Auxiliary diagnosis method and device based on medical diagnosis model, electronic equipment and storage medium
By employing a medical diagnostic model-based assisted diagnostic method, utilizing the Transformer disease prediction network and a large language model for feature extraction and processing of electronic medical records, the problem of data inconsistency in electronic health records is resolved, thereby improving the accuracy and completeness of assisted diagnosis.
Patent Information
- Application Number
- CN202411175260.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-26
- Publication Date
- 2024-11-12
AI Technical Summary
Existing electronic health record analysis methods suffer from inconsistent, incomplete, and erroneous data due to the wide range of data sources and diverse input methods, which affects the accuracy of assisted diagnosis.
We employ an auxiliary diagnostic method based on a medical diagnostic model. By acquiring electronic medical records, we perform feature extraction and processing, clean and standardize the data using an improved Transformer disease prediction network and a large language model, and combine traditional Chinese and Western medicine knowledge for diagnosis.
It achieves automated data cleaning and standardization, improves data integrity and accuracy, avoids errors in auxiliary diagnostic information, and helps doctors make correct diagnoses.
Smart Images

Figure CN118942682A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of medical information processing technology, and more specifically, to an auxiliary diagnosis method, device, electronic device and storage medium based on a medical diagnosis model. Background Art
[0002] With the development of computer technology, computer-based medical assisted diagnosis technology has become increasingly mature, and the most common method now is the electronic health record analysis method. Electronic health records are a digital medical record system that records patients' health information, including medical history, diagnosis results, treatment plans, medication status, laboratory test results, etc. This solution manages and analyzes electronic health records through a computer system, which can provide important decision-making support for doctors.
[0003] The electronic health record analysis method mainly includes processes such as electronic data collection and storage, data processing and analysis, and clinical decision support. Through the processing and analysis of electronic health records, a patient's medical history report can be generated to help doctors understand the patient's long-term health status. It can also recommend possible diagnostic results and treatment plans and other auxiliary diagnostic information based on a large amount of historical data and medical knowledge base, as well as remind doctors of medication precautions. The electronic health record analysis method can greatly improve the efficiency and accuracy of medical work. Doctors can quickly access the patient's complete health record through the system without having to look through paper medical records. At the same time, the system's automated analysis and reminder functions can reduce human errors and improve the accuracy of diagnosis and treatment. The electronic health record system also supports data sharing and collaboration.
[0004] However, after studying the existing electronic health record analysis methods, the inventors of this application found that due to the wide range of data sources and diverse input methods, the data in the electronic health records often contain inconsistencies, incompleteness, and erroneous information, which will make the subsequent auxiliary diagnostic information prone to errors, thereby misleading doctors to make incorrect diagnostic results. Summary of the invention
[0005] In view of this, the present application provides an auxiliary diagnosis method, device, electronic device and storage medium based on a medical diagnosis model, which are used to provide doctors with accurate auxiliary diagnosis information to help doctors make correct diagnosis results.
[0006] In order to achieve the above objectives, the proposed solution is as follows:
[0007] An auxiliary diagnosis method based on a medical diagnosis model is applied to electronic equipment, and the auxiliary diagnosis method comprises the following steps:
[0008] Obtaining an electronic medical record, which carries part or all of the historical health data, personal medical records, medical consultation data, and current physical examination information of a patient to be diagnosed;
[0009] Performing feature extraction processing on the electronic medical record to obtain feature data;
[0010] The electronic medical record and the characteristic data are processed based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of the patient to be diagnosed. The medical diagnosis model includes an improved Transformer disease prediction network and a large language model obtained by supervised micro-debugging of a large medical model.
[0011] Optionally, the step of performing feature extraction processing on the electronic medical record to obtain feature data comprises the following steps:
[0012] Extracting features of the content carried by the electronic medical record based on a machine learning algorithm to obtain medical record feature data;
[0013] Feature extraction is performed on existing medical knowledge to obtain expert knowledge data, wherein the feature data includes case feature data and expert knowledge data.
[0014] Optionally, the expert knowledge data includes knowledge representation based on a vector space model, knowledge representation based on association rules, and knowledge representation based on domain knowledge.
[0015] Optionally, the auxiliary diagnosis and treatment information includes disease prediction information and diagnosis analysis information.
[0016] Optionally, the processing of the electronic medical record and the characteristic data based on the medical diagnosis model comprises the steps of:
[0017] Processing the electronic medical record and the characteristic data based on the improved Transformer disease prediction network to obtain the disease prediction information;
[0018] The electronic medical record and the characteristic data are processed based on the large language model to obtain the diagnostic analysis information.
