AI auxiliary decision-making method and system for department-specific medical treatment

Abstract

The invention discloses a department-specific medical AI auxiliary decision-making method and system. The method comprises the following steps: acquiring electronic medical record data of a patient; based on the event triggering mechanism and the patient electronic medical record data, extracting static features, dynamic trend features and feature association degrees, and obtaining patient portrait vectors; constructing a vector knowledge base according to department dimension hierarchy, performing semantic matching retrieval on patient portrait vectors to obtain a retrieval result, and further generating candidate diagnoses, candidate doctor advices and candidate consultation departments; and screening and calibrating the candidate diagnoses and the candidate medical advices in sequence to generate standardized diagnoses and standardized medical advices, calibrating the candidate consultation departments to generate standardized consultation departments, and finally outputting a standardized recommendation result. According to the method, a dynamic patient portrait is generated, a department specific vector knowledge base is constructed, and standardized diagnosis, doctor advice and consultation suggestions are generated through multi-model collaborative recommendation, so that medical AI aid decision-making with precise clinical adaptation, illusion reduction and efficiency improvement is realized.

Description

Technical Field

[0001] This invention relates to the fields of medical data processing and artificial intelligence technology, and in particular to a department-specific medical AI-assisted decision-making method and system. Background Technology

[0002] With the deepening of medical informatization, the integration of artificial intelligence technology and clinical diagnosis and treatment is becoming increasingly close. AI-assisted decision-making systems, by analyzing and processing patient medical record data, provide doctors with core support such as diagnostic suggestions and medical order recommendations. They have become an important tool for improving the efficiency and accuracy of diagnosis and treatment, and their application in clinical scenarios is becoming more and more widespread.

[0003] The core technologies of existing medical AI-assisted decision-making systems mostly revolve around medical record data parsing and single-model reasoning: by extracting basic information such as diagnosis and symptoms from a patient's single medical record to generate a static summary, relying on a general medical knowledge base, using a single model (such as GPT-3.5) for decision reasoning, and finally outputting treatment suggestions in natural language form.

[0004] However, existing technologies have significant shortcomings in practical applications: First, patient profiles are static, and existing technologies are only based on single medical record data, which cannot integrate dynamic data such as changes in symptoms and trends in test indicators after the patient is admitted to the hospital in real time, resulting in recommendations that lag behind the development of the disease.

[0005] Secondly, the recommendations lack clinical applicability. Existing technologies use general medical knowledge bases without considering departmental specific medication habits and clinical pathways. The recommended drugs and tests may exceed the department's access scope or be inconsistent with doctors' clinical practice habits, resulting in high implementation costs.

[0006] Third, the risk of AI illusion is prominent. Existing single-model decision-making lacks an effective verification mechanism, which is prone to "illusionary recommendations", such as recommending non-existent drugs or omitting contraindications, and lacks an effective verification mechanism.

[0007] Fourth, medical orders are not standardized. Existing technologies mostly output natural language suggestions, without specifying key information such as dosage, frequency, and route of administration. Doctors need to manually convert these into standardized medical orders, resulting in low diagnostic and treatment efficiency. Therefore, there is an urgent need for a medical decision support solution that can dynamically update patient status, adapt to the clinical realities of departments, reduce the risk of AI illusions, and output standardized medical orders to address the shortcomings of existing technologies. Summary of the Invention

