Admission assessment sheet automatic generation system based on ai large model

Abstract

The application discloses an admission assessment form automatic generation system based on an AI large model, relates to the technical field of admission assessment form generation, and has the technical scheme that after a user inputs information through a voice input mode, a voice file is generated; the voice file is transmitted to a voice recognition model for processing and is converted into initial text; the initial text is transmitted to a large language model, and a target title and an answer specification requirement corresponding to the target title are matched from a knowledge base in association with the content of the initial text; the large language model generates formatted answer content according to the target title, the answer specification requirement and the understanding of the initial text and fills the target form; and rule data of a target medical institution and instance data related to a specific patient are acquired, wherein the rule data comprises department diagnosis and treatment data, clinical path variation adaptation data and patient cross-department referral association data; and the effect is that the admission assessment form is more targeted and applicable.

Description

Technical Field

[0001] This invention relates to the field of hospital admission assessment form generation technology, and more specifically, to an automatic hospital admission assessment form generation system based on an AI large model. Background Technology

[0002] In the daily clinical work of the medical industry, the admission assessment form is a crucial medical document recording a patient's basic condition, past medical history, and symptoms upon admission. The quality and efficiency of its completion directly affect the effectiveness of subsequent treatment. However, traditional automated admission assessment form generation systems rely primarily on manual completion by medical staff, a tedious and time-consuming process. Patient information must be entered item by item, including basic personal information, past medical history, and descriptions of recent symptoms, all meticulously written. Different medical staff have different writing habits and varying understandings and uses of medical terminology, resulting in a lack of standardized format and content in admission assessment forms. This inconsistency hinders the sharing of medical information between different departments and medical institutions, creating difficulties for subsequent medical data analysis based on admission assessment forms, as non-standardized data is difficult to effectively integrate and analyze. Summary of the Invention

[0003] To address the shortcomings of existing technologies, the purpose of this invention is to provide an automatic hospital admission assessment form generation system based on an AI large model.

[0004] To achieve the above objectives, the present invention provides the following technical solution: An AI-based big data model-driven automatic admission assessment form generation system includes: Matching module: After the user inputs information via voice, an audio file is generated; the audio file is sent to the speech recognition model for processing and converted into initial text; the initial text is sent to the large language model, which then matches the target questions associated with the content of the initial text and the corresponding answer specifications from the knowledge base; Filling module: The large language model generates formatted answer content and fills it into the target form based on the target question, answer specification requirements, and understanding of the initial text; Acquisition Module: Acquires rule data of the target medical institution and instance data related to specific patients. The rule data includes departmental diagnosis and treatment data, clinical pathway variation adaptation data, and patient cross-departmental referral association data. The instance data includes patient diagnosis and treatment full-link data and historical form filling error tracing data. Analysis module: Processes and analyzes rule data and instance data to generate an initial version of the target form; Verification module: Based on predefined answer specifications, the initial version is verified for compliance, consistency and risk, and then a list of verification issues and optimization solutions are generated; Output module: Automatically corrects the problems in the initial version based on the checklist and optimization plan, and synchronously updates the dynamic rule base of the form generation, finally outputting the target form that conforms to the specifications.

[0005] Preferably, the rule data and instance data are processed and analyzed to obtain an initial version of the target form, specifically including the following steps: Dynamic adaptation rules are established based on the departmental diagnosis and treatment data and clinical pathway variation adaptation data; field inheritance and supplementation rules for processing referral information are formulated based on the patient cross-departmental referral association data. By processing dynamic adaptation rules, field inheritance, supplementary rules, and patient diagnosis and treatment data across the entire process, a structured diagnosis and treatment information graph is generated. The system calls a dynamic rule base to generate an initial version of the target form by filtering and extracting field data that matches the current department's characteristics, clinical pathway variations, and referral status from the diagnosis and treatment information graph.

[0006] Preferably, the dynamic adaptation rules, field inheritance, supplementary rules, and patient diagnosis and treatment data are processed to generate a structured diagnosis and treatment information graph, which specifically includes the following steps; The dynamic adaptation rules, field inheritance and supplementary rules are integrated to form a dynamic rule library for form generation; The data from the entire patient diagnosis and treatment process is processed in a time-series and structured manner, and a structured diagnosis and treatment information map is generated through event extraction and cross-departmental data association.

[0007] Preferably, the departmental diagnosis and treatment data includes department-specific diagnosis and treatment items, mandatory record items corresponding to characteristic diagnosis and treatment, and statistical standards for departmental diagnosis and treatment data; The clinical pathway variation adaptation data includes fields that need to be added due to adjustments in treatment plans, record item adaptation rules applicable to special cases, and logical association update rules between fields after pathway variation.

