Case analysis method and device, equipment, storage medium and program product

Abstract

The invention provides a case analysis method and device, equipment, a storage medium and a program product, and the method comprises the steps: obtaining to-be-analyzed case information and medical model information, carrying out the case keyword extraction and feature vector extraction of the to-be-analyzed case information, and obtaining a case keyword set and an initial feature vector; determining a screening analysis knowledge base cluster according to the medical model information and the case keyword set, and inputting the initial feature vector and the screening analysis knowledge base cluster into a preset analysis model for case information screening analysis to determine a case analysis result; according to the method, the knowledge base subset suitable for the current case is screened by combining the medical model information and the case keyword set, the analysis precision and the efficiency of the analysis process are improved, higher-level semantic understanding and logical reasoning are carried out through the feature vectors, the accuracy and reliability of case analysis are remarkably improved, and the user experience is improved. And the quality and the credibility of the answers are improved, so that the user experience is improved, and the practicability and the reliability are enhanced.

Description

Technical Field

[0001] The present application relates to the field of medical analysis, and in particular to a case analysis method, device, equipment, storage medium and program product. Background Art

[0002] In medical education, case analysis is essential for cultivating clinical thinking and practical ability. Through real or simulated case studies, theoretical knowledge can be applied to practical situations, which can not only enhance the understanding of diseases, but also lay a solid foundation for future clinical work. However, the traditional case analysis teaching method often has limited case resources, which directly affects the opportunity to contact various diseases. Secondly, the analysis process lacks interactivity due to the one-way explanation method used in traditional classrooms. Finally, the traditional method relies on experience and judgment, lacks systematic data analysis support based on a reasonable computing system, and is difficult to achieve accurate evaluation and feedback. With the development of artificial intelligence technology, generative question-answering technology has brought new opportunities for medical case analysis. It can automatically generate detailed answers based on the input questions, which can not only provide rich case resources, but also provide a more diverse, interactive and personalized learning experience.

[0003] However, in actual applications, the question-answering generation model sometimes generates case analysis results that do not match the case information due to the description content. For example, it fails to correctly understand the key symptoms, medical history or laboratory test results in the case and provides an incorrect diagnostic suggestion or treatment plan. In addition, the generation of answer information may be inaccurate, which may affect the understanding of the disease and the formation of diagnostic thinking, and cannot be effectively assisted when faced with complex cases, thereby affecting the efficiency and quality of diagnosis and treatment, and reducing the practicality and reliability of the case analysis assistance process. Summary of the invention

[0004] The purpose of the embodiments of the present invention is to provide a case analysis method, device, equipment, storage medium and program product to solve the above technical problems.

[0005] The present invention provides a case analysis method, which includes: obtaining case information to be analyzed and medical model information; performing case keyword extraction on the case information to be analyzed to obtain a case keyword set, and performing feature vector extraction on the case information to be analyzed to obtain an initial feature vector; determining a screening analysis knowledge base cluster based on the medical model information and the case keyword set; inputting the initial feature vector and the screening analysis knowledge base cluster into a preset analysis model to perform case information screening analysis to obtain vector information to be processed, so as to determine a case analysis result based on the vector information to be processed.

[0006] In one embodiment of the present invention, the preset analysis model includes a similarity screening model and a classification model, and inputting the initial feature vector and the screening analysis knowledge base cluster into the preset analysis model for case information screening analysis includes: inputting the initial feature vector and the screening analysis knowledge base cluster into the similarity screening model for feature vector screening to obtain a first screening text vector set; inputting the initial feature vector and the first screening text vector set into the classification model for classification screening to obtain a second screening text vector set, and determining the second screening text vector set as the vector information to be processed.

[0007] In one embodiment of the present invention, the initial feature vector and the screening analysis knowledge base cluster are input into the similarity screening model for feature vector screening, including: performing similarity calculation based on the initial feature vector and the text vector of each medical model knowledge base in the screening analysis knowledge base cluster to obtain feature similarity parameters of each medical model knowledge base; determining the text vector of the medical model knowledge base whose feature similarity parameter is greater than or equal to a preset similarity threshold as the first screening text vector to obtain a first screening text vector set.

[0008] In one embodiment of the present invention, the initial feature vector and the first screening text vector set are input into a classification model for classification screening to obtain a second screening text vector set, including: inputting the first screening text vector set into a feature extraction layer for classification feature extraction to obtain a screening text feature vector set, and inputting the screening text feature vector into a fully connected layer; the fully connected layer predicts the correlation between the initial feature vector and each screening text feature vector in the screening text feature vector set based on a preset activation function to obtain the correlation probability of each screening text feature vector in the screening text feature vector set; the correlation probability of each screening text feature vector in the screening text feature vector set is input into a normalization layer for normalization processing to obtain the association probability of each screening text feature vector in the screening text feature vector set, and a preset number of screening text feature vectors with the highest association probability are determined as the second screening text vector set; the feature extraction layer, the fully connected layer and the normalization layer all belong to the classification model.

[0009] In one embodiment of the present invention, determining a screening analysis knowledge base cluster based on the medical model information and the case keyword set includes: matching corresponding associated knowledge identifiers according to the medical model information; determining a medical model knowledge base with the same associated knowledge identifier in a preset analysis knowledge base cluster according to the associated knowledge identifier, and determining the medical model knowledge base with the same associated knowledge identifier as a target analysis knowledge base cluster, the medical model knowledge base includes at least one associated knowledge identifier, and the target analysis knowledge base cluster includes at least two medical model knowledge bases; screening is performed in the target analysis knowledge base cluster according to the case keyword set to obtain a screening analysis knowledge base cluster.

[0010] In one embodiment of the present invention, the target analysis knowledge base cluster is screened according to the case keyword set to obtain the screened analysis knowledge base cluster, including: respectively determining the overlap parameters of each medical model knowledge base in the target analysis knowledge base cluster and the case keyword set; and eliminating the medical model knowledge bases in the target analysis knowledge base cluster whose overlap parameters are less than a preset screening threshold to obtain the screened analysis knowledge base cluster.

