Medical aid decision-making information processing method and device, and storage medium

Abstract

The invention provides a medical aid decision information processing method and device and a storage medium, and relates to the technical field of computers. The method comprises the steps of generating a question statement for a target medical application scene according to content information of the target medical application scene in response to a calling operation of the target medical application scene for a clinical medical decision application; according to the question statement, performing matching by adopting a preset clinical medical knowledge base to obtain target knowledge content matched with the question statement; filling a cue word template of the target medical application scene with the target knowledge content and the question statement to generate a target cue word of the target medical application scene; according to the target prompt word, calling a preset large language model to obtain reply prompt content of the target medical application scene; and displaying the reply prompt content of the target medical application scene in a display interface of the clinical medical decision application. The reply prompt content is obtained according to the clinical medical knowledge base, and the answer accuracy of the large language model in the medical field is guaranteed.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and more specifically, to a method, device, and storage medium for processing medical decision-making assistance information. Background Art

[0002] The Large Language Model (LLM) is a deep learning model trained on massive amounts of text data. It not only generates natural language text but also deeply understands the meaning of the text and can handle various natural language tasks such as text summarization, question answering, and translation.

[0003] However, current large language models have exposed some problems. For example, existing large language models can only provide general answers. This approach is effective for answering general questions in common fields, but it cannot guarantee the accuracy of answers in highly professional fields, especially the medical field. Summary of the Invention

[0004] In response to the above-mentioned deficiencies in the prior art, the present application provides a method, device and storage medium for processing medical decision-making assistance information to solve the problems existing in the prior art.

[0005] The technical solutions adopted in the embodiments of this application are as follows:

[0006] In a first aspect, an embodiment of the present application provides a method for processing medical decision-making assistance information, the method comprising:

[0007] In response to a call operation of a target medical application scenario for a clinical medical decision application, generating a question statement for the target medical application scenario based on content information of the target medical application scenario;

[0008] According to the question statement, a preset clinical medical knowledge base is used for matching to obtain target knowledge content that matches the question statement;

[0009] Filling the target knowledge content and the question sentence into the prompt word template of the target medical application scenario to generate the target prompt word of the target medical application scenario;

[0010] According to the target prompt word, calling a preset large language model to obtain the response prompt content of the target medical application scenario;

[0011] The reply prompt content of the target medical application scenario is displayed in the display interface of the clinical medical decision-making application.

[0012] In one embodiment, before filling the target knowledge content and the question statement into the prompt word template of the target medical application scenario and generating the target prompt word of the target medical application scenario, the method further includes:

[0013] According to the identifier of the target medical application scenario, the application tool of the clinical medical decision application is called to obtain the prompt word template of the target medical application scenario from the prompt word templates of multiple medical application scenarios.

[0014] In one embodiment, the step of matching the question statement with a preset clinical medical knowledge base to obtain target knowledge content that matches the question statement includes:

[0015] Performing vector encoding on the question sentence to obtain a question sentence vector of the question sentence;

[0016] Performing similarity matching on the question vector and knowledge vectors of a plurality of preset knowledge contents in the preset clinical medical knowledge base to obtain a matching result;

[0017] According to the matching result, a plurality of candidate knowledge contents whose similarities meet preset conditions are determined from the plurality of preset knowledge contents as the target knowledge contents.

[0018] In one embodiment, before performing similarity matching between the question vector and knowledge vectors of a plurality of preset knowledge contents in the preset clinical medical knowledge base and obtaining a matching result, the method further includes:

[0019] Processing knowledge documents in a preset medical field to obtain a plurality of knowledge vectors of first knowledge contents;

[0020] The knowledge vectors of the plurality of first knowledge contents are stored in the preset clinical medical knowledge base.

[0021] In one embodiment, the processing of the knowledge documents in the preset medical field to obtain a plurality of knowledge vectors of the first knowledge content includes:

[0022] Classifying the knowledge documents in the preset medical field into medical topics to obtain multiple medical topics;

[0023] Splitting the knowledge documents under each medical topic to obtain medical documents under each medical topic;

[0024] Vector encoding is performed on the medical documents under the medical topics to obtain knowledge vectors of the multiple first knowledge contents.

