CN118942682A - Auxiliary diagnosis method, device, electronic device and storage medium based on medical diagnosis model - Google Patents
Auxiliary diagnosis method, device, electronic device and storage medium based on medical diagnosis model Download PDFInfo
- Publication number
- CN118942682A CN118942682A CN202411175260.2A CN202411175260A CN118942682A CN 118942682 A CN118942682 A CN 118942682A CN 202411175260 A CN202411175260 A CN 202411175260A CN 118942682 A CN118942682 A CN 118942682A
- Authority
- CN
- China
- Prior art keywords
- medical
- data
- model
- electronic
- diagnosis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H10/00—ICT specially adapted for the handling or processing of patient-related medical or healthcare data
- G16H10/60—ICT specially adapted for the handling or processing of patient-related medical or healthcare data for patient-specific data, e.g. for electronic patient records
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Public Health (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Molecular Biology (AREA)
- Epidemiology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Databases & Information Systems (AREA)
- Pathology (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The application discloses an auxiliary diagnosis method, an auxiliary diagnosis device, electronic equipment and a storage medium based on a medical diagnosis model, wherein the method and the device are applied to the electronic equipment, and particularly acquire an electronic medical record; performing feature extraction processing on the electronic medical record to obtain feature data; and processing the electronic medical record and the characteristic data based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of a patient to be diagnosed, wherein the medical diagnosis model comprises an improved transducer disease prediction network and a large-scale language model obtained by performing supervised micro-debugging processing on the large-scale medical model. According to the technical scheme, automatic data cleaning and standardization can be realized based on a natural language processing technology of a large language model, inconsistent data can be identified and corrected, and the integrity and accuracy of the data are improved. Therefore, errors of auxiliary diagnosis information obtained later can be avoided, and doctors can be helped to avoid making wrong diagnosis results.
Description
Technical Field
The present application relates to the field of medical information processing technology, and more particularly, to an auxiliary diagnosis method, apparatus, electronic device, and storage medium based on a medical diagnosis model.
Background
With the development of computer technology, the medical auxiliary diagnosis technology based on computer technology is mature, and an electronic health record analysis method is more common. Electronic health records are a digital medical record system that records patient health information including medical history, diagnostic results, treatment regimens, medication, laboratory test results, and the like. The scheme manages and analyzes the electronic health records through the computer system, and can provide important decision support for doctors.
The electronic health record analysis method mainly comprises the processes of electronic data collection and storage, data processing and analysis, clinical decision support and the like. The medical history report of the patient can be generated through processing and analyzing the electronic health record, so that doctors can be helped to know the long-term health condition of the patient, and possible diagnosis results, treatment schemes and other auxiliary diagnosis information can be recommended according to a large amount of historical data and a medical knowledge base, and the doctor is reminded of taking medicine. The electronic health record analysis method can greatly improve the efficiency and accuracy of medical work. A doctor can quickly access the complete health record of the patient through the system without turning over the paper medical record. Meanwhile, the automatic analysis and reminding functions of the system can reduce human errors and improve the accuracy of diagnosis and treatment. Electronic health record systems also support sharing and collaboration of data.
However, the inventor of the present application finds that, after researching the existing electronic health record analysis method, the scheme has the problems of inconsistent, incomplete and error information of the data in the electronic health record due to wide data sources and various input modes, which can cause the auxiliary diagnosis information obtained later to be easy to make errors, so that the doctor can be misled to make an error diagnosis result.
Disclosure of Invention
In view of the above, the present application provides an auxiliary diagnosis method, apparatus, electronic device and storage medium based on a medical diagnosis model for providing accurate auxiliary diagnosis information to a doctor to help the doctor to make a correct diagnosis result.
In order to achieve the above object, the following solutions have been proposed:
an auxiliary diagnosis method based on a medical diagnosis model is applied to electronic equipment, and comprises the following steps:
Acquiring an electronic medical record, wherein the electronic medical record carries part or all of historical health data, personal diagnosis and treatment records, medical inquiry 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;
And processing the electronic medical record and the characteristic data based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of the patient to be diagnosed, wherein the medical diagnosis model comprises an improved transducer disease prediction network and a large-scale language model obtained by performing supervised micro-debugging processing on the large-scale medical model.
Optionally, the feature extraction processing is performed on the electronic medical record to obtain feature data, including the steps of:
Performing feature extraction on the content carried by the electronic medical record based on a machine learning algorithm to obtain medical record feature data;
and extracting features of the existing medical knowledge to obtain expert knowledge data, wherein the feature data comprises case feature data and expert knowledge data.
Optionally, the expert knowledge data includes a knowledge representation based on a vector space model, a knowledge representation based on association rules, and a knowledge representation based on domain knowledge.
Optionally, the auxiliary diagnosis and treatment information includes disease prediction information and diagnosis analysis information.