[0011] An embodiment of the present invention also provides a case analysis device, which includes: an analysis information acquisition module, used to acquire case information to be analyzed and medical model information; a key information extraction module, used to extract case keywords from the case information to be analyzed to obtain a case keyword set, and to extract feature vectors from the case information to be analyzed to obtain an initial feature vector; a target knowledge base determination module, used to determine a screening analysis knowledge base cluster based on the medical model information and the case keyword set; and a case feature analysis module, used to input the initial feature vector and the screening analysis knowledge base cluster into a preset analysis model to perform case information screening analysis to obtain vector information to be processed, so as to determine the case analysis result based on the vector information to be processed.

[0012] An embodiment of the present invention also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the electronic device implements the case analysis method as described in any one of the above embodiments.

[0013] An embodiment of the present invention further provides a computer-readable storage medium on which computer-readable instructions are stored. When the computer-readable instructions are executed by a processor of a computer, the computer executes the case analysis method as described in any one of the above embodiments.

[0014] An embodiment of the present invention further provides a computer program product, including a computer program, which, when executed by a processor, implements the case analysis method as described in any one of the above embodiments.

[0015] The present invention provides a case analysis method, device, equipment, storage medium and program product, which obtains case information and medical model information to be analyzed, and extracts case keywords and feature vectors for the case information to be analyzed to obtain a case keyword set and an initial feature vector, determines a screening analysis knowledge base cluster according to the medical model information and the case keyword set, and inputs the initial feature vector and the screening analysis knowledge base cluster into a preset analysis model to perform case information screening analysis to determine the case analysis result; the present application extracts keywords from the case, combines the medical model information and the case keyword set to dynamically select the knowledge base subset that best suits the current case, thereby improving the accuracy of the analysis, reducing the amount of data required to be processed in subsequent analysis, and improving the efficiency of the analysis process, converting unstructured case text into feature vectors for higher-level semantic understanding and logical reasoning, and generating more detailed analysis information, which not only improves the depth of the analysis, but also ensures the scientificity and rationality of the output results, significantly improves the accuracy and reliability of the case analysis, reduces the situation where the generated results do not match the case information, and also improves the quality and credibility of the answer information, which not only improves the user experience, but also enhances the practicality and reliability, making it more suitable for medical education and clinical decision support scenarios.

[0016] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present application, and together with the specification, are used to explain the principles of the present application. Obviously, the drawings described below are only some embodiments of the present application, and for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work. In the drawings: Figure 1 is a schematic diagram of an exemplary system architecture shown in an exemplary embodiment of the present application; Figure 2 is a flow chart of a case analysis method shown in an exemplary embodiment of the present application; Figure 3 It is a flowchart of a specific case analysis method shown by an exemplary embodiment; Figure 4 is a schematic diagram of a specific case analysis auxiliary process architecture shown in an exemplary embodiment; Figure 5 is a schematic diagram of a case analysis device shown in an exemplary embodiment of the present application; Figure 6It is a structural diagram of a computer system of an electronic device shown in an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0018] The following will describe the embodiments of the present invention with reference to the accompanying drawings and specific embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be understood that the preferred embodiments are only for illustrating the present invention, not for limiting the scope of protection of the present invention.

[0019] It should be noted that the illustrations provided in the following embodiments are only used to illustrate the basic concept of the present invention in a schematic manner, and thus the illustrations only show components related to the present invention rather than being drawn according to the number, shape and size of components in actual implementation. In actual implementation, the type, quantity and proportion of each component may be changed arbitrarily, and the component layout may also be more complicated.

[0020] In the following description, numerous details are discussed to provide a more thorough explanation of the embodiments of the present invention. However, it is obvious to those skilled in the art that the embodiments of the present invention can be implemented without these specific details. In other embodiments, well-known structures and devices are shown in the form of block diagrams rather than in detail to avoid making the embodiments of the present invention difficult to understand.

[0021] The "and / or" mentioned in this application describes the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the associated objects before and after are in an "or" relationship.

[0022] Figure 1 It is a schematic diagram of an exemplary system architecture shown in an exemplary embodiment of the present application.

[0023] Reference Figure 1As shown, the system architecture may include a terminal device 110 and a server 120. The server 120 obtains case information to be analyzed and medical model information through the terminal device 110, performs case keyword extraction on the case information to be analyzed, obtains a case keyword set, and performs feature vector extraction on the case information to be analyzed, obtains an initial feature vector, and the server 120 determines a screening analysis knowledge base cluster according to the medical model information and the case keyword set, inputs the initial feature vector and the screening analysis knowledge base cluster into a preset analysis model to perform case information screening analysis, obtains vector information to be processed, and determines the case analysis result according to the vector information to be processed. The above-mentioned server 120 refers to a device that provides back-end services for the target application in the terminal, and can be a server cluster composed of at least one or more of physical servers, virtual servers, cloud servers, etc., which is mainly composed of big data analysis servers and artificial intelligence / machine learning servers, and can specifically include multiple knowledge bases and trained question and answer generation models; the above-mentioned terminal device 110 includes but is not limited to mobile phones, tablets, computers, laptops and other devices, and the terminal device is installed with one or more target applications. The target application here can refer to an application with data processing (such as information retrieval, intelligent question and answer) functions. For example, the target application can include an independent application, a web application, a mini-program in a host application, etc. Specifically, the target application can refer to a browser application, a video playback application, a social application, a content publishing application, an intelligent question and answer application, a game application, etc.

