Language Model: A large language model based on gynecologic oncology specialty medical corpus (such as NCCN / CSCO Gynecologic Oncology Guidelines, chemotherapy regimen manuals, specific drug instructions, psychosocial research literature).

[0078] The model's construction method includes: base model selection, construction of a multi-source specialty corpus (containing 50,000 to 100,000 high-quality corpora), supervised fine-tuning training (using LoRA technology with a learning rate of 2e-5 to 5e-5), model evaluation, and a continuous learning mechanism to ensure the model's question-answering accuracy and medical safety in the field of gynecologic oncology.

[0079] (3) Multimodal sentiment analysis model: A multimodal sentiment analysis model integrating text, speech and visual features is constructed. The model encodes text features through BERT, acoustic features through convolutional neural networks, and visual features through temporal convolutional networks. After concatenating the three types of features, the model outputs the emotion classification results (calm, mild anxiety, high anxiety, fear, anger) and confidence scores through a fully connected network, thereby achieving accurate identification of the user's mental and physical state.

[0080] (4)Psychosomatic Support Module for Gynecological Oncology: Integrate the above large language model and multimodal sentiment analysis model. By analyzing the text / speech input by patients, specifically identify psychological stress signals related to gynecological tumor treatment (such as keywords related to "fertility", "menopause", "hair loss", "spousal relationship", etc. and corresponding emotional expressions), and actively trigger a structured psychological counseling content library that matches the medical scenario (such as counseling plans for anxiety about preserving fertility function, cognitive-behavioral intervention materials for coping with body image changes), to achieve seamless integration of medical information support and psychological intervention.

[0081] (5)Hierarchical Response Mechanism and Stress Threshold Judgment: Based on the comprehensive stress score (S = α·T + β·V + γ·A) output by the multimodal sentiment analysis model, where S represents the intensity of the comprehensive psychological stress signal, T, V, and A respectively represent the feature scores of text, speech, and facial expressions, and α, β, and γ represent the weight coefficients corresponding to each modality; set hierarchical response thresholds: Mild state (0.6 < S ≤ 0.75): Call the general encouragement corpus and output soothing statements; Moderate to severe state (S > 0.75): Activate the precise content trigger engine, match and push in-depth explanations and coping strategy content directly related to the current troubled theme.

[0082] (6)Gynecological Oncology Full-cycle Closed-loop Management System: This system reads the electronic medical record data of patients' gynecological oncology specialties定向 through a secure data interface (HL7 / FHIR) authorized by the hospital.

[0083] Its core intelligent reminder and tracking engine is built with a gynecological tumor treatment plan parser, which can automatically identify key events such as "Day 1 of TP regimen chemotherapy", "Bevacizumab maintenance treatment", "15th pelvic radiotherapy", etc., and automatically generate and dynamically adjust personalized reminder queues throughout the hospital and outside the hospital based on this.

[0084] (7)Knowledge Bundled Push and Digital Roadmap Construction Method: The reminder content not only includes time points, but also bundles and pushes specialty education knowledge strongly related to them (for example, when pushing the reminder of "Take palbociclib today", synchronously attach the key point video of "Monitoring and prevention of neutropenia" unique to this drug).

[0085] At the same time, the system structurally records all interaction events (including question content, pushed knowledge point IDs, and patients' confirmation status of reminders), forming a traceable digital roadmap for gynecological tumor patient education for efficacy and compliance correlation analysis.

[0086] In specific implementation, this embodiment can be implemented through a mobile robot entity or a fixed terminal device (such as a tablet computer) with a touch interaction screen, microphone, speaker, and camera: 1. AI Core Processing Layer 1.1 The process of constructing the large language model and fine-tuning it in a specialized manner is as follows: (1) Selection of base model We selected a publicly available pre-trained large language model with strong Chinese semantic understanding capabilities (such as ChatGLM3-6B, Baichuan2-7B, or Qwen-7B) as the base model.

[0087] The base model has completed large-scale pre-training on a general corpus and possesses basic language understanding and generation capabilities.

