Gynecological tumor intelligent inquiry and risk evaluation method

Abstract

The invention relates to an intelligent inquiry and risk evaluation method for gynecological tumors. The method comprises the following steps: acquiring symptom information of a consultant and tumor medical history information of relatives in a plurality of generations of families of the consultant; constructing a tumor genetic family diagram by using the tumor medical history information, and then evaluating genetic risks of consultants to generate genetic consultation suggestions; constructing a symptom semantic network based on a medical ontology, and converting the symptom information into standard symptom data; calling a corresponding hierarchical inquiry protocol according to the standard symptom data to generate an intelligent diagnosis result; and integrating the genetic risk assessment result and the intelligent diagnosis result and outputting a diagnosis suggestion. Gynecological tumor inquiry consultation and risk evaluation can be automatically completed, and meanwhile a clinical research form is generated and inserted into a hospital electronic medical record system. The whole process does not need manual operation of doctors, the workload of information collection and input of the doctors is greatly reduced, and the doctors can concentrate on the core diagnosis and treatment process.

Description

Technical Field

[0001] The present invention relates to the field of smart medical technology, and in particular to a method for intelligent diagnosis and risk assessment of gynecological tumors. Background Art

[0002] In the clinical diagnosis and treatment of gynecological oncology, the traditional consultation model has long relied on a linear, face-to-face information collection process between doctors and patients. Doctors are required to engage in meticulous communication with patients within limited clinic time, meticulously inquiring about various relevant information. This process consumes considerable time and effort to ensure the accuracy and completeness of the information provided. After the consultation, doctors must manually enter the results into the electronic medical record system, further consuming valuable diagnostic and treatment time and, to a certain extent, limiting the efficiency and accessibility of medical services.

[0003] While current mainstream electronic health record systems enable digital storage of diagnostic and treatment data, they still suffer from numerous deficiencies in gynecological oncology settings. These include data fragmentation across different subsystems (e.g., multimodal data such as pathology images (DICOM), laboratory reports (HL7 V2), and genetic testing (FASTQ)), a lack of unified semantic mapping, low levels of data structuring and standardization, and low efficiency and high error rates in manual data transcription. Furthermore, existing paper or electronic forms are unable to capture the temporal and spatial evolution of symptoms in real time, such as the dynamic correlation between pain intensity and menstrual cycles and the continuous fluctuation trends of tumor markers. These factors, to a certain extent, limit the efficiency, quality, and accessibility of medical services. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a gynecological tumor consultation and risk assessment method based on an intelligent platform, which can automatically complete gynecological tumor consultation and risk assessment, and at the same time generate a clinical research form and insert it into the hospital electronic medical record system.

[0005] The technical solution adopted by the present invention to solve the technical problem is to provide a method for intelligent diagnosis and risk assessment of gynecological tumors, including the following steps:

[0006] Collect information on the consultant's symptoms and the cancer history of relatives within several generations of their family;

[0007] Using the tumor medical history information to construct a tumor genetic pedigree, and then assessing the genetic risk of the consultant to generate genetic counseling recommendations;

[0008] constructing a symptom semantic network based on medical ontology and converting the symptom information into standard symptom data;

[0009] Retrieving the corresponding hierarchical consultation protocol based on the standard symptom data to generate an intelligent diagnosis result;

[0010] The genetic risk assessment results and the intelligent diagnosis results are combined to output a diagnostic suggestion.

[0011] Furthermore, the genetic risk assessment of the consultant is achieved by conducting a comprehensive risk assessment using a professional assessment model including HBOC risk judgment logic, LS risk assessment logic and PJS risk assessment logic.

[0012] Furthermore, the genetic counseling recommendations include:

[0013] Generate the first consultation recommendation for those who carry BRCA1 / BRCA2 gene mutations;

[0014] Generate second consultation recommendations for those who carry genes related to Lynch syndrome;

[0015] A third consultation recommendation is generated for those who carry the STK11 gene mutation.

[0016] Furthermore, the first consultation recommendation includes risk prevention and screening for gynecological tumors, including ovarian cancer and breast cancer; the second consultation recommendation includes risk prevention and screening for tumors, including endometrial cancer and colorectal cancer; and the third consultation recommendation includes risk prevention and screening for diseases, including ovarian cancer, breast cancer, and gastrointestinal hamartomatous polyps.

[0017] Furthermore, the risk prevention and screening for gynecological tumors, including ovarian cancer and breast cancer, includes:

[0018] Strengthen gynecological cancer screening;

[0019] For those who wish to have children, preventive salpingo-oophorectomy is performed after childbirth;

[0020] For those who are not planning to have children for the time being, oral contraceptives;

[0021] Immediate family members undergo relevant genetic testing.

[0022] Furthermore, the risk prevention and screening for tumors including endometrial cancer and colorectal cancer includes:

[0023] For female patients, endometrial biopsy and colonoscopy should be performed every 1-2 years starting at age 20-25, or 5-10 years earlier than the age of the earliest cancer diagnosis in the family.