[0019] Optionally, the processing of the electronic medical record and the characteristic data based on the medical diagnosis model further includes the steps of:
[0020] The electronic medical record, the characteristic data and the disease prediction information are processed based on the large language model to obtain the diagnostic analysis information.
[0021] Optionally, the data used in the supervised micro-debugging process includes part or all of open source general data, medical consultation data, human feedback data, and medical domain knowledge.
[0022] Optionally, the medical field knowledge includes modern medical knowledge and traditional medical knowledge.
[0023] An auxiliary diagnosis device based on a medical diagnosis model is applied to an electronic device, and the auxiliary diagnosis device comprises:
[0024] A medical record acquisition module is configured to acquire an electronic medical record, wherein the electronic medical record carries part or all of the historical health data, personal medical records, medical consultation data, and current physical examination information of a patient to be diagnosed;
[0025] A feature extraction module is configured to perform feature extraction processing on the electronic medical record to obtain feature data;
[0026] The data processing module is configured to process the electronic medical record and the feature data based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of the patient to be diagnosed. The medical diagnosis model includes an improved Transformer disease prediction network and a large language model obtained by supervised micro-debugging of a large medical model.
[0027] An electronic device, comprising at least one processor and a memory connected to the processor, wherein:
[0028] The memory is used to store computer programs or instructions;
[0029] The processor is used to execute the computer program or instruction so that the electronic device implements the auxiliary diagnosis method as described above.
[0030] A computer-readable storage medium is applied to an electronic device, wherein the storage medium carries one or more computer programs, and the one or more computer programs can be executed by the electronic device, so that the electronic device can implement the auxiliary diagnosis method as described above.
[0031] It can be seen from the above technical solutions that the present application discloses an auxiliary diagnosis method, device, electronic device and storage medium based on a medical diagnosis model. The method and device are applied to electronic devices, specifically to obtain an electronic medical record, which carries part or all of the historical health data, personal medical records, medical consultation data and current physical examination information of a patient to be diagnosed; feature extraction processing is performed on the electronic medical record to obtain feature data; the electronic medical record and feature data are processed based on the medical diagnosis model to obtain auxiliary diagnosis and treatment information of the patient to be diagnosed, and the medical diagnosis model includes an improved Transformer disease prediction network and a large language model obtained by supervised micro-debugging of a large medical model. The technical solution in this application can realize automatic data cleaning and standardization based on the natural language processing technology of the large language model, can identify and correct inconsistent data, and improve the integrity and accuracy of the data. Thereby, it is possible to avoid errors in the auxiliary diagnosis information obtained subsequently, and it is also possible to help doctors avoid making wrong diagnosis results. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0033] Figure 1 A flowchart of an auxiliary diagnosis method based on a medical diagnosis model according to an embodiment of the present application;
[0034] Figure 2 A schematic diagram of an improved Transformer disease prediction model according to an embodiment of the present application;
[0035] Figure 3 A schematic diagram of an application example of a large language model according to an embodiment of the present application;
[0036] Figure 4a A schematic diagram of an inference result of a large language model in an embodiment of the present application;
[0037] Figure 4b A schematic diagram of another inference result of a large language model in an embodiment of the present application;
[0038] Figure 4c A schematic diagram of another inference result of a large language model in an embodiment of the present application;
[0039] Figure 4d A schematic diagram of another inference result of a large language model in an embodiment of the present application;
[0040] Figure 4e A schematic diagram of another inference result of a large language model in an embodiment of the present application;
[0041] Figure 5 A block diagram of an auxiliary diagnosis device based on a medical diagnosis model according to an embodiment of the present application;
[0042] Figure 6 A block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION
[0043] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions 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. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0044] Different medical diagnosis tasks require different algorithms. Small models are suitable for processing structured data and predicting specific diseases, while large models are usually computationally complex and resource-intensive, and perform better in complex medical knowledge understanding and disease diagnosis reasoning. In order to achieve effective processing of diverse medical tasks, this application proposes an algorithm solution that combines large models with small models. On the one hand, they can complement each other's advantages and enhance the accuracy and robustness of auxiliary diagnosis; on the other hand, they can reduce overall resource consumption and efficient diagnosis by assigning specific tasks to models that occupy different computing resources.
[0045] Figure 1 This is a flow chart of an auxiliary diagnosis method based on a medical diagnosis model according to an embodiment of the present application.
[0046] like Figure 1 As shown, the auxiliary diagnosis method provided by the present application is applied to an electronic device for auxiliary diagnosis based on a medical diagnosis model, where the medical diagnosis model includes the large model and the small model mentioned above. The electronic device can be understood as a computer, server or cloud platform with data computing and information processing capabilities, and the auxiliary diagnosis method includes the following specific steps:
[0047] S1. Obtain electronic medical records.