[0088] (2) Construction of a gynecologic oncology corpus Construct a high-quality specialist corpus for fine-tuning, with corpus sources including but not limited to: Authoritative clinical guidelines: NCCN Clinical Practice Guidelines for Gynecologic Oncology and CSCO Guidelines for the Diagnosis and Treatment of Gynecologic Oncology (Chinese version), extracting structured content on diagnosis, staging, treatment principles, and follow-up recommendations; Specialty drug instructions: Drug instructions for commonly used gynecologic oncology drugs (such as paclitaxel, carboplatin, olaparib, bevacizumab, etc.), focusing on the indications, dosage, adverse reactions and management; Nursing operating procedures: the hospital's and industry's standard gynecologic oncology nursing routines, health education manuals, and complication management procedures; Typical question-and-answer pairs: Compiled by senior nurses and doctors in the department, based on frequently asked questions from patients in real clinical work (such as "What should I do if my white blood cell count is low after chemotherapy?" "Can I have sex after a total hysterectomy?") and their standard answers, a question-and-answer pair dataset was constructed, and each question-and-answer pair was reviewed by two people; Literature on psychosocial support: academic literature related to psychosocial intervention for gynecologic oncology patients, a database of psychological counseling scripts, and materials on cognitive behavioral intervention.

[0089] The above corpus was anonymized (patient privacy information was removed) and structured annotated to form a training dataset containing approximately 50,000 to 100,000 high-quality specialist corpus entries.

[0090] (3) Fine-tuning the training process Specialized training of the pedestal model is performed using supervised fine-tuning. The specific steps include: Data formatting: Converting text from a specialized corpus into a format suitable for fine-tuning, such as "instruction-input-output" triples. For example: Instruction: "Please explain the causes and coping methods of leukopenia in simple and easy-to-understand language." Input: (Patient's current treatment information, optional) Output: "White blood cells are the guardians of our body... (standard answer)" Training parameter settings: Low-rank adaptation or adapter fine-tuning techniques are employed to train only a small number of newly added parameters while keeping the main parameters of the base model unchanged. This reduces computational resource requirements and prevents catastrophic forgetting. Typical parameter settings are as follows: ① Learning rate: 2e-5 to 5e-5; ② Batch size: 4-8; ③ Number of training epochs: 3-5; ④ Optimizer: AdamW.

[0091] Training process: Formatted training data is input into the model, and the model parameters are updated through backpropagation, enabling the model to learn the terminology, question-and-answer patterns, and expression styles specific to the field of gynecologic oncology. During training, the model accuracy is evaluated on a reserved validation set after each epoch, and training stops when the validation set loss no longer decreases.

[0092] (4) Model evaluation and optimization After training, the model is evaluated in the following ways: Automatic evaluation: Calculate metrics such as accuracy, recall, and BLEU score on the reserved test set to ensure that the model's answers to specialist questions meet the required quality standards; Manual evaluation: Senior nurses and doctors in the department conduct blind evaluations of 100-200 typical questions and answers generated by the model. The evaluation dimensions include medical accuracy, ease of understanding, and safety. The pass rate must reach more than 95%. Adversarial testing: Input boundary questions (such as "Is it okay if I stop treatment?" or "Can I increase the dosage of this medicine myself?") to verify whether the model will output unsafe content, and set up a safety fallback mechanism (such as triggering a fixed response of "It is recommended that you consult your attending physician").

[0093] (5) Deployment and Continuous Learning After fine-tuning, the model is deployed to a local server or a trusted cloud platform and made available to the robot via API. Simultaneously, the system retains an incremental learning mechanism: high-quality question-and-answer pairs from real patient-robot interactions, verified by nurses, are periodically anonymized and fed back into the training dataset, enabling continuous model optimization and knowledge updates.

[0094] 1.2 Precise Question Answering and Mind-Body Support Module This module is the core interactive unit of the AI ​​core processing layer. It calls the large language model that has been fine-tuned in 1.1 and works in conjunction with the psychosomatic state recognition results of the user interaction layer to achieve a seamless integration of medical information support and psychological intervention.

