Language optimization method and device for medical image diagnosis report and medium
By conducting grammar examination, professional term verification and expression consistency examination of medical imaging diagnostic reports, combined with natural language processing technology and medical knowledge base, and using BERT model for text optimization, the problem of irregular reporting language is solved and the ability of reporting is improved Readability and accuracy save doctors' review time.
Patent Information
- Application Number
- CN202411459464.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-18
- Publication Date
- 2025-02-14
AI Technical Summary
The irregular language in medical imaging diagnostic reports leads to grammatical errors, improper use of professional terms and inconsistent expression, affecting the readability, accuracy and professionalism of the report, and causing troubles to follow-up medical decision-making and medical research.
By collecting text of medical imaging diagnostic reports, preprocessing and grammar examinations are performed, combining medical imaging diagnostic knowledge base and historical reports, the natural language processing library is used to perform part-of-speech annotation, professional term verification and expression consistency examinations, and finally using the Transformers model The BERT model is text normalized and optimized.
Improve the readability, accuracy and standardization of medical imaging diagnostic reports, reduce misunderstandings and error diagnosis caused by language problems, provide a reliable data foundation for subsequent data statistical analysis and medical research, and save doctors' review reports time improves work efficiency.
Smart Images
Figure CN119443042A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of artificial intelligence, and in particular to a method, device and medium for language optimization of medical imaging diagnostic reports. Background Art
[0002] Medical imaging diagnosis is a key link in the field of modern medicine, and its diagnostic report is the core carrier of medical information transmission and communication. However, due to the combined influence of factors such as doctors' high workload, time pressure, differences in personal language habits, and uneven professional levels, the phenomenon of non-standard language in diagnostic reports is common. Specifically, grammatical errors lead to unclear meaning, such as inconsistent subject and predicate, incomplete sentence components, etc.; improper use of professional terms, such as misuse, abuse or fabricated terms, weakens the professionalism and authority of the report; inconsistent expressions, the same disease is described differently in different reports, which brings great trouble to subsequent medical decision-making, peer communication and medical research.
[0003] At present, although some medical information systems have managed and standardized diagnostic reports to a certain extent, there are still obvious deficiencies in the depth and accuracy of language optimization and standardization. Most systems can only perform simple format checks and basic vocabulary corrections, lack in-depth understanding of the professional content and complex context unique to medical imaging diagnosis, and find it difficult to achieve comprehensive and accurate optimization and standardization of report language.
[0004] Therefore, how to optimize the language of medical imaging diagnostic reports more accurately is an issue that needs to be addressed urgently. Summary of the invention
[0005] The present application provides a language optimization method for medical imaging diagnosis reports. By performing grammar detection and expression consistency detection on the medical imaging diagnosis report, and verifying and optimizing the language of the report according to the detection results, a more accurate medical imaging diagnosis report can be obtained.
[0006] In order to achieve the above object, the present invention adopts the following technical scheme:
[0007] In a first aspect, the present invention provides a method for optimizing the language of a medical imaging diagnostic report, the method comprising:
[0008] Collecting the medical imaging diagnosis report to be processed, extracting the text in the medical imaging diagnosis report, and preprocessing the text;
[0009] Collecting a medical imaging diagnosis knowledge base, and using a pre-trained recurrent neural network in combination with the medical imaging diagnosis knowledge base to extract and verify professional terms in the text;
[0010] Using a first natural language processing library to perform part-of-speech tagging and grammar checking on the text to obtain a grammar checking result;
[0011] Collecting historical medical imaging diagnosis reports and standard templates, comparing the text with the description of the same diagnosis in the historical medical imaging diagnosis reports and standard templates based on text similarity, and obtaining a description inspection result;
[0012] The text is normalized and optimized using a second natural language processing library according to the grammar check result, the expression check result and the professional terminology.
[0013] In a preferred example of the present application, it can be further configured that the preprocessing of the text includes:
[0014] Using regular expressions to remove punctuation marks, numbers, and special characters from the text, converting the text into a unified encoding format to obtain a cleaned text;
[0015] The Python jieba library is used to segment the cleaned text.
[0016] In a preferred example of the present application, it can be further configured to use a pre-trained recurrent neural network combined with a medical imaging diagnosis knowledge base to extract and verify professional terms in the text, including:
[0017] Convert text into vectors using a pre-trained word vector model, such as Word2Vec or GloVe;
[0018] In combination with a medical imaging diagnosis knowledge base, a pre-trained recurrent neural network is used to identify and extract professional vocabulary and terminology in the vector, and the recurrent neural network includes an LSTM.