[0024] Maintain a healthy weight, increase exercise, quit smoking and limit alcohol consumption;

[0025] Family members undergo relevant genetic testing.

[0026] Furthermore, the risk prevention and screening for diseases including ovarian cancer, breast cancer, and gastrointestinal hamartomatous polyps include:

[0027] Regular gastrointestinal endoscopy to monitor polyps and surgically remove them if necessary;

[0028] Strengthen gynecological cancer screening;

[0029] For those who wish to have children, preventive salpingo-oophorectomy is performed after childbirth;

[0030] For those who are not planning to have children for the time being, oral contraceptives;

[0031] Immediate family members undergo relevant genetic testing.

[0032] Furthermore, the generating of intelligent diagnosis results by retrieving corresponding hierarchical consultation protocols based on the standard symptom data includes:

[0033] Screening out several key symptoms from the standard symptom data;

[0034] Obtaining a diagnosis and treatment guideline path tree, and retrieving nodes matching the key symptoms therefrom to obtain several preliminary diagnosis paths;

[0035] A preliminary judgment result is generated according to the proportion of the matched nodes in the preliminary diagnosis path.

[0036] Furthermore, generating a preliminary judgment result according to the proportion of the matched nodes in the preliminary diagnosis path includes:

[0037] Calculating the proportion of the matched nodes in each of the preliminary diagnosis paths;

[0038] If there is a preliminary diagnostic pathway with a proportion greater than or equal to the set threshold, the disease screening corresponding to the pathway is determined to be positive;

[0039] Otherwise, if any of the set emergency warning words is matched, the consultant is judged to be a high-risk screening subject.

[0040] Furthermore, the method also includes encapsulating the standard symptom data into structured data that complies with the HL7 FHIR standard in JSON-LD format, and embedding the encapsulated structured data into the hospital EMR system using the SMART on FHIR protocol.

[0041] Furthermore, the method also includes the step of analyzing and obtaining the evolution trajectory of the consultant's symptoms over time based on the standard symptom data, and then generating a symptom heat map.

[0042] Furthermore, the method also includes a step of following up the consultant based on the genetic risk assessment and the intelligent diagnosis results.

[0043] Furthermore, the tumor medical history information includes basic information, tumor medical history and death information.

[0044] Furthermore, the use of the tumor medical history information to construct a tumor genetic pedigree diagram includes:

[0045] Define family members with unique member IDs, and standardize the collected data into a family member data structure with member ID as the primary key to obtain a family member database;

[0046] Based on the family member database, starting from the consultant, the generation coefficient is decreased upwards and increased downwards, the kinship of all family members is polled and the corresponding family member nodes are matched to establish a tree-structured kinship network;

[0047] Traverse the kinship network to draw out each family member node, identify the family member nodes with a history of cancer, and draw relationship lines between nodes based on kinship to obtain the consultant's cancer genetic pedigree.

[0048] Furthermore, polling the kinship of all family members and matching corresponding family member nodes to establish a kinship network with a tree structure includes:

[0049] Create a set of nodes to be matched and initialize them to all family member nodes;

[0050] Establish a node set for each generation and move the consultant from the node set to be matched to the node set of this generation;

[0051] According to the order of the consultant's spouse, the consultant's children, the consultant's siblings and their spouses, the consultant's siblings' children, the consultant's parents, the consultant's parents' siblings, and the consultant's parents' parents, the family member nodes that match the current kinship relationship are retrieved from the set of nodes to be matched. During the retrieval, the family members are matched according to the arrangement order of the family members in the node set of the corresponding generation, and the matched family member nodes are moved to the node set of the corresponding generation according to the matching order to obtain a tree-structured kinship network.

[0052] Furthermore, the process of searching for family member nodes that match the current kinship relationship from the set of nodes to be matched is carried out in the order of the consultant's spouse, the consultant's children, the consultant's siblings and their spouses, the consultant's siblings' children, the consultant's parents, the consultant's parents' siblings, and the consultant's parents' parents, and matching the family member nodes according to the order in which the family members are arranged in the node set of the corresponding generation during the search, and moving the matched family member nodes to the node set of the corresponding generation in the matching order, including:

[0053] Search for the spouse and children of the consultant from the set of nodes to be matched, move the spouse to the current generation node set and place it to the left of the consultant, and move the children to the child generation node set;

[0054] Find the consultant's siblings and the spouse of each sibling from the set of nodes to be matched, and move each sibling and their spouse to the current generation node set adjacent to each other and place them to the right of the consultant;

[0055] Traverse each sibling of the consultant in the current generation node set, search for the children of each sibling in the to-be-matched node set in order, place the family member nodes that are siblings of each other adjacently, and move the children of each sibling to the descendant node set in the search order and place them to the right of the consultant's children;

[0056] Find the parents of the consultant from the set of nodes to be matched, move their parents to the set of parent nodes and place the mother to the right of the father;

[0057] Find the siblings of both parents of the consultant from the set of nodes to be matched, move the siblings of the father to the parent node set and place them on the left side of the father, and move the siblings of the mother to the parent node set and place them on the right side of the mother;

[0058] The parents of both parents of the consultant are searched from the set of nodes to be matched, and their grandparents and maternal grandparents are moved to the ancestor node set in the search order.