[0008] Preferably, dynamic adaptation rules are established based on the departmental diagnosis and treatment data and clinical pathway variation adaptation data, specifically including the following steps: Extract the required records corresponding to specific treatment items from the department's diagnosis and treatment data, and mark them as core fields; Extract fields that need to be added due to adjustments in treatment protocols or special cases from the clinical pathway variation adaptation data, and mark them as variation adaptation fields; Build a mapping relationship and set dynamically adjustable attributes for the fields in the mapping relationship so that the trigger form fields are automatically updated when the relevant clinical pathway variation type changes.

[0009] Preferably, based on the patient inter-departmental referral association data, rules for field inheritance and supplementation for processing referral information are formulated, specifically including the following steps: Based on the aforementioned patient cross-departmental referral data, the fields are categorized into inherited fields and supplementary validation fields; The rules for field inheritance and supplementation are as follows: inherited fields are extracted from referral data and populated into the current form; supplementary verification fields are verified and supplemented in accordance with the clinical pathway variation requirements of the current department. When integrating and forming a dynamic rule base for form generation, define data types and fill format examples for each field.

[0010] Preferably, the patient's entire diagnosis and treatment data is processed in a time-series and structured manner, and a structured diagnosis and treatment information graph is generated through event extraction and cross-departmental data association. Specifically, this includes the following steps: The data from the entire patient diagnosis and treatment process is processed to obtain information on event types, event attributes, and event results. The event type, event attributes, event result information, and data of the current diagnosis and treatment stage are correlated and matched; and a diagnosis and treatment information graph is constructed based on the hierarchical structure.

[0011] Preferably, the data from the entire patient diagnosis and treatment process is processed to obtain event type, event attributes, and event result information, specifically including the following steps: Time series analysis was used to segment and label the entire patient care process data according to the time nodes of diagnosis and treatment. The medical event extraction model extracts event type, event attributes, and event outcome information from the data at each time point.

[0012] Preferably, the initial version of the target form is generated by calling the form generation dynamic rule base to filter and extract field data that matches the current department characteristics, clinical pathway variations, and referral status from the diagnosis and treatment information graph. This specifically includes the following steps: Based on the target department identifier and clinical pathway variation type, the required core fields and variation adaptation fields are selected by calling dynamic adaptation rules; If the patient has a referral record, the directly inherited fields will be automatically filled after the field inheritance and supplementation rules are invoked, and a prompt will be made to add fields that need to be verified. Using the patient's unique identifier as an index, the field data extracted from the diagnosis and treatment information graph is filled into the target form; The required fields for missing data in the diagnosis and treatment information graph are marked, and the format of the entered data is automatically adjusted according to the format example of the form-generated dynamic rule base to obtain the initial version of the target form.

[0013] Preferably, the patient interdepartmental referral data includes the referral diagnosis, recommended treatment plan, pre-assessment opinion of the receiving department, and key examination results during the referral process.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This invention's matching module supports user input via voice, enhancing the convenience of information collection. The generated voice file is converted into initial text by a speech recognition model, and a large language model matches the target question and answer specifications from a knowledge base, ensuring the accuracy and standardization of information extraction and laying a solid foundation for subsequent form generation. Based on the target question, answer specifications, and understanding of the initial text, formatted answer content is generated and filled into the target form. This automates and intelligently transforms information into forms, not only improving the efficiency of form generation but also standardizing colloquial expressions into medical professional terminology, ensuring the professionalism and consistency of the form content. The acquisition module can comprehensively acquire rule data from the target medical institution and instance data related to specific patients. Rule data covers departmental diagnosis and treatment data, clinical pathway variation adaptation data, and patient cross-departmental referral correlation data. Instance data includes patient diagnosis and treatment data across the entire chain and historical form filling error tracing data. This allows subsequent form generation to fully integrate the medical institution's characteristic diagnosis and treatment standards, clinical pathway changes, and the actual situation of patient referrals, making the admission assessment form more targeted and applicable. Meanwhile, historical data on form errors can provide a reference for avoiding similar mistakes and improve form quality. The initial version of the target form is generated by processing and analyzing rule-based and instance data. By integrating and analyzing multi-source data, the value of the data is fully explored, ensuring that the initial form not only conforms to the overall rules of the medical institution but also accurately reflects the individual patient's treatment situation, providing a relatively complete foundation for subsequent verification and correction.