[0024] Schematically, the server 120 obtains the case information and medical model information to be analyzed through the terminal device 110, and performs case keyword extraction and feature vector extraction on the case information to be analyzed to obtain a case keyword set and an initial feature vector, determines a screening analysis knowledge base cluster based on the medical model information and the case keyword set, and inputs the initial feature vector and the screening analysis knowledge base cluster into a preset analysis model to perform case information screening analysis to determine the case analysis result; the present application improves the accuracy of the analysis by extracting keywords from the case, and dynamically selects the knowledge base subset that best suits the current case in combination with the medical model information and the case keyword set, and reduces the amount of data required to be processed in subsequent analysis, thereby improving the efficiency of the analysis process, converting unstructured case text into feature vectors for higher-level semantic understanding and logical reasoning, and generating more detailed analysis information, which not only improves the depth of the analysis, but also ensures the scientificity and rationality of the output results, significantly improves the accuracy and reliability of the case analysis, reduces the mismatch between the generated results and the case information, and also improves the quality and credibility of the answer information, which not only improves the user experience, but also enhances practicality and reliability, making it more suitable for medical education and clinical decision support scenarios.

[0025] Figure 2is a flowchart of a case analysis method shown in an exemplary embodiment of the present application. The case analysis method can be Figure 1 The implementation environment may be implemented in other implementation environments, and the above implementation environments are not specifically limited here. Figure 2 As shown, the flowchart of the case analysis method includes at least steps S210 to S240, which are described in detail as follows: In step S210, case information to be analyzed and medical model information are obtained.

[0026] In one embodiment of the present application, the main application scenario of the present application is set as a diagnostic auxiliary analysis method in the medical teaching process and the clinical diagnosis process. By integrating advanced natural language processing (NLP), machine learning and artificial intelligence technologies, it is intended to provide medical students, residents and clinicians with an intelligent, interactive case-assisted analysis method to help learn medical knowledge and improve clinical skills more effectively. For example, teachers can upload or select classic cases for students to learn online, and students can interact with auxiliary diagnosis, ask questions and get instant feedback, simulate real doctor-patient dialogues, practice questioning skills, or in other application scenarios, when doctors see patients, they can input information such as medical history, symptoms, and test results, recommend appropriate treatment plans based on the patient's individual characteristics (such as age, gender, allergy history, etc.), and provide relevant clinical guidelines and the latest research progress to help doctors confirm or rule out certain diseases.

[0027] In one embodiment of the present application, the case information to be analyzed can be the content of the case picture uploaded by the user, the case text content input by the user, or the voice about the case input by the user, or a combination of at least two of the above methods. For example, the user can upload a picture containing case information, such as a medical record card, an examination report, an imaging picture, etc., and then extract the text information in the picture through optical character recognition (OCR) technology or other image feature processing technology, and perform medical image analysis on the imaging picture to extract key features; the user can directly enter the text description of the case, including but not limited to the patient's chief complaint, the occurrence and development process of symptoms, physical examination results, past health conditions and medical history, personal living habits and experiences, family history, etc.; the user can use the voice input function to orally narrate the patient's case information, convert the voice into text, and perform natural language processing (NLP) to extract key information; the user can also choose to use multiple input methods at the same time, such as uploading pictures and supplementing text descriptions, or uploading relevant image materials after recording voice, and comprehensively process these different forms of information to ensure that all details of the case are captured comprehensively and accurately. For example, the medical model selected by the user is the TCM model, and the input case information to be analyzed includes: basic information such as the patient's name, gender, age, contact information, the patient's current most concerned symptoms or problems such as headache, cough, abdominal pain, etc., the time when the patient's symptoms first appeared, the duration, factors that aggravate or relieve the development process, the patient's body temperature, blood pressure, heart rate, respiratory rate, abnormalities found in physical examination, etc., the patient's past medical history, surgical history, allergy history, chronic disease management, etc., and information related to the patient's personal living habits and experiences such as eating habits, work and rest patterns, work environment, exercise, smoking and drinking history, etc., whether the immediate family members have hereditary diseases or other important medical history conditions. For the image content, use OCR technology to extract text; for imaging images, perform medical image analysis; for voice content, use voice recognition technology to convert it into text; all text content is standardized to ensure that the format is unified to extract key features to obtain the case information to be analyzed.

[0028] In one embodiment of the present application, medical model information refers to a specific knowledge system and methodology used to guide and assist medical diagnosis, treatment, and research, which can be classified according to different medical fields, theoretical bases, and application scopes. For example, based on medical theories, there are traditional Chinese medicine models, covering traditional Chinese medicine diagnosis, traditional Chinese medicine treatment, acupuncture, massage, health care, etc.; and Western medicine models, covering disease diagnosis, drug treatment, surgical treatment, rehabilitation treatment, etc. In some other embodiments, different models can also be classified based on pathological types. For example, the pulmonology model focuses on the structure and function of the respiratory system, including the diagnosis and treatment of respiratory diseases such as chronic obstructive pulmonary disease (COPD), asthma, pneumonia, lung cancer, etc.; the cardiology model studies the physiological and pathological mechanisms of the heart and blood vessel systems, covering the prevention, diagnosis, treatment, and rehabilitation of cardiovascular diseases; the oncology model studies the occurrence, development, and metastasis mechanisms of tumors, including the diagnosis, staging, treatment, and prognosis evaluation of various malignant tumors; the neurology model studies the structure and function of the nervous system, including the diagnosis and treatment of nervous system diseases such as stroke, epilepsy, Parkinson's disease, Alzheimer's disease, etc.; the infectious disease model studies the transmission routes, infection mechanisms, and prevention and control measures of pathogens, including the prevention, diagnosis, treatment, and public health management of various infectious diseases; the pediatrics model studies common and frequently-occurring diseases in pediatrics, including neonatology, pediatrics internal medicine, pediatrics surgery, child health care, etc.

[0029] It should be noted that the above examples of medical model information are only exemplary examples listed in the embodiments of the present application to illustrate the content of medical model information, and do not mean that the medical model information of the present application only includes the above information. In the process of specific application, the specific setting of medical model information can be carried out according to the specific diagnosis classification direction involved in medical colleges and hospitals. It can include larger classifications or more detailed classifications, and the specific content of medical model information is not limited here.

[0030] In step S220, case keyword extraction is performed on the case information to be analyzed to obtain a set of case keywords, and feature vector extraction is performed on the case information to be analyzed to obtain an initial feature vector.