[0048] The electronic medical record refers to part or all of the patient's historical health data, personal medical records, medical consultation data, and current physical examination information generated manually based on data entry. In addition, the above electronic medical records can also be automatically generated based on a certain electronic platform. The current physical examination information refers to the body information obtained recently or currently through certain technical means, such as medical images, ultrasound test results, blood routine data, etc.
[0049] S2. Extract features from the electronic medical records to obtain feature data.
[0050] That is, the information in the electronic medical record is characterized to obtain characteristic data that conforms to the model input habits. The characteristic data in this application includes the medical record characteristic book generated from the electronic medical record, and also includes expert instruction data that is not derived from the electronic medical record. Specifically, the characteristic data is obtained by the following method:
[0051] First, feature extraction is performed on the content of the electronic medical record based on the machine learning algorithm to obtain case feature data. The main function of the machine learning algorithm is to learn and predict diseases from data such as electronic medical records and doctor-patient conversations. It includes various classification, regression, clustering and other models that can automatically learn from data and apply them to new cases.
[0052] Then or at the same time, feature extraction is performed on the existing medical knowledge to obtain expert knowledge data. The expert knowledge data includes but is not limited to knowledge representation based on vector space model, knowledge representation based on association rules and knowledge representation based on domain knowledge. The sources of expert knowledge include medical literature, clinical guidelines, etc. This knowledge is incorporated into the system through certain rules and reasoning methods to assist decision-making and judgment in the diagnosis process.
[0053] Knowledge representation based on the vector space model refers to the use of mathematical identifiers by the vector space model (VSM) to represent text content. In the recommendation decision technology, in order to make the unstructured text information describing the item computable, the vector space model is often used for algebraic transformation. The more commonly used text vectorization representation method is term frequency-inverse document frequency (TF-IDF). This method believes that the importance of a word is positively correlated with the frequency of its appearance in the document and negatively correlated with the frequency of its appearance in the entire language. TF-IDF can mine keywords in documents and apply it to mine the correlation between the recommended object and the corpus features.
[0054] In the knowledge representation based on association rules, drugs are defined as "items", drug sets are defined as "item sets", and the full item set is recorded as: I = {i1, i2, ..., i m}, "k-item set" refers to a set of k drugs, such as {aspirin} is a 1-item set, {aspirin, ibuprofen} is a 2-item set. A patient's historical diagnosis and treatment sequence is recorded as a "transaction" T, which can contain multiple drugs, indicating the list of drugs prescribed by the doctor during a visit to the doctor. Association rules are used to mine relevant information from a transaction set, expressed as: X=>Y, where and This paper uses indicators such as co-occurrence probability, confidence, support, expected credibility, and lift to evaluate the effectiveness of association rules.
[0055] The following is a detailed introduction to the calculation process of these five association rule metrics. Co-occurrence probability was first used in the embedding representation of natural language processing to count the co-occurrences between words and enhance the contextual semantic expression. Assuming a represents a word and b represents the context, the co-occurrence probability is calculated as the ratio of the number of times a appears in the context of b to the number of times any A appears in the context of b. In natural language processing, context refers to other words that are no more than m (m≤10) away from the current word.
[0056] The confidence of association rule A=>B indicates the probability of item set B appearing at the same time when item set A appears, that is, the ratio of the number of elements in the intersection of item set A and item set B to the number of elements in item set A. Confidence is used to evaluate the accuracy of association rules, indicating the degree of inevitability of B occurring under the premise of A occurring. The support of association rule A=>B indicates the ratio of the number of elements in the intersection of item set A and item set B to the total number of elements in the transaction set count(T).
[0057] Support is used to evaluate the importance of association rules and indicates the universality of the current rule among all rules. The expected credibility of item set A is usually defined by the probability of item set A appearing, that is, the ratio of the number of elements in item set A to the number of elements in the total transaction set T. The improvement of association rule A=>B also describes the probability of b appearing when a appears. The difference from confidence is that the improvement also introduces the probability of a and b appearing separately. The improvement is the ratio of confidence to expected credibility, indicating whether this rule promotes or weakens the probability of item set B appearing, as shown in formula (7). The improvement is used to evaluate the effectiveness of association rules. When the improvement is less than 1, it means that there is a negative correlation between a and b. When the improvement is greater than 1, it means that there is a positive correlation between a and b.
[0058] Knowledge representation based on domain knowledge, for example, for TCM, the domain knowledge of TCM mainly includes three categories: congenital constitution estimation based on the patient's date of birth, numerical segmentation features derived from the patient's gender and age, and symptom similarity matching derived from standard symptoms. The following is the logic of generating representations of three types of TCM domain knowledge.
[0059] a) The logic of deriving congenital constitution from date of birth: Traditional Chinese medicine believes that there is a certain correlation between a person's illness and his or her congenital constitution, and the sensitivity of an individual to disease varies with different constitutions. Factors affecting individual constitution include age, gender, congenitality, living environment and other factors. Scholars have paid attention to the relationship between congenital constitution and date of birth. The present invention uses the congenital constitution calculation mechanism to calculate multiple characteristics such as "congenital weak organs", "dominant qi and guest qi", "yin and yang and five elements" according to the individual's date of birth. The date of birth affects a person's congenital constitution, which is mainly affected by the objective living environment of time and space on the newborn. In addition, the congenital constitution will also change with the changes in the acquired environment.
[0060] b) Gender and age segmentation logic unique to TCM: TCM theory believes that individuals have growth cycles, and the physical manifestations and conditions of individuals in different cycles will vary greatly. The Yellow Emperor's Classic of Internal Medicine describes the impact of gender on the life cycle of individuals. Individuals with different sex ratios have different growth cycles, based on the principle of "eight years for males and seven years for females", that is, a growth cycle of eight years for males and seven years for females. This paper discretizes the ages of men and women using the piecewise function of "eight years for males and seven years for females" to obtain age characteristics from the perspective of TCM.
[0061] c) Medical case symptom semantic similarity matching logic: Traditional Chinese medicine is an empirical science, and similar medical cases play an important auxiliary role in the diagnosis and treatment process. Doctors can quickly diagnose the current patient's condition based on the prescriptions in existing medical cases and give guidance based on individual differences.
[0062] S3. Process electronic medical records and feature data based on medical diagnosis models.
[0063] That is, the electronic medical records and feature data are processed based on the medical diagnosis model including the large model and the small model to obtain the auxiliary diagnosis and treatment information of the patient. The small model here refers to the improved Transformer disease prediction network, and the large model refers to the large medical model of the supervised micro-debugging process, or it is called a large language model. By processing the electronic medical records and feature data, auxiliary diagnosis and treatment information including basic prediction information and diagnostic analysis information is obtained. Specifically, the processing is achieved by the following means:
[0064] On the one hand, the electronic medical records and feature data are processed based on the improved Transformer disease prediction network to obtain disease prediction information. The data of this network includes 18,625 data sets, 16 features, and 1 label. In the disease prediction model, the label is the disease type, and the features are shown in the following table:
[0065]
[0066]
[0067] On the other hand, electronic medical records and feature data are processed based on large language models to obtain diagnostic analysis information.
[0068] On the other hand, based on the disease prediction information, the electronic medical records, feature data and basic prediction information can be processed based on the large language model to obtain the diagnostic analysis information.
[0069] The small model in this application is the specific structure of the improved Transformer disease prediction network. Figure 2 As shown. Transformer is a deep learning model architecture for natural language processing. It processes input data through a self-attention mechanism, can capture long-distance dependencies in the data, and is suitable for processing sequence data such as text. Compared with traditional recurrent neural networks and convolutional neural networks, the Transformer model performs well in processing long texts, translation, text generation and other tasks. It is the basic architecture of large language models such as GPT and BERT. A specific disease prediction process is as follows:
[0070]
[0071] The large model provided in this application refers to a large language model (Large Language Model) obtained after supervised micro-tuning of the medical large model based on fine-tuning data, such as Baichuan-7B, Ziya-LLaMA-13B-Pretrain-v1, ChatGLM-6B, LLaMA-7B, LLaMA-7B, Bloom-7B, Ziya-LLaMA-13B, baichuan-7B, ChatGLM-6B, B loomz-7B, ChatGLM-6B, ChatGLM-6B, baichuan, LLaMA, ChatGLM and B loomz, etc.
[0072] The fine-tuning data here includes open source general data, medical consultation data, human feedback data, and medical domain knowledge. Open source general data is a publicly available data set that can be used to train and improve models; medical consultation data refers to data related to medical consultation, which may include medical history, symptom descriptions, etc.; human feedback data is evaluation or correction from people, which helps to improve the accuracy and reliability of the model. Medical domain knowledge includes but is not limited to modern medical knowledge including Western medicine, and traditional medical knowledge including Chinese medicine, Mongolian medicine, Tibetan medicine, etc.