[0095] (1) Question and Answer Processing Flow When a user asks a question via voice or text, this module performs the following steps: Processing step 1: Receive user input; Technical Implementation 1: Obtain the text or speech-to-text content input by the user from the user interaction layer; Processing step 2. Context construction; Technical Implementation 2: Extract the current patient's personalized tags (disease, treatment stage, complications, etc.) from the patient's records and combine them with the user input to form a complete context; Processing step 3. Model invocation; Technical Implementation 3: Fine-tune the large language model by taking the constructed context input; Processing step 4. Response generation; Technical Implementation 4: After the model generates the answer, it is output after post-processing (such as sensitive word filtering and length control).

[0096] (2) Psychosomatic support synergy mechanism This module is linked in real time with the mind-body state recognition results of the user interaction layer: Normal state: Only provide medical information support, and maintain a professional yet accessible answer style; For mild anxiety: In addition to the medical answer, preface with a reassuring statement, such as "I understand your concerns about this issue, let me explain it to you in detail..."; For moderate to severe anxiety / fear: prioritize triggering psychosomatic support responses, first providing psychological guidance content, and then pushing in-depth explanations and coping strategies directly related to the distressing topic.

[0097] (3) Safety and quality control mechanism Medical accuracy assurance: After the model answers, the system automatically compares the similarity with the standard answers in the structured knowledge base. If the similarity is lower than the threshold, manual review is triggered. Safety backup: When the model recognizes that the user input involves extreme content such as self-harm or suicide, it will forcibly output a fixed safety message: "Your situation is very important. We suggest that you contact your attending physician immediately or call the psychological assistance hotline XXX-XXXX-XXXX". Interaction log recording: Each question and answer records user input, model output, psychosomatic state recognition results, response time, etc., and stores them in the patient file and interaction log library.

[0098] 1.3 Intent Recognition Module This module is responsible for understanding the user's true needs, mapping natural language questions to task categories that the system can perform, and providing accurate intent labels for subsequent knowledge retrieval and content generation.

[0099] Example: When a patient asks "What should I do if I feel nauseous after chemotherapy?", the module will recognize their intent as "searching for ways to manage side effects".

[0100] 1.4 Knowledge Retrieval and Generation Module This module is responsible for accurately retrieving relevant information from a structured knowledge base and, by combining intent recognition results and personalized tags, generating answers that meet user needs.

[0101] 1.5 Personalization Engine The engine continuously receives patient data from the hospital information system or data entered through initial interviews, maintaining a dynamic profile for each patient to ensure that every interaction and content push is tailored to them.

[0102] 2. User Interaction Layer This layer includes a touchscreen, microphone, speaker, and camera. Patients can select functions via the touchscreen (such as "My Education," "Ask a Question," and "Medication Reminder") or interact with the robot directly via voice. The camera assists in capturing, recognizing, and analyzing emotions.

[0103] The specific workflow for implementing mind-body state recognition in the user interaction layer is as follows: (1) Multimodal signal acquisition: Receive text input, speech-to-text content and anonymized video emoticon data with consent from the user (captured by camera, processed locally and not uploaded); (2) Extraction of special emotional features: The text information is compared and weighted with the locally constructed gynecological tumor psychosomatic keyword feature library. This feature library not only contains general emotional words, but also focuses on including words related to the treatment consequences of gynecological tumors such as 'ovarian function', 'menopausal symptoms', 'lymphedema', and 'chemotherapy', as well as their common emotional expression patterns. (3) Mental and physical state determination and response: The above features are comprehensively analyzed using a pre-trained multimodal sentiment analysis model (such as a model that integrates BERT and visual features).

[0104] The specific analysis process of the multimodal sentiment analysis model is as follows: ①Text feature encoding: The user input text is encoded into a 768-dimensional text feature vector using BERT-base-Chinese; ② Speech feature encoding: The Mel frequency cepstral coefficients extracted from the speech signal are encoded into a 256-dimensional acoustic feature vector through a convolutional neural network; ③ Visual feature encoding: The sequence of key facial expression points is encoded into a 128-dimensional visual feature vector through a temporal convolutional network; ④ Feature fusion: The above three types of features are concatenated into a 1152-dimensional fusion feature, which is then input into a fully connected network to output the sentiment classification result and confidence score.

[0105] When a user is determined to be in a state of "treatment-related high anxiety" or "disturbed by specific side effects", this submodule not only calls up the encouragement corpus, but also prioritizes triggering in-depth explanations and coping strategies that are directly related to the current distress topic and have been reviewed by medical experts.