[0059] Furthermore, when drawing a tumor genetic pedigree, the distance between any two adjacent generations is set to a fixed value.

[0060] Furthermore, when drawing a tumor genetic pedigree diagram, the node corresponding to the consultant is used as the benchmark, and the node spacing is dynamically allocated according to the number of nodes of family members of the same generation.

[0061] Furthermore, the family member data structure includes a member ID, a basic information field, a tumor history field, a death information field and a kinship field; the kinship field includes a spouse member ID, a child member ID, a parent member ID and a sibling member ID.

[0062] Furthermore, the family member node adopts the family member data structure.

[0063] Furthermore, the collection of the consultant's symptom information and the tumor medical history information of his / her family members in several generations is achieved by constructing an AI intelligent body to simulate a doctor to conduct automated medical interviews with the consultant.

[0064] Furthermore, during automated medical consultations, the AI ​​agent dynamically analyzes whether the consultant's relatives have related disease factors through hybrid retrieval technology combined with a customized genetic knowledge base, and conducts in-depth medical consultations on the relative branches with related disease factors based on the analysis results.

[0065] Furthermore, the family member database is obtained by the AI ​​agent through large-model JSON structured processing and Function Calling technology, combined with prompt word engineering technology, to understand professional tumor information from natural text information and extract effective fields that are conducive to structuring, and assemble them into JSON structured data for specified gynecological tumors.

[0066] Beneficial effects

[0067] Due to the adoption of the above-mentioned technical solution, the present invention has the following advantages and positive effects compared with the existing technology: the present invention collects the consultant's symptom information by constructing an AI intelligent platform, and constructs a symptom semantic network based on medical ontology to convert the collected symptom information into standard symptom data, and then retrieves the corresponding hierarchical consultation protocol according to the standard symptom data to generate intelligent diagnosis results, which can provide professional and accurate answers; in addition, through the standardized packaging of data, clinical research forms can be automatically generated, which completely liberates doctors from the tedious and inefficient work of manually collecting information and entering electronic medical records, allowing doctors to focus more on the core diagnosis and treatment process, effectively improving the utilization efficiency of medical resources and the quality of medical services; the present invention also collects the tumor medical history information of relatives in several generations of the consultant's family and draws a tumor genetic pedigree chart, and then uses the genetic pedigree chart to perform tumor genetic risk assessment, thereby realizing self-service tumor genetic risk assessment and greatly improving the efficiency and quality of medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic diagram of a process of an embodiment of the present invention;

[0069] Figure 2 This is a schematic diagram of the medical consultation process according to an embodiment of the present invention;

[0070] Figure 3 is a tumor pedigree diagram according to an embodiment of the present invention;

[0071] Figure 4 It is a schematic diagram of the intelligent follow-up process of an embodiment of the present invention. DETAILED DESCRIPTION

[0072] Below in conjunction with specific embodiment, further set forth the present invention.Should be understood that these embodiments are only used to illustrate the present invention and are not used in limiting the scope of the present invention.In addition, should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall equally within the scope limited by the appended claims of the application.

[0073] The embodiment of the present invention relates to a gynecological tumor intelligent diagnosis and risk assessment method based on an artificial intelligence platform, such as Figure 1 As shown, the following steps are included:

[0074] Collect information on the consultant's symptoms and the cancer history of relatives within several generations of their family;

[0075] Using the tumor medical history information to construct a tumor genetic pedigree, and then assessing the genetic risk of the consultant to generate genetic counseling recommendations;

[0076] Construct a symptom semantic network based on medical ontology and convert the collected symptom information into standard symptom data;

[0077] Based on the standard symptom data, the corresponding hierarchical consultation protocol is retrieved to generate intelligent diagnosis results;

[0078] A diagnostic recommendation is generated by combining the genetic risk assessment result and the intelligent diagnosis result.

[0079] Among them, the consultant’s relevant information can be realized by building an AI intelligent platform to simulate the doctor to conduct automated diagnosis of the consultant. Figure 2 As shown, the intelligent agent platform provides a specially designed mobile application interface, allowing users to use the system without obstacles at any time and any place. Users can conveniently input text through their mobile phones to start a conversation with the artificial intelligence agent, and can be equipped with a voice-to-text function to reduce the difficulty of user input. After receiving the text input by the user, the intelligent agent platform immediately starts the semantic analysis program, and through in-depth analysis of the semantics of the input text, it determines whether the user has the intention to undergo risk screening, and clarifies whether the user is seeking consultation for his or her own health status or for related consultation on behalf of others. Based on the determined user intentions, the intelligent agent platform provides users with corresponding services, namely risk screening process guidance or gynecological tumor consultation services.