[0073] Western medicine data comes from open source data of medical big models, including situational dialogues, knowledge questions and answers, etc. Scenario dialogue data includes disease description, etiology analysis, disease diagnosis, pathological diagnosis, drug dosage, medication recommendations, medical knowledge, prognosis assessment, medical advice, medical advice, preventive measures, treatment plans, indicator interpretation, etc., as well as multiple rounds of patient information reasoning and diagnosis data. Knowledge questions and answers come from medical databases, covering 28 departments by providing specific medical consensus and clinical guidelines texts. Medical knowledge question and answer data includes knowledge questions and answers based on clinical guidelines and medical consensus, knowledge questions and answers based on physician qualification examination questions, real doctor-patient questions and answers, and knowledge questions and answers based on structured medical graphs.
[0074] like Figure 3 As shown, this embodiment demonstrates the working process of a large language model. Among them, professional medical data is used as prompts to guide it to generate more accurate results; the extraction step is to extract key information from the input, such as diagnosis, symptoms, treatment methods and preventive measures; "Input, Instruction, Output" is the process of the large language model receiving input, following instructions and generating output. A and B in the large language model represent the two main components or stages in the model. Q|K|V refers to query, key and value, which are commonly used terms in natural language processing to describe how the model processes and understands text information. The dialog box on the right shows a question and answer about stomachache and eating apples. This shows that the model can understand and respond to specific health problems and provide relevant suggestions.
[0075] It can be seen from the above technical solution that this embodiment provides an auxiliary diagnosis method based on a medical diagnosis model, which is applied to electronic devices, specifically to obtain an electronic medical record, which carries part or all of the historical health data, personal medical records, medical consultation data and current physical examination information of a patient to be diagnosed; feature extraction processing is performed on the electronic medical record to obtain feature data; based on the medical diagnosis model, the electronic medical record and feature data are processed to obtain auxiliary diagnosis and treatment information of the patient to be diagnosed, and the medical diagnosis model includes an improved Transformer disease prediction network and a large language model obtained by supervised micro-debugging of a large medical model. The technical solution in this application can realize automatic data cleaning and standardization based on the natural language processing technology of the large language model, can identify and correct inconsistent data, and improve the integrity and accuracy of the data. Thereby, errors in the auxiliary diagnosis information obtained subsequently can be avoided, which can also help doctors avoid making wrong diagnosis results.
[0076] The inventors of this application obtained the following results after actual verification based on the general large models of chatglm-6b and Qwen-7b deployed on GPU: Figures 4a to 4e From the reasoning results, it can be seen that the present application perfectly achieves the invention purpose of the present application.
[0077] This application proposes the following innovative points:
[0078] 1) A modeling and prediction scheme that combines a large model with a small model is proposed, that is, a multi-level prediction model architecture, a linkage system of a large model and multiple small models. The large model is used for global feature extraction and preliminary prediction, and the small model is used for refined prediction in a specific field. The large model is responsible for comprehensive processing of a large amount of heterogeneous data and generating high-level feature representations, while the small model focuses on specific diseases or specific diagnostic tasks and uses the features generated by the large model for further analysis. First, in terms of data processing and feature extraction, the large model uses the deep learning technology Transformer to process massive medical data and extract global features. Based on the features generated by the large model, the small model combines professional knowledge in a specific field to make refined predictions.
[0079] Secondly, in terms of model linkage mechanism, the data transmission and linkage mechanism between the large model and the small model ensures efficient flow of information between different models. The selection and adjustment of the small model are guided by the preliminary prediction results of the large model, thereby improving the accuracy and efficiency of the overall prediction. In addition, in terms of adaptive learning and optimization, an adaptive learning method is proposed, which enables the large model and the small model to be dynamically adjusted and optimized according to different diagnostic tasks and data characteristics.
[0080] Describe the model's training and updating mechanisms in detail to ensure that the system improves its prediction performance through continuous learning and improvement.
[0081] 2) Both TCM and Western medicine auxiliary diagnosis can be supported, and TCM and Western medicine can be linked. The auxiliary system supports both TCM and Western medicine diagnosis, and can process and analyze diagnostic data under the two medical systems. The system can integrate TCM's four diagnostics (looking, smelling, asking, and cutting) data with Western medicine's examination, laboratory tests and other modern medical data. A method of combining the medical knowledge graphs of TCM and Western medicine is proposed to build a comprehensive medical knowledge base. In the diagnosis process, a large model is used to integrate the characteristics of TCM and Western medicine to generate a unified diagnostic recommendation. Small models are used to conduct joint diagnosis of TCM and Western medicine for specific diseases to improve the comprehensiveness and accuracy of diagnosis.
[0082] 3) Integrate the representation of medical knowledge into data and propose methods for medical knowledge representation, including the systematic representation of knowledge such as theories of traditional Chinese medicine and Western medicine, diagnostic criteria, and treatment plans. By embedding medical knowledge into data representation technology, the diagnostic model can understand and use professional knowledge for prediction. Based on the knowledge-driven diagnostic model, intelligent diagnosis is performed by combining medical knowledge and patient data.
[0083] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the 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 implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square 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 square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart 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.
[0084] Although operations are depicted in a particular order, this should not be understood as requiring that the operations be performed in the particular order shown, or in a sequential order.Multitasking and parallel processing may be advantageous under certain circumstances.
[0085] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.
[0086] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as C or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer.
[0087] Figure 5 This is a block diagram of an auxiliary diagnosis device based on a medical diagnosis model according to an embodiment of the present application.
[0088] like Figure 5 As shown, the auxiliary diagnosis device provided by the present application is applied to an electronic device for auxiliary diagnosis based on a medical diagnosis model, where the medical diagnosis model includes the large model and the small model mentioned above. The electronic device can be understood as a computer, server or cloud platform with data computing and information processing capabilities, and the auxiliary diagnosis device includes a medical record acquisition module 10, a feature extraction module 20 and a data processing module 30.
[0089] The case acquisition module is used to obtain electronic medical records.
[0090] The electronic medical record refers to part or all of the patient's historical health data, personal medical records, medical consultation data, and current physical examination information generated manually based on data entry. In addition, the above electronic medical records can also be automatically generated based on a certain electronic platform. The current physical examination information refers to the body information obtained recently or currently through certain technical means, such as medical images, ultrasound test results, blood routine data, etc.
[0091] The feature extraction module is used to extract features from electronic medical records to obtain feature data.
[0092] That is, the information in the electronic medical record is characterized to obtain characteristic data that conforms to the model input habits. The characteristic data in this application includes the medical record characteristic book generated from the electronic medical record, and also includes expert instruction data that is not derived from the electronic medical record. Specifically, the characteristic data is obtained by the following method:
[0093] First, feature extraction is performed on the content of the electronic medical record based on the machine learning algorithm to obtain case feature data. The main function of the machine learning algorithm is to learn and predict diseases from data such as electronic medical records and doctor-patient conversations. It includes various classification, regression, clustering and other models that can automatically learn from data and apply them to new cases.
[0094] Then or at the same time, feature extraction is performed on the existing medical knowledge to obtain expert knowledge data. The expert knowledge data includes but is not limited to knowledge representation based on vector space model, knowledge representation based on association rules and knowledge representation based on domain knowledge. The sources of expert knowledge include medical literature, clinical guidelines, etc. This knowledge is incorporated into the system through certain rules and reasoning methods to assist decision-making and judgment in the diagnosis process.
[0095] The data processing module is used to process electronic medical records and feature data based on the medical diagnosis model.
[0096] That is, the electronic medical records and feature data are processed based on the medical diagnosis model including the large model and the small model to obtain the auxiliary diagnosis and treatment information of the patient. The small model here refers to the improved Transformer disease prediction network, and the large model refers to the large medical model of the supervised micro-debugging process, or it is called a large language model. By processing the electronic medical records and feature data, auxiliary diagnosis and treatment information including basic prediction information and diagnostic analysis information is obtained. Specifically, the processing is achieved by the following means:
[0097] On the one hand, the electronic medical records and feature data are processed based on the improved Transformer disease prediction network to obtain disease prediction information. The data of this network includes 18,625 data sets, 16 features, and 1 label. In the disease prediction model, the label is the disease type, and the features are shown in the following table:
[0098]
[0099]
[0100] On the other hand, electronic medical records and feature data are processed based on large language models to obtain diagnostic analysis information.
[0101] On the other hand, based on the disease prediction information, the electronic medical records, feature data and basic prediction information can be processed based on the large language model to obtain the diagnostic analysis information.
[0102] The small model in this application is the specific structure of the improved Transformer disease prediction network. Figure 2 As shown in the figure. Transformer is a deep learning model architecture for natural language processing. It processes input data through a self-attention mechanism, can capture long-distance dependencies in the data, and is suitable for processing sequence data such as text. Compared with traditional recurrent neural networks and convolutional neural networks, the Transformer model performs well in processing long texts, translation, text generation and other tasks. It is the basic architecture of large language models such as GPT and BERT.
[0103] The large model provided in this application refers to a large language model (Large Language Model) obtained after supervised micro-tuning of the medical large model based on fine-tuning data, such as Baichuan-7B, Ziya-LLaMA-13B-Pretrain-v1, ChatGLM-6B, LLaMA-7B, LLaMA-7B, Bloom-7B, Ziya-LLaMA-13B, baichuan-7B, ChatGLM-6B, B loomz-7B, ChatGLM-6B, ChatGLM-6B, baichuan, LLaMA, ChatGLM and B loomz, etc.
[0104] The fine-tuning data here includes open source general data, medical consultation data, human feedback data, and medical domain knowledge. Open source general data is a publicly available data set that can be used to train and improve models; medical consultation data refers to data related to medical consultation, which may include medical history, symptom descriptions, etc.; human feedback data is evaluation or correction from people, which helps to improve the accuracy and reliability of the model. Medical domain knowledge includes but is not limited to modern medical knowledge including Western medicine, and traditional medical knowledge including Chinese medicine, Mongolian medicine, Tibetan medicine, etc.
[0105] Western medicine data comes from open source data of medical big models, including situational dialogues, knowledge questions and answers, etc. Scenario dialogue data includes disease description, etiology analysis, disease diagnosis, pathological diagnosis, drug dosage, medication recommendations, medical knowledge, prognosis assessment, medical advice, medical advice, preventive measures, treatment plans, indicator interpretation, etc., as well as multiple rounds of patient information reasoning and diagnosis data. Knowledge questions and answers come from medical databases, covering 28 departments by providing specific medical consensus and clinical guidelines texts. Medical knowledge question and answer data includes knowledge questions and answers based on clinical guidelines and medical consensus, knowledge questions and answers based on physician qualification examination questions, real doctor-patient questions and answers, and knowledge questions and answers based on structured medical graphs.
[0106] like Figure 3As shown, this embodiment demonstrates the working process of a large language model. Among them, professional medical data is used as prompts to guide it to generate more accurate results; the extraction step is to extract key information from the input, such as diagnosis, symptoms, treatment methods and preventive measures; "Input, Instruction, Output" is the process of the large language model receiving input, following instructions and generating output. A and B in the large language model represent the two main components or stages in the model. Q|K|V refers to query, key and value, which are commonly used terms in natural language processing to describe how the model processes and understands text information. The dialog box on the right shows a question and answer about stomachache and eating apples. This shows that the model can understand and respond to specific health problems and provide relevant suggestions.
[0107] It can be seen from the above technical solution that this embodiment provides an auxiliary diagnosis device based on a medical diagnosis model, which is applied to electronic equipment, specifically to obtain an electronic medical record, which carries part or all of the historical health data, personal medical records, medical consultation data and current physical examination information of a patient to be diagnosed; feature extraction processing is performed on the electronic medical record to obtain feature data; based on the medical diagnosis model, the electronic medical record and feature data are processed to obtain auxiliary diagnosis and treatment information of the patient to be diagnosed, and the medical diagnosis model includes an improved Transformer disease prediction network and a large language model obtained by supervised micro-debugging of a large medical model. The technical solution in this application can realize automatic data cleaning and standardization based on the natural language processing technology of the large language model, can identify and correct inconsistent data, and improve the integrity and accuracy of the data. Thereby, it is possible to avoid errors in the auxiliary diagnosis information obtained subsequently, and it can also help doctors avoid making wrong diagnosis results.
[0108] The units involved in the embodiments described in the present disclosure may be implemented by software or hardware. The name of a unit does not limit the unit itself in some cases. For example, the first acquisition unit may also be described as a "unit for acquiring at least two Internet Protocol addresses".
[0109] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.
[0110] Figure 6A block diagram of an electronic device according to an embodiment of the present application.
[0111] Reference below Figure 6 , which shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present disclosure. The terminal device in the embodiments of the present disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. The electronic device is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.
[0112] The electronic device may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory ROM 602 or a program loaded from an input device 606 into a random access memory RAM 603. In the RAM, various programs and data required for the operation of the electronic device are also stored. The processing device, ROM, and RAM are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.
[0113] Typically, the following devices may be connected to the I / O interface: input devices including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 608 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 609. The communication device 609 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows an electronic device with various devices, it should be understood that it is not required to implement or have all the devices shown. More or fewer devices may be implemented or have alternatively.
[0114] The present application also provides a computer-readable storage medium embodiment.
[0115] The computer-readable storage medium is applied to an electronic device and carries one or more computer programs. When the one or more computer programs are executed by the electronic device, the electronic device obtains an electronic medical record, which carries part or all of the historical health data, personal medical records, medical consultation data and current physical examination information of a patient to be diagnosed; performs feature extraction processing on the electronic medical record to obtain feature data; processes the electronic medical record and feature data based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of the patient to be diagnosed, and the medical diagnosis model includes an improved Transformer disease prediction network and a large language model obtained by supervised micro-debugging of a large medical model. The technical solution in this application can realize automatic data cleaning and standardization based on the natural language processing technology of the large language model, can identify and correct inconsistent data, and improve the integrity and accuracy of the data. Thereby, it is possible to avoid errors in the auxiliary diagnosis information obtained subsequently, and it can also help doctors avoid making wrong diagnosis results.
[0116] It should be noted that the computer-readable medium disclosed above 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, but not limited to, 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 or 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.
[0117] In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, 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 combination with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.
[0118] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.
[0119] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.
[0120] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.
[0121] The technical solution provided by the present invention is introduced in detail above. Specific examples are used in this article to illustrate the principle and implementation mode of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation mode and application scope. In summary, the content of this specification should not be understood as limiting the present invention.
Claims
1. An auxiliary diagnosis method based on a medical diagnosis model, applied to electronic equipment, It is characterized in that The auxiliary diagnosis method comprises the steps of: Obtaining an electronic medical record, which carries part or all of the historical health data, personal medical records, medical consultation data, and current physical examination information of a patient to be diagnosed; Performing feature extraction processing on the electronic medical record to obtain feature data; The electronic medical record and the characteristic data are processed based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of the patient to be diagnosed. The medical diagnosis model includes an improved Transformer disease prediction network and a large language model obtained by supervised micro-debugging of a large medical model.
2. The auxiliary diagnosis method according to claim 1, It is characterized in that The step of extracting features from the electronic medical record to obtain feature data comprises the following steps: Extracting features of the content carried by the electronic medical record based on a machine learning algorithm to obtain medical record feature data; Feature extraction is performed on existing medical knowledge to obtain expert knowledge data, wherein the feature data includes case feature data and expert knowledge data.
3. The auxiliary diagnosis method according to claim 2, It is characterized in that The expert knowledge data includes knowledge representation based on a vector space model, knowledge representation based on association rules and knowledge representation based on domain knowledge.
4. The auxiliary diagnosis method according to claim 1, It is characterized in that The auxiliary diagnosis and treatment information includes disease prediction information and diagnosis analysis information.
5. The auxiliary diagnosis method according to claim 4, It is characterized in that The processing of the electronic medical record and the characteristic data based on the medical diagnosis model comprises the steps of: Processing the electronic medical record and the characteristic data based on the improved Transformer disease prediction network to obtain the disease prediction information; The electronic medical record and the characteristic data are processed based on the large language model to obtain the diagnostic analysis information.
6. The auxiliary diagnosis method according to claim 5, It is characterized in that The processing of the electronic medical record and the characteristic data based on the medical diagnosis model further includes the steps of: The electronic medical record, the characteristic data and the disease prediction information are processed based on the large language model to obtain the diagnostic analysis information.
7. The auxiliary diagnosis method according to claim 1, It is characterized in that The data used in the supervised micro-debugging process includes part or all of open source general data, medical consultation data, human feedback data, and medical domain knowledge.
8. The auxiliary diagnosis method according to claim 7, It is characterized in that The medical field knowledge includes modern medical knowledge and traditional medical knowledge.
9. An auxiliary diagnosis device based on a medical diagnosis model, applied to electronic equipment, It is characterized in that The auxiliary diagnosis device comprises: A medical record acquisition module is configured to acquire an electronic medical record, wherein the electronic medical record carries part or all of the historical health data, personal medical records, medical consultation data, and current physical examination information of a patient to be diagnosed; A feature extraction module is configured to perform feature extraction processing on the electronic medical record to obtain feature data; The data processing module is configured to process the electronic medical record and the feature data based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of the patient to be diagnosed. The medical diagnosis model includes an improved Transformer disease prediction network and a large language model obtained by supervised micro-debugging of a large medical model.
10. An electronic device, It is characterized in that The electronic device comprises at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs or instructions; The processor is used to execute the computer program or instruction so that the electronic device implements the auxiliary diagnosis method according to any one of claims 1 to 8.
11. A computer-readable storage medium, applied to an electronic device, It is characterized in that The storage medium carries one or more computer programs, and the one or more computer programs can be executed by the electronic device, so that the electronic device can implement the auxiliary diagnosis method according to any one of claims 1 to 8.
Citation Information
Cited By
Automatic iteration method and system of medical identification and diagnosis model and storage medium
CN119851965A
Pancreatic cancer prediction method and system based on local and global confusion weighted pruning
CN121839088A