[0015] The validation module performs compliance, consistency, and risk checks on the initial version based on predefined answer specifications, generating a list of validation issues and optimization plans. This step can promptly identify problems in the initial form, such as whether it complies with medical standards, whether the data is inconsistent, and whether there are potential diagnostic and treatment risks, thus avoiding errors in advance and ensuring the accuracy and security of the form.

[0016] The output module automatically corrects issues in the initial version based on the checklist and optimization solutions, and synchronously updates the dynamic rule base for form generation, ultimately outputting a target form that conforms to the specifications. The automatic correction function further improves the efficiency of form generation and reduces the workload of manual correction; while the synchronous update of the dynamic rule base for form generation allows the experience and rules from this correction to be accumulated, continuously optimizing the system and ensuring the quality of subsequently generated forms continues to improve, forming a virtuous cycle. Overall, this greatly improves the efficiency, accuracy, and standardization of admission assessment form generation, providing more reliable information support for clinical diagnosis and treatment. Attached Figure Description

[0017] Figure 1 A schematic diagram illustrating the steps of the automatic generation system for admission assessment forms based on a large AI model proposed in this invention; Figure 2 This is a schematic diagram illustrating the steps of obtaining the initial version of the target form in the AI-based large model-based automatic generation system for admission assessment forms proposed in this invention. Figure 3 This is a schematic diagram in an embodiment of the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0019] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0020] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0021] Reference Figures 1-3 As shown.

[0022] The embodiments further illustrate the automatic generation system for admission assessment forms based on AI large models proposed in this invention.

[0023] An AI-based big data model-driven automatic admission assessment form generation system includes: Matching module: After the user inputs information via voice, an audio file is generated; the audio file is sent to the speech recognition model for processing and converted into initial text; the initial text is sent to the large language model, which then matches the target questions associated with the content of the initial text and the corresponding answer specifications from the knowledge base; Fill module: The large language model generates formatted answer content based on the target question, answer specification requirements, and understanding of the initial text, and fills it into the target form; Users of this application input information via voice, including the patient's basic information, medical history, and symptoms. Once input is complete, a corresponding audio file is generated. This audio file is then sent to a speech recognition model, which processes the speech signal into initial text. This initial text is then sent to a large language model, which, relying on its vast knowledge base, performs semantic analysis on the initial text to identify key information. This allows the model to match target questions related to the initial text content and also retrieves the corresponding answer specifications, such as the answer format and required key information points.

[0024] The large language model obtains the target question and answer specifications, and performs structured processing on the original information after understanding the initial text. It transforms colloquial expressions into standardized medical terminology; for example, "recently experiencing dizziness" becomes the chief complaint: "paroxysmal dizziness." Complex information is logically broken down; for instance, "having coronary heart disease" and "currently taking aspirin and statins" are broken down into "past medical history," "coronary heart disease," "aspirin," and "statins." The large language model generates answer content that conforms to the format requirements, and then accurately fills this content into the target form, thus automatically generating the admission assessment form. The entire process, leveraging AI technology, automates and intelligently generates standardized forms from voice input, thereby improving the efficiency and accuracy of admission assessment form generation.

[0025] Acquisition Module: Acquires rule data of the target medical institution and instance data related to specific patients. Rule data includes departmental diagnosis and treatment data, clinical pathway variation adaptation data, and patient cross-departmental referral data. Instance data includes patient diagnosis and treatment full-chain data and historical form filling error tracing data. Analysis module: Processes and analyzes rule data and instance data to generate an initial version of the target form; Verification module: Based on predefined answer specifications, the initial version is verified for compliance, consistency and risk, and then a list of verification issues and optimization solutions are generated; Output module: Automatically corrects the problems in the initial version based on the checklist and optimization plan, and synchronously updates the form to generate a dynamic rule base, finally outputting the target form that conforms to the specifications.

[0026] The verification module of this application conducts a comprehensive and detailed check on the initial version of the form generated by the filling module based on predefined answer specifications. These predefined answer specifications cover various standards in the medical field, such as terminology usage guidelines, data format requirements, and diagnostic logic principles. The verification module mainly includes compliance, consistency, and risk checks. Compliance checks verify whether the form content conforms to relevant regulations in the medical industry and the internal system requirements of medical institutions, such as whether there is any misuse of medical terminology or missing necessary medical record items. Consistency checks ensure that the information in different parts of the form matches each other without contradictions, such as whether the patient's medical history and current diagnosis are logically consistent. Risk checks focus on identifying missing or incorrect information that may pose risks to subsequent diagnosis and treatment, such as omitting drug allergy history, which could lead to danger during medication. After this series of checks, the verification module generates a detailed checklist of verification issues, clearly listing various problems in the form, and also develops targeted optimization plans to guide how to correct these problems.