Optionally, the processing the electronic medical record and the feature data based on the medical diagnosis model includes the steps of:
processing the electronic medical record and the characteristic data based on the improved transducer disease prediction network to obtain the disease prediction information;
And processing the electronic medical record and the characteristic data based on the large language model to obtain the diagnosis analysis information.
Optionally, the processing the electronic medical record and the feature data based on the medical diagnosis model further includes the steps of:
And processing the electronic medical record, the characteristic data and the disease prediction information based on the large language model to obtain the diagnosis analysis information.
Optionally, the data used in the supervised micro debug process includes part or all of open source general data, medical consultation data, human feedback data, and medical domain knowledge.
Optionally, the medical domain knowledge includes modern medical knowledge and traditional medical knowledge.
An auxiliary diagnostic apparatus based on a medical diagnostic model, applied to an electronic device, comprising:
The medical record acquisition module is configured to acquire an electronic medical record, wherein the electronic medical record carries part or all of historical health data, personal diagnosis and treatment records, medical inquiry data and current physical examination information of a patient to be diagnosed;
the 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 characteristic 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 comprises an improved transducer disease prediction network and a large-scale language model obtained by performing supervised micro-debugging processing on the large-scale medical model.
An electronic device comprising at least one processor and a memory coupled to the processor, wherein:
the memory is used for storing a computer program or instructions;
the processor is configured to execute the computer program or instructions to cause the electronic device to implement the auxiliary diagnostic method as described above.
A computer-readable storage medium for application to an electronic device, the storage medium carrying one or more computer programs executable by the electronic device to enable the electronic device to implement an auxiliary diagnostic method as described above.
From the above technical solution, the present application discloses an auxiliary diagnosis method, apparatus, electronic device and storage medium based on a medical diagnosis model, where the method and apparatus are applied to electronic devices, specifically, obtain electronic medical records, where the electronic medical records carry part or all of historical health data, personal diagnosis and treatment records, medical inquiry 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; and processing the electronic medical record and the characteristic data based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of a patient to be diagnosed, wherein the medical diagnosis model comprises an improved transducer disease prediction network and a large-scale language model obtained by performing supervised micro-debugging processing on the large-scale medical model. According to the technical scheme, automatic data cleaning and standardization can be realized based on a natural language processing technology of a large language model, inconsistent data can be identified and corrected, and the integrity and accuracy of the data are improved. Therefore, errors of auxiliary diagnosis information obtained later can be avoided, and doctors can be helped to avoid making wrong diagnosis results.
Drawings
In order to more clearly illustrate the embodiments of the application or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of an auxiliary diagnostic method based on a medical diagnostic model according to an embodiment of the present application;
FIG. 2 is a schematic diagram of an improved transducer disease prediction model according to an embodiment of the present application;
FIG. 3 is a schematic diagram of an application example of a large language model according to an embodiment of the present application;
FIG. 4a is a schematic diagram of an inference result of a large language model in an embodiment of the present application;
FIG. 4b is a schematic diagram of another inference result of a large language model in an embodiment of the present application;
FIG. 4c is a schematic diagram of yet another inference result of a large language model in an embodiment of the present application;
FIG. 4d is a schematic diagram of yet another inference result of a large language model in an embodiment of the present application;
FIG. 4e is a schematic diagram of yet another inference result of a large language model in an embodiment of the present application;
FIG. 5 is a block diagram of an auxiliary diagnostic apparatus based on a medical diagnostic model according to an embodiment of the present application;
fig. 6 is a block diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Different algorithms are required for different medical diagnostic tasks, small models are suitable for the processing of structured data and the prediction of specific diseases, while large models are generally complex in calculation and high in resource consumption, and perform better in complex medical knowledge understanding and disease diagnosis pushing. In order to realize effective treatment of diversified medical tasks, the application provides an algorithm scheme of a large-model linkage small model, on one hand, the advantages of the large-model linkage small model can be mutually complemented, and the accuracy and the robustness of auxiliary diagnosis are enhanced; on the other hand, the overall resource consumption and the efficient diagnosis can be reduced by assigning specific tasks to models occupying different computing resources.
Fig. 1 is a flowchart of an auxiliary diagnosis method based on a medical diagnosis model according to an embodiment of the present application.
As shown in fig. 1, the auxiliary diagnosis method provided by the application is applied to an electronic device for performing auxiliary diagnosis based on a medical diagnosis model, wherein the medical diagnosis model comprises the large model and the small model. The electronic device can be understood as a computer, a server or a cloud platform with data computing capability and information processing capability, and the auxiliary diagnosis method comprises the following specific steps:
s1, acquiring an electronic medical record.
The electronic medical record refers to part or all of historical health data, personal diagnosis and treatment records, medical inquiry data and current physical examination information of a patient, which are generated based on data input in a manual mode, and the electronic medical record can be automatically generated based on a certain electronic platform. The current physical examination information refers to physical information which is acquired recently or currently through a certain technical means, such as medical images, ultrasonic detection results, blood routine data and the like.