[0080] For users who explicitly request risk screening, the system will automatically trigger a prompt to guide them through the process of collecting information about several generations of their family and creating a cancer genetic pedigree. This information includes basic information about relatives, their cancer history, and any deaths.

[0081] In order to achieve accurate identification and management of family members, the system assigns a unique member ID to each family member and uses the member ID as the primary key to standardize the collected information according to specific rules, ultimately forming a family member data structure. Through this process, the system builds a comprehensive and standardized family member database. The family member database is processed by the AI ​​intelligent platform through large-model JSON structured processing and Function Calling technology, combined with prompt word engineering technology, to understand professional tumor information from natural text information, extract valid fields that are conducive to structuring, and assemble them into JSON structured data for specified gynecological tumors.

[0082] In some preferred embodiments, the family member data structure includes multiple key fields, specifically member ID, basic information field, cancer medical history field, death information field, and kinship field. The kinship field is further broken down into spouse ID, offspring ID, parent ID, and sibling ID. Through this meticulous and comprehensive data structure design, the system can accurately record and reflect the complex kinship relationships between family members, providing a solid data foundation for subsequent risk assessment.

[0083] Based on the established family member database, with the screened person as the starting point, the generation coefficient is decreased upwards and the generation coefficient is increased downwards, and the kinship of all family members in the database is polled one by one. In this process, the system matches the corresponding family member nodes and builds a tree-structured kinship network based on this. After completing the construction of the kinship network, the system further traverses the network and draws out each family member node. For family member nodes with a history of cancer, the system will make special marks and draw the relationship lines between the nodes based on the accurate kinship, thereby generating a clear and complete tumor genetic pedigree diagram for the screened person, such as Figure 3 shown.

[0084] More specifically, when drawing a tumor genetic pedigree chart, the system follows strict setting rules to ensure the standardization and readability of the pedigree chart. Among them, the spacing between any two adjacent generations is set to a fixed value. The setting of this fixed value has been carefully considered to ensure that the pedigree chart is clear in visual presentation and that the hierarchical relationship between different generations is reflected. At the same time, with the node corresponding to the screened person as the reference point, the system dynamically and reasonably allocates the node spacing based on the actual number of nodes of family members of the same generation. This dynamic allocation mechanism can make full use of the drawing space, avoid nodes being too dense or sparse, and enable the pedigree chart to maintain a good display effect at different family sizes.

[0085] After the pedigree is drawn, the AI ​​system automatically reads the key information in the pedigree and calls the pre-set assessment model to perform risk assessment. In this implementation, a comprehensive application of professional assessment models including HBOC risk judgment logic, LS risk assessment logic, and PJS risk assessment logic is selected to provide targeted genetic counseling recommendations after a comprehensive comprehensive risk assessment. The assessment results are then organized and converted into natural language conclusions that are easy for users to understand, providing users with an intuitive and easy-to-understand risk assessment report. Specifically, it includes:

[0086] For those who have not yet undergone genetic testing, we recommend appropriate genetic testing based on the comprehensive assessment results;

[0087] For those who have undergone genetic testing, different suggestions are given for different genes, such as:

[0088] BRCA1 / BRCA2 genes: If the patient is found to carry the BRCA1 / BRCA2 gene mutation, the risk of ovarian cancer, breast cancer and other gynecological tumors is significantly increased. Patients are advised to strengthen cancer screening, such as a breast clinical examination and breast MRI every 6-12 months starting at the age of 25-28; and annual transvaginal ultrasound and serum CA125 testing to screen for ovarian cancer. For patients who wish to have children, preventive salpingo-oophorectomy after childbirth can be considered to reduce the risk of ovarian cancer. If children are not planned for the time being, oral contraceptives can be considered to reduce the risk, but the side effects of the drugs need to be weighed. In addition, it is recommended that their immediate family members also undergo relevant genetic testing to detect potential risks early.

[0089] Lynch syndrome-related genes such as MLH1, MSH2, MSH6, and PMS2: Patients who carry mutations in these genes have an increased risk of developing various cancers, including endometrial cancer and colorectal cancer. For female patients, it is recommended that endometrial biopsies and colonoscopies be performed every 1-2 years, starting at the age of 20-25, or 5-10 years earlier than the age of the earliest cancer diagnosis in the family. In terms of lifestyle, maintaining a healthy weight, increasing exercise, quitting smoking, and limiting alcohol consumption are encouraged to reduce the risk of cancer. Considering the genetic nature of the disease, it is recommended that family members of the patient also undergo genetic testing to enable early detection and intervention.