[0025] In one embodiment, the method further comprises:

[0026] Processing the knowledge document uploaded by the user to obtain a plurality of knowledge vectors of the second knowledge content;

[0027] The knowledge vectors of the plurality of second knowledge contents are stored in the preset clinical medical knowledge base.

[0028] In one embodiment, the processing of the knowledge document uploaded by the user to obtain a plurality of knowledge vectors of the second knowledge content includes:

[0029] Classifying the knowledge document into medical topics to obtain multiple medical topics;

[0030] Splitting the knowledge documents under each medical topic to obtain the knowledge documents under each medical topic;

[0031] Vector encoding is performed on the knowledge documents under the medical topics to obtain knowledge vectors of the multiple second knowledge contents.

[0032] In one embodiment, the target medical application scenario is a patient risk warning application scenario, a diagnosis preservation application scenario, a medical order issuance application scenario, a medical knowledge retrieval application scenario, or a medical technical report interpretation application scenario.

[0033] In the second aspect, an embodiment of the present application provides a computer device, comprising: a processor, a storage medium and a bus, wherein the storage medium stores program instructions executable by the processor. When the computer device is running, the processor and the storage medium communicate through the bus, and the processor executes the program instructions to implement the method for processing medical auxiliary decision-making information described in any of the above embodiments.

[0034] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for processing medical auxiliary decision-making information described in any of the above embodiments is implemented.

[0035] The beneficial effects of the present application are as follows: the present application provides a method, device, and storage medium for processing medical decision-making assistance information. The method includes responding to a call operation of a target medical application scenario for a clinical medical decision application, generating a question statement for the target medical application scenario based on content information of the target medical application scenario; matching the question statement with a preset clinical medical knowledge base to obtain target knowledge content that matches the question statement; filling the target knowledge content and the question statement into a prompt word template for the target medical application scenario to generate a target prompt word for the target medical application scenario; calling a preset large language model based on the target prompt word to obtain reply prompt content for the target medical application scenario; and displaying the reply prompt content for the target medical application scenario in a display interface of the clinical medical decision application.

[0036] Among them, the question statements for the target medical application scenarios are matched using a preset clinical medical knowledge base. Since the clinical medical knowledge base stores medical-related knowledge, this application obtains reply prompts based on the clinical medical knowledge base, ensuring the accuracy of the large language model's answers in the medical field. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.

[0038] Figure 1 This is a flowchart of a method for processing medical decision-making assistance information provided in one embodiment of the present application;

[0039] Figure 2 A schematic diagram of a display page of a clinical medical decision-making application provided in one embodiment of the present application;

[0040] Figure 3 This is a second flow chart of a method for processing medical decision-making assistance information provided in an embodiment of the present application;

[0041] Figure 4 This is a flowchart of a method for processing medical decision-making assistance information according to an embodiment of the present application;

[0042] Figure 5 This is a fourth flow chart of a method for processing medical decision-making assistance information provided in one embodiment of the present application;

[0043] Figure 6 This is a fifth flow chart of a method for processing medical decision-making assistance information provided in one embodiment of the present application;

[0044] Figure 7 Flowchart 6 of the method for processing medical decision-making assistance information provided in one embodiment of the present application;

[0045] Figure 8 A schematic diagram of the structure of a device for processing medical decision-making assistance information provided in one embodiment of the present application;

[0046] Figure 9 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0048] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for protection, but merely represents selected embodiments of the present application. All other embodiments obtained by persons of ordinary skill in the art based on the embodiments in the present application without creative work are within the scope of protection of the present application.

[0049] In addition, the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0050] It should be noted that, in the absence of conflict, the features in the embodiments of this application can be combined with each other.

[0051] ChatGPT, an intelligent conversational system based on a pre-trained large language model launched by OpenAI, an American artificial intelligence company, has sparked a wave of online interest and quickly become one of the fastest-growing applications in history. However, existing large language models have exposed several issues, including knowledge timeliness, difficulty in updating knowledge, uncontrolled answer content, inability to privately deploy, and data security. These issues have prevented them from meeting the needs of large and medium-sized enterprises and users in knowledge-intensive industries. Furthermore, existing large language models can only provide general answers, which are effective for general questions in common domains. However, they cannot guarantee accuracy in highly specialized fields, particularly medicine. For example, they may force answers to unfamiliar questions, inevitably resulting in obvious errors or inconsistencies in the knowledge base, commonly known as large language model "hallucinations." This is essentially due to the large language models' lack of integration with medical domain knowledge.

