Medical data processing method and system, medium and equipment

Abstract

The invention provides a medical data processing method and system, a medium and equipment, and the method comprises the steps: carrying out the entity recognition of medical text information, and obtaining a named entity; mapping the named entity and the standardized terms in the standardized term library by using a preset rule library to obtain a mapping result; obtaining an auditing result for auditing the mapping result; and training a neural network model by using the auditing result and the mapping result, extracting a mapping rule from the trained neural network model by using a rule extraction algorithm, and updating the extracted mapping rule to a rule base. The machine learning algorithm is used for learning and optimizing the mapping result of the rule library, the term mapping rule is continuously updated to update the standardized term library, new term changes can be adapted, and the flexibility and adaptability of term standardization are improved.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a medical data processing method, system, medium and equipment. Background Art

[0002] Medical terminology refers to professional terms in the medical field, which can be used to represent various things, phenomena and processes such as diseases, drugs, surgical operations and inspections. With the rapid development of medical science and technology, new diseases, new treatments and emerging medical technologies are constantly emerging, the number of medical terms has increased dramatically, and the forms of expression are becoming increasingly diverse. Since medical terminology is indispensable in expressing medical information in clinical information systems, and each hospital has differences in actual use, this will cause the same meaning of medical terminology to be expressed in a variety of ways, which will bring great difficulties to subsequent medical record data analysis and other work. Therefore, it is very necessary to standardize medical terminology.

[0003] At present, the standardization of medical terminology mainly relies on static rule sets, which are usually formulated by professional medical organizations or standardization agencies and contain a series of guidelines on term definition, usage specifications and classification. Static rule sets did provide strong support for the standardized use of medical terminology in the early stage, but with the explosive growth of medical knowledge and the complex changes in terminology, its limitations such as poor adaptability, lack of flexibility and limited accuracy gradually emerged, making it difficult to adapt to the rapid development and diversified expression of medical terminology. Summary of the invention

[0004] In view of the defects in the prior art, an object of the present invention is to provide a medical data processing method, system, medium and device.

[0005] According to a medical data processing method provided by the present invention, the method includes: performing entity recognition on medical text information to obtain named entities; using a preset rule library to map the named entities with standardized terms to obtain mapping results; obtaining audit results of the mapping results; training a neural network model with the audit results and the mapping results, extracting mapping rules from the trained neural network model using a rule extraction algorithm, and updating the extracted mapping rules to the rule library.

[0006] It is not difficult to understand that this application obtains named entities in medical texts through entity recognition, and then maps the recognized named entities to standardized terms, thus achieving preliminary standardization of named entities. Then, a machine learning algorithm is used to learn and optimize the mapping results of the rule base, and the term mapping rules are continuously updated to update the standardized term base, which can adapt to new term changes and improve the flexibility and adaptability of term standardization.

[0007] Furthermore, the entity recognition of the medical text information to obtain the named entities includes: converting each character of the medical text information into a word vector and a word vector to which each character belongs, and splicing the word vector and the word vector to obtain a spliced ​​vector; determining a spliced ​​feature vector by performing feature extraction on the spliced ​​vector, and obtaining a probability that each character corresponds to a different label based on the spliced ​​feature vector; determining a label sequence of the medical text information based on the probability that each character corresponds to a different label; and determining the named entities in the medical text information based on the label sequence of the medical text information.

[0008] Furthermore, the method of using a preset rule base to map the named entity with the standardized terms in the standardized terminology library includes at least one of the following: calculating the string similarity between the named entity and the medical term in the standard terminology library, and if it is less than a first preset value, mapping the named entity to the medical term; calculating the similarity between the named entity and the keywords of the medical term in the standard terminology library, and if the similarity meets a second preset value, mapping the named entity to the keywords of the medical term in the standard terminology library; comparing the named entity with the medical vocabulary, and if there is a corresponding standard term in the medical vocabulary, mapping the named entity to the corresponding standard term.

[0009] Furthermore, the calculation formula of the string similarity rule is as follows:

[0010]

[0011] Wherein, S(X,Y) represents the distance between the named entity and the medical term in the standard terminology library, X\Y represents the number of characters of the medical term (Y) in the named entity (X) but not in the standard terminology library, Y\X represents the number of characters of the medical term (Y) in the standard terminology library but not in the named entity (X), X∩Y represents the number of characters of the medical term (Y) in both the named entity (X) and the standard terminology library, and α and β are two parameter coefficients.