[0027] The output module automatically corrects the initial form version based on the checklist and optimization solutions provided by the validation module. It supplements missing information, corrects errors, and adjusts non-compliant formats according to the optimization solutions. During this automatic correction process, the output module simultaneously updates the dynamic rule base for form generation. This dynamic rule base records various rules and experiences from the form generation process. By continuously incorporating validation results and correction strategies, the rule base becomes more comprehensive, better supporting the generation of subsequent forms and preventing similar problems from recurring. Finally, the output module outputs the corrected target form that meets all specification requirements, providing accurate and reliable information for clinical diagnosis and treatment.

[0028] The initial version of the target form is generated by processing and analyzing the rule data and instance data, specifically including the following steps: Dynamic adaptation rules are established based on departmental diagnosis and treatment data and clinical pathway variation adaptation data; field inheritance and supplementation rules are formulated based on patient cross-departmental referral association data to process referral information. By processing dynamic adaptation rules, field inheritance, supplementary rules, and patient diagnosis and treatment data across the entire process, a structured diagnosis and treatment information graph is generated. The system calls a dynamic rule base to generate an initial version of the target form by filtering and extracting field data that matches the current department's characteristics, clinical pathway variations, and referral status from the diagnosis and treatment information graph.

[0029] This application first establishes dynamic adaptation rules based on departmental diagnosis and treatment data and clinical pathway variation adaptation data. For example, a cardiology department may have its own specific myocardial enzyme spectrum testing items, and the clinical pathway for patients with acute myocardial infarction may have special variations, such as requiring more frequent monitoring of myocardial enzyme indicators. Based on this departmental diagnosis and treatment data and clinical pathway variation adaptation data, corresponding dynamic adaptation rules are established to automatically adjust the field settings related to myocardial enzyme monitoring in the form when encountering patients with acute myocardial infarction. At the same time, based on patient cross-departmental referral correlation data, rules for field inheritance and supplementation for processing referral information are formulated. For example, when a patient is referred from surgery to internal medicine, key information about the surgical history needs to be inherited into the internal medicine admission assessment form; this is field inheritance. Internal medicine needs to supplement some assessment items that are not available in surgery and are specific to internal medicine diagnosis and treatment, such as the functional test results of the internal medicine system; this is field supplementation. The corresponding rules clarify which fields need to be inherited and which need to be supplemented.

[0030] By processing dynamic adaptation rules, field inheritance and supplementation rules, and patient treatment end-to-end data, a structured treatment information graph is generated. The patient treatment end-to-end data includes all treatment-related information from the patient's initial visit to their current hospital admission. These data are integrated and organized using rules to construct a clearly hierarchical and logically structured graph, presenting various treatment information such as basic patient information, past medical history, examination results, and treatment processes in a structured format.

[0031] The system invokes a dynamic rule base to generate forms, filtering and extracting field data from the treatment information graph that matches the current department's characteristics, clinical pathway variations, and referral status, thereby generating an initial version of the target form. For example, if the current department is cardiology, treating myocardial diseases, and the patient has clinical pathway variations and referral status, then the system filters out fields related to cardiology-specific treatments, fields adapted to clinical pathway variations, and fields related to referral inheritance and supplementation from the treatment information graph. This data is then extracted and filled into the corresponding positions in the form, thus forming an initial version of the admission assessment form that meets the needs of the current treatment scenario.

[0032] The dynamic adaptation rules, field inheritance, supplementary rules, and patient diagnosis and treatment data are processed to generate a structured diagnosis and treatment information graph, which includes the following steps; Integrate dynamic adaptation rules, field inheritance and supplementary rules to form a dynamic rule library for form generation; The data from the entire patient diagnosis and treatment process is processed in a time-series and structured manner, and a structured diagnosis and treatment information map is generated through event extraction and cross-departmental data association.

[0033] Departmental diagnosis and treatment data includes department-specific diagnosis and treatment items, mandatory record items corresponding to characteristic diagnosis and treatment, and statistical standards for departmental diagnosis and treatment data; Clinical pathway variation adaptation data includes fields that need to be added due to adjustments in treatment protocols, record item adaptation rules applicable to special cases, and logical association update rules between fields after pathway variation.