[0019] In a preferred example of the present application, it can be further configured that the text is grammatically checked using a first natural language processing library, including:
[0020] The Python NLTK library is used to check whether the statements in the text are consistent with the subject and predicate.
[0021] In a preferred example of the present application, it can be further configured to include:
[0022] Use CoreNLP Parser to perform grammatical analysis on the text to obtain the complexity of each sentence in the text;
[0023] Whether the sentence is too complex is determined by comparing a preset threshold with the complexity, and the determination result is added to the grammar check result.
[0024] In a preferred example of the present application, it can be further configured that the second natural language processing library is a BERT model based on the Transformers model.
[0025] In a second aspect, the present application provides a language optimization device for a medical imaging diagnostic report, the device comprising:
[0026] A data acquisition module is used to acquire the medical imaging diagnosis report to be processed, extract the text in the medical imaging diagnosis report, and pre-process the text;
[0027] A term acquisition module, used to acquire a medical imaging diagnosis knowledge base, and use a pre-trained recurrent neural network in combination with the medical imaging diagnosis knowledge base to extract and verify professional terms in the text;
[0028] A grammar checking module, used to perform part-of-speech tagging and grammar checking on the text using a first natural language processing library to obtain a grammar checking result;
[0029] A statement checking module is used to collect historical medical imaging diagnosis reports and standard templates, compare the text with the statements of the same diagnosis in the historical medical imaging diagnosis reports and standard templates based on text similarity, and obtain statement checking results;
[0030] A normalization module is used to normalize and optimize the text using a second natural language processing library according to the grammar check result, the expression check result and the professional terminology.
[0031] In a third aspect, the present application provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the language optimization method for medical imaging diagnostic reports as described in any one of the above items are implemented.
[0032] In a fourth aspect, the present application provides a computer-readable storage medium having a program stored thereon, wherein when the program is executed by a processor, the language optimization method for a medical imaging diagnostic report as described in any one of the above items is implemented.
[0033] In a fifth aspect, the present application provides a computer program product, comprising computer instructions, which, when executed by a processor, implement the steps of the language optimization method for medical imaging diagnostic reports as described in any one of the above items.
[0034] In summary, compared with the prior art, the technical solution provided in the embodiment of the present application has at least the following beneficial effects:
[0035] The present application provides a language optimization method for medical imaging diagnostic reports. On the one hand, by checking and optimizing the consistency of grammar, professional terms and expressions, the readability, accuracy and standardization of medical imaging diagnostic reports are improved, and misunderstandings and misdiagnoses caused by language problems in the reports are reduced. This provides a data basis for subsequent data statistical analysis and sharing, which is conducive to the development and progress of the medical field. On the other hand, this method can be applied to medical information systems, saving doctors time in reviewing reports and improving their work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 A flowchart of a method for language optimization of a medical imaging diagnostic report provided in one embodiment of the present application.
[0037] Figure 2 A structural diagram of a language optimization device for a medical imaging diagnostic report provided in one embodiment of the present application. DETAILED DESCRIPTION
[0038] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0039] In one embodiment of the present application, a language optimization of a medical imaging diagnosis report is provided. Figure 1 As shown, the method includes:
[0040] S100: Collecting a medical imaging diagnosis report to be processed, extracting text from the medical imaging diagnosis report, and preprocessing the text;
[0041] Specifically, the medical imaging diagnosis reports to be processed are collected from the medical information system, the text in the report is extracted, and then the text is preprocessed by cleaning, word segmentation, etc.
[0042] In a preferred embodiment, the preprocessing of the text includes:
[0043] Using regular expressions to remove punctuation marks, numbers, and special characters from the text, converting the text into a unified encoding format to obtain a cleaned text;
[0044] The Python jieba library is used to segment the cleaned text.
[0045] In this embodiment, the accuracy of subsequent grammar checking and expression consistency checking can be significantly improved by text preprocessing. The Python jieba library has excellent performance and flexibility, supports multiple word segmentation modes and algorithms, and can improve the accuracy and efficiency of word segmentation of the cleaned text.
[0046] S200: collecting a medical imaging diagnosis knowledge base, and using a pre-trained recurrent neural network in combination with the medical imaging diagnosis knowledge base to extract and verify professional terms in the text;
[0047] Specifically, the recurrent neural network may be a long short-term memory network (LSTM), a gated recurrent unit (GRU), etc. Before inputting into the recurrent neural network, the text is first vectorized.
[0048] In a preferred embodiment, a pre-trained recurrent neural network is used in combination with a medical imaging diagnosis knowledge base to extract and verify professional terms in the text, including:
[0049] Convert text into vectors using a pre-trained word vector model, such as Word2Vec or GloVe;
[0050] In combination with a medical imaging diagnosis knowledge base, a pre-trained recurrent neural network is used to identify and extract professional vocabulary and terminology in the vector, and the recurrent neural network includes an LSTM.
[0051] In specific implementation, the Word2Vec is a language model that learns semantic knowledge from a large amount of text in an unsupervised manner. The GloVe (Global Vectors for Word Representation) is a word embedding method that aims to represent words through word vectors so that these vectors can contain as much semantic and grammatical information as possible. The LSTM (Long Short-Term Memory) is a long short-term memory network, which is a time recurrent neural network.
[0052] S300: Use a first natural language processing library to perform part-of-speech tagging and grammar checking on the text to obtain a grammar checking result;
[0053] Specifically, the first natural language processing library may be a Python NLTK library.
[0054] In a preferred embodiment, the text is grammatically checked using a first natural language processing library, including:
[0055] The Python NLTK library is used to check whether the statements in the text are consistent with the subject and predicate.
[0056] S400: collecting historical medical imaging diagnosis reports and standard templates, and comparing the text with the description of the same diagnosis in the historical medical imaging diagnosis reports and standard templates based on text similarity to obtain a description inspection result;
[0057] Specifically, by comparing the text with historical reports and authoritative standard templates, it is checked whether the same diagnosis is expressed consistently in different reports. When using text similarity methods for comparison, cosine similarity, Jaccard similarity, etc. can be used to compare and analyze the same diagnosis expressions in different reports.
[0058] S500: using a second natural language processing library to normalize and optimize the text according to the grammar check result, the expression check result and the professional terminology.
[0059] Specifically, a pre-trained language model, i.e., the second natural language processing model, such as GPT-3 or BERT, is used to generate optimized text in combination with a fine-tuning strategy formulated according to the grammar check result, the expression check result, and the professional terminology. At the same time, rule-based methods and statistical models are used to standardize and unify terms and expressions.
[0060] In a preferred embodiment, the second natural language processing library is a BERT model based on the Transformers model. The BERT (Bidirectional Encoder Representations from Transformers) model is a bidirectional, context-based embedding model. In this embodiment, the BERT model has a significant effect on the relational feature extraction between the data characters of the text and its performance improvement.
[0061] In this embodiment, on the one hand, by checking and optimizing the consistency of grammar, professional terms and expressions, the readability, accuracy and standardization of medical imaging diagnosis reports are improved, misunderstandings and misdiagnoses caused by language problems in the reports are reduced, and a data basis is provided for subsequent data statistical analysis and sharing, which is conducive to the development and progress of the medical field. On the other hand, this method can be applied to medical information systems, saving doctors time in reviewing reports and improving their work efficiency.
[0062] In some embodiments, it also includes:
[0063] Use CoreNLP Parser to perform grammatical analysis on the text to obtain the complexity of each sentence in the text;
[0064] Whether the sentence is too complex is determined by comparing a preset threshold with the complexity, and the determination result is added to the grammar check result.
[0065] In specific implementation, the CoreNLP Parser is a natural language analysis library that integrates all natural language processing tools and can judge the complexity of sentences.
[0066] In this embodiment, the complexity of the sentence is determined by CoreNLP Parser, which can improve the accuracy of the grammar check result and avoid misjudgment caused by overly complex sentences.
[0067] This application also provides a language optimization device for medical imaging diagnostic reports, see Figure 2 As shown, the device comprises:
[0068] The data acquisition module 100 is used to acquire the medical imaging diagnosis report to be processed, extract the text in the medical imaging diagnosis report, and pre-process the text;
[0069] A term acquisition module 200 is used to acquire a medical imaging diagnosis knowledge base, and use a pre-trained recurrent neural network in combination with the medical imaging diagnosis knowledge base to extract and verify professional terms in the text;
[0070] A grammar checking module 300 is used to perform part-of-speech tagging and grammar checking on the text using a first natural language processing library to obtain a grammar checking result;
[0071] The expression checking module 400 is used to collect historical medical imaging diagnosis reports and standard templates, compare the text with the expressions of the same diagnosis in the historical medical imaging diagnosis reports and standard templates based on text similarity, and obtain expression checking results;
[0072] The normalization module 500 is used to normalize and optimize the text using a second natural language processing library according to the grammar check result, the expression check result and the professional terminology.
[0073] The functional implementation of each module in the above-mentioned language optimization device for medical image diagnostic reports corresponds to the steps in the above-mentioned embodiment of the language optimization method for medical image diagnostic reports, and its functions and implementation processes will not be repeated here one by one.
[0074] The present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the language optimization method for medical imaging diagnostic reports as described in any of the above embodiments are implemented.
[0075] The present application also provides a computer-readable storage medium, on which a program is stored, wherein the computer-readable storage medium refers to a carrier for storing data, which may include but is not limited to a floppy disk, an optical disk, a hard disk, a flash memory, a USB flash drive, and / or a memory stick, etc., and the computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The working process, working details, and technical effects of the computer-readable storage medium provided in this embodiment can be found in the above embodiment of a language optimization method for a medical imaging diagnostic report, which will not be repeated here.
[0076] The application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the language optimization method for medical imaging diagnostic reports as described in any of the above embodiments.
[0077] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0078] The technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. The above-mentioned embodiments only express several implementation methods of the present application, and their descriptions are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, without departing from the concept of the present application, several deformations and improvements can be made, which all belong to the protection scope of the present application. Therefore, the protection scope of the patent of this application shall be based on the attached claims.
Claims
1. A method for language optimization of medical imaging diagnostic reports, characterized in that: include: Collecting the medical imaging diagnosis report to be processed, extracting the text in the medical imaging diagnosis report, and preprocessing the text; Collecting a medical imaging diagnosis knowledge base, and using a pre-trained recurrent neural network in combination with the medical imaging diagnosis knowledge base to extract and verify professional terms in the text; Using a first natural language processing library to perform part-of-speech tagging and grammar checking on the text to obtain a grammar checking result; Collecting historical medical imaging diagnosis reports and standard templates, comparing the text with the description of the same diagnosis in the historical medical imaging diagnosis reports and standard templates based on text similarity, and obtaining a description inspection result; The text is normalized and optimized using a second natural language processing library according to the grammar check result, the expression check result and the professional terminology.
2. The language optimization method for medical imaging diagnostic reports according to claim 1, characterized in that: The preprocessing of the text includes: Using regular expressions to remove punctuation marks, numbers, and special characters from the text, converting the text into a unified encoding format to obtain a cleaned text; The Python jieba library is used to segment the cleaned text.
3. The language optimization method for medical imaging diagnostic reports according to claim 2, characterized in that: Use a pre-trained recurrent neural network combined with a medical imaging diagnosis knowledge base to extract and verify the professional terms in the text, including: Convert text into vectors using a pre-trained word vector model, such as Word2Vec or GloVe; In combination with a medical imaging diagnosis knowledge base, a pre-trained recurrent neural network is used to identify and extract professional vocabulary and terminology in the vector, and the recurrent neural network includes an LSTM.
4. The language optimization method for medical imaging diagnostic reports according to claim 2, characterized in that: Using a first natural language processing library to perform grammar checking on the text includes: The Python NLTK library is used to check whether the statements in the text are consistent with the subject and predicate.
5. The language optimization method for medical imaging diagnostic reports according to claim 4, characterized in that: Also includes: Use CoreNLP Parser to perform grammatical analysis on the text to obtain the complexity of each sentence in the text; Whether the sentence is too complex is determined by comparing a preset threshold with the complexity, and the determination result is added to the grammar check result.
6. The language optimization method for medical imaging diagnostic reports according to claim 1, characterized in that: The second natural language processing library is a BERT model based on the Transformers model.
7. A language optimization device for medical imaging diagnostic reports, characterized in that: include: A data acquisition module is used to acquire the medical imaging diagnosis report to be processed, extract the text in the medical imaging diagnosis report, and pre-process the text; A term acquisition module, used to acquire a medical imaging diagnosis knowledge base, and use a pre-trained recurrent neural network in combination with the medical imaging diagnosis knowledge base to extract and verify professional terms in the text; A grammar checking module, used to perform part-of-speech tagging and grammar checking on the text using a first natural language processing library to obtain a grammar checking result; A statement checking module is used to collect historical medical imaging diagnosis reports and standard templates, compare the text with the statements of the same diagnosis in the historical medical imaging diagnosis reports and standard templates based on text similarity, and obtain statement checking results; A normalization module is used to normalize and optimize the text using a second natural language processing library according to the grammar check result, the expression check result and the professional terminology.
8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the language optimization method for medical imaging diagnostic reports according to any one of claims 1 to 6 are implemented.
9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program, wherein when the program is executed by a processor, the language optimization method for a medical imaging diagnostic report according to any one of claims 1 to 6 is implemented.
10. A computer program product comprising computer instructions, characterized in that: When executed by a processor, the computer instructions implement the steps of the language optimization method for medical imaging diagnostic reports described in claims 1 to 6.