[0012] It is not difficult to understand that the present application constructs a rule set including synonyms, abbreviations and standardized expressions, uses a rule base including string similarity rules, keyword mapping rules and vocabulary rules to map the identified terms with the standardized terminology library, and uses a string similarity algorithm (such as Levenshtein distance, Jaccard similarity, etc.) to calculate the similarity between the terms to be identified and the terms in the standardized terminology library; by setting a suitable threshold, the terms to be identified with a high similarity to the standardized terms can be identified, thereby improving the recognition accuracy.

[0013] Furthermore, the method for training the neural network model includes: dividing the mapping results and the audit results into a training set and a verification set, training the neural network model through the training set, verifying the trained neural network model through the verification set, and adjusting the parameters of the neural network model according to the verification results.

[0014] Furthermore, the rule extraction algorithm includes a decision tree algorithm.

[0015] Furthermore, the decision tree algorithm includes: constructing a decision tree according to the audit results and the mapping results, taking the output of the neural network model as the training target of the decision tree, and obtaining a trained decision tree; starting from the root node of the decision tree, recursively traversing each subtree of the decision tree, and obtaining the mapping rules by connecting the conditional judgments of the subtrees.

[0016] It is not difficult to understand that this application integrates professional, accurate and efficient screening of manual review into the process of medical terminology standardization, which not only greatly improves the standardization accuracy of terminology, ensures the consistency and standardization of terminology use, but also significantly optimizes the circulation efficiency and understanding depth of medical information. This process effectively reduces communication barriers and medical risks caused by terminology differences, promotes the steady improvement of medical quality and higher patient satisfaction, and lays a solid foundation for the progress and development of the medical field.

[0017] A medical data processing system provided according to the present invention comprises:

[0018] The term recognition module is used to perform entity recognition on medical text information and obtain named entities;

[0019] A term mapping module is used to map the named entity with the standardized term in the standardized term base using a preset rule base to obtain a mapping result;

[0020] A feedback module, used to obtain an audit result of the mapping result;

[0021] The learning optimization module uses the audit results and mapping results to train the neural network model, uses the rule extraction algorithm to extract the mapping rules from the trained neural network model, and updates the extracted mapping rules to the rule base.

[0022] According to a computer-readable storage medium storing a computer program provided by the present invention, the steps of the medical data processing method are implemented when the computer program is executed by a processor.

[0023] An electronic device provided according to the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the steps of the medical data processing method are implemented when the computer program is executed by the processor.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] The present invention innovatively integrates machine learning algorithms into the rule base, and realizes the dynamic update of term mapping rules and the instant expansion of the standardized term base by continuously learning and optimizing the term mapping results. This mechanism can not only keenly capture and adapt to the latest changes in terminology, but also significantly enhance the flexibility and adaptability of term standardization, making a great contribution to medical communication and scientific research exploration.

[0026] Other beneficial effects of the present invention will be explained in the specific implementation manner through the introduction of specific technical features and technical solutions. Through the introduction of these technical features and technical solutions, those skilled in the art should be able to understand the beneficial technical effects brought about by the technical features and technical solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0028] Figure 1 The figure is a flow chart of the method of the present invention.

[0029] Figure 2 It is a schematic diagram of the process of updating the rule base in the present invention.

[0030] Figure 3 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0031] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several changes and improvements can also be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0032] The present application provides a medical data processing method, which specifically includes performing entity recognition on medical text information to obtain named entities; using a preset rule library to map the named entities with standardized terms in a standardized term library to obtain mapping results; reviewing the mapping results to obtain review results; training a neural network model with the review results and the mapping results, extracting mapping rules from the trained neural network model, and updating the extracted mapping rules to the rule library.

[0033] It is not difficult to understand that this application uses a deep learning model to identify medical text information to obtain named entities, maps the identified named entities with standardized terms in the standardized terminology library, inputs the terms that have not been successfully identified or mapped into the neural network model for training through manual review, and finally updates the rule base based on the trained neural network model.

[0034] Reference Figure 1 and Figure 2 As shown, a medical data processing method comprises:

[0035] Step S1: Perform entity recognition on medical text information to obtain named entities.

[0036] These medical text information can come from public medical literature, case reports, electronic health records (EHR), etc. In this application, a deep learning model can be used for recognition, and the deep learning model is trained by the entity data set corresponding to the medical professional dictionary, so that the deep learning model can be more accurate in identifying named entities in medical text information.