[0034] This application first integrates dynamic adaptation rules and field inheritance and supplementation rules to form a dynamic rule library for form generation. The dynamic adaptation rules and field inheritance and supplementation rules are formulated by combining departmental diagnosis and treatment data with patient cross-departmental referral data. For example, if a cardiology department has a coronary angiography procedure, the corresponding mandatory record items are the angiography results and intraoperative conditions, along with the department's statistical standards for diagnosis and treatment data, such as the statistical range and calculation method of myocardial enzyme indicators. Clinical pathway variation adaptation data is also crucial. For instance, a new field for emergency PCI postoperative care has been added for patients with acute myocardial infarction. There are applicable record item adaptation rules for myocardial infarction patients with diabetes, and update rules exist for the logical relationship between blood glucose monitoring and myocardial enzyme changes after pathway variation. These are all incorporated into the dynamic adaptation rules. The field inheritance and supplementation rules clarify which information should be directly inherited and which needs to be supplemented during cross-departmental referrals.

[0035] The entire patient care process involves time-series and structured processing of data. Time-series processing organizes the data according to the order of treatment time. For example, a patient might have an electrocardiogram (ECG) at a community hospital, then a coronary CT scan at a tertiary hospital, and subsequently undergo PCI surgery; each time point in the process is sequentially arranged. Event extraction is then used to extract event types, attributes, and outcomes from the data at these time points. Simultaneously, cross-departmental data association is performed, linking the patient's treatment data across different departments to generate a structured infographic of the patient's care. This infographic clearly presents key information throughout the entire treatment process, from onset of illness to hospital admission, providing a comprehensive and organized data source for subsequent form generation.

[0036] Dynamic adaptation rules are established based on departmental diagnosis and treatment data and clinical pathway variation adaptation data, specifically including the following steps: Extract the required records corresponding to specific treatment items from the department's diagnosis and treatment data, and mark them as core fields; Extract fields that need to be added due to adjustments in treatment protocols or special cases from the clinical pathway variation adaptation data, and mark them as variation adaptation fields; Build a mapping relationship and set dynamically adjustable attributes for the fields in the mapping relationship so that the trigger form fields are automatically updated when the relevant clinical pathway variation type changes.

[0037] This application first extracts the required records corresponding to specific treatment items from the department's clinical data and marks them as core fields. For example, if the cardiology department has a specific treatment item called coronary angiography, then the required records for contrast agent usage and the degree of vascular stenosis during the angiography process are marked as core fields. These fields are essential key information that the department cannot do without when conducting clinical assessments.

[0038] Fields added due to adjustments in treatment protocols or special cases are extracted from clinical pathway variation adaptation data and marked as variation adaptation fields. For example, when the treatment protocol for acute myocardial infarction patients is adjusted from routine drug therapy to emergency interventional therapy, fields for intraoperative stent selection and postoperative antithrombotic regimen are added; for special myocardial infarction cases with severe arrhythmias, fields for arrhythmia type and control status are added. These newly added fields are all considered variation adaptation fields.

[0039] A mapping relationship is established between core fields and variant adaptation fields, and dynamic adjustment attributes are set for fields within the mapping relationship. When the variant type of a relevant clinical pathway changes, the triggering form fields are automatically updated. For example, when a clinical pathway changes from a routine treatment pathway for acute myocardial infarction to an interventional treatment pathway for acute myocardial infarction, a variant adaptation field for intraoperative stent selection is automatically added to the form based on the set dynamic adjustment attributes. Simultaneously, it is ensured that the newly added field forms a reasonable association with the core fields of the coronary angiography results. This allows the form to reflect changes in information recording needs due to the specificity of the treatment plan or case in a timely and accurate manner, thus providing dynamic and precise field support for the generation of medical forms.

[0040] Based on patient cross-departmental referral data, rules for field inheritance and supplementation for processing referral information were developed, including the following steps: Based on the patient's cross-departmental referral data, the fields are categorized into inherited fields and supplementary validation fields; The rules for field inheritance and supplementation are as follows: inherited fields are extracted from referral data and populated into the current form; supplementary verification fields are verified and supplemented in accordance with the clinical pathway variation requirements of the current department. When integrating and forming a dynamic rule base for form generation, define data types and fill format examples for each field.

[0041] This application first categorizes relevant fields into inherited fields and supplementary validation fields based on the associated data of patients' interdepartmental referrals. For example, when a patient is referred from surgery to internal medicine, the patient's basic identity information and past surgical history already recorded by surgery are inherited fields; while the patient's cardiac function status and internal medicine-specific examination indicators that the internal medicine department needs to further confirm for its own diagnostic and treatment needs are supplementary validation fields.