[0052] Therefore, an embodiment of the present application provides a method for processing medical decision-making assistance information, which can solve the above problems. The method can be generated by any computer device with computing and processing capabilities. The computer device can be, for example, a terminal-oriented computer device or a back-end server.

[0053] The following is a specific example description of the processing method of medical auxiliary decision-making information provided by this application through multiple examples in combination with the accompanying drawings.

[0054] Figure 1 This is a flow chart of a method for processing medical decision-making assistance information provided in one embodiment of the present application, such as Figure 1 As shown, the method includes:

[0055] S101. Respond to a call operation of a target medical application scenario for a clinical medical decision-making application, and generate a question statement for the target medical application scenario based on content information of the target medical application scenario.

[0056] Clinical medical decision-making applications refer to clinical decision support systems (CDSS). The target medical application scenarios can be patient risk warning application scenarios, diagnosis preservation application scenarios, medical order issuance application scenarios, medical knowledge retrieval application scenarios, or medical technical report interpretation application scenarios.

[0057] Optionally, the target medical application scenarios may also include: smart diagnosis and treatment recommendation and treatment plan recommendation scenarios, patient health education scenarios, medical record content quality control scenarios, medical record generation scenarios, report generation scenarios, knowledge question and answer scenarios, and other knowledge content generation scenarios related to medical scenarios.

[0058] For example, in the diagnosis preservation application scenario and the medical order issuance application scenario, doctors can make clinical decisions; in the smart diagnosis and treatment recommendation and treatment plan recommendation scenarios, medical record generation scenarios, and medical record content quality control scenarios, doctors can write medical record documents; in the report generation scenario and the medical technical report interpretation application scenario, doctors can review examination and test reports; in the patient risk warning application scenario, the patient health education scenario, the medical knowledge review application scenario, and the knowledge question and answer scenario, doctors can view patient medical records.

[0059] It should also be noted that when different medical personnel (such as doctors and nurses) use the clinical medical decision-making application, the prompt word templates generated in the same scenario are also different.

[0060] Figure 2 This is a schematic diagram of a display page of a clinical medical decision-making application provided in one embodiment of the present application, for example Figure 2As shown, when the target medical application scenario is a patient risk warning application scenario, the user clicks on the preset control (such as the CDSS Copilot robot control) on the clinical medical decision-making application, which is deemed to have received a call operation for the clinical medical decision-making application from the target medical application scenario. Then, based on the content information of the target medical application scenario, a question statement for the target medical application scenario is generated. The users of this application refer to medical personnel. Medical personnel can call out the clinical medical decision-making application at any time in business systems such as doctor stations or computer clients, and trigger various medical application scenarios to generate question statements for the target medical application scenario.

[0061] The question statement is related to the target medical application scenario. For example, if the target medical application scenario is a patient risk warning application scenario, the question statement may be "You are a medical expert. Based on ${abnormal indicator data} and ${warning template}, generate a risk warning message for the doctor." If the target medical application scenario is a medical order issuance application scenario, the question statement may be "What is the clinical significance of ${instructional test item}? What are the precautions?" In actual operation, different question statements can be set according to actual needs, and this embodiment does not limit this.

[0062] S102: According to the question statement, a preset clinical medical knowledge base is used for matching to obtain target knowledge content that matches the question statement.

[0063] The preset clinical medical knowledge base stores knowledge content in the medical field. After obtaining the question statement, the preset clinical medical knowledge base can be used to perform matching based on the question statement to obtain target knowledge content that matches the question statement.

[0064] Specifically, the knowledge content is stored in the preset clinical medical knowledge base in the form of knowledge vectors. After obtaining the question statement, the question statement can be converted into a vector form, and then the vector of the question statement is matched with the knowledge vector stored in the preset clinical medical knowledge base for semantic similarity. Multiple candidate knowledge fragments whose semantic similarity with the vector of the question statement is greater than or equal to a preset threshold are obtained from the preset clinical medical knowledge base as the target knowledge content matching the question statement.

[0065] For example, the number of candidate knowledge fragments can be set to 4, the candidate knowledge fragments are sorted according to semantic similarity, and the 4 candidate knowledge fragments with the highest similarity are selected as the target knowledge content matching the question statement.