[0090] STK11 gene (associated with PJS syndrome): Carrying a STK11 gene mutation increases the risk of developing gastrointestinal hamartomatous polyps and gynecological cancers such as breast and ovarian cancer. Regular colonoscopy is recommended to monitor polyps and, if necessary, surgical removal. Gynecological cancer screening should refer to the screening protocol for individuals with BRCA gene mutations, but should be adjusted based on individual circumstances. Regular physical examinations are recommended, and patients should closely monitor their health and seek medical attention promptly if any unusual symptoms occur. Inform patients of the importance of genetic testing for their family members to identify potential risks early and implement preventive measures.

[0091] Throughout the genetic counseling process, the AI ​​platform provides information to clients in a variety of ways. For example, an intelligent voice assistant provides detailed interpretation of genetic counseling recommendations; genetic test results, risk assessments, and counseling recommendations are presented in an illustrated electronic report format for easy access at any time; and for complex medical content, online or offline Q&A sessions with professional doctors are arranged to ensure that clients fully understand the relevant information and make appropriate health decisions.

[0092] For users who still need consultation, the system will provide them with professional gynecological tumor consultation services. Through a carefully designed multi-round dialogue mechanism, it will guide patients to accurately and thoroughly describe their main symptoms. For example, the following symptoms may be included:

[0093] Bleeding pattern, including but not limited to the timing of bleeding during the menstrual cycle, changes in bleeding volume, and whether bleeding is regular;

[0094] Pain characteristics, such as the location of pain (lower abdomen, pelvis, or other specific area), the nature of the pain (dull, sharp, stabbing, etc.), the frequency and duration of pain attacks, etc.;

[0095] Abnormal discharge characteristics, including color (whether it is yellow, green, bloody, etc.), odor (whether it has a strange smell, foul odor, etc.), and texture (thin, thick, dreg-like, etc.);

[0096] Body mass changes, which involve the increase or decrease trend and magnitude of body mass over a certain period of time;

[0097] Accompanying symptoms, such as fever, fatigue, frequent urination, urgency, and other related symptoms.

[0098] Natural language questions entered by users are transformed through a carefully designed prompt word project, allowing them to be submitted to the large model in a manner that is easier for the model to understand and process. To accurately process natural language expressions, this implementation utilizes medical ontology to construct a symptom semantic network. This effectively maps the symptoms described by patients in natural language to the ICD-11 standard terminology system, providing a standardized and normalized data foundation for subsequent diagnostic analysis.

[0099] Based on the acquired structured symptom data, this embodiment activates the hierarchical consultation protocol through Function Calling technology and analyzes and obtains the diagnosis results.

[0100] In the intelligent consultation system for gynecological tumors of the present invention, the hierarchical consultation protocols include primary protocols, advanced protocols and critical value protocols, each of which has clear triggering conditions and has close logical connections with each other.

[0101] Primary Protocol: The agent invokes the "Guidelines for the Diagnosis and Treatment of Gynecological Oncology" path tree and strictly implements mandatory symptom combination screening. This precise traversal of the guideline path tree ensures comprehensive and systematic screening of key symptom combinations, providing a solid foundation for preliminary diagnosis. The primary protocol trigger logic: When the client's standard symptom data meets the trigger criteria for general screening, the primary protocol is initiated. Its core operation is symptom path tree screening. The specific implementation process is based on a pre-defined medical ontology system and authoritative gynecological oncology diagnosis and treatment guidelines, such as the NCCN Guidelines and the ESGO Guidelines. First, a symptom keyword library and a symptom combination rule library are established, which serve as the basis for processing the standardized symptom data. The system uses keyword matching to accurately extract core symptom terms from the user's description and then meticulously compares these symptom terms against a pre-set high-sensitivity symptom combination table. If the number or combination of key symptoms does not meet the established screening criteria, and no key symptom combination is matched, nor are there any critical warning words, the system determines that the general screening trigger criteria have been met and activates the primary protocol. During the primary protocol execution phase, the agent will call upon the path tree of the "Guidelines for the Diagnosis and Treatment of Gynecological Oncology" and strictly follow the regulations to screen for required symptom combinations. This process will comprehensively and systematically sort out key symptom combinations, providing a solid and reliable foundation for subsequent preliminary diagnosis.

[0102] Advanced Protocol: Based on the strength of correlation between symptoms, the agent dynamically loads a library of imaging features (following FIGO staging criteria) and a tumor marker decision matrix (such as CA125 / HE4 / ROMA index). When symptoms are closely correlated and indicate the possibility of a specific disease, relevant imaging features and tumor marker data are promptly incorporated to facilitate a more in-depth and accurate diagnosis. Triggering Logic for the Advanced Protocol: When the system, during in-depth analysis of standard symptom data, identifies significant features suggestive of a tumor, the Advanced Protocol is activated. Specifically, this occurs when the correlation strength between associated symptoms extracted from the standard symptom data exceeds a set threshold and these associated symptoms strongly indicate the possibility of a specific disease. In this case, the system proactively retrieves relevant imaging features and tumor marker data for the specific disease, such as CA125 index, HE4 index, and ROMA index, leveraging this critical information to assist in diagnosis and initiate the Advanced Protocol. During the Advanced Protocol's execution, the agent dynamically loads a library of imaging features based on FIGO staging criteria and a tumor marker decision matrix based on the strength of correlation between symptoms. By integrating these resources, a more in-depth and accurate diagnosis is achieved, striving to clarify the specific disease situation.