[0042] These fields are processed according to the rules for field inheritance and supplementation. For inherited fields, the relevant information is directly extracted from the referral data and populated into the current department's form. For example, the patient's basic identity information is extracted and automatically filled into the corresponding position on the internal medicine admission assessment form. For supplementary verification fields, verification and supplementation are carried out in conjunction with the clinical pathway variation requirements of the current department. For example, if the internal medicine clinical pathway involves the special management of patients with heart failure, the supplementary verification field will be checked to see if the indicator exists in the referral data. If not, it needs to be tested and recorded.

[0043] Finally, when integrating and forming the dynamic rule base for form generation, data types and filling format examples are defined for each field. For example, the inherited field "Surgical History" is a text data type, and the filling format example is "Appendectomy performed on May 10, 2023, with good postoperative recovery"; the supplementary validation field "BNP Level" is a numerical data type, and the filling format example is "350 pg / mL". When generating forms, referral-related fields are processed according to these rules, ensuring the effective transmission of referral information while meeting the current department's diagnostic and treatment needs, thus enabling forms such as admission assessment forms to comprehensively and accurately reflect the patient's condition.

[0044] The entire patient diagnosis and treatment process involves time-series and structured processing of data, followed by the generation of a structured diagnosis and treatment information graph through event extraction and cross-departmental data association. This process includes the following steps: The data from the entire patient diagnosis and treatment process is processed to obtain information on event types, event attributes, and event results. The event type, event attributes, event result information, and data of the current diagnosis and treatment stage are correlated and matched; and a diagnosis and treatment information graph is constructed based on the hierarchical structure.

[0045] This application first processes the data from the entire patient diagnosis and treatment process, extracting event type, event attribute, and event outcome information. For example, when a patient undergoes coronary angiography, the coronary angiography itself is the event type; the type of contrast agent used and the operation duration are event attributes; and the 70% stenosis of the left anterior descending artery after the angiography is the event outcome information.

[0046] This approach associates and matches event types, event attributes, and event outcomes with data from the patient's current treatment stage. For example, if a patient is currently in the preparation stage for coronary intervention, the event information from coronary angiography is linked to their treatment needs. A hierarchical information graph is constructed. This hierarchy progresses from basic patient information to past medical history, and then to different levels of examinations, diagnoses, and treatments. In the information graph, basic patient information is at the most fundamental level. Coronary angiography, as an examination event, has its event type, attributes, and results presented at the examination level, and is linked to the current treatment stage (preparation stage for coronary intervention). This clearly demonstrates the patient's treatment process and key information from examination to subsequent treatment preparation, allowing medical staff to comprehensively and intuitively understand the patient's treatment status and providing strong data support for subsequent treatment decisions and form generation.

[0047] The entire patient diagnosis and treatment process involves processing data to obtain event type, event attributes, and event outcome information. This includes the following steps: Time series analysis was used to segment and label the entire patient care process data according to the time nodes of diagnosis and treatment. The medical event extraction model extracts event type, event attributes, and event outcome information from the data at each time point.

[0048] This application first employs a time-series approach, dividing and labeling the patient's entire treatment process data according to the time nodes of diagnosis and treatment. For example, if a patient undergoes a routine blood test on May 10, 2024, a cardiac ultrasound on May 12, 2024, and starts taking antihypertensive medication on May 15, 2024, these different time points are the treatment time nodes. Based on these time nodes, the entire treatment process data is divided into different segments, and events occurring within each time segment are labeled to clarify whether they are examination events or treatment events.

[0049] The medical event extraction model extracts event type, event attributes, and event result information from the data at each time point. Taking a routine blood test on May 10, 2024 as an example, the medical event extraction model would extract the event type as "routine blood test" from the test data at that time point; the event attributes would include the testing institution and testing equipment; and the event result information would be the specific results of various indicators in the routine blood test, such as a white blood cell count of 5.2 × 10⁻⁶. 9 / L, red blood cell count 4.5×10 12For example, an event where a patient started taking antihypertensive medication on May 15, 2024, is classified as "medication therapy." Event attributes include the drug name and dosage. The outcome might subsequently show a decrease in blood pressure from 160 / 100 mmHg to 135 / 85 mmHg after medication. This method extracts key information from each time point in the patient's entire treatment process, providing a detailed and structured data foundation for subsequent work such as constructing a treatment information graph.

[0050] The initial version of the target form is generated by calling the dynamic rule base for form generation, filtering and extracting field data that matches the current department's characteristics, clinical pathway variations, and referral status from the diagnosis and treatment information graph. This process includes the following steps: Based on the target department identifier and clinical pathway variation type, the required core fields and variation adaptation fields are selected by calling dynamic adaptation rules; If the patient has a referral record, the directly inherited fields will be automatically filled after the field inheritance and supplementation rules are invoked, and a prompt will be made to add fields that need to be verified. Using the patient's unique identifier as an index, the field data extracted from the diagnosis and treatment information graph is filled into the target form; The required fields for missing data in the diagnosis and treatment information graph are marked, and the format of the entered data is automatically adjusted according to the format example of the form-generated dynamic rule base to obtain the initial version of the target form.