[0008] To address the aforementioned issues, this invention proposes a department-specific medical AI-assisted decision-making method and system. This invention generates dynamic patient profiles, constructs a department-specific vector knowledge base, and generates standardized diagnoses, medical orders, and consultation suggestions through multi-model collaborative recommendations, thereby achieving precise clinical adaptation, reducing illusions, and improving efficiency in medical AI-assisted decision-making.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: In a first aspect, the present invention provides a department-specific medical AI-assisted decision-making method, comprising the following steps: Obtain patient electronic medical record data; Based on the event triggering mechanism and patient electronic medical record data, static features, dynamic trend features, and feature correlation are extracted, and the static features, dynamic trend features, and feature correlation are fused to generate patient profile vectors and obtain patient profile keywords. A vector knowledge base is constructed hierarchically according to departmental dimensions. Based on the vector knowledge base, semantic matching retrieval is performed on patient profile vectors to obtain retrieval results. Based on patient profile keywords and retrieval results, candidate diagnoses, candidate medical orders, and candidate consultation departments are generated. The candidate diagnoses and candidate medical orders are sequentially screened and calibrated to generate standardized diagnoses and standardized medical orders. The candidate consultation departments are calibrated to generate standardized consultation departments. Finally, standardized recommendation results containing standardized diagnoses, standardized medical orders, and standardized consultation departments are output.

[0010] As an optional implementation, the patient's electronic medical record data includes basic patient information, dynamic diagnosis and treatment information, and department configuration information; wherein the basic patient information includes name, age, gender, past medical history, and surgical history; the dynamic diagnosis and treatment information includes the latest diagnosis, symptom records, laboratory test results, executed medical orders, and execution feedback; the department configuration information includes a list of commonly used drugs in the target department, a list of commonly used laboratory test items, and a clinical pathway template.

[0011] As an alternative implementation, based on the patient's electronic medical record data, static features are extracted using the BERT-BiLSTM-CRF model, including standardized mappings of diagnosis name and surgical name; dynamic trend features are extracted using a temporal convolutional network, including slope and variance of test indicators; and feature correlation is calculated using an attention mechanism.

[0012] As an optional implementation, the vector knowledge base constructed hierarchically by department includes a core layer, an evidence-based layer, a rule layer, and an update unit; the core layer includes instructions for commonly used drugs in the department, operating procedures for laboratory tests, and in-hospital clinical pathways; the evidence-based layer includes core journal literature from UpToDate clinical consultants, the Chinese Clinical Guidelines Database, and PubMed; the rule layer includes departmental expert consensus, rules on contraindications for combined drug use, and rules for dose adjustment; and the update unit updates guidelines / literature through web crawling.

[0013] As an alternative implementation, the event triggering mechanism includes setting a minimum update interval, and triggering an update if a key indicator mutation occurs during the interval; wherein, the key indicator mutation is a doubling of creatinine or a sudden rise in blood sugar.

[0014] As an alternative implementation method, the specific method for generating standardized medical orders is as follows: First, a medical order embedding vector library is constructed. Candidate medical orders are retrieved through FAISS search to obtain similar medical orders. Valid candidate medical orders are then selected through screening. Next, based on the valid candidate medical orders and the patient's electronic medical record data, a personalized fit score is obtained, and highly fit valid candidate medical orders are selected. Finally, the department's historical medical order templates are retrieved, and the statistical mode of dosage, frequency, and usage is extracted to generate standardized medical orders.

[0015] Secondly, the present invention provides a department-specific medical AI-assisted decision-making system, comprising the following modules: The data acquisition module is configured to acquire patient electronic medical record data; The patient profile generation module is configured to: extract static features, dynamic trend features, and feature correlation based on event triggering mechanism and patient electronic medical record data; and fuse static features, dynamic trend features, and feature correlation to generate patient profile vectors and obtain patient profile keywords. The recommendation module is configured to: construct a vector knowledge base hierarchically by department; perform semantic matching retrieval on patient profile vectors based on the vector knowledge base to obtain retrieval results; generate candidate diagnoses, candidate medical orders, and candidate consultation departments based on patient profile keywords and retrieval results; sequentially screen and calibrate the candidate diagnoses and candidate medical orders to generate standardized diagnoses and standardized medical orders; calibrate the candidate consultation departments to generate standardized consultation departments; and finally output standardized recommendation results containing standardized diagnoses, standardized medical orders, and standardized consultation departments.

[0016] Thirdly, the present invention provides an electronic device, including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, complete the department-specific medical AI-assisted decision-making method described in the first aspect.

[0017] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the department-specific medical AI-assisted decision-making method described in the first aspect.