[0031] In one embodiment of the present application, the TF-IDF (Term Frequency-Inverse Document Frequency) algorithm is used for case keyword extraction to evaluate the relative importance of each word in the current case. TF-IDF not only considers the frequency of the word in the current text but also its universality in the entire corpus, thus avoiding misjudging overly common words (such as "de", "shi") as keywords.

[0032] In some other embodiments, keyword extraction can also be based on rule-based methods, statistical methods, and machine learning methods. Among them, the rule-based method is to use a predefined medical terminology list and rule base to identify common medical terms and phrases. For example, based on a built-in terminology list containing common disease names (such as "hypertension", "diabetes", "pneumonia"), symptoms (such as "headache", "cough", "fever"), signs (such as "accelerated heart rate", "shortness of breath"), test results (such as "increased white blood cell count"), etc., extraction is performed, wherein the terminology list can be customized and expanded according to different medical fields (such as traditional Chinese medicine, Western medicine, and specialist medicine) to ensure that a wide range of medical knowledge is covered; or a series of rules are defined to identify specific medical expressions or patterns, such as "the patient's chief complaint is...", "the patient has a history of...", "laboratory tests show..." and other common medical sentence patterns, and then extract abbreviations, synonyms, and variant forms to ensure that the same concept in different forms can be recognized.

[0033] Secondly, the statistical method is to calculate the frequency of each word or phrase in the text and select words with higher frequency as candidate keywords. The most basic method includes word segmentation of the input text and counting the number of occurrences of each word or phrase. High-frequency words usually represent important information in the text, but some non-key nouns need to be manually removed, which will reduce efficiency to a certain extent. There are also machine learning-based methods, which refer to the use of trained machine learning models (such as TF-IDF, Word2Vec, BERT, etc.) to conduct in-depth analysis of the text and extract keywords with high weights. The TF-IDF model uses a machine learning-based TF-IDF model to further optimize the extraction of keywords. Through training, the weight of words can be automatically adjusted to improve the accuracy of keyword extraction; the Word2Vec model is a word vector model based on a neural network that can map words to a multidimensional space and capture the semantic relationship between words. By analyzing the vector representation of words, potential keywords related to the case can be identified. For example, even if "chest tightness" does not appear directly in the text, it can be identified as a keyword based on its similarity to "chest pain"; BERT (Bidirectional Encoder Representations from The BERT model is a bidirectional encoder model that can understand the contextual meaning of words in sentences. The pre-trained BERT model can more accurately identify complex medical terms and phrases. For example, BERT can understand the association between the various symptoms in "the patient complained of chest tightness, accompanied by palpitations and shortness of breath" and extract "chest tightness", "palpitations" and "shortness of breath" as keywords. Through the comprehensive application of the above methods, key medical terms and phrases can be efficiently and accurately extracted from multimodal case information to form a structured keyword set, providing a solid foundation for subsequent case analysis, knowledge base screening and diagnostic assistance.

[0034] In one embodiment of the present application, the process of extracting feature vectors of case information to be analyzed is to process the case information based on an embedding model, with the aim of converting multimodal case data (such as text, pictures, and voice) into a unified embedding vector representation, which can capture the semantic information of the text. The embedding models include but are not limited to Erlangshen-SimCSE, bge, Moka Massive Mixed Embedding (M3E), etc.

[0035] Among them, Erlangshen-SimCSE is a Chinese sentence embedding model based on the Transformer architecture, which can generate high-quality sentence-level embedding vectors. It optimizes by contrastive learning, making similar sentences closer in the embedding space and dissimilar sentences farther apart; bge is a general embedding model that is suitable for multiple languages ​​and fields, and can be fine-tuned to adapt to specific medical fields to provide more accurate embedding representations; Moka Massive Mixed Embedding (M3E) is suitable for processing large-scale mixed data. It combines multiple embedding methods and can have good embedding vector extraction in the medical field. The preprocessed case information to be analyzed is input into the selected embedding model to generate the corresponding embedding vector. Each embedding vector is a high-dimensional real number vector that can represent the semantic information of the text. In some implementable environments, if the input contains data of multiple modalities (such as pictures, voices, and texts), the embedding vectors of each modality can be generated separately, and then combined into a comprehensive embedding vector through fusion strategies (such as weighted average, splicing, attention mechanism, etc.), which helps to make full use of different types of data and improve the expressive power of the embedding vector.

[0036] In step S230, a screening and analysis knowledge base cluster is determined based on the medical model information and the case keyword set.

[0037] In one embodiment of the present application, a target analysis knowledge base cluster is determined in a preset analysis knowledge base cluster based on medical model information, the target analysis knowledge base cluster includes at least two medical model knowledge bases, and the target analysis knowledge base cluster is screened according to a case keyword set to obtain a screened analysis knowledge base cluster.

[0038] In one embodiment of the present application, determining a target analysis knowledge base cluster in a preset analysis knowledge base cluster based on medical model information includes matching corresponding associated knowledge identifiers according to the medical model information, determining a medical model knowledge base with the same associated knowledge identifier in the preset analysis knowledge base cluster based on the associated knowledge identifier, and determining the medical model knowledge base with the same associated knowledge identifier as the target analysis knowledge base cluster, wherein the medical model knowledge base includes at least one associated knowledge identifier.

[0039] In one embodiment of the present application, the medical model information is one or more medical models selected by the user through the system interface. For example, the user can select "Chinese medicine model", "Western medicine model" or "respiratory model", etc. According to the medical model information, the knowledge identification management module is called to obtain the associated knowledge identification corresponding to the selected medical model information. The knowledge identification management module includes at least one mapping table, which records the relationship between each medical model and its corresponding knowledge identification. For example, the traditional Chinese medicine model corresponds to the first knowledge identifier, the western medicine model corresponds to the second knowledge identifier, the respiratory disease model corresponds to the third knowledge identifier..., query and obtain the corresponding knowledge identifier from the mapping table. For example, if the user selects "traditional Chinese medicine model", the "first knowledge identifier" will be obtained as the corresponding associated knowledge identifier, and all medical model knowledge bases with the same associated knowledge identifier will be searched in the preset analysis knowledge base cluster. Each medical model knowledge base contains at least one associated knowledge identifier, which is used to identify the medical field and content scope to which it belongs. The medical model knowledge base with the same associated knowledge identifier is determined as the target analysis knowledge base cluster. For example, if the user selects "traditional Chinese medicine model" and obtains the "first knowledge identifier", all medical model knowledge bases with the "first knowledge identifier" will be found in the preset analysis knowledge base cluster and used as the target analysis knowledge base cluster.