[0106] Tiered response mechanism: ① Mild state (overall stress score 0.6-0.75): Call the general encouragement corpus to output reassuring statements; ② Moderate to severe state (overall stress score > 0.75): Activate the precise content triggering engine, and match and push in-depth explanations and coping strategies from the structured psychological counseling content library based on the identified core stress themes.

[0107] Examples of matching rules are shown in Table 1 below: Table 1. Example table of matching rules:

[0108] Complete interactive example: The patient typed: "I haven't had my period for three months since I started chemotherapy. I'm only 32 years old. Does this mean I'll never be able to have children? I'm so scared." ① Keyword matching: Detected "after chemotherapy", "no menstruation", "giving birth", "fear" - identified the theme as "fertility anxiety"; ②Emotion determination: The text sentiment analysis output "high anxiety" (confidence level 0.82); ③ Overall stress score: S=0.5×0.9+0.3×0.82+0.2×0.70=0.836>0.75, triggering a moderate to severe response; ④ Content Trigger: First, output reassuring statements: "I understand your concerns after reading your description. Chemotherapy can indeed affect ovarian function, but everyone's recovery is different; let me introduce some relevant information to you." Then, push a 3-minute popular science video, "Ovarian Function Protection and Fertility Preservation after Chemotherapy," and an online consultation portal for the "Reproductive Medicine Center."

[0109] 3. Data and Service Layer (1) Structured knowledge base: It stores standardized educational materials (text, pictures, and videos) for different diseases, stages, and symptoms, which are reviewed by medical experts.

[0110] (2) Patient records and interaction logs: record each patient's personalized tags and all interaction history with the robot for effect evaluation and scientific research analysis.

[0111] (3) Intelligent reminders and tracking: The system retrieves structured treatment data of authorized patients from the hospital's HIS / PACS / LIS system at regular intervals through a read-only, encrypted HL7 or FHIR interface.

[0112] The engine's built-in gynecologic oncology treatment protocol parser can identify key events such as 'TP regimen chemotherapy day 1', 'bevacizumab maintenance therapy', and 'pelvic radiotherapy session 15', and automatically generate task queues accordingly.

[0113] For medication reminders, the engine will link to the hospital's pharmaceutical knowledge base and embed specialized usage instructions for the drug in the reminder message (such as 'Doxorubicin: Urine may turn red after taking the drug, which is normal').

[0114] For follow-up examination reminders, the relevant examination precautions database is used (e.g., 'fasting and abstaining from water are required before a follow-up pelvic MRI examination').

[0115] All push notifications and patient feedback (such as 'medication taken' confirmations) are encrypted and stored, and linked to the patient's anonymized ID and treatment stage tag to form a longitudinal dataset for research analysis. Example of workflow: A newly admitted cervical cancer chemotherapy patient uses the robot for the first time.

[0116] The system labels patients with tags such as "cervical cancer", "first chemotherapy", and "no serious complications" by reading electronic medical records or entering information from nurses.

[0117] After the patient logs in, the robot first pushes a customized "Welcome and First Chemotherapy Instructions" video for her. Then, the patient asks a question via voice: "What will happen if my white blood cell count is low?" The AI ​​accurately recognizes the question, extracts information from its knowledge base, and generates an answer: "Hello, white blood cells are the guardians of our body. If the number is low, our resistance will decrease, making us more susceptible to infection."

[0118] But don't worry too much, we have many ways to deal with it, such as getting injections to increase white blood cell count, paying attention to nutrition, and preventing infection.

[0119] Your current blood test results are under close monitoring, and we will address any changes promptly. It should be noted that, compared with existing standardized, one-dimensional education and outreach programs, this embodiment has the following significant technical advantages: Improving the accuracy and compliance of information delivery: By combining the technologies of "personalized content generation engine" and "precise question and answer module", the problems of information loss and comprehension bias are solved, enabling patients to obtain easily understandable and highly relevant information, thereby significantly improving treatment compliance and self-management ability.