[0051] This application first uses the target department identifier and clinical pathway variation type to apply dynamic adaptation rules to filter out the required core fields and variation adaptation fields. For example, for cardiology patients with acute myocardial infarction, the core fields include myocardial enzyme spectrum indicators and electrocardiogram results. If the patient is a special case with diabetes, the variation adaptation fields include dynamic blood glucose monitoring data.

[0052] If the patient has a referral record, the field inheritance and supplementation rules are applied, automatically filling in the directly inherited fields while prompting the patient to supplement any fields that need verification. For example, if the patient was referred from a primary care hospital, the patient's history of hypertension from the primary care hospital is a directly inherited field, which will be automatically filled into the current form; while the cardiac function classification is a field that needs verification, prompting the medical staff in the current department to verify and supplement it.

[0053] Using the patient's unique identifier as an index, relevant field data is extracted from the medical information graph and populated into the target form. The medical information graph integrates the entire medical data chain from patient visit to hospital admission, locating various relevant data for each patient through their unique identifier.

[0054] The system marks required fields for missing data in the medical information graph and automatically adjusts the format of the entered data based on the format example of the dynamic rule base generated by the form, thus obtaining the initial version of the target form. For example, if the required field of admission time is missing in the graph, it is marked; for blood pressure records, the original 130 and 85 are automatically adjusted to 130 / 85 mmHg according to the format example of systolic / diastolic blood pressure in the rule base.

[0055] The data associated with patient referrals across departments includes the referral diagnosis, recommended treatment plan, pre-assessment opinion from the receiving department, and key examination results during the referral process.

[0056] For example, a patient may visit the internal medicine department of a primary care hospital due to symptoms of chest tightness and chest pain. After examination, the primary care hospital may refer the patient to a higher-level hospital and diagnose them with suspected coronary artery disease, but not rule out acute coronary syndrome. The hospital may also provide recommended treatment options, such as suggesting referral to the cardiology department of a higher-level hospital for further coronary angiography to clarify the diagnosis and develop a precise treatment plan.

[0057] The receiving department formulates a preliminary assessment based on the referral information provided by the primary care hospital. For example, if the patient is an elderly male with a history of hypertension, and the primary care hospital's examination results strongly suggest coronary artery disease, further examinations such as coronary angiography are needed to assess the degree of vascular stenosis and determine whether interventional treatment is necessary. Key examination results during the referral process, such as an electrocardiogram showing ST-segment elevation with a convex shape and elevated troponin I in the cardiac enzyme spectrum, are important components of the referral-related data.

[0058] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0059] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An automatic admission assessment form generation system based on an AI large-scale model, characterized in that, include: Matching module: Generates voice files after users input information via voice input; The audio file is sent to a speech recognition model for processing and then converted into initial text. After the initial text is sent to the large language model, it is matched with the target questions associated with the content of the initial text and the corresponding answer specifications from the knowledge base. Filling module: The large language model generates formatted answer content based on the target question, answer specification requirements, and understanding of the initial text, and fills it into the target form; Acquisition Module: Acquires rule data of the target medical institution and instance data related to specific patients. The rule data includes departmental diagnosis and treatment data, clinical pathway variation adaptation data, and patient cross-departmental referral association data. The instance data includes patient diagnosis and treatment full-link data and historical form filling error tracing data. Analysis module: Processes and analyzes rule data and instance data to generate an initial version of the target form; Verification module: Based on predefined answer specifications, the initial version is verified for compliance, consistency and risk, and then a list of verification issues and optimization solutions are generated; Output module: Automatically corrects the problems in the initial version based on the checklist and optimization plan, and synchronously updates the dynamic rule base of the form generation, finally outputting the target form that conforms to the specifications.

2. The automatic admission assessment form generation system based on AI large model according to claim 1, characterized in that, The initial version of the target form is generated by processing and analyzing the rule data and instance data, specifically including the following steps: Dynamic adaptation rules are established based on the departmental diagnosis and treatment data and clinical pathway variation adaptation data; field inheritance and supplementation rules for processing referral information are formulated based on the patient cross-departmental referral association data. By processing dynamic adaptation rules, field inheritance, supplementary rules, and patient diagnosis and treatment data across the entire process, a structured diagnosis and treatment information graph is generated. The system calls a dynamic rule base to generate an initial version of the target form by filtering and extracting field data that matches the current department's characteristics, clinical pathway variations, and referral status from the diagnosis and treatment information graph.