S2, extracting features of the electronic medical record to obtain feature data.
And carrying out characteristic processing on the information in the electronic medical record to obtain characteristic data conforming to the model input habit. The feature data in the application comprises a medical record feature book generated from the electronic medical record and expert indication data not derived from the electronic medical record. Specifically, the characteristic data is obtained by the following method:
Firstly, extracting characteristics of the content of the electronic medical record based on a machine learning algorithm to obtain case characteristic data. The machine learning algorithm is mainly used for learning and predicting diseases from data such as electronic medical records and dialogues of doctors and patients, and comprises various models such as classification, regression, clustering and the like, and can automatically learn from the data and be applied to new cases.
And then or simultaneously, extracting the characteristics of the prior medical knowledge to obtain expert knowledge data. The expert knowledge data includes, but is not limited to, a knowledge representation based on a vector space model, a knowledge representation based on association rules, and a knowledge representation based on domain knowledge. Expert knowledge sources include medical literature, clinical guidelines, and the like. This knowledge is incorporated into the system by certain rules and reasoning methods to aid decisions and judgments in the diagnostic process.
Based on knowledge representation of the vector space model, it is directed to vector space model (Vector Space Model, VSM) to represent text content using mathematical identifiers. In the recommended decision technique, in order to make the unstructured text information describing the object computable, a vector space model is often used to make algebraic transformations. The more text vectorized representation method used is Term frequency-inverse document frequency (TF-IDF). The method considers that the importance of a word is positively correlated with the frequency with which it appears in a document and negatively correlated with the frequency with which it appears in the overall language. The TF-IDF may mine keywords in the document, applying it to mine correlations between the object to be recommended and features of the corpus.
In the knowledge representation based on association rules, a drug is defined as an "item", a drug set is defined as an "item set", and a full item set is noted as: i= { I 1,i2,...,im }, the "k-term set" refers to a set containing k drugs, for example { aspirin } is a 1-term set, { aspirin, ibuprofen } is a 2-term set. A patient's historic diagnosis and treatment sequence, denoted as a "business" T, may contain a plurality of drugs, representing a list of drugs given by the doctor during a visit by the patient,The association rule is to mine relevant information from a transaction set, expressed as x= > Y, whereAnd is also provided withIndexes such as co-occurrence probability, confidence, support, expected reliability, promotion and the like are used for evaluating the validity of the association rule.
The calculation process of these five association rule metrics is described in detail below. Co-occurrence probability is applied to Embedding representation of natural language processing at the earliest, and the co-occurrence situation between words is counted, so that the context semantic expression is enhanced. Assuming that a represents a word and b represents a context, the co-occurrence probability is calculated as the ratio of the number of occurrences of a in the context of b to the number of occurrences of any one a in the context of b. In natural language processing, context refers to other words that are no more than m (m.ltoreq.10) away from the current word.
The confidence level of the association rule a= > B indicates the probability that the item set B appears simultaneously when the item set a appears, i.e. the ratio of the number of elements in the intersection of the item set a and the item set B to the number of elements in the item set a is calculated. Confidence is used to evaluate the accuracy of the association rule, indicating how much certainty B will occur if a occurs. The support degree of the association rule a= > B represents the ratio of the number of elements in the intersection of the item set a and the item set B to the total transaction set element number count (T).
The support is used to evaluate the importance of the association rule, indicating the universality of the current rule among all rules. The desired confidence level of item set a is typically defined by the probability that item set a will appear, i.e., the number of elements of item set a is proportional to the number of elements of the total transaction set T. The improvement degree of the association rule A= > B is also used for describing the occurrence probability of B when a occurs, and the difference between the improvement degree and the confidence degree is that the improvement degree also introduces the occurrence probability of each of a and B. The improvement is the ratio of confidence to desired confidence, indicating whether this rule has a probability of promoting or weakening the occurrence of item set B, as shown in equation (7). The degree of improvement is used to evaluate the validity of the association rule, and when the degree of improvement is less than 1, it indicates that there is a negative correlation between a and b, and when the degree of improvement is greater than 1, it indicates that there is a positive correlation between a and b.
Knowledge representation based on domain knowledge, for example, for traditional Chinese medicine, domain knowledge of traditional Chinese medicine mainly includes three types: based on the birth date of the patient derived congenital constitution, based on the sex age of the patient derived numerical segmentation features, and based on standard-grade symptomatic similarity matching. The following is the logic of the knowledge generation characterization of three fields of traditional Chinese medicine.