[0066] S103: Fill the target knowledge content and question sentence into the prompt word template of the target medical application scenario to generate the target prompt word of the target medical application scenario.

[0067] The prompt word template for each medical application scenario can be a pre-set template format. After obtaining the target knowledge content that matches the question statement, the target knowledge content and the question statement are filled into the prompt word template for the target medical application scenario to generate the target prompt word for the target medical application scenario.

[0068] For example, in the application scenario of prescribing medical orders, the prompt word template is in the form of: "Clinical knowledge: ${knowledge content}; Please answer the following questions based on the above clinical knowledge. The question is: What is the clinical significance of ${test item}? What are the precautions?"

[0069] In a medical order prescribing scenario, a clinician orders a test item, such as "serum calcitonin measurement." Based on the test item content, a question is generated: "What is the clinical significance of serum calcitonin measurement? What are the precautions?" This question is then vectorized and matched against a clinical knowledge vector database for semantic similarity. The topK target knowledge items with a high degree of match exceeding a set similarity threshold are obtained. For example, if topK = 4, the current question yields four target knowledge items: ["Precautions: Generally, fasting venous blood is collected in the morning or mid-morning," "Clinical Significance: 1. Increased levels are seen in patients with medullary thyroid cancer and renal failure. 2. Decreased levels are seen in patients who have undergone total thyroidectomy," "Reference Range: Serum calcitonin: <8.8 ng / L for men; <5.8 ng / L for women," and "Testing Methods: Serum calcitonin is typically tested using radioimmunoassay, while dynamic serum calcitonin testing typically uses semi-quantitative solid-phase immunochromatography and immunochemiluminescence, the latter of which offers higher accuracy and sensitivity."].

[0070] Based on the matched target knowledge content and question statements, the prompt word template was reassembled to obtain the following new template: "Clinical Knowledge: ${Precautions\nUsually, fasting venous blood is collected in the morning or mid-afternoon.\n\nClinical Significance\n1. Increased levels may be seen in patients with medullary thyroid cancer and renal failure.\n2. Decreased levels may be seen in patients who have undergone total thyroidectomy.\n\nReference Range\nSerum calcitonin: Males: <8.8 ng / L; Females: <5.8 ng / L.\n\nTesting Methods\nSerum calcitonin is typically tested using radioimmunoassay, while dynamic serum calcitonin is typically tested using semiquantitative solid-phase immunochromatography and immunochemiluminescence, the latter of which has higher accuracy and sensitivity.} Please answer the following questions based on the above clinical knowledge. The question is: What is the clinical significance of ${serum calcitonin determination}? What are the precautions?"

[0071] S104: Based on the target prompt word, call the preset large language model to obtain the response prompt content of the target medical application scenario.

[0072] After receiving the target prompt for the target medical application scenario, the preset large language model is called. Based on the target prompt, the preset large language model generates a response prompt for the target medical application scenario: "The clinical significance of serum calcitonin measurement includes: 1. Elevated levels are seen in patients with medullary thyroid cancer and renal failure. 2. Decreased levels are seen in patients who have undergone total thyroidectomy. Note: Venous blood is generally collected in the morning or mid-morning on an empty stomach."

[0073] S105. Displaying reply prompt content of the target medical application scenario in the display interface of the clinical medical decision-making application.

[0074] After obtaining the reply prompt content of the target medical application scenario, the reply prompt content of the target medical application scenario can be displayed in the display interface of the clinical medical decision-making application.

[0075] In summary, this embodiment uses a preset clinical medical knowledge base to match question statements for the target medical application scenario. Since the clinical medical knowledge base stores medical-related knowledge, this application obtains reply prompt content based on the clinical medical knowledge base, ensuring the accuracy and reliability of the large language model's answers in the medical field.

[0076] In one embodiment, the prompt word template for each medical application scenario can be customized. For example, the prompt word template can be designed according to the structure of "role + task + requirement + prompt" according to each medical application scenario, and the input and output parameters and output length of the prompt word template can be constrained. Then, the set template is stored according to the identification of its corresponding target medical application scenario for future use.

[0077] Then, in S103, before filling the target knowledge content and question statement into the prompt word template of the target medical application scenario and generating the target prompt word of the target medical application scenario, the method of the present application may also include: calling the application tool (CDSS knowledge assistant) of the clinical medical decision-making application according to the identification of the target medical application scenario, and obtaining the prompt word template of the target medical application scenario from the prompt word templates of multiple medical application scenarios.