[0040] In some other embodiments, if there are multiple medical models in the medical model information (such as "traditional Chinese medicine model" and "western medicine model"), the associated knowledge identifier corresponding to each model is obtained respectively, and the knowledge base with the corresponding knowledge identifier is determined in the preset analysis knowledge base cluster, and these knowledge bases are merged into a comprehensive target analysis knowledge base cluster to support multidisciplinary joint analysis.

[0041] In one embodiment of the present application, screening in the target analysis knowledge base cluster according to the case keyword set includes respectively determining the overlap parameters of each medical model knowledge base in the target analysis knowledge base cluster and the case keyword set, and eliminating the medical model knowledge bases in the target analysis knowledge base cluster whose overlap parameters are less than a preset screening threshold, to obtain a screened analysis knowledge base cluster.

[0042] In an embodiment of the present application, the term frequency-inverse document frequency (TF-IDF) method is used to determine the screening analysis knowledge base cluster that has a high overlap with the case keyword set in the target analysis knowledge base cluster. The term frequency-inverse document frequency method is used to capture the keyword matching degree between each medical model knowledge base and the case keyword set in the target analysis knowledge base cluster, that is, if the documents in a medical model knowledge base contain more keywords in the case keyword set, then the medical model knowledge base is considered to be strongly related to the case information. Pre-screening the knowledge base by the TF-IDF method is faster than directly screening the knowledge base by similarity, thereby increasing the speed of case information analysis and improving user experience.

[0043] In step S240, the initial feature vector and the screening analysis knowledge base cluster are input into a preset analysis model to perform case information screening analysis to obtain the vector information to be processed, so as to determine the case analysis result according to the vector information to be processed.

[0044] In one embodiment of the present application, the preset analysis model includes at least one of a similarity screening model and a classification model.

[0045] In one embodiment of the present application, the initial feature vector and the screening analysis knowledge base cluster are input into the similarity screening model for feature vector screening to obtain a first screening text vector set, the initial feature vector and the first screening text vector set are input into the classification model for classification screening to obtain a second screening text vector set, and the second screening text vector set is determined as the vector information to be processed.

[0046] In one embodiment of the present application, the initial feature vector and the screening analysis knowledge base cluster are input into the similarity screening model for feature vector screening, including performing similarity calculation based on the initial feature vector and the text vector of each medical model knowledge base in the screening analysis knowledge base cluster to obtain the feature similarity parameter of each medical model knowledge base, and determining the text vector of the medical model knowledge base whose feature similarity parameter is greater than or equal to a preset similarity threshold as the first screening text vector to obtain a first screening text vector set.

[0047] In one embodiment of the present application, the initial feature vector is a multidimensional feature representation extracted from the case information to be analyzed, the screening and analysis knowledge base cluster includes at least one medical model knowledge base, each knowledge base contains a large number of documents and their corresponding text vectors, and the similarity screening model is used to calculate the similarity between the initial feature vector and the text vector in the knowledge base and screen out text vectors with higher correlation according to a preset threshold, and the first screened text vector set is a text vector set that is highly correlated with the case information obtained after similarity screening.

[0048] The initial feature vector is compared with the text vector of each medical model knowledge base in the screening analysis knowledge base cluster one by one, that is, for each document in each medical model knowledge base, the similarity between its text vector and the initial feature vector is calculated. The similarity calculation method includes at least one of the following algorithms: cosine similarity algorithm, Euclidean distance algorithm, Manhattan distance algorithm, etc. For example, in the embodiment of the present application, the cosine similarity algorithm is used to calculate the similarity between the text vectors of all its documents and the initial feature vector for each medical model knowledge base, and obtain a feature similarity parameter, which reflects the relevance of the knowledge base as a whole to the case information. The text vector of the medical model knowledge base whose feature similarity parameter is greater than or equal to the preset similarity threshold is determined as the first screening text vector, wherein the preset similarity threshold is used to screen out text vectors that are highly relevant to the case information. The selection of the threshold can be adjusted according to the actual application scenario. In the embodiment of the present application, the value is usually between 0.7 and 1.0, depending on the accuracy requirements and data distribution of the system.

[0049] Based on the above similarity screening process, on the one hand, this application places similarity screening before classification semantic screening, which can efficiently complete the data screening task with less data comparison calculation, and can quickly exclude irrelevant knowledge bases in the early stage, effectively filter out low-similarity case text data, so that the amount of feature data that the classification model needs to process when performing semantic screening is greatly reduced. Compared with directly performing semantic feature screening and searching on the full database, this strategy can comprehensively improve the response speed and overall efficiency of the question-answering system, greatly shorten the user waiting time, and thus improve the quality and satisfaction of medical services; on the other hand, the similarity screening process has the ability to accurately screen out text vectors that are highly relevant to case information from a huge medical model knowledge base. This ensures that when the classification model performs semantic screening, it only analyzes the knowledge base content that is most closely related to the case information, effectively avoiding the interference of irrelevant information, and the screening results obtained are more in line with the actual situation of the case, thereby improving the accuracy of diagnosis and treatment recommendations, so that users can receive more accurate and personalized medical recommendations.