[0103] Critical Value Protocol: The intelligent agent matches the NCCN alert vocabulary in real time and immediately initiates emergency response procedures for red alert symptoms with serious warning significance, such as "progressive abdominal distension" and "cachexia." Upon detecting such critical symptoms, appropriate measures are swiftly implemented to protect the patient's life. Triggering Logic for the Critical Value Protocol: Once a high-risk symptom alert is triggered, the critical value protocol is immediately activated. The system has pre-built a symptom alert vocabulary for symptoms with serious warning significance. When searching standard symptom data, if it finds a symptom included in the symptom alert vocabulary, such as "progressive abdominal distension" or "cachexia" that are classified as red alert symptoms, the system reacts swiftly, initiating the emergency response process and entering the critical value protocol state. Under the guidance of the critical value protocol, the system swiftly implements a series of measures to protect the patient's life, including promptly notifying the relevant physician to ensure the rapid intervention of professional medical personnel and arranging emergency examinations to obtain more detailed information as soon as possible, providing strong support for subsequent treatment. Interrelationships between the protocols: The three protocols have a clear progressive and complementary relationship. The primary protocol, as the foundation of the entire screening process, primarily performs preliminary symptom screening for most inquiries. If suspicious signs are detected during the primary protocol screening process, but the existing information cannot provide a definitive diagnosis, the system will trigger the advanced protocol based on the specific circumstances. Advanced protocols incorporate additional diagnostic evidence, such as imaging features and tumor marker data, to provide a more in-depth analysis and assessment. The critical value protocol, however, serves as a crucial emergency safeguard within the entire system. Regardless of whether the system is screening in the primary or advanced protocol, the critical value protocol is immediately activated upon detection of critical symptoms. This protocol prioritizes patient safety, ensuring that patients with urgent and severe symptoms receive timely and effective treatment. For example, if a warning word is detected during the primary protocol symptom tree screening, the system will immediately terminate the primary protocol and initiate the critical value protocol. Similarly, if no significant abnormalities are detected during the primary protocol screening, but critical symptoms emerge during the advanced protocol analysis of associated symptoms, the critical value protocol will also be swiftly activated to maximize patient safety.

[0104] In actual application scenarios, once the consultant enters symptom information and family medical history, the system will quickly initiate two parallel analysis processes. The pedigree analysis module will conduct in-depth mining of family medical history data to determine the consultant's genetic risk level; at the same time, the consultation module will conduct a detailed analysis of the symptom information to generate intelligent diagnostic results. Subsequently, the comprehensive assessment module will integrate these two results and, based on preset scientific rules, ultimately generate personalized diagnostic recommendations including a follow-up plan. For consultants with high genetic risks and suspected tumor symptoms, the system will efficiently arrange high-level examinations for them and promptly refer them to specialists for consultation to ensure that they receive the most professional diagnosis and treatment. For consultants with low genetic risks but mild symptoms, the system will develop a reasonable routine follow-up plan for them and provide targeted health management advice to help them maintain health and prevent the occurrence and development of diseases.

[0105] The pedigree chart and medical interview results are not isolated from each other, but are closely intertwined and complementary, providing key support for the accurate diagnosis and risk assessment of gynecological tumors.

[0106] Pedigree charts are primarily used to determine whether a client is at high genetic risk, which in turn influences screening priorities and emergency alert mechanisms. Specifically, if the system's in-depth analysis of the pedigree chart reveals multiple family members with gynecological or other related cancers, and these cases exhibit a specific inheritance pattern, the client is identified as being at high genetic risk. Once identified as high risk, subsequent screening strategies are adjusted, significantly increasing the priority level. For example, more frequent checkups, such as regular gynecological ultrasounds and tumor marker testing, may be scheduled to detect potential lesions. Alternatively, more in-depth testing, such as genetic testing, may be employed to further clarify whether the client carries the relevant disease-causing genes. In extreme cases, if the risk assessment indicates an extremely high risk, the system will immediately trigger an emergency alert, notifying both the physician and the client, prompting them to take effective intervention measures as soon as possible to avoid delays in treatment.

[0107] The results of the consultation are based on a detailed analysis of the consultant's symptom information. First, the system will convert the symptom information described by the consultant into standard symptom data by constructing a symptom semantic network, laying the foundation for subsequent accurate diagnosis. Afterwards, based on these standard symptom data, the system will call the corresponding hierarchical consultation protocol to generate intelligent diagnostic results. For example, in the process of analyzing symptom data, if the consultant is found to have symptoms such as "progressive abdominal distension" that have critical warning significance, the system will preliminarily determine that the consultant is a high-risk screening subject; if the symptom combination covers a certain proportion of the required nodes in a specific path tree, such as 80%, it will be preliminarily judged as a positive screening result for the corresponding disease.

[0108] The genetic risk assessment results of the pedigree chart are combined with the intelligent diagnosis results obtained from the medical consultation to generate a more comprehensive and accurate preliminary intelligent diagnosis, genetic risk assessment, and screening or follow-up recommendations. For example, if the pedigree chart shows that the consultant belongs to a high-risk group for genetic diseases, and at the same time, during the medical consultation, it is found that the consultant has symptoms of suspected tumors, then after comprehensive consideration, the system may directly recommend that the consultant undergo a comprehensive tumor examination. In addition to routine imaging examinations such as pelvic CT and MRI, it will also focus on testing relevant tumor markers such as CA125 and HE4 to more accurately judge the condition. Conversely, if the pedigree chart shows a low genetic risk, but some warning symptoms appear during the medical consultation, the system may recommend that the consultant conduct regular follow-up observations, pay close attention to changes in symptoms, and provide corresponding health management suggestions based on the actual situation, such as lifestyle adjustments and regular self-examination guidance.

[0109] Furthermore, this implementation leverages the advanced capabilities of Search-Augmented Generation (RAG) to empower the large model with a vertical domain knowledge base of gynecological oncology. With this specialized knowledge base, the large model can provide professional and accurate responses to user inquiries.

[0110] The RAG function relies on its built-in gynecological tumor knowledge base. Under the Retrieval Augmentation Generation (RAG) framework, the system of the present invention integrates three knowledge sources to improve the accuracy and comprehensiveness of diagnosis:

[0111] A localized gynecological oncology knowledge map covers the latest guidelines from international authoritative organizations such as FIGO, NCCN, and ESGO. By integrating cutting-edge medical knowledge, it provides the latest and most authoritative knowledge support for diagnosis.

[0112] The Evidence-Based Medicine Evidence Database enables data exchange through the clinical decision support system interface. Leveraging a wealth of clinical evidence from evidence-based medicine, it provides a scientific and reliable basis for diagnosis and treatment decisions.

[0113] The personalized treatment rule engine dynamically generates personalized treatment plans based on key information such as the patient's age, fertility needs, and genetic test results. This fully considers individual patient differences and enables precision medicine.

[0114] Compared to common large-scale models, this knowledge base empowers the system with unique vertical expertise. When users ask questions about gynecological tumors, the system leverages the expertise within the local knowledge base to quickly and accurately search and match information, providing more targeted and professional responses and effectively improving their consultation experience and effectiveness.

[0115] In some preferred implementations, a multi-dimensional CRF form can also be constructed synchronously during the consultation process. The specific operations are as follows:

[0116] Encapsulate structured medical consultation data in JSON-LD format to ensure data compliance with the HL7 FHIR standard. Utilizing standardized data formats facilitates data storage, transmission, and sharing.

[0117] Using the SMART on FHIR protocol, CRFs are seamlessly embedded into the hospital's EMR system, enabling field-level data mapping. This ensures that consultation data can be efficiently integrated with the hospital's existing systems, improving the continuity of medical processes and data availability.

[0118] Furthermore, the system can automatically extract the temporal symptom evolution trajectory and generate a visual symptom heat map. By dynamically analyzing the changes in symptoms over time, it can visually present the development trend of the disease and assist doctors in making more accurate judgments.

[0119] In addition, an intelligent system of human-machine collaboration can be built to integrate AI agents into the follow-up of gynecological cancer patients, which can not only give full play to the efficiency advantages of AI but also retain the core position of doctors in complex decision-making.

[0120] like Figure 4 As shown, the AI ​​intelligent agent platform can be responsible for the following follow-up plans, including: standardized data collection, initial symptom screening and warning, medication reminder supervision, psychological status monitoring, daily notification information sending, etc.

[0121] However, AI is prohibited from automatically performing complex and sensitive follow-up tasks, such as adjusting treatment plans, diagnosing recurrence / metastasis, and making decisions about end-of-life care. For these tasks, AI can only provide recommendations to doctors, who then make the final decision.

[0122] During the AI ​​follow-up process, multimodal input data can be supported, such as: speech recognition, dialect adaptation, incision photo infection grading assessment, and judgment of the patient's language and emotional depression tendency.

[0123] In this embodiment, the human-machine collaborative follow-up may include one or more of the following tasks:

[0124] (1) Intelligent agents work automatically:

[0125] Automatically formulate follow-up plans and templates

[0126] Recommend clinical trials based on the patient's financial ability (NLP analysis program informed consent form)

[0127] Generate a personalized health education plan (such as targeted drug use and grapefruit taboo reminders)

[0128] Predict disease trends based on test indicators, physical signs data, examination reports, etc.