3. The automatic admission assessment form generation system based on AI large model according to claim 2, characterized in that, The dynamic adaptation rules, field inheritance, supplementary rules, and patient diagnosis and treatment data are processed to generate a structured diagnosis and treatment information graph, which includes the following steps; The dynamic adaptation rules, field inheritance and supplementary rules are integrated to form a dynamic rule library for form generation; The data from the entire patient diagnosis and treatment process is processed in a time-series and structured manner, and a structured diagnosis and treatment information map is generated through event extraction and cross-departmental data association.

4. The automatic generation system for admission assessment forms based on AI large-scale models according to claim 3, characterized in that, The departmental diagnosis and treatment data includes department-specific diagnosis and treatment items, mandatory record items corresponding to characteristic diagnosis and treatment, and statistical standards for departmental diagnosis and treatment data; The clinical pathway variation adaptation data includes fields that need to be added due to adjustments in treatment plans, record item adaptation rules applicable to special cases, and logical association update rules between fields after pathway variation.

5. The automatic generation system for admission assessment forms based on AI large-scale models according to claim 4, characterized in that, Dynamic adaptation rules are established based on the departmental diagnosis and treatment data and clinical pathway variation adaptation data, specifically including the following steps: Extract the required records corresponding to specific treatment items from the department's diagnosis and treatment data, and mark them as core fields; Extract fields that need to be added due to adjustments in treatment protocols or special cases from the clinical pathway variation adaptation data, and mark them as variation adaptation fields; Build a mapping relationship and set dynamically adjustable attributes for the fields in the mapping relationship so that the trigger form fields are automatically updated when the relevant clinical pathway variation type changes.

6. The automatic generation system for admission assessment forms based on AI large-scale models according to claim 5, characterized in that, Based on the aforementioned patient cross-departmental referral data, rules for field inheritance and supplementation for processing referral information are formulated, specifically including the following steps: Based on the aforementioned patient cross-departmental referral data, the fields are categorized into inherited fields and supplementary validation fields; The rules for field inheritance and supplementation are as follows: inherited fields are extracted from referral data and populated into the current form; supplementary verification fields are verified and supplemented in accordance with the clinical pathway variation requirements of the current department. When integrating and forming a dynamic rule base for form generation, define data types and fill format examples for each field.

7. The automatic generation system for admission assessment forms based on AI large-scale models according to claim 6, characterized in that, The entire patient diagnosis and treatment process involves time-series and structured processing of data, followed by the generation of a structured diagnosis and treatment information graph through event extraction and cross-departmental data association. This process includes the following steps: The data from the entire patient diagnosis and treatment process is processed to obtain information on event types, event attributes, and event results. The event type, event attributes, event result information, and data of the current diagnosis and treatment stage are correlated and matched; and a diagnosis and treatment information graph is constructed based on the hierarchical structure.

8. The automatic generation system for admission assessment forms based on AI large-scale models according to claim 7, characterized in that, The entire patient diagnosis and treatment process involves processing data to obtain event type, event attributes, and event outcome information. This includes the following steps: Time series analysis was used to segment and label the entire patient care process data according to the time nodes of diagnosis and treatment. The medical event extraction model extracts event type, event attributes, and event outcome information from the data at each time point.

9. The automatic generation system for admission assessment forms based on AI large-scale models according to claim 8, characterized in that, The initial version of the target form is generated by calling the dynamic rule base for form generation, filtering and extracting field data that matches the current department's characteristics, clinical pathway variations, and referral status from the diagnosis and treatment information graph. This process includes the following steps: Based on the target department identifier and clinical pathway variation type, the required core fields and variation adaptation fields are selected by calling dynamic adaptation rules; If the patient has a referral record, the directly inherited fields will be automatically filled after the field inheritance and supplementation rules are invoked, and a prompt will be made to add fields that need to be verified. Using the patient's unique identifier as an index, the field data extracted from the diagnosis and treatment information graph is filled into the target form; The required fields for missing data in the diagnosis and treatment information graph are marked, and the format of the entered data is automatically adjusted according to the format example of the form-generated dynamic rule base to obtain the initial version of the target form.

10. The automatic generation system for admission assessment forms based on AI large models according to claim 9, characterized in that, The patient cross-departmental referral data includes the referral diagnosis, recommended treatment plan, pre-assessment opinion of the receiving department, and key examination results during the referral process.

Images