[0050] In one embodiment of the present application, the classification model includes at least one of the following: a feature extraction layer, a fully connected layer, and a normalization layer. For example, inputting the initial feature vector and the first screening text vector set into the classification model for classification screening includes inputting the first screening text vector set into the feature extraction layer for classification feature extraction to obtain a screening text feature vector set, and inputting the screening text feature vector into the fully connected layer, the fully connected layer predicts the correlation between the initial feature vector and each screening text feature vector in the screening text feature vector set based on a preset activation function to obtain the correlation probability of each screening text feature vector in the screening text feature vector set, and the correlation probability of each screening text feature vector in the screening text feature vector set is input into the normalization layer for normalization processing to obtain the association probability of each screening text feature vector in the screening text feature vector set, and the preset number of screening text feature vectors with the highest association probability are determined as the second screening text vector set. In some other preferred embodiments, the classification model may not include a feature extraction layer, and the initial feature vector and the first screening text vector set are used as the screening text feature vector set for input into the fully connected layer.

[0051] In one embodiment of the present application, the initial feature vector and the first screened text vector set are respectively input into the feature extraction layer. The feature extraction layer uses a neural network structure (such as a convolutional neural network CNN, a recurrent neural network RNN ​​or a Transformer) to perform classification feature extraction on the input feature vector, capture key information and patterns in the feature vector, and obtain a screened text feature vector set.

[0052] In one embodiment of the present application, after the set of filtered text feature vectors is input into the fully connected layer, the fully connected layer predicts the correlation between the embedded feature vector and each filtered text feature vector based on a preset activation function (such as ReLU, Sigmoid, Tanh and other activation functions), and outputs the correlation probability between each filtered text feature vector and the embedded feature vector. The correlation probability reflects the degree of correlation between each filtered text feature vector and the case information. The closer the value is to 1, the higher the correlation.

[0053] In one embodiment of the present application, the correlation probability of each filtered text feature vector in the filtered text feature vector set is input into the normalization layer and the correlation probability is normalized using the softmax function so that the sum of all correlation probabilities is 1. The normalized probability value is called the association probability, which indicates the relative importance of each filtered text feature vector among all candidate texts. The above association probability reflects the relative degree of association between each filtered text feature vector and the case information. The closer the value is to 1, the higher the association.

[0054] In some implementable environments, a preset number (such as the first 5, the first 10, etc.) can be set according to user needs or configurations to indicate the number of most relevant text vectors that are ultimately retained. The associated probability of each screening text feature output by the classification model is arranged from high to low, and the screening text features ranked in the front by the preset number are selected to be added to the second screening text vector set.

[0055] It is not difficult to understand that the text vectors most relevant to the case information are further screened out from the first screened text vector set, so as to more accurately capture the correlation between the text vector and the case information, ensure that the final screened text vector is the most relevant, improve the accuracy of diagnosis and treatment recommendations, capture the complex patterns and key information in the feature vectors, generate higher-level embedded feature vectors and screened text feature vectors, and enhance the text expression ability.

[0056] In one embodiment of the present application, the vector information to be processed is input into a selected large language model, and the language understanding and generation capabilities of the large language model are applied to accurately capture key information in the case text, such as symptom descriptions, medical history records, etc. to determine the case analysis results. Among them, the large language model (Large Language Model) has a high degree of versatility and flexibility, and can be widely used in a variety of downstream tasks, including but not limited to text processing, semantic understanding, machine translation, robot question answering, and knowledge graph construction. Specifically, the large language model in the embodiment of the present application includes but is not limited to: (1) Embeddings from Language Models (ELMO) based on language models, which can capture the different meanings of words in different contexts by training deep bidirectional language models to improve the accuracy of case analysis; (2) Transformer-based pre-trained language model (Bidirectional Encoder Representations from Transformers, BERT), which uses the bidirectional encoder of the Transformer architecture to pre-train the language model. In case analysis, the BERT model can deeply understand the semantic information of the case text and effectively improve the depth and breadth of the analysis; (3) Transformer-based autoregressive language model (Generative Pre-Trained Transformer, GPT), which learns the generation rules of language through autoregression. In case analysis, it can be used to generate case summaries, predict disease development trends, and other tasks, providing more comprehensive and detailed information support. This embodiment introduces a large language model to process the vector information to be processed and determine the case analysis results, thereby providing strong data support for subsequent medical diagnosis, treatment recommendations, etc.

[0057] Based on the above embodiment of applying a large language model to determine case analysis results, the following comparative embodiment is further discussed: Among them, the comparative example 1 is based on the application of similarity screening, classification model and other screening means in the aforementioned content of this application to accurately capture the key information in the case text, such as disease type, symptom description, etc. The obtained vector information to be processed provides more accurate data input for the subsequent large language model analysis, so that the large language model can conduct in-depth case analysis. Due to the accuracy of the previous screening and classification, the large language model can more accurately understand the semantic information of the case text and generate more accurate and detailed case analysis results. Comparative Example 2 is a single screening means in the prior art. The existing large language model conversion process only uses simple screening means (such as keyword matching or simple semantic similarity calculation) to screen case texts, which often cannot accurately capture the complex semantic information in the case text, resulting in low accuracy of the screening results. Therefore, the large language model will be interfered by irrelevant information when processing these texts, resulting in the generated case analysis results being inaccurate or detailed.

[0058] By comparing Example 1 with Example 2, it can be seen that the screening process of the present application combines a variety of screening technical means, which can more efficiently screen case texts while improving the accuracy of screening results. Compared with traditional technical means, the solution of the present application reduces the interference of irrelevant data, and the large language model can focus more on processing information related to the target case, thereby generating more accurate and reliable case analysis results.