A) Logic for deriving congenital constitution at birth date: the Chinese medicine theory considers that a disease of a person has certain association with the congenital constitution, and the sensitivity degree of the life individuals to the disease is different according to the constitution. Factors affecting the constitution of an individual include various factors such as age, sex, congenital, living environment, etc. There have been scholars in the interest of the relation between congenital constitutions and birth date. The invention uses the congenital constitution reckoning mechanism to reckon the characteristics of 'congenital weak viscera', 'main qi passenger qi', 'yin-yang five elements' and the like according to the birth date of an individual. The birth date is the congenital constitution of the user, and is mainly influenced by the objective living environment in time space. In addition, congenital constitutions change with the change of the acquired environment.
B) Gender age segmentation logic specific to the traditional Chinese medicine field: the theory of traditional Chinese medicine considers that living individuals have growth periods, and the physical manifestations and the disease conditions of individuals with different periods are greatly different. The influence of sex on the life cycle of individuals is described in Huangdi's Nei Jing, individuals with different sex ratios have different growth cycles, and the principle of ' men eight women seven ' is based on that the male individuals are one growth cycle for eight years and the female individuals are one growth cycle for seven years. The male and female ages are discretized by using the piecewise functions of 'male, eight, female and seven', respectively, so that the age characteristics under the view angle of traditional Chinese medicine are obtained.
C) Medical case symptom semantic similarity matching logic: traditional Chinese medicine is an experience science, and similar medical cases play an important auxiliary role in diagnosis and treatment processes. The doctor can quickly diagnose the illness state of the current patient according to the prescription of the existing medical records and gives guidance by combining individual differences.
And S3, processing the electronic medical record and the characteristic data based on the medical diagnosis model.
The electronic medical record and the characteristic data are processed based on a medical diagnosis model comprising a large model and a small model, and auxiliary diagnosis and treatment information of a patient is obtained. The small model herein refers to an improved transducer disease prediction network, and the large model refers to a large medical model that will be subjected to the supervised micro-debugging process, or referred to as a large language model. And processing the electronic medical record and the characteristic data to obtain auxiliary diagnosis and treatment information comprising basic prediction information and diagnosis analysis information. Specifically, the processing is realized by the following means:
In one aspect, the electronic medical record and the feature data are processed based on the improved transducer disease prediction network to obtain disease prediction information. The data of the network comprises 18625 data sets, 16 characteristics and 1 label, and in a disease prediction model, the label is a disease type, and the characteristics are shown in the following table:
on the other hand, the electronic medical record and the characteristic data are processed based on the large language model, so that diagnosis analysis information is obtained.
In still another aspect, the diagnosis analysis information can be obtained by processing the electronic medical record, the feature data and the basic prediction information based on the large language model on the basis of obtaining the disease prediction information.
The specific structure of the small model, i.e., the improved transducer disease prediction network, of the present application is shown in FIG. 2. The transducer is a deep learning model architecture for natural language processing, and is capable of capturing long-distance dependency relationships in data by processing input data through a self-attention mechanism (self-attention), and is suitable for processing sequence data such as text. Compared with the traditional cyclic neural network and convolutional neural network, the transducer model is excellent in processing tasks such as long text, translation, text generation and the like. It is the infrastructure of large language models such as GPT and BERT. One specific disease prediction scheme is as follows:
The large model provided by the application refers to a large language model (Large Language Model) obtained by performing supervised micro-debugging processing on a medical large model based on fine tuning data, and the medical large model is Baichuan-7B、Zi ya-LLaMA-13B-Pretrain-v1、ChatGLM-6B、LLaMA-7B、LLaMA-7B、Bloom-7B、Ziya-LLaMA-13B、bai chuan-7B、ChatGLM-6B、B loomz-7B、ChatGLM-6B、ChatGLM-6B、bai chuan、LLaMA、ChatGLM, B loomz and the like.
The fine tuning data here includes open source general data, medical consultation data, human feedback data, medical domain knowledge, and the like. Open source generic data is a publicly available dataset that can be used to train and refine models; medical consultation data refers to data related to medical consultation, possibly including medical history, symptom descriptions, etc.; human feedback data is an assessment or correction from a person that helps to improve the accuracy and reliability of the model. The medical domain knowledge includes, but is not limited to, modern medical knowledge including western medicine, and traditional medical knowledge including traditional Chinese medicine, mongolian medicine, tibetan medicine, and the like.
Western medicine data is derived from open source data of a medical large model, and comprises scene dialogue, knowledge questions and answers and the like. The situational dialogue data includes disease descriptions, etiology analysis, disease diagnosis, pathological diagnosis, drug dosage, drug administration advice, medical knowledge, prognosis evaluation, medical advice, preventive measures, treatment schemes, index interpretation and the like, and also multiple rounds of patient information reasoning and diagnosis data. The knowledge questions and answers originate from a medical database and cover 28 departments by providing specific medical consensus and clinical guideline text. The medical knowledge question and answer data comprises a knowledge question and answer based on clinical guidelines and medical consensus, a knowledge question and answer based on physician qualification questions, a true doctor-patient question and a knowledge question and answer based on a structured medical map.