[0120] Achieving a unified standardization and personalization in the missionary process: This invention uses AI technology to solidify best practices into a standard knowledge base, while using algorithms to achieve personalized content distribution, solving the pain point of inconsistent quality in traditional oral missionary work and ensuring the homogeneous output of high-quality missionary work.

[0121] Extending the boundaries of nursing services and improving efficiency and safety: Robots take on most of the repetitive and common educational and question-and-answer tasks, freeing nurses from repetitive work and allowing them to focus on more complex and personalized care. At the same time, all interaction records are electronic and traceable, leaving objective evidence of the medical process and reducing the risk of disputes.

[0122] Empowering specialized research and integrated psychosomatic care: The system automatically collects and structures full-cycle interactive data (digital pathmap), providing quantitative datasets for specialized nursing research on the effectiveness of education for gynecologic oncology patients and factors influencing adherence. Furthermore, the unique "psychosomatic support module" directly addresses the specific psychosocial needs of gynecologic oncology patients, achieving a leap from general emotional comfort to specialized, scenario-based psychosocial support, thus improving patients' overall medical experience and treatment confidence.

[0123] This embodiment, through the above-described scheme, synchronizes the patient's in-hospital treatment data via a secure data interface, utilizes a pre-set treatment plan parser to identify key treatment nodes, and automatically generates reminders for medication and follow-up examinations. It captures the push timestamps of the personalized educational content, the patient's viewing time, and interactive click behavior in real time as output records. Simultaneously, it collects the patient's confirmation of the reminder tasks, extension requests, or side effect reports as feedback data. Using the patient's unique anonymous identifier as an index, it structures and maps the patient's interaction data, key treatment nodes, output records, and feedback data according to a preset time sequence, constructing a visualized digital educational path map for the patient. This approach ensures the continuity and standardization of out-of-hospital treatment management by securely synchronizing treatment data and automatically identifying key treatment nodes to generate reminders. Furthermore, by collecting educational output records and patient feedback data, and using the patient's unique anonymous identifier as an index to structure and correlate multi-source data according to a time sequence, it constructs a visualized digital educational path map. This allows for the full-cycle quantitative tracking and traceable management of patient treatment adherence and educational effectiveness while protecting patient privacy, providing data support for subsequent efficacy evaluation and plan optimization.

[0124] Accordingly, the present invention further provides an intelligent health education device for gynecological tumor patients.

[0125] Reference Figure 7 , Figure 7 This is a functional block diagram of the first embodiment of the intelligent health education device for gynecological tumor patients of the present invention.

[0126] In the first embodiment of the intelligent health education device for gynecological tumor patients of the present invention, the intelligent health education device for gynecological tumor patients includes: The identification and analysis module 10 is used to acquire the patient's electronic medical record information and patient interaction data, identify the gynecological tumor disease type and treatment stage label based on the electronic medical record information, and specifically identify psychological stress signals related to gynecological tumor treatment based on the patient interaction data, and generate a current psychosomatic profile of the patient.

[0127] The solution generation module 20 is used to match and dynamically generate personalized education content and psychological counseling solutions from a structured knowledge base based on the current psychosomatic state profile, and output them to the patient through an interactive terminal.

[0128] The reminder association module 30 is used to automatically generate medication and follow-up reminder tasks based on the patient's treatment data, and to structurally associate the output records of the personalized education content and the patient's feedback data on the reminder tasks to form the patient's education digital path map.

[0129] The steps for implementing each functional module of the intelligent health education device for gynecological tumor patients can be referred to in the various embodiments of the intelligent health education method for gynecological tumor patients of the present invention, and will not be repeated here.

[0130] Furthermore, this embodiment of the invention also proposes a storage medium storing an intelligent health education program for gynecological tumor patients. When the intelligent health education program for gynecological tumor patients is executed by a processor, it performs the operations described in the above embodiment of the intelligent health education method for gynecological tumor patients.

[0131] Those skilled in the art will understand that all or part of the steps in the methods described above can be implemented by a program instructing related hardware. The program is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium is a computer-readable storage medium, including: USB flash drive, mobile hard drive, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, and other media that can store program code.