[0059] Please refer to Figure 3 , Figure 3 FIG. 1 is a flow chart of a specific case analysis method shown in an exemplary embodiment. Figure 3As shown, in S301, in response to the first operation of the user, a knowledge identifier is obtained, wherein the first operation of the user is a process of obtaining medical model information, and the knowledge identifier is determined according to the medical model information, and the knowledge identifier here is equivalent to the associated knowledge identifier in the above embodiment; in S302, in response to the second operation of the user, case information is obtained, and the case information here is equivalent to the case information to be analyzed in the above embodiment; in S303, based on the knowledge identifier, a second knowledge base cluster (target analysis knowledge base cluster) matching the knowledge identifier is determined in the first knowledge base cluster (preset analysis knowledge base cluster); in S304, a third knowledge base cluster (screening analysis knowledge base cluster) having a large overlap with the case information keywords in the second knowledge base cluster (target analysis knowledge base cluster) is determined by the word frequency-inverse document frequency method; S In 305, the case information is processed to generate an embedding vector (consistent with the initial feature vector in the above embodiment); in S306, the text vector and the embedding vector (consistent with the initial feature vector in the above embodiment) of the knowledge base (medical model knowledge base) of the third knowledge base cluster (screening analysis knowledge base cluster) are calculated by a similarity algorithm to determine the first screening text vector and obtain a first screening text vector set; in S307, the input first screening text vector and the embedding vector are classified by a classification model to generate a classification result, and the second screening text vector is confirmed according to the classification result to obtain a second screening text vector set; in S308, the second screening text vector (consistent with the vector information to be processed in the above embodiment) is input into the large language model to determine the case analysis result. It should be noted that steps S305-S306 can also be before S304, and steps S305-S306 and S304 can also be executed in parallel, and the first screening text vector is determined by weighting the results of the similarity algorithm and the word frequency-inverse document frequency method.

[0060] Please refer to Figure 4 , Figure 4 is a schematic diagram of a specific case analysis auxiliary process architecture shown in an exemplary embodiment. Figure 4As shown, the embedding model is used to convert the input case information into an embedding vector, the similarity algorithm is used to calculate the similarity between the embedding vector and the text vector stored in the knowledge base, and the first screening text vector whose similarity meets the threshold condition is screened out, the classification model is used to classify the input first screening text vector, and confirm the second screening text vector according to the classification result, and the large language model is used to determine the case analysis result corresponding to the case information according to the second screening text vector. According to the knowledge identifier obtained from the case information, the screening analysis knowledge base cluster is extracted in the knowledge base cluster, and the text vector of the screening analysis knowledge base cluster is output, and the knowledge base cluster includes multiple knowledge bases. At the same time, the case information is embedded in the embedding model. The similarity calculation method is used to calculate the converted vector and the previous text vector to obtain the first screening text vector, and then the classification model is used to obtain the second screening text vector. Finally, the second screening text vector is analyzed and converted in the large language model to obtain the case analysis result.

[0061] The present invention provides a case analysis method, device, equipment, storage medium and program product, which obtains case information and medical model information to be analyzed, and extracts case keywords and feature vectors for the case information to be analyzed to obtain a case keyword set and an initial feature vector, determines a screening analysis knowledge base cluster according to the medical model information and the case keyword set, and inputs the initial feature vector and the screening analysis knowledge base cluster into a preset analysis model to perform case information screening analysis to determine the case analysis result; the present application extracts keywords from the case, combines the medical model information and the case keyword set to dynamically select the knowledge base subset that best suits the current case, thereby improving the accuracy of the analysis, reducing the amount of data required to be processed in subsequent analysis, and improving the efficiency of the analysis process, converting unstructured case text into feature vectors for higher-level semantic understanding and logical reasoning, and generating more detailed analysis information, which not only improves the depth of the analysis, but also ensures the scientificity and rationality of the output results, significantly improves the accuracy and reliability of the case analysis, reduces the situation where the generated results do not match the case information, and also improves the quality and credibility of the answer information, which not only improves the user experience, but also enhances the practicality and reliability, making it more suitable for medical education and clinical decision support scenarios.

[0062] The following describes an embodiment of the device of the present application, which can be used to execute the case analysis method in the above embodiment of the present application. For details not disclosed in the embodiment of the device of the present application, please refer to the embodiment of the case analysis method in the above embodiment of the present application.

[0063] Figure 5 is a schematic diagram of a case analysis device shown in an exemplary embodiment of the present application. The device can be applied to Figure 2 The method implementation process shown in the figure can be based on Figure 1 The present invention is executed in the implementation environment shown in the figure, and may also be applicable to other exemplary implementation environments and specifically configured in other devices. This embodiment does not limit the implementation environment to which the apparatus is applicable.

[0064] like Figure 5 As shown, the exemplary case analysis device includes: an analysis information acquisition module 501, a key information extraction module 502, a target knowledge base determination module 503 and a case feature analysis module 504.

[0065] Among them, the analysis information acquisition module 501 is used to obtain the case information and medical model information to be analyzed; the key information extraction module 502 is used to extract case keywords for the case information to be analyzed, obtain a case keyword set, and extract feature vectors for the case information to be analyzed, and obtain an initial feature vector; the target knowledge base determination module 503 is used to determine the screening analysis knowledge base cluster according to the medical model information and the case keyword set; the case feature analysis module 504 is used to input the initial feature vector and the screening analysis knowledge base cluster into the preset analysis model to perform case information screening analysis, and obtain the vector information to be processed, so as to determine the case analysis result according to the vector information to be processed.

[0066] An embodiment of the present application also provides an electronic device, comprising: one or more processors; a storage device for storing one or more programs, when the one or more programs are executed by one or more processors, the electronic device implements the case analysis method provided in the above-mentioned embodiments.

[0067] Figure 6 is a schematic diagram of the structure of a computer system of an electronic device according to an exemplary embodiment of the present application. It should be noted that: Figure 6 The computer system 600 of the electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0068] like Figure 6As shown, computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to the program stored in read-only memory (ROM) 602 or the program loaded from the storage part to random access memory (RAM) 603, such as executing the method in the above embodiment. In RAM 603, various programs and data required for system operation are also stored. CPU 601, ROM 602 and RAM 603 are connected to each other through a bus. I / O interface 605 is also connected to bus 604, wherein I / O interface 605 refers to input / output (Input / Output) interface.

[0069] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section performs communication processing via a network such as the Internet. A drive is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed so that a computer program read therefrom is installed into the storage section 608 as needed.

[0070] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication section 609, and / or installed from a removable medium 611. When the computer program is executed by a central processing unit (CPU) 601, various functions defined in the system of the present application are executed.