[0078] Figure 3 This is a second flow chart of a method for processing medical decision-making assistance information provided in an embodiment of the present application, as shown in FIG. Figure 3 As shown, in S102, according to the question statement, a preset clinical medical knowledge base is used for matching to obtain target knowledge content that matches the question statement, which may include:

[0079] S201. Perform vector encoding on the question sentence to obtain a question vector of the question sentence.

[0080] After obtaining the question sentence, the text of the question sentence is vector-encoded using the Embedding model to obtain the question sentence vector of the question sentence.

[0081] S202: Perform similarity matching on the question vector and knowledge vectors of multiple preset knowledge contents in a preset clinical medical knowledge base to obtain a matching result.

[0082] Then, similarity matching is performed on the question vector and the knowledge vectors of multiple preset knowledge contents in the preset clinical medical knowledge base to obtain a matching result, and a semantic similarity ranking between the knowledge vectors of the multiple preset knowledge contents and the question vector is obtained.

[0083] S203 : According to the matching results, determine, from the plurality of preset knowledge contents, a plurality of candidate knowledge contents whose similarities meet preset conditions as target knowledge contents.

[0084] Finally, based on the matching results, four candidate knowledge contents whose similarities meet preset conditions are determined from multiple preset knowledge contents as target knowledge contents.

[0085] The data source of the preset clinical medical knowledge base can be the knowledge literature in the preset medical field. Figure 4 This is a flowchart of a method for processing medical auxiliary decision-making information provided in an embodiment of the present application, as shown in FIG. Figure 4 As shown, before performing similarity matching between the question vector and the knowledge vectors of multiple preset knowledge contents in the preset clinical medical knowledge base in S202 and obtaining a matching result, the method of the present application may further include:

[0086] S301: Process knowledge documents in a preset medical field to obtain a plurality of knowledge vectors of first knowledge content.

[0087] This embodiment can automatically acquire knowledge documents in a preset medical field, and process the acquired knowledge documents in the preset medical field using an Embedding model to obtain a plurality of knowledge vectors of the first knowledge content.

[0088] Among them, the embedding model is a technology that can convert words, phrases, or entire sentences into vector form. These vectors can capture rich semantic meaning, allowing computers to process text like numbers. In the era of large language models, due to the limited amount of input text that large models can handle, in this case, text must be divided into smaller blocks or "segments" before the large language model can process them individually. However, this segmentation may lead to problems with the contextual coherence and overall coherence of the output, thereby reducing the quality of the generated text.

[0089] The Embedding model represents words and phrases as high-dimensional vectors, which allows the language model to encode the contextual information of the input text in a compact and efficient manner. The model can then use this contextual information to generate more coherent and contextually appropriate output text, even if the input text is divided into multiple segments. This embodiment uses an open source Embedding model (such as BAAI / bge-large-zh) to encode a collection of knowledge documents in a pre-determined medical field, forming a vector set of medical documents as the knowledge vector of the first knowledge content.

[0090] S302: Store the knowledge vectors of the plurality of first knowledge contents into a preset clinical medical knowledge base.

[0091] Then, the knowledge vectors of the plurality of first knowledge contents are vectorized and stored in a vector database by way of subject classification and document index construction for easy query.

[0092] Figure 5 This is a fourth flow chart of a method for processing medical decision-making assistance information provided in an embodiment of the present application, as shown in FIG. Figure 5 As shown, the process of processing the knowledge documents in the preset medical field in S301 to obtain a plurality of knowledge vectors of the first knowledge content may include:

[0093] S401. Classify the knowledge documents in the preset medical field into medical topics to obtain multiple medical topics.

[0094] The pre-defined medical knowledge documents in this example are primarily sourced from medical clinical pathways, expert guidelines and consensus, medical specialty books, medical clinical operation manuals, nursing operation manuals, and surgical operation manuals. They cover 10,575 disease knowledge items, 729 examination knowledge items, 2,299 laboratory knowledge items, 1,557 surgical procedures, 229 nursing procedures, 1,444 clinical pathways, and 577 guideline consensus items. The acquired data undergoes refined deduplication and cleaning preprocessing, and these documents are categorized by multiple medical themes, including disease, examination, testing, blood use, medication, surgical procedures, and nursing procedures.