[0129] (2) Doctor supervision interface:

[0130] Real-time display: patient anxiety index, abnormal signs cluster analysis, doctor-patient dialogue keyword cloud

[0131] (3) Authorizing AI to automatically process scenarios:

[0132] Issue routine blood test application forms (only for maintenance patients who have been stable for more than 3 months); adjust antiemetic dosages (automatically calculated based on the MASCC score sheet); and approve transportation subsidy applications (based on the medical insurance policy knowledge base).

[0133] (4) Emergency takeover plan:

[0134] When the agent recognizes that the patient mentions keywords such as "suicidal tendencies":

[0135] Immediately start human doctor video intervention

[0136] Contact preset emergency contacts simultaneously

[0137] Automatically save all interaction data for future reference.

Claims

1. A method for intelligent diagnosis and risk assessment of gynecological tumors, characterized by: The following steps are involved: Collect information on the consultant's symptoms and the cancer history of relatives within several generations of their family; Using the tumor medical history information to construct a tumor genetic pedigree, and then assessing the genetic risk of the consultant to generate genetic counseling recommendations; constructing a symptom semantic network based on medical ontology and converting the symptom information into standard symptom data; Retrieving the corresponding hierarchical consultation protocol based on the standard symptom data to generate an intelligent diagnosis result; The genetic risk assessment results and the intelligent diagnosis results are combined to output a diagnostic suggestion.

2. The method according to claim 1, characterized in that The genetic risk assessment of the consultant is achieved by conducting a comprehensive risk assessment using a professional assessment model including HBOC risk judgment logic, LS risk assessment logic and PJS risk assessment logic.

3. The method according to claim 2, characterized in that Genetic counseling recommendations include: Generate the first consultation recommendation for those who carry BRCA1 / BRCA2 gene mutations; Generate second consultation recommendations for those who carry genes related to Lynch syndrome; A third consultation recommendation is generated for those who carry the STK11 gene mutation.

4. The method according to claim 3, characterized in that The first consultation recommendation includes risk prevention and screening for gynecological tumors, including ovarian cancer and breast cancer; the second consultation recommendation includes risk prevention and screening for tumors, including endometrial cancer and colorectal cancer; the third consultation recommendation includes risk prevention and screening for diseases, including ovarian cancer, breast cancer, and gastrointestinal hamartomatous polyps.

5. The method according to claim 4, characterized in that The risk prevention and screening for gynecological cancers, including ovarian and breast cancer, include: Strengthen gynecological cancer screening; For those who wish to have children, preventive salpingo-oophorectomy is performed after childbirth; For those who are not planning to have children for the time being, oral contraceptives; Immediate family members undergo relevant genetic testing.

6. The method according to claim 4, characterized in that The cancer risk prevention and screening for endometrial cancer and colorectal cancer include: For female consultants, start from the age of 20-25, or 5-10 years earlier than the age at which the earliest cancer in the family was diagnosed. Endometrial biopsy and colonoscopy every 1–2 years; Maintain a healthy weight, increase exercise, quit smoking and limit alcohol consumption; Family members undergo relevant genetic testing.

7. The method according to claim 4, characterized in that The risk prevention and screening for diseases including ovarian cancer, breast cancer, and gastrointestinal hamartomatous polyps include: Regular gastrointestinal endoscopy to monitor polyps and surgically remove them if necessary; Strengthen gynecological cancer screening; For those who wish to have children, preventive salpingo-oophorectomy is performed after childbirth; For those who are not planning to have children for the time being, oral contraceptives; Immediate family members undergo relevant genetic testing.

8. The method according to claim 1, characterized in that The step of retrieving the corresponding hierarchical consultation protocol based on the standard symptom data and generating an intelligent diagnosis result includes: Screening out several key symptoms from the standard symptom data; Obtaining a diagnosis and treatment guideline path tree, and retrieving nodes matching the key symptoms therefrom to obtain several preliminary diagnosis paths; A preliminary judgment result is generated according to the proportion of the matched nodes in the preliminary diagnosis path.

9. The method according to claim 8, characterized in that Generating a preliminary judgment result according to the proportion of the matched nodes in the preliminary diagnosis path includes: Calculating the proportion of the matched nodes in each of the preliminary diagnosis paths; If there is a preliminary diagnostic pathway with a proportion greater than or equal to the set threshold, the disease screening corresponding to the pathway is determined to be positive; Otherwise, if any of the set emergency warning words is matched, the consultant is judged to be a high-risk screening subject.

10. The method according to claim 1, characterized in that The method also includes encapsulating the standard symptom data into structured data that complies with the HL7 FHIR standard in JSON-LD format, and embedding the encapsulated structured data into the hospital EMR system using the SMART on FHIR protocol.

Images