As shown in fig. 3, this embodiment illustrates the operation of a large language model (Large Language Model). Wherein professional medical data is used as prompt to guide it to generate more accurate results; the extracting step is to extract key information such as diagnosis, symptoms, treatment methods and preventive measures from the input; "Input, I nstruction, output" is the process by which a large language model receives Input, follows instructions, and produces output. A and B in the large language model represent two main components or phases in the model. Q|K|V refers to queries (Query), keys (Key), and values (value), which are terms commonly used in natural language processing to describe how a model processes and understands textual information. The right dialog box shows a question and answer regarding stomach ache and eating apples. This suggests that the model can understand and respond to specific health problems and provide relevant advice.
From the above technical solution, it can be seen that the present embodiment provides an auxiliary diagnosis method based on a medical diagnosis model, where the method is applied to an electronic device, specifically, an electronic medical record is obtained, where the electronic medical record carries part or all of historical health data, a personal diagnosis record, medical inquiry 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; and processing the electronic medical record and the characteristic data based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of a patient to be diagnosed, wherein the medical diagnosis model comprises an improved transducer disease prediction network and a large-scale language model obtained by performing supervised micro-debugging processing on the large-scale medical model. According to the technical scheme, automatic data cleaning and standardization can be realized based on a natural language processing technology of a large language model, inconsistent data can be identified and corrected, and the integrity and accuracy of the data are improved. Therefore, errors of auxiliary diagnosis information obtained later can be avoided, and doctors can be helped to avoid making wrong diagnosis results.
The inventor of the present application obtains the reasoning results shown in fig. 4a to 4e after carrying out actual verification based on the common large models of chatglm-6b and Qwen-7b deployed by the GPU, and from the perspective of the results, the present application perfectly achieves the present application.
The application provides the following innovation points:
1) And a modeling prediction scheme of a large model linkage small model, namely a multi-level prediction model framework, and a linkage system of the large model and a plurality of small models, is provided. The large model is used for global feature extraction and preliminary prediction, and the small model is used for fine prediction in the subdivision field. The large model is responsible for comprehensively processing a large amount of heterogeneous data to generate a high-level characteristic representation, while the small model is focused on a specific disease or a specific diagnosis task, and the characteristics generated by the large model are utilized for further analysis. Firstly, in terms of data processing and feature extraction, a large model processes massive medical data by using a deep learning technology transducer to extract global features. On the basis of the characteristics generated by the large model, the small model is combined with professional knowledge in a specific field to carry out fine prediction.
Secondly, in the aspect of a model linkage mechanism, data transmission and linkage mechanisms between a large model and a small model ensure that information flows between different models efficiently. The preliminary prediction result of the large model is used for guiding the selection and adjustment of the small model, so that the accuracy and efficiency of overall prediction are improved. In addition, in the aspects of self-adaptive learning and optimization, a self-adaptive learning method is provided, so that a large model and a small model can be dynamically adjusted and optimized according to different diagnosis tasks and data characteristics.
Detailed description of training and updating mechanisms of the model ensures that the system promotes predictive performance in constant learning and improvement.
2) The auxiliary diagnosis of traditional Chinese medicine and western medicine can be supported, and the traditional Chinese medicine and the western medicine are linked. And simultaneously, the auxiliary system for diagnosis of traditional Chinese medicine and western medicine is supported, and diagnosis data under two medical systems can be processed and analyzed. The system can integrate four-diagnosis (inspection, smelling, asking and cutting) data of traditional Chinese medicine and modern medical data such as examination and assay of Western medicine. A method for combining the medical knowledge maps of traditional Chinese medicine and Western medicine is provided, and a comprehensive medical knowledge base is constructed. In the diagnosis process, the characteristics of traditional Chinese medicine and Western medicine are fused by using a large model, and unified diagnosis suggestions are generated. And the small model is used for carrying out the combined diagnosis of Chinese and Western medicine for specific diseases, so that the comprehensiveness and accuracy of diagnosis are improved.
3) The representation of knowledge in the medical field is integrated into data, and a method for representing the medical knowledge is provided, wherein the method comprises systematic representation of knowledge such as theories, diagnosis standards, treatment schemes and the like of traditional Chinese medicine and Western medicine. By embedding medical knowledge into the data representation, the diagnostic model is enabled to understand and make use of expertise for prediction. Based on the knowledge-driven diagnostic model, intelligent diagnosis is performed in combination with medical knowledge and patient data.
The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
Although operations are depicted in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order. In certain circumstances, multitasking and parallel processing may be advantageous.
It should be understood that the various steps recited in the method embodiments of the present disclosure may be performed in a different order and/or performed in parallel. Furthermore, method embodiments may include additional steps and/or omit performing the illustrated steps. The scope of the present disclosure is not limited in this respect.