[0037] In an embodiment of the present application, the deep learning model may be BioBERT or PubMedBERT. Taking BioBERT as an example, although BioBERT is a pre-trained language representation model for the biomedical field, the named entities contained in the latest medical text information are not recorded in the standardized terminology library and are difficult to be accurately identified, so BioBERT needs to be pre-trained.

[0038] Specifically, pre-training can be to first initialize BioBERT using BERT's weights. The BERT needs to be pre-trained on a general domain corpus to learn the ability to recognize general terms. Then BioBERT is pre-trained on a biomedical domain corpus to learn the ability to recognize biomedical domain terms. In this way, the ability to recognize the latest named entities is obtained.

[0039] Specifically, the deep learning model may include: an embedding layer, a feature extraction layer, and a fully connected layer; the embedding layer is used to convert each character of the medical document into a word vector and a word vector to which each character belongs. The feature extraction layer is used to concatenate the word vector and the word vector to obtain a concatenated vector. The fully connected layer is used to process the concatenated vector through an activation function to obtain the probability of different labels in the sequence annotation corresponding to each character, thereby obtaining a label sequence of the medical document, and determining the named entity in the medical document based on the label sequence of the medical document.

[0040] Step S2: Use a preset rule base to map the named entity with the standardized terms in the standardized term base to obtain a mapping result.

[0041] Step S201: Use one or more rules in a preset rule base to perform term mapping.

[0042] Specifically, the preset rule base includes: string similarity rules, keyword mapping rules and / or vocabulary rules.

[0043] In one embodiment, string similarity rules are used for mapping. The string similarity rules are based on the edit distance between two names for mapping identification; if the edit distance between two names is less than a predefined threshold, they can be considered to represent the same medical term, as shown below:

[0044]

[0045] Where S(X,Y) represents the distance between the named entity and the medical term in the standardized terminology, X\Y represents the number of characters in the medical term (Y) in the named entity (X) that are not in the standardized terminology, Y\X represents the number of characters in the medical term (Y) in the standardized terminology that are not in the named entity (X), X∩Y represents the number of characters in the medical term (Y) in both the named entity (X) and the standardized terminology, and α and β are two parameter coefficients.

[0046] Assume that the named entity identified from the medical text information is "amoxicillin", and the standard term in the standardized terminology library is "amoxicillin".

[0047] According to the above formula, X\Y=1, Y\X=1, X∩Y=3; set α and β to 0.5, then:

[0048] S(X, Y)=3 / (3+1)=0.75;

[0049] If they are two exactly the same terms, then X\Y and Y\X are both 0, and the calculation result is 1, which means that they are exactly the same when the distance is 1, and the closer to 1, the more they have the same meaning.

[0050] The keyword mapping rules are based on whether the keywords contained in the identified terms are identical or similar to the keywords contained in the terms in the standardized term base. The named entity "Amoxicillin" is identified from the medical text information and attempts to map it with the terms in the standardized term base. During the mapping process, the system analyzes the term "Amoxicillin" and searches for the term "Amoxicillin" in the mapped standardized term base that contains the same or similar keywords. The identified terms are related to +

[0051] The terms in the standardized terminology base contain the same keywords and can be considered to represent the same medical term, that is, the identified terms are mapped to the terms in the standardized terminology base.

[0052] The vocabulary rule determines whether two names represent the same medical term based on whether there is a corresponding standard term in the medical vocabulary. The named entity "Amoxicillin" is identified from the medical text information, and the system then queries the medical vocabulary, maps the queried "Amoxicillin" to it, and outputs it as a standardized result.

[0053] Step S202: output mapping results.

[0054] The mapping results include mapped results and unmapped results. The mapped results may include erroneous mapping results, and the unmapped results may also include omitted mapping results, so they need to be reviewed through step S3.

[0055] Step S3: Obtain the audit result of the mapping result.

[0056] Specifically, the accuracy of the mapping result in step S2 can be determined by manual review. The review result can be: modifying the mapped result and remapping the unmapped result to obtain a new mapped result. The method for reviewing the mapping result includes:

[0057] Step S301: Display the mapping result of step S2 through a user interaction interface for review by professionals.

[0058] Step S302: obtaining the audit result through the user interaction interface, the audit result including: for the mapped result, obtaining the input revised mapping information; for the unmapped result, obtaining the input new mapping information.

[0059] The modified mapping information is information or instructions for replacing a standardized term in a mapped result with another standardized term. According to the information or instructions, the corresponding mapped result can be updated.