[0132] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0133] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0134] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A gynecological tumor patient intelligent health education method, characterized in that, The intelligent health education method for gynecologic oncology patients includes: The system acquires the patient's electronic medical record information and patient interaction data, identifies the gynecological tumor disease type and treatment stage label based on the electronic medical record information, and specifically identifies psychological stress signals related to gynecological tumor treatment based on the patient interaction data, thereby generating a current psychosomatic profile of the patient. Based on the current psychosomatic state profile, personalized educational content and psychological counseling plans are dynamically generated from a structured knowledge base and output to the patient through an interactive terminal. Based on the patient's treatment data, medication and follow-up reminders are automatically generated, and the output records of the personalized education content and the patient's feedback data on the reminders are structurally linked to form the patient's digital education path map. 2.The intelligent health education and propaganda method for gynecological tumor patients of claim 1, wherein, The process involves acquiring the patient's electronic medical record information and patient interaction data, identifying the gynecological tumor disease type and treatment stage label based on the electronic medical record information, and specifically identifying psychological stress signals related to gynecological tumor treatment based on the patient interaction data to generate a current psychosomatic profile of the patient, including: The patient's electronic medical record is obtained from the hospital information system through a secure encrypted interface. Natural language processing technology is used to perform structured parsing of the electronic medical record to identify the disease type and TNM stage of the gynecological tumor and determine the current treatment stage label. The patient's voice information is collected in real time through the microphone of the interactive terminal, and the patient's facial expression video data is captured through the camera of the interactive terminal. The facial expression video data is anonymized and desensitized to obtain the target facial expression video data after anonymization and desensitization. The patient's text information is collected in real time through the input module of the interactive terminal. The voice information, the target facial expression video data and the facial expression information are used as patient interaction data. The patient interaction data is semantically compared with a pre-set gynecological tumor psychosomatic keyword feature library, and a multimodal emotion analysis model is used to perform weighted calculations on voice tone and facial micro-expressions to specifically identify psychological stress signals closely related to gynecological tumor treatment. The disease type, treatment stage label, and psychological stress signal are fused and mapped in multiple dimensions to construct a current psychosomatic profile of the patient that includes both physiological treatment progress and psychological stress dimensions. 3.The intelligent health education and propaganda method for gynecological tumor patients of claim 1, wherein, The process of matching and dynamically generating personalized educational content and psychological counseling plans from a structured knowledge base based on the current psychosomatic state profile, and then outputting them to the patient via an interactive terminal, includes: The current psychosomatic state profile of the patient is deconstructed to extract disease stage labels, treatment plan labels, and psychological stress type labels, and a multidimensional feature vector for retrieval is constructed. In the structured knowledge base, semantic matching and weight calculation are performed using the multidimensional feature vectors to locate the standard nursing education material node corresponding to the disease stage label, the treatment cycle guidance node corresponding to the treatment plan label, and the psychological counseling strategy node matching the psychological stress type label. Using natural language generation and multimedia synthesis technologies, the located standard nursing education material nodes, treatment cycle guidance nodes, and psychological counseling strategy nodes are dynamically assembled and semantically fused to generate personalized education content and psychological counseling plans that are adapted to the patient's current cognitive level and emotional resilience, and then output to the patient through an interactive terminal.

4. The intelligent health education method for gynecological tumor patients as described in claim 3, characterized in that, The process involves deconstructing the patient's current psychosomatic state profile to extract disease stage labels, treatment plan labels, and psychological stress type labels, and constructing a multidimensional feature vector for retrieval, including: Access the data structure of the patient's current psychosomatic state profile, parse the physiological treatment dimension data mapped from the electronic medical record information in the data structure, and extract the disease stage label representing the degree of tumor progression and the treatment plan label representing the current medical intervention from the physiological treatment dimension data; The psychological state dimension data formed by patient interaction data analysis in the data structure is parsed, and psychological stress type labels representing the patient's emotional focus are extracted from the psychological state dimension data; The disease stage label, treatment plan label, and psychological stress type label are encoded and combined according to a preset dimensional structure to construct a multidimensional feature vector containing physiological treatment features and psychological state features. The multidimensional feature vector serves as a retrieval index to locate knowledge nodes in the structured knowledge base that match the patient's current state.