[0095] S402: Split the knowledge documents under each medical topic into their contents to obtain medical documents under each medical topic.

[0096] According to the divided medical topics, knowledge documents are loaded separately, the text content of the documents is read, and the text is split according to characters, length and semantics to form a themed medical document collection.

[0097] Specifically, knowledge documents include semi-structured data documents (csv, xlsx, json) and unstructured data documents (txt, pdf, docx, doc). Combined with the text splitting method optimized based on Chinese punctuation, paragraph semantics and maximum length paragraph combination, the knowledge documents can be disassembled by topic. After disassembly, each knowledge document is divided into multiple knowledge fragments, which is conducive to the large language model to screen knowledge fragments.

[0098] S403 : Perform vector encoding on the medical documents under each medical topic to obtain a plurality of knowledge vectors of the first knowledge content.

[0099] Finally, the Embedding model is used to vectorize the medical documents under each medical topic and store them in the medical knowledge vector library. Specifically, the open source Embedding model (such as BAAI / bge-large-zh) is used to encode the disassembled medical documents to form a medical document vector set, and the medical document vector set is vectorized and stored by subject classification and document index construction through the Faiss vector database to obtain multiple knowledge vectors of the first knowledge content.

[0100] The data source of the preset clinical medical knowledge base can also be knowledge documents manually uploaded by users. Figure 6 This is a flowchart of a method for processing medical decision-making assistance information provided in an embodiment of the present application, as shown in FIG. Figure 6 As shown, the method of the present application may further include:

[0101] S501: Process the knowledge document uploaded by the user to obtain a plurality of knowledge vectors of the second knowledge content.

[0102] Users can upload knowledge documents through the knowledge document upload platform, and the Embedding model can process the knowledge documents uploaded by users to obtain knowledge vectors of multiple second knowledge contents.

[0103] S502: Store the knowledge vectors of the plurality of second knowledge contents into a preset clinical medical knowledge base.

[0104] Then, the knowledge vectors of the plurality of second knowledge contents are vectorized and stored in a vector database by classifying them by subject and constructing a document index for easy query.

[0105] Figure 7 This is a flowchart of a method for processing medical decision-making assistance information provided in an embodiment of the present application, as shown in FIG. Figure 7 As shown, the processing of the knowledge document uploaded by the user in S501 to obtain multiple knowledge vectors of the second knowledge content may include:

[0106] S601: Classify the knowledge documents into medical topics to obtain multiple medical topics.

[0107] After obtaining the knowledge documents uploaded by users through the knowledge document upload platform, these knowledge documents are subjected to fine-grained deduplication and cleaning pre-processing, and these knowledge documents are classified according to multiple medical topics such as diseases, examinations, tests, blood use, medication, surgical operations, nursing operations, etc.

[0108] S602: Split the content of the knowledge documents under each medical topic to obtain the knowledge documents under each medical topic.

[0109] According to the divided medical topics, knowledge documents are loaded separately, the text content of the knowledge documents is read, and the text of the knowledge documents is split according to characters, length and semantics to form a collection of knowledge documents with topics.

[0110] Specifically, knowledge documents include semi-structured data documents (csv, xlsx, json) and unstructured data documents (txt, pdf, docx, doc). Combined with the text splitting method optimized based on Chinese punctuation, paragraph semantics and maximum length paragraph combination, the knowledge documents uploaded by users can be disassembled by topic. After disassembly, each knowledge document is divided into multiple knowledge fragments, which is conducive to the large language model to screen knowledge fragments.

[0111] S603: Perform vector encoding on the knowledge documents under each medical topic to obtain a plurality of knowledge vectors of the second knowledge content.

[0112] Finally, the Embedding model is used to vector encode the knowledge documents under each medical topic to obtain multiple knowledge vectors of the second knowledge content and store them in the medical knowledge vector library.

[0113] In summary, this application provides a method for processing medical decision-making assistance information, which has the following advantages:

[0114] 1. Improves the reliability and interpretability of large language models in intelligent dialogue in the medical field, while enhancing the application value of clinical medical decision-making applications (CDSS) in clinical decision-making business.