[0060] The newly created mapping information may be information or instructions for creating a new standardized term in the standardized term base and mapping the named entity to the new standardized term, or information or instructions for directly mapping the named entity to an existing standardized term in the standardized term base. Based on the information or instructions, a new mapped result may be obtained.

[0061] Step S303: According to the revised mapping information and the newly created mapping information, the corresponding named entity is mapped to the corresponding standardized term in the standardized term library to obtain the audit result.

[0062] Step S4: train the neural network model with the audit results and the mapping results, extract the mapping rules from the trained neural network model using the rule extraction algorithm, and update the extracted mapping rules to the rule base.

[0063] Specifically, the audit results and the mapped results in the mapping results are divided into a training set and a verification set. The training set is used to train the neural network model, and the verification set is used to verify the trained neural network model. The parameters of the neural network model are adjusted according to the verification results until the accuracy of the neural network model in mapping professional terms reaches the preset requirements. Then, the mapping rules learned by the neural network model are extracted from the neural network model through the rule extraction algorithm and stored in the rule base. Thereafter, when the neural network model obtains new mapped results, it can learn to update its internal mapping rules. The mapping rules can be extracted from the neural network model by timed extraction and replace the historical mapping rules in the rule base.

[0064] The method for extracting mapping rules may be a decision tree algorithm. The decision tree algorithm constructs a decision tree based on the audit results and the mapping results, takes the output of the neural network model as the training target of the decision tree, and obtains a trained decision tree; starting from the root node of the decision tree, recursively traverses each subtree of the decision tree, and obtains the mapping rules by connecting the conditional judgments of the subtrees.

[0065] In summary, this application innovatively integrates machine learning algorithms into the rule base, and realizes the dynamic update of term mapping rules and the instant expansion of the standardized term base by continuously learning and optimizing the term mapping results; this mechanism can not only keenly capture and adapt to the latest changes in terminology, but also significantly enhance the flexibility and adaptability of term standardization, laying a more solid and intelligent foundation for medical communication and scientific research exploration.

[0066] like Figure 3 As shown, a medical data processing system is used to execute the above-mentioned medical data processing method. The system includes: a term recognition module, a term mapping module, a learning optimization module, a feedback module, a user interaction and display module, a system integration and coordination module and a monitoring module.

[0067] The term recognition module uses models such as BioBERT to obtain recognized text information and perform entity recognition, and transmits the recognized named entities to the term mapping module; the selected model should preferably be good at handling the medical field;

[0068] A term mapping module is used to receive the named entities identified by the term recognition module, and map the named entities with the standardized term library based on a preset rule library, and output the mapping results to the learning optimization module;

[0069] A feedback module, which is used to collect manual review results of the mapping results and transmit them to the learning optimization module;

[0070] The learning optimization module receives the mapping results of the term mapping module and the review results of the feedback module, divides the mapping results and the review results into a training set and a verification set, trains the neural network model through the training set, verifies the trained neural network model through the verification set, and adjusts the parameters of the neural network model according to the verification results; extracts mapping rules from the trained neural network model through the rule extraction algorithm, and stores the extracted mapping rules in the rule base to complete the update of the rule base.

[0071] The user interaction and display module is used to display structured data according to external requests. For example, analysis results can be displayed through charts and reports, and customized queries and visual displays can be supported to improve user experience.

[0072] The system integration and coordination module is used for interface standardization and data synchronization mechanism between modules, and performs integration testing and stress testing to ensure the coordinated work of various modules in the system and is responsible for transmitting the system's integrated data and feedback structure data.

[0073] The monitoring module is used to monitor and analyze the system's performance indicators and deploy the system. If a system anomaly is detected, an alarm message will be issued; at the same time, the system's operating data and maintenance logs will be collected.

[0074] It should be understood that the embodiments of the present invention can be implemented in hardware, software or a combination thereof. Software can be stored in a volatile or non-volatile storage device (regardless of whether the data is deletable or rewritable) like a ROM, in a memory like a RAM, a memory chip, a device or an integrated circuit, or in a storage medium on which data can be optically or magnetically recorded and from which data can be read by a machine (e.g., a computer), such as a compact disc (CD), a digital versatile disc (DVD), a disk, or a tape. Moreover, the embodiments of the present invention can be implemented in a computer or portable terminal having a controller and a memory, and the memory is an example of a non-temporary computer-readable storage medium suitable for storing one or more programs including commands for implementing the embodiments of the present invention. Therefore, the present invention includes a program having a code for implementing the device or method defined by the claims, and a machine-readable non-temporary storage medium storing the program. The program can be transferred electrically by a medium such as a communication signal transmitted via a wired or wireless connection, and the program and its equivalent are included in the present invention.

[0075] The device may receive a program from a program providing device connected via a cable or wirelessly and store the program. The program providing device may include a program containing commands for implementing the embodiments of the present invention, a memory for storing information required by the embodiments of the present invention, a communication module for communicating with the mobile device via a cable or wirelessly, and a controller for transmitting the program to the device automatically or upon request of the device.

[0076] Those skilled in the art know that, in addition to realizing the system and its various devices, modules, and units provided by the present invention in a purely computer-readable program code, it is entirely possible to realize the same functions in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers by logically programming the method steps. Therefore, the system and its various devices, modules, and units provided by the present invention can be considered as a hardware component, and the devices, modules, and units included therein for realizing various functions can also be regarded as structures within the hardware component; the devices, modules, and units for realizing various functions can also be regarded as both software modules for realizing the method and structures within the hardware component.

[0077] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. In the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A medical data processing method, characterized in that: include: Perform entity recognition on medical text information to obtain named entities; Mapping named entities with standardized terms in a standardized term base using a preset rule base to obtain a mapping result; Obtaining an audit result of the mapping result; The neural network model is trained with the audit results and mapping results, the mapping rules are extracted from the trained neural network model using a rule extraction algorithm, and the extracted mapping rules are updated to the rule base.

2. The medical data processing method according to claim 1, characterized in that: The entity recognition of medical text information to obtain named entities includes: Convert each character of the medical text information into a word vector and a word vector to which each character belongs, and concatenate the word vector and the word vector to obtain a concatenated vector; Determine a splicing feature vector by performing feature extraction on the splicing vector, and obtain the probability that each character corresponds to a different label according to the splicing feature vector; Determining a label sequence of the medical text information according to the probability that each character corresponds to different labels; A named entity in the medical text information is determined based on the tag sequence of the medical text information.

3. The medical data processing method according to claim 1, characterized in that: The method of mapping the named entity with the standardized term in the standardized term base using the preset rule base includes at least one of the following: Calculating the string similarity between the named entity and the medical term in the standard term library, and if the string similarity is less than a first preset value, mapping the named entity to the medical term; Calculating the similarity between the named entity and the keywords of the medical term in the standard term library, and if the similarity satisfies a second preset value, mapping the named entity to the keywords of the medical term in the standard term library; The named entity is compared with a medical vocabulary, and if a corresponding standard term exists in the medical vocabulary, the named entity is mapped to the corresponding standard term.

4. The medical data processing method according to claim 3, characterized in that: The calculation formula for calculating the string similarity between the named entity and the medical term in the standard term library is as follows: Wherein, S(X,Y) represents the distance between the named entity and the medical term in the standard terminology library, X\Y represents the number of characters of the medical term (Y) in the named entity (X) but not in the standard terminology library, Y\X represents the number of characters of the medical term (Y) in the standard terminology library but not in the named entity (X), X∩Y represents the number of characters of the medical term (Y) in both the named entity (X) and the standard terminology library, and α and β are two parameter coefficients.

5. The medical data processing method according to claim 1, characterized in that: Methods for training neural network models include: The mapping results and audit results are divided into a training set and a validation set. The neural network model is trained with the training set, and the trained neural network model is validated with the validation set. The parameters of the neural network model are adjusted according to the validation results.

6. The medical data processing method according to claim 1, characterized in that: The rule extraction algorithm includes a decision tree algorithm.

7. The medical data processing method according to claim 6, characterized in that: The decision tree algorithm includes: A decision tree is constructed based on the audit results and mapping results, and the output of the neural network model is used as the training target of the decision tree to obtain a trained decision tree; Starting from the root node of the decision tree, recursively traverse each subtree of the decision tree, and obtain the mapping rules by connecting the conditional judgments of the subtrees.

8. A medical data processing system, characterized in that: include: The term recognition module is used to perform entity recognition on medical text information and obtain named entities; A term mapping module is used to map the named entity with the standardized term in the standardized term base using a preset rule base to obtain a mapping result; A feedback module, used to obtain an audit result of the mapping result; The learning optimization module uses the audit results and mapping results to train the neural network model, uses the rule extraction algorithm to extract the mapping rules from the trained neural network model, and updates the extracted mapping rules to the rule base.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the medical data processing method according to any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the computer program is executed by a processor, the steps of the medical data processing method according to any one of claims 1 to 7 are implemented.

Images