[0071] It should be noted that the computer-readable medium shown in the embodiment of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable computer program is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. A computer program contained on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0072] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. Among them, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0073] In the corresponding drawings of the above embodiments, connecting lines may indicate the connection relationship between the various components to indicate more constituent signal paths (constituent_signalpath) and / or one or more ends of some lines may have arrows to indicate the main information flow direction. Connecting lines, as a mark, are not a limitation to the scheme itself, but the use of these lines in combination with one or more exemplary embodiments helps to connect circuits or logic units more easily. Any represented signal (determined by design requirements or preferences) may actually include one or more signals that can be transmitted in either direction and implemented with any appropriate type of signal scheme.

[0074] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.

[0075] Another aspect of the present application further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. The computer-readable storage medium may be included in the electronic device described in the above embodiment, or may exist independently without being assembled into the electronic device.

[0076] An embodiment of the present application further provides a computer program product, including a computer program, which, when executed by a processor, implements the case analysis method as described in any one of the above embodiments.

[0077] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0078] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the implementation methods of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation methods of the present application.

[0079] It should be noted that the present application can be used in many general or special computing system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0080] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the embodiments disclosed herein. The present application is intended to cover any variations, uses or adaptations of the present application, which follow the general principles of the present application and include common knowledge or customary technical means in the art that are not disclosed in the present application.

[0081] It should be understood that the above content of the present application is only a preferred exemplary embodiment of the present application and is not intended to limit the implementation scheme of the present application. Ordinary technicians in this field can easily make corresponding changes or modifications based on the main concept and spirit of the present application. Therefore, the scope of protection of the present application should be the scope of protection required by the claims.

Claims

1. A case analysis method, characterized in that: The case analysis method comprises: Obtain case information and medical model information to be analyzed; Extracting case keywords from the case information to be analyzed to obtain a case keyword set, and extracting feature vectors from the case information to be analyzed to obtain an initial feature vector; Determine a screening and analysis knowledge base cluster according to the medical model information and the case keyword set; The initial feature vector and the screening analysis knowledge base cluster are input into a preset analysis model to perform case information screening analysis to obtain the vector information to be processed, so as to determine the case analysis result according to the vector information to be processed.

2. The case analysis method according to claim 1, characterized in that: The preset analysis model includes a similarity screening model and a classification model, and inputting the initial feature vector and the screening analysis knowledge base cluster into the preset analysis model to perform case information screening analysis includes: Inputting the initial feature vector and the screening analysis knowledge base cluster into a similarity screening model to perform feature vector screening to obtain a first screening text vector set; The initial feature vector and the first screened text vector set are input into a classification model for classification screening to obtain a second screened text vector set, and the second screened text vector set is determined as vector information to be processed.

3. The case analysis method according to claim 2, characterized in that: Inputting the initial feature vector and the screening analysis knowledge base cluster into the similarity screening model to perform feature vector screening includes: Perform similarity calculation based on the initial feature vector and the text vector of each medical model knowledge base in the screening and analysis knowledge base cluster to obtain feature similarity parameters of each medical model knowledge base; The text vectors of the medical model knowledge base whose feature similarity parameters are greater than or equal to a preset similarity threshold are determined as first screening text vectors to obtain a first screening text vector set.

4. The case analysis method according to claim 2, characterized in that: The initial feature vector and the first screening text vector set are input into a classification model for classification screening, and a second screening text vector set is obtained, comprising: Inputting the first screening text vector set into the feature extraction layer for classification feature extraction to obtain a screening text feature vector set, and inputting the screening text feature vector into the fully connected layer; The fully connected layer predicts the correlation between the initial feature vector and each of the filtered text feature vectors in the filtered text feature vector set based on a preset activation function to obtain the correlation probability of each of the filtered text feature vectors in the filtered text feature vector set; Inputting the correlation probability of each screening text feature vector in the screening text feature vector set into the normalization layer for normalization processing to obtain the correlation probability of each screening text feature vector in the screening text feature vector set, and determining a preset number of screening text feature vectors with the highest correlation probability as the second screening text vector set; The feature extraction layer, fully connected layer and normalization layer all belong to the classification model.

5. The case analysis method according to claim 1, characterized in that: Determining the screening and analysis knowledge base cluster according to the medical model information and the case keyword set includes: Matching corresponding associated knowledge identifiers according to the medical model information; Determine a medical model knowledge base with the same associated knowledge identifier in a preset analysis knowledge base cluster according to the associated knowledge identifier, and determine the medical model knowledge base with the same associated knowledge identifier as a target analysis knowledge base cluster, wherein the medical model knowledge base includes at least one associated knowledge identifier, and the target analysis knowledge base cluster includes at least two medical model knowledge bases; The target analysis knowledge base cluster is screened according to the case keyword set to obtain a screened analysis knowledge base cluster.

6. The case analysis method according to claim 5, characterized in that: The target analysis knowledge base cluster is screened according to the case keyword set, and the screening analysis knowledge base cluster includes: Respectively determining the overlap parameters between each medical model knowledge base in the target analysis knowledge base cluster and the case keyword set; The medical model knowledge bases in the target analysis knowledge base cluster whose overlap parameters are less than the preset screening threshold are eliminated to obtain a screening analysis knowledge base cluster.

7. A case analysis device, characterized in that: The case analysis device comprises: An analysis information acquisition module is used to obtain case information to be analyzed and medical model information; A key information extraction module is used to extract case keywords from the case information to be analyzed to obtain a case keyword set, and to extract feature vectors from the case information to be analyzed to obtain an initial feature vector; A target knowledge base determination module, used to determine a screening and analysis knowledge base cluster according to the medical model information and the case keyword set; The case feature analysis module is used to input the initial feature vector and the screening analysis knowledge base cluster into a preset analysis model to perform case information screening analysis to obtain the vector information to be processed, so as to determine the case analysis result based on the vector information to be processed.

8. An electronic device, characterized in that: It comprises a processor, a memory and a communication bus; the communication bus is used to connect the processor and the memory; the processor is used to execute the computer program stored in the memory to implement the case analysis method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program is used to enable a computer to execute the case analysis method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images