Computer program code for carrying out operations of the present disclosure may be written in one or more programming languages, including, but not limited to, an object oriented programming language such as Java, smalltalk, C ++ and conventional procedural programming languages, such as the C-language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of remote computers, the remote computer may be connected to the user computer through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external computer.
Fig. 5 is a block diagram of an auxiliary diagnostic apparatus based on a medical diagnostic model according to an embodiment of the present application.
As shown in fig. 5, the auxiliary diagnostic apparatus provided by the present application is applied to an electronic device for performing auxiliary diagnosis based on a medical diagnostic model including the above-described large model and small model. The electronic device can be understood as a computer, server or cloud platform having data computing and information processing capabilities, and the auxiliary diagnostic device includes a medical record acquisition module 10, a feature extraction module 20, and a data processing module 30.
The case acquisition module is used for acquiring the electronic medical record.
The electronic medical record refers to part or all of historical health data, personal diagnosis and treatment records, medical inquiry data and current physical examination information of a patient, which are generated based on data input in a manual mode, and the electronic medical record can be automatically generated based on a certain electronic platform. The current physical examination information refers to physical information which is acquired recently or currently through a certain technical means, such as medical images, ultrasonic detection results, blood routine data and the like.
The feature extraction module is used for extracting features of the electronic medical record to obtain feature data.
And carrying out characteristic processing on the information in the electronic medical record to obtain characteristic data conforming to the model input habit. The feature data in the application comprises a medical record feature book generated from the electronic medical record and expert indication data not derived from the electronic medical record. Specifically, the characteristic data is obtained by the following method:
Firstly, extracting characteristics of the content of the electronic medical record based on a machine learning algorithm to obtain case characteristic data. The machine learning algorithm is mainly used for learning and predicting diseases from data such as electronic medical records and dialogues of doctors and patients, and comprises various models such as classification, regression, clustering and the like, and can automatically learn from the data and be applied to new cases.
And then or simultaneously, extracting the characteristics of the prior medical knowledge to obtain expert knowledge data. The expert knowledge data includes, but is not limited to, a knowledge representation based on a vector space model, a knowledge representation based on association rules, and a knowledge representation based on domain knowledge. Expert knowledge sources include medical literature, clinical guidelines, and the like. This knowledge is incorporated into the system by certain rules and reasoning methods to aid decisions and judgments in the diagnostic process.
The data processing module is used for processing the electronic medical record and the characteristic data based on the medical diagnosis model.
The electronic medical record and the characteristic data are processed based on a medical diagnosis model comprising a large model and a small model, and auxiliary diagnosis and treatment information of a patient is obtained. The small model herein refers to an improved transducer disease prediction network, and the large model refers to a large medical model that will be subjected to the supervised micro-debugging process, or referred to as a large language model. And processing the electronic medical record and the characteristic data to obtain auxiliary diagnosis and treatment information comprising basic prediction information and diagnosis analysis information. Specifically, the processing is realized by the following means:
In one aspect, the electronic medical record and the feature data are processed based on the improved transducer disease prediction network to obtain disease prediction information. The data of the network comprises 18625 data sets, 16 characteristics and 1 label, and in a disease prediction model, the label is a disease type, and the characteristics are shown in the following table:
on the other hand, the electronic medical record and the characteristic data are processed based on the large language model, so that diagnosis analysis information is obtained.
In still another aspect, the diagnosis analysis information can be obtained by processing the electronic medical record, the feature data and the basic prediction information based on the large language model on the basis of obtaining the disease prediction information.
The specific structure of the small model, i.e., the improved transducer disease prediction network, of the present application is shown in FIG. 2. The transducer is a deep learning model architecture for natural language processing, and is capable of capturing long-distance dependency relationships in data by processing input data through a self-attention mechanism (self-attention), and is suitable for processing sequence data such as text. Compared with the traditional cyclic neural network and convolutional neural network, the transducer model is excellent in processing tasks such as long text, translation, text generation and the like. It is the infrastructure of large language models such as GPT and BERT.
The large model provided by the application refers to a large language model (Large Language Model) obtained by performing supervised micro-debugging processing on a medical large model based on fine tuning data, and the medical large model is Baichuan-7B、Zi ya-LLaMA-13B-Pretrain-v1、ChatGLM-6B、LLaMA-7B、LLaMA-7B、Bloom-7B、Ziya-LLaMA-13B、bai chuan-7B、ChatGLM-6B、B loomz-7B、ChatGLM-6B、ChatGLM-6B、bai chuan、LLaMA、ChatGLM, B loomz and the like.
The fine tuning data here includes open source general data, medical consultation data, human feedback data, medical domain knowledge, and the like. Open source generic data is a publicly available dataset that can be used to train and refine models; medical consultation data refers to data related to medical consultation, possibly including medical history, symptom descriptions, etc.; human feedback data is an assessment or correction from a person that helps to improve the accuracy and reliability of the model. The medical domain knowledge includes, but is not limited to, modern medical knowledge including western medicine, and traditional medical knowledge including traditional Chinese medicine, mongolian medicine, tibetan medicine, and the like.