[0018] Fifthly, the present invention provides a computer program product, including a computer program, which, when executed by a processor, implements the department-specific medical AI-assisted decision-making method described in the first aspect.

[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention generates and updates patient profile vectors through an event-triggered mechanism, integrating dynamic patient condition data in real time. This solves the problem of delayed recommendation results caused by static patient profiles in existing technologies, ensuring that recommendation results are synchronized with the progression of the disease. Furthermore, this invention constructs a hierarchical RAG knowledge base according to departmental dimensions, combining department-specific data for semantic matching retrieval and result filtering and calibration. The recommended content aligns with departmental medication habits and clinical pathways, reducing implementation costs.

[0020] This invention verifies candidate results through a multi-stage screening and calibration process, effectively avoiding the AI ​​illusion risk of single-model decision-making, reducing issues such as missed contraindications, and improving the accuracy of treatment recommendations. Furthermore, this invention directly outputs recommended results including standardized diagnoses, standardized medical orders, and standardized consultation departments, eliminating the need for manual conversion by doctors, significantly reducing operational costs and improving clinical efficiency.

[0021] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0022] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0023] Figure 1 This is a flowchart of the dynamic patient profile vector generation in Embodiment 1 of the present invention; Figure 2 This is a flowchart of the multi-model collaborative recommendation in Embodiment 1 of the present invention; Figure 3 This is a module architecture diagram of the department-specific medical AI-assisted decision-making system in Embodiment 2 of the present invention. Detailed Implementation

[0024] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0025] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0026] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0027] Example 1 like Figure 1 As shown, this embodiment provides a department-specific medical AI-assisted decision-making method, using 100 patients with acute myocardial infarction in the cardiology department of a specialized hospital as test subjects. The method specifically includes the following steps: S1: Obtain patient electronic medical record data.

[0028] In step S1, patient electronic medical record data is obtained from the hospital's HIS system, LIS system, and EMR system. The patient electronic medical record data includes: Basic patient information includes, but is not limited to, name, age, gender, past medical history, and surgical history; Dynamic medical information includes, but is not limited to, the latest diagnosis (ICD-10 code), symptom records (including onset time and duration), test results (including indicator values ​​and test time), executed medical orders and execution feedback; Departmental Configuration Information: Top 100 list of commonly used drugs in the target department (including generic name, indications, and contraindications), list of commonly used laboratory tests, and clinical pathway template.

[0029] S2: Based on an event-triggered mechanism and patient electronic medical record data, static features, dynamic trend features, and feature correlation are extracted. These features are then fused to generate a patient profile vector and obtain patient profile keywords. The event-triggered mechanism generates and updates the patient profile vector, integrating dynamic patient condition data in real time. This solves the problem of delayed recommendation results caused by static patient profiles in existing technologies, ensuring that recommendation results are synchronized with the development of the patient's condition.

[0030] In step S2, in response to the medical order signature event, the patient profile is generated and updated through the event triggering mechanism, such as... Figure 1 As shown, it specifically includes: The system subscribes to "medical order signature" and "test result entry" events in the hospital's HIS system, setting a minimum update interval (24 hours in this example). If a critical indicator undergoes a sudden change during this period (such as doubling of creatinine or a sudden increase in blood glucose ≥3 mmol / L), an emergency update is triggered. The system then determines whether the patient profile vector update conditions (medical order signature event trigger or critical indicator change) are met. If met, static features, dynamic trend features, and feature correlation are updated, and the patient profile vector is updated. If not met, no update is performed. This event-driven profile update mechanism integrates patient condition change data in real time, ensuring that recommendations are synchronized with disease progression, reducing the recommendation lag by 80% compared to static profile systems.

[0031] Based on the patient's electronic medical record data, static features are extracted using the BERT-BiLSTM-CRF model. In this embodiment, static features include standardized mappings of diagnosis name and surgical name. Dynamic trend features are extracted using a temporal convolutional network (TCN), which includes the slope and variance of test indicators over the past 7 days. The attention mechanism is used to calculate the feature correlation (such as the mutual information between "pneumonia" and "chest X-ray patchy shadow"). By fusing static features, dynamic trend features, and feature correlation, a patient profile vector is generated, resulting in 1024-dimensional patient profile keywords.

[0032] S3: Construct a vector knowledge base hierarchically by department. Based on the vector knowledge base, perform semantic matching retrieval on patient profile vectors to obtain retrieval results. Based on patient profile keywords and retrieval results, generate candidate diagnoses, candidate medical orders, and candidate consultation departments. Filter and calibrate the candidate diagnoses and candidate medical orders in sequence to generate standardized diagnoses and standardized medical orders. calibrate the candidate consultation departments to generate standardized consultation departments. Finally, output standardized recommendation results containing standardized diagnoses, standardized medical orders, and standardized consultation departments.

[0033] In step S3, as Figure 2 As shown, a vector knowledge base is constructed hierarchically according to departmental dimensions, including: Core layer: Instructions for commonly used drugs in the department (such as aspirin and ticagrelor PDF structured analysis), operating procedures for laboratory tests (such as troponin and electrocardiogram), and in-hospital clinical pathways; Evidence-based layer: UpToDate clinical consultant, Chinese Clinical Guidelines Database (CGAC), and core journal articles with an impact factor ≥ 5 in the past 5 years published in PubMed; Rule level: Departmental expert consensus, rules on contraindications for combined medication (such as "cephalosporins + alcohol"), and rules for dose adjustment (such as reducing drug dose by 30% for patients with renal insufficiency). Update Unit: Guidelines / literature updates are crawled monthly, and department heads manually review new rules monthly, labeling them as "strongly recommended" or "weakly recommended".

[0034] Knowledge bases are built for each department, incorporating departmental medication habits and clinical pathways. Recommended drugs and tests are all within the department's access scope. Standardized medical orders can be directly imported into the HIS system, reducing the average adjustment time for doctors from 15 minutes to 4.8 minutes, improving efficiency by 68%.

[0035] In step S3, based on the vector knowledge base, semantic matching retrieval is performed on the patient profile vector to obtain retrieval results, including relevant literature evidence, drug / test list, and clinical pathway rules. Using the Zhongyang Vertical Domain Medical Domain Large Model (72B parameters), patient profile keywords and retrieval results are input to generate candidate diagnoses (including ICD-10 codes), candidate medical orders (including drugs / tests), and candidate consultation departments.

[0036] The specific method for generating standardized medical orders is as follows: First, a medical order embedding vector library is constructed (using the bge-large-zh-v1.5 vector generation tool from the Zhongyang Vertical Domain Medical Field). Candidate medical orders are retrieved using FAISS search to obtain the Top 5 similar medical orders. Valid candidate medical orders are obtained by filtering out similar medical orders with a similarity of ≥0.85. Then, using the Zhongyang Vertical Medical Domain (14B parameters), based on valid candidate medical orders and patient electronic medical record data (valid candidate medical orders + key features such as patient liver and kidney function / allergy history), a personalized suitability score (0-100 points) is obtained, and highly suitable valid candidate medical orders (those with scores ≥90 points) are selected. Finally, the department's historical medical order templates are retrieved, and the statistical mode of dosage, frequency, and usage is extracted to generate standardized medical orders. A four-level verification mechanism of "large model generation - vector library filtering - small model calibration - historical data verification" is adopted, with an AI illusion rate ≤5%, a reduction of 67.2% compared to a single model system, and a contraindication omission rate ≤0.3%.

[0037] The method for generating standardized diagnoses is the same as that for generating standardized medical orders. However, it's important to note that the vector library construction differs. In the process of generating standardized diagnoses, a vector library of all diagnoses in the hospital is constructed, containing all diagnoses the hospital can perform. After obtaining candidate consultation departments, they are matched against the list of departments that can be invited for consultation within the current department. Only when a match is found are standardized consultation departments obtained (and recommended).

[0038] The final output includes standardized diagnoses, standardized medical orders, and standardized recommendations from standardized consultation departments, including: New diagnostic list (including ICD-10 codes and levels of evidence); Add a list of medical orders (including generic name of the drug / name of the test item, dosage, frequency, route of administration / sampling method); List of discontinued medical orders (including reasons for discontinuation); Consultation recommendations (including recommended consultation departments and consultation priorities); The standardized recommendation results can be directly imported into the hospital's HIS system for doctors to review and confirm.

[0039] Example 2 This embodiment provides a department-specific medical AI-assisted decision-making system, such as... Figure 3 As shown, it includes the following modules: The data acquisition module is configured to acquire patient electronic medical record data; The patient profile generation module is configured to: extract static features, dynamic trend features, and feature correlation based on event triggering mechanism and patient electronic medical record data; and fuse static features, dynamic trend features, and feature correlation to generate patient profile vectors and obtain patient profile keywords. The recommendation module is configured to: construct a vector knowledge base hierarchically by department; perform semantic matching retrieval on patient profile vectors based on the vector knowledge base to obtain retrieval results; generate candidate diagnoses, candidate medical orders, and candidate consultation departments based on patient profile keywords and retrieval results; sequentially screen and calibrate the candidate diagnoses and candidate medical orders to generate standardized diagnoses and standardized medical orders; calibrate the candidate consultation departments to generate standardized consultation departments; and finally output standardized recommendation results containing standardized diagnoses, standardized medical orders, and standardized consultation departments.

[0040] It should be noted that the above modules correspond to the steps in Embodiment 1, and the examples and application scenarios implemented by the above modules and their corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should also be noted that the above modules can be executed in a computer system as part of the system.

[0041] In further embodiments, the following is also provided: An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the method described in Embodiment 1. For brevity, further details are omitted here.

[0042] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0043] A computer-readable storage medium for storing computer instructions that, when executed by a processor, perform the method of Embodiment 1.

[0044] The method in Example 1 can be directly executed by a hardware processor, or it can be executed by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0045] A computer program product includes a computer program that, when executed by a processor, implements the method in Embodiment 1.

[0046] The present invention also provides at least one computer program product tangibly stored on a non-transitory computer-readable storage medium. The computer program product includes computer-executable instructions, such as instructions included in program modules, which execute in a device on a target real or virtual processor to perform the processes / methods described above. Typically, program modules include routines, programs, libraries, objects, classes, components, data structures, etc., that perform specific tasks or implement specific abstract data types. In various embodiments, the functionality of program modules can be combined or divided among program modules as needed. The machine-executable instructions for the program modules can execute within a local or distributed device. In a distributed device, the program modules can reside in both local and remote storage media.

[0047] The computer program code used to implement the methods of the present invention may be written in one or more programming languages. This computer program code may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the computer or other programmable data processing device, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a computer, partially on a computer, as a stand-alone software package, partially on a computer and partially on a remote computer, or entirely on a remote computer or server.

[0048] In the context of this invention, computer program code or related data may be carried by any suitable carrier to enable a device, apparatus, or processor to perform the various processes and operations described above. Examples of carriers include signals, computer-readable media, and the like. Examples of signals may include electrical, optical, radio, sound, or other forms of propagation signals, such as carrier waves, infrared signals, etc.

[0049] Those skilled in the art will recognize that the units and algorithm steps described in conjunction with the embodiments herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0050] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A department-specific medical AI-assisted decision-making method, characterized in that, Includes the following steps: Obtain patient electronic medical record data; Based on the event triggering mechanism and patient electronic medical record data, static features, dynamic trend features, and feature correlation are extracted, and the static features, dynamic trend features, and feature correlation are fused to generate patient profile vectors and obtain patient profile keywords. A vector knowledge base is constructed hierarchically according to departmental dimensions. Based on the vector knowledge base, semantic matching retrieval is performed on patient profile vectors to obtain retrieval results. Based on patient profile keywords and retrieval results, candidate diagnoses, candidate medical orders, and candidate consultation departments are generated. The candidate diagnoses and candidate medical orders are sequentially screened and calibrated to generate standardized diagnoses and standardized medical orders. The candidate consultation departments are calibrated to generate standardized consultation departments. Finally, standardized recommendation results containing standardized diagnoses, standardized medical orders, and standardized consultation departments are output.

2. The department-specific medical AI-assisted decision-making method as described in claim 1, characterized in that, The patient's electronic medical record data includes basic patient information, dynamic diagnosis and treatment information, and department configuration information. The basic patient information includes name, age, gender, past medical history, and surgical history. The dynamic diagnosis and treatment information includes the latest diagnosis, symptom records, laboratory test results, executed medical orders, and execution feedback. The department configuration information includes a list of commonly used drugs in the target department, a list of commonly used laboratory test items, and a clinical pathway template.

3. The department-specific medical AI-assisted decision-making method as described in claim 1, characterized in that, Based on the patient's electronic medical record data, static features were extracted using the BERT-BiLSTM-CRF model, including standardized mappings of diagnosis name and surgical name; dynamic trend features were extracted using a temporal convolutional network, including slope and variance of test indicators. Calculate feature correlation using attention mechanism.

4. The department-specific medical AI-assisted decision-making method as described in claim 1, characterized in that, The vector knowledge base, constructed hierarchically by department, includes a core layer, an evidence-based layer, a rule layer, and an update unit. The core layer includes instructions for commonly used drugs in the department, operating procedures for laboratory tests, and in-hospital clinical pathways. The evidence-based layer includes core journal literature from UpToDate clinical consultants, the Chinese Clinical Guidelines Database, and PubMed. The rule layer includes departmental expert consensus, rules on contraindications for combined drug use, and rules for dose adjustment. The update unit updates guidelines / literature through web crawling.

5. The department-specific medical AI-assisted decision-making method as described in claim 1, characterized in that, The event triggering mechanism includes setting a minimum update interval. If a key indicator changes drastically during this period, an update is triggered. The key indicator changes include doubling of creatinine and a sudden rise in blood sugar.

6. The department-specific medical AI-assisted decision-making method as described in claim 1, characterized in that, The specific method for generating standardized medical orders is as follows: First, a medical order embedding vector library is constructed. Candidate medical orders are retrieved through FAISS search to obtain similar medical orders. Valid candidate medical orders are then selected through screening. Then, based on the valid candidate medical orders and the patient's electronic medical record data, a personalized fit score is obtained, and highly fit valid candidate medical orders are selected. Finally, the department's historical medical order templates were retrieved, and the statistical mode of dosage, frequency, and usage was extracted to generate standardized medical orders.

7. A department-specific medical AI-assisted decision-making system, characterized in that, include: The data acquisition module is configured to acquire patient electronic medical record data; The patient profile generation module is configured to: extract static features, dynamic trend features, and feature correlation based on event triggering mechanism and patient electronic medical record data; and fuse static features, dynamic trend features, and feature correlation to generate patient profile vectors and obtain patient profile keywords. The recommendation module is configured to: construct a vector knowledge base hierarchically by department; perform semantic matching retrieval on patient profile vectors based on the vector knowledge base to obtain retrieval results; generate candidate diagnoses, candidate medical orders, and candidate consultation departments based on patient profile keywords and retrieval results; sequentially screen and calibrate the candidate diagnoses and candidate medical orders to generate standardized diagnoses and standardized medical orders; calibrate the candidate consultation departments to generate standardized consultation departments; and finally output standardized recommendation results containing standardized diagnoses, standardized medical orders, and standardized consultation departments.

8. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the department-specific medical AI-assisted decision-making method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, Used to store computer instructions, which, when executed by a processor, complete the department-specific medical AI-assisted decision-making method as described in any one of claims 1-6.

10. A computer program product, characterized in that, The invention includes a computer program that, when executed by a processor, implements the department-specific medical AI-assisted decision-making method according to any one of claims 1-6.

Images