[0115] 2. In medical scenarios, we've specialized the content of large model answers in two ways. First, we use structured filtering and vector matching to locate relevant knowledge, improving the professionalism of large model Q&A. Second, by clearly locating the source of knowledge, we reduce the requirements for logical reasoning in the large model, improve the reliability of knowledge Q&A, and ensure that the knowledge content can be traced.

[0116] 3. The preset clinical medical knowledge base can be updated by automatically acquiring knowledge documents in the preset medical field and knowledge documents uploaded by users. The knowledge vectors stored in the preset clinical medical knowledge base can be updated in real time, ensuring the timeliness of the knowledge vectors.

[0117] The following continues to explain the devices, equipment and storage media for executing the method for processing medical auxiliary decision-making information provided by any of the above embodiments of the present application. The specific implementation process and the technical effects produced are the same as those of the corresponding method embodiments mentioned above. For the sake of brief description, for the parts not mentioned in the following embodiments, please refer to the corresponding content in the method embodiments.

[0118] Optionally, the present application also provides a device for processing medical decision-making assistance information. Figure 8 A schematic diagram of the structure of a device for processing medical decision-making assistance information provided in one embodiment of the present application is shown in FIG. Figure 8 As shown, the device includes:

[0119] The generating module 10 is used to respond to the calling operation of the target medical application scenario for the clinical medical decision application, and generate a question statement for the target medical application scenario according to the content information of the target medical application scenario.

[0120] The matching module 20 is used to perform matching based on the question statement using a preset clinical medical knowledge base to obtain target knowledge content that matches the question statement.

[0121] The filling module 30 is used to fill the target knowledge content and question sentences into the prompt word template of the target medical application scenario to generate the target prompt word of the target medical application scenario.

[0122] The acquisition module 40 is used to call a preset large language model according to the target prompt word to obtain the reply prompt content of the target medical application scenario.

[0123] The display module 50 is used to display the reply prompt content of the target medical application scenario in the display interface of the clinical medical decision-making application.

[0124] Optionally, the acquisition module 40 is further configured to call an application tool of a clinical medical decision application according to the identifier of the target medical application scenario, and acquire a prompt word template of the target medical application scenario from prompt word templates of multiple medical application scenarios.

[0125] Optionally, the matching module 20 is also used to perform vector encoding on the question statement to obtain a question vector of the question statement; perform similarity matching on the question vector and the knowledge vectors of multiple preset knowledge contents in the preset clinical medical knowledge base to obtain a matching result; and based on the matching result, determine multiple candidate knowledge contents whose similarities meet preset conditions from the multiple preset knowledge contents as target knowledge contents.

[0126] Optionally, the acquisition module 40 is further configured to process knowledge documents in a preset medical field to obtain a plurality of knowledge vectors of the first knowledge content; and store the plurality of knowledge vectors of the first knowledge content in a preset clinical medical knowledge base.

[0127] Optionally, the acquisition module 40 is also used to classify the knowledge documents in the preset medical field into medical topics to obtain multiple medical topics; to split the content of the knowledge documents under each medical topic to obtain medical documents under each medical topic; and to vector encode the medical documents under each medical topic to obtain multiple knowledge vectors of the first knowledge content.

[0128] Optionally, the acquisition module 40 is further configured to process the knowledge document uploaded by the user to obtain a plurality of knowledge vectors of the second knowledge content; and store the plurality of knowledge vectors of the second knowledge content in a preset clinical medical knowledge base.

[0129] Optionally, the acquisition module 40 is also used to classify the knowledge documents by medical topics to obtain multiple medical topics; perform content splitting on the knowledge documents under each medical topic to obtain knowledge documents under each medical topic; and perform vector encoding on the knowledge documents under each medical topic to obtain multiple knowledge vectors of the second knowledge content.

[0130] The above-mentioned device is used to execute the method provided in the above-mentioned embodiment. Its implementation principle and technical effect are similar and will not be repeated here.

[0131] The above modules can be one or more integrated circuits configured to implement the above methods, such as one or more application-specific integrated circuits (ASICs), one or more microprocessors, or one or more field programmable gate arrays (FPGAs). For another example, when a module is implemented by scheduling program code through a processing element, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0132] Figure 9 A schematic diagram of the structure of a computer device provided in one embodiment of the present application is shown in FIG. Figure 9As shown, an embodiment of the present application also provides a computer device, including: a processor 100, a storage medium 200 and a bus 300, wherein the storage medium stores program instructions executable by the processor. When the computer device is running, the processor and the storage medium communicate through the bus, and the processor executes the program instructions to implement the method for processing medical auxiliary decision information described in any of the above embodiments.

[0133] An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the method for processing medical auxiliary decision-making information described in any of the above embodiments is implemented.

[0134] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0135] The units described as separate components may or may not be physically separate, and 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 these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0136] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0137] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor (English: processor) to perform some steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (English: Read-Only Memory, abbreviated: ROM), a random access memory (English: Random Access Memory, abbreviated: RAM), a disk or an optical disk, and other media that can store program code.

[0138] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A method for processing medical decision-making assistance information, characterized in that: The method comprises: In response to a call operation of a target medical application scenario for a clinical medical decision application, generating a question statement for the target medical application scenario based on content information of the target medical application scenario; According to the question statement, a preset clinical medical knowledge base is used for matching to obtain target knowledge content that matches the question statement; Filling the target knowledge content and the question sentence into the prompt word template of the target medical application scenario to generate the target prompt word of the target medical application scenario; According to the target prompt word, calling a preset large language model to obtain the response prompt content of the target medical application scenario; The reply prompt content of the target medical application scenario is displayed in the display interface of the clinical medical decision-making application.

2. The method according to claim 1, characterized in that Before filling the target knowledge content and the question statement into the prompt word template of the target medical application scenario to generate the target prompt word of the target medical application scenario, the method further includes: According to the identifier of the target medical application scenario, the application tool of the clinical medical decision application is called to obtain the prompt word template of the target medical application scenario from the prompt word templates of multiple medical application scenarios.

3. The method according to claim 1, characterized in that According to the question statement, matching is performed using a preset clinical medical knowledge base to obtain target knowledge content that matches the question statement, including: Performing vector encoding on the question sentence to obtain a question sentence vector of the question sentence; Performing similarity matching on the question vector and knowledge vectors of a plurality of preset knowledge contents in the preset clinical medical knowledge base to obtain a matching result; According to the matching result, a plurality of candidate knowledge contents whose similarities meet preset conditions are determined from the plurality of preset knowledge contents as the target knowledge contents.

4. The method according to claim 3, characterized in that Before performing similarity matching between the question vector and knowledge vectors of a plurality of preset knowledge contents in the preset clinical medical knowledge base to obtain a matching result, the method further includes: Processing knowledge documents in a preset medical field to obtain a plurality of knowledge vectors of first knowledge contents; The knowledge vectors of the plurality of first knowledge contents are stored in the preset clinical medical knowledge base.

5. The method according to claim 4, characterized in that The processing of the knowledge documents in the preset medical field to obtain a plurality of knowledge vectors of the first knowledge content includes: Classifying the knowledge documents in the preset medical field into medical topics to obtain multiple medical topics; Splitting the knowledge documents under each medical topic to obtain medical documents under each medical topic; Vector encoding is performed on the medical documents under the medical topics to obtain knowledge vectors of the multiple first knowledge contents.

6. The method according to claim 4, characterized in that The method further comprises: Processing the knowledge document uploaded by the user to obtain a plurality of knowledge vectors of the second knowledge content; The knowledge vectors of the plurality of second knowledge contents are stored in the preset clinical medical knowledge base.

7. The method according to claim 6, characterized in that The processing of the knowledge document uploaded by the user to obtain a plurality of knowledge vectors of the second knowledge content includes: Classifying the knowledge document into medical topics to obtain multiple medical topics; Splitting the knowledge documents under each medical topic to obtain the knowledge documents under each medical topic; Vector encoding is performed on the knowledge documents under the medical topics to obtain knowledge vectors of the multiple second knowledge contents.

8. The method according to claim 1, characterized in that The target medical application scenarios include patient risk warning application scenarios, diagnosis preservation application scenarios, medical order issuance application scenarios, medical knowledge retrieval application scenarios, or medical technical report interpretation application scenarios.

9. A computer device, characterized in that: include: A processor, a storage medium and a bus, wherein the storage medium stores program instructions executable by the processor. When the computer device is running, the processor and the storage medium communicate through the bus, and the processor executes the program instructions to implement the method for processing medical auxiliary decision information as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, which, when executed by a processor, implements the method for processing medical auxiliary decision-making information according to any one of claims 1 to 8.

Images