Western medicine data is derived from open source data of a medical large model, and comprises scene dialogue, knowledge questions and answers and the like. The situational dialogue data includes disease descriptions, etiology analysis, disease diagnosis, pathological diagnosis, drug dosage, drug administration advice, medical knowledge, prognosis evaluation, medical advice, preventive measures, treatment schemes, index interpretation and the like, and also multiple rounds of patient information reasoning and diagnosis data. The knowledge questions and answers originate from a medical database and cover 28 departments by providing specific medical consensus and clinical guideline text. The medical knowledge question and answer data comprises a knowledge question and answer based on clinical guidelines and medical consensus, a knowledge question and answer based on physician qualification questions, a true doctor-patient question and a knowledge question and answer based on a structured medical map.
As shown in fig. 3, this embodiment illustrates the operation of a large language model (Large Language Model). Wherein professional medical data is used as prompt to guide it to generate more accurate results; the extracting step is to extract key information such as diagnosis, symptoms, treatment methods and preventive measures from the input; "Input, I nstruction, output" is the process by which a large language model receives Input, follows instructions, and produces output. A and B in the large language model represent two main components or phases in the model. Q|K|V refers to queries (Query), keys (Key), and values (value), which are terms commonly used in natural language processing to describe how a model processes and understands textual information. The right dialog box shows a question and answer regarding stomach ache and eating apples. This suggests that the model can understand and respond to specific health problems and provide relevant advice.
From the above technical solution, it can be seen that the present embodiment provides an auxiliary diagnosis device based on a medical diagnosis model, where the device is applied to an electronic apparatus, specifically, obtains an electronic medical record, where the electronic medical record carries part or all of historical health data, a personal diagnosis record, medical inquiry 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; and processing the electronic medical record and the characteristic data based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of a patient to be diagnosed, wherein the medical diagnosis model comprises an improved transducer disease prediction network and a large-scale language model obtained by performing supervised micro-debugging processing on the large-scale medical model. According to the technical scheme, automatic data cleaning and standardization can be realized based on a natural language processing technology of a large language model, inconsistent data can be identified and corrected, and the integrity and accuracy of the data are improved. Therefore, errors of auxiliary diagnosis information obtained later can be avoided, and doctors can be helped to avoid making wrong diagnosis results.
The units involved in the embodiments of the present disclosure may be implemented by means of software, or may be implemented by means of hardware. The name of the unit does not in any way constitute a limitation of the unit itself, for example the first acquisition unit may also be described as "unit acquiring at least two internet protocol addresses".
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: a Field Programmable Gate Array (FPGA), an Application Specific Integrated Circuit (ASIC), an Application Specific Standard Product (ASSP), a system on a chip (SOC), a Complex Programmable Logic Device (CPLD), and the like.
Fig. 6 is a block diagram of an electronic device according to an embodiment of the present application.
Referring now to fig. 6, a schematic diagram of an electronic device suitable for use in implementing embodiments of the present disclosure is shown. The terminal devices in the embodiments of the present disclosure may include, but are not limited to, mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and the like, and stationary terminals such as digital TVs, desktop computers, and the like. The electronic device is merely an example and should not impose any limitations on the functionality and scope of use of embodiments of the present disclosure.
The electronic device may comprise a processing means (e.g. a central processor, a graphics processor, etc.) 601, which may perform various suitable actions and processes according to programs stored in a read only memory ROM602 or loaded from an input means 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 by bus 604. An input/output (I/O) interface 605 is also connected to bus 604.
In general, the following devices may be connected to the I/O interface: input devices including, for example, touch screens, touch pads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; an output device 607 including, for example, a Liquid Crystal Display (LCD), a speaker, a vibrator, and the like; storage 608 including, for example, magnetic tape, hard disk, etc.; and a communication device 609. The communication means 609 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. While an electronic device having various means is shown in the figures, it is to be understood that not all of the illustrated means are required to be implemented or provided. More or fewer devices may be implemented or provided instead.
The application also provides a computer-readable storage medium embodiment.
The computer readable storage medium is applied to the electronic equipment and carries one or more computer programs, and when the one or more computer programs are executed by the electronic equipment, the electronic equipment acquires an electronic medical record, and the electronic medical record carries part or all of historical health data, personal diagnosis and treatment records, medical inquiry 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; and processing the electronic medical record and the characteristic data based on a medical diagnosis model to obtain auxiliary diagnosis and treatment information of a patient to be diagnosed, wherein the medical diagnosis model comprises an improved transducer disease prediction network and a large-scale language model obtained by performing supervised micro-debugging processing on the large-scale medical model. According to the technical scheme, automatic data cleaning and standardization can be realized based on a natural language processing technology of a large language model, inconsistent data can be identified and corrected, and the integrity and accuracy of the data are improved. Therefore, errors of auxiliary diagnosis information obtained later can be avoided, and doctors can be helped to avoid making wrong diagnosis results.