5. The intelligent health education method for gynecological tumor patients as described in claim 3, characterized in that, The method employs natural language generation and multimedia synthesis technologies to dynamically assemble and semantically fuse the located standard nursing education material nodes, treatment cycle guidance nodes, and psychological counseling strategy nodes. This generates personalized education content and psychological counseling plans adapted to the patient's current cognitive level and emotional resilience, and outputs them to the patient through an interactive terminal, including: The pre-trained natural language generation model is invoked to dynamically adjust the vocabulary complexity and sentence structure of the educational and guidance text to be generated based on the current psychosomatic state profile of the patient. Multimedia synthesis technology is then used to convert the generated text content into speech or combine it with animated videos for presentation. During the dynamic assembly process, the standard nursing education material nodes and the treatment cycle guidance nodes are logically sorted according to the treatment timeline, and the psychological counseling strategy nodes are inserted as an emotional buffer layer between the key information nodes. The semantic fusion algorithm eliminates expression conflicts between different nodes, and dynamically adjusts the tone and information density of the output content according to the psychological stress signal to generate personalized educational content and psychological counseling plans in the form of text, audio and video, which are then output to the patient through the interactive terminal.

6. The intelligent health education method for gynecological tumor patients as described in claim 1, characterized in that, The system automatically generates medication and follow-up reminders based on the patient's treatment data, and structures and correlates the output records of the personalized education content with the patient's feedback data on the reminders to form the patient's digital education path map, including: By synchronizing patients' in-hospital treatment data through a secure data interface, and using a pre-set treatment plan parser to identify key treatment nodes, reminders for medication and follow-up examinations are automatically generated. The timestamp of the personalized education content push, the patient's viewing time, and interactive click behavior will be captured in real time and recorded as output records; Simultaneously, the patient's confirmation of the reminder task, extension application, or side effect report information is collected as feedback data; using the patient's unique anonymous identifier as an index, the patient interaction data, the key treatment nodes, the output records, and the feedback data are structured, stored, and mapped in a pre-defined time series to construct a visualized digital educational path map for the patient.

7. The intelligent health education method for gynecological tumor patients as described in claim 6, characterized in that, The process involves synchronizing the patient's in-hospital treatment data through a secure data interface, using a pre-set treatment plan parser to identify key treatment nodes, and automatically generating reminders for medication and follow-up examinations, including: The patient's in-hospital treatment data is synchronized through a secure data interface. The in-hospital treatment data includes the start date of the chemotherapy cycle, the duration of targeted drug maintenance therapy, and the postoperative follow-up examination time. Based on the start date of the chemotherapy cycle, the duration of the targeted drug maintenance therapy, and the postoperative follow-up examination time, a medication and follow-up reminder task is generated, which includes the drug name, dosage, side effect warnings, and pre-examination preparations, and is pushed out on time via mobile or robot terminals.

8. A smart health education device for gynecological tumor patients, characterized in that, The intelligent health education device for gynecologic oncology patients includes: The identification and analysis module is used to acquire the patient's electronic medical record information and patient interaction data, identify the gynecological tumor disease type and treatment stage label based on the electronic medical record information, and specifically identify psychological stress signals related to gynecological tumor treatment based on the patient interaction data, and generate a current psychosomatic profile of the patient. The solution generation module is used to match and dynamically generate personalized education content and psychological counseling solutions from a structured knowledge base based on the current psychosomatic state profile, and output them to the patient through an interactive terminal. The reminder association module is used to automatically generate medication and follow-up reminder tasks based on the patient's treatment data, and to structurally associate the output records of the personalized education content and the patient's feedback data on the reminder tasks to form the patient's education digital path map.

9. A smart health education device for gynecological tumor patients, characterized in that, The intelligent health education device for gynecological tumor patients includes: a memory, a processor, and an intelligent health education program for gynecological tumor patients stored in the memory and executable on the processor. The intelligent health education program for gynecological tumor patients is configured to implement the steps of the intelligent health education method for gynecological tumor patients as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores an intelligent health education program for gynecological tumor patients. When the intelligent health education program for gynecological tumor patients is executed by the processor, it implements the steps of the intelligent health education method for gynecological tumor patients as described in any one of claims 1 to 7.

Images