It should be noted that the computer readable medium described in the present disclosure may be a computer readable signal medium or a computer readable storage medium, or any combination of the two. The computer readable storage medium can be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples of the computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
In the context of this disclosure, a computer-readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present disclosure, however, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, with the computer-readable program code embodied therein. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination of the foregoing. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: electrical wires, fiber optic cables, RF (radio frequency), and the like, or any suitable combination of the foregoing.
In this specification, each embodiment is described in a progressive manner, and each embodiment is mainly described by differences from other embodiments, and identical and similar parts between the embodiments are all enough to be referred to each other.
While preferred embodiments of the present invention have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiment and all such alterations and modifications as fall within the scope of the embodiments of the invention.
Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or terminal. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or terminal device that comprises the element.
The foregoing has outlined rather broadly the more detailed description of the invention in order that the detailed description of the invention that follows may be better understood, and in order that the present principles and embodiments may be better understood; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present invention, the present description should not be construed as limiting the present invention in view of the above.
Claims (11)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411175260.2A CN118942682A (en) | 2024-08-26 | 2024-08-26 | Auxiliary diagnosis method, device, electronic device and storage medium based on medical diagnosis model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411175260.2A CN118942682A (en) | 2024-08-26 | 2024-08-26 | Auxiliary diagnosis method, device, electronic device and storage medium based on medical diagnosis model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN118942682A true CN118942682A (en) | 2024-11-12 |
Family
ID=93344653
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411175260.2A Pending CN118942682A (en) | 2024-08-26 | 2024-08-26 | Auxiliary diagnosis method, device, electronic device and storage medium based on medical diagnosis model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118942682A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119851965A (en) * | 2025-03-18 | 2025-04-18 | 国家超级计算天津中心 | Automatic iteration method, system and storage medium of medical diagnosis model |
-
2024
- 2024-08-26 CN CN202411175260.2A patent/CN118942682A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119851965A (en) * | 2025-03-18 | 2025-04-18 | 国家超级计算天津中心 | Automatic iteration method, system and storage medium of medical diagnosis model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11942221B2 (en) | Disambiguation of ambiguous portions of content for processing by automated systems | |
| US11182550B2 (en) | Cognitive building of medical condition base cartridges based on gradings of positional statements | |
| US11200968B2 (en) | Verifying medical conditions of patients in electronic medical records | |
| CN111801741B (en) | Adverse drug reaction analysis | |
| US10699215B2 (en) | Self-training of question answering system using question profiles | |
| JP6700283B2 (en) | A medical differential diagnosis device adapted to determine an optimum sequence of diagnostic tests for identifying a disease state by adopting a diagnostic validity standard | |
| US10971254B2 (en) | Medical condition independent engine for medical treatment recommendation system | |
| CN113015977A (en) | Deep learning based diagnosis and referral of diseases and conditions using natural language processing | |
| US20210390435A1 (en) | Analytics Framework for Selection and Execution of Analytics in a Distributed Environment | |
| US11495332B2 (en) | Automated prediction and answering of medical professional questions directed to patient based on EMR | |
| US10892056B2 (en) | Artificial intelligence based alert system | |
| US12106054B2 (en) | Multi case-based reasoning by syntactic-semantic alignment and discourse analysis | |
| US11847411B2 (en) | Obtaining supported decision trees from text for medical health applications | |
| Shen et al. | Enhancing ontology-driven diagnostic reasoning with a symptom-dependency-aware Naïve Bayes classifier | |
| US11990216B2 (en) | Detecting discrepancies between clinical notes and administrative records | |
| US20180211008A1 (en) | Assist Selection of Provider/Facility for Surgical Procedures Based on Frequency of Procedure, History of Complications, and Cost | |
| CN114139546A (en) | Formulating problems using differences between sets of syntax trees and differences between sets of semantic trees | |
| US11170895B2 (en) | Olfactory cognitive diagnosis | |
| CN118571495B (en) | Inner ear disease diagnosis model construction method, system, medium, product and terminal | |
| Perkins et al. | Improving clinical Documentation with artificial intelligence: A systematic review | |
| CN118841158A (en) | Treatment scheme recommendation method, treatment scheme recommendation device, computer equipment and storage medium | |
| CN118942682A (en) | Auxiliary diagnosis method, device, electronic device and storage medium based on medical diagnosis model | |
| Sequí-Sabater et al. | Artificial intelligence in rheumatology research: what is it good for? | |
| Kim et al. | Automatic diagnosis of medical conditions using deep learning with Symptom2VEC | |
| Pillai et al. | Improving postsurgical fall detection for older Americans using LLM-driven analysis of clinical narratives |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |