Automatic coding method and system for surgical operation records

Abstract

The invention discloses an automatic coding method and system for surgical operation records. The method comprises the following steps: acquiring and preprocessing the surgical operation records of a patient; selecting an AI coding function, and generating an operation coding result and a corresponding evidence fragment based on the operation coding model; checking the coding result and the corresponding evidence fragment in the evidence interaction module, carrying out error correction on the coding result and the evidence fragment of multi-coding, error coding and missing coding, and marking an error reason; manually adjusting the sequence of the evidence fragments; the typical surgical operation records are collected, and collection reasons are described; and storing the coding result, the evidence fragment, the audited coding result, the audited evidence fragment, the collected surgical operation record and the collected reason in a database module. According to the invention, the artificial coding burden can be reduced, the medical insurance settlement efficiency is improved, and the model performance is optimized through continuous learning.

Description

Technical Field

[0001] The present invention relates to the technical field of medical event recognition, and in particular to a method and system for automatically encoding surgical operation records. Background Art

[0002] In the medical field, the standardization of medical terminology is an important demand in the medical field, including the standardization of disease terms, surgical operation terms, and drug names, which is of great significance to medical research, medical insurance payment, and public hospital performance evaluation. However, due to the differences in the terms used by different medical institutions and clinicians, the diagnosis and treatment information in various types of clinical medical data generally has the problem of non-standard and non-uniform terms, making the archiving and classification of data very complicated and difficult, consuming a lot of manpower and material resources, and it is difficult to mine or analyze it with the help of the latest information technologies such as big data and artificial intelligence.

[0003] In order to solve this problem, automatic coding technology has become a new solution in the scenario of medical record digitization. With the development of artificial intelligence technology, more and more people have begun to try to apply natural language processing technology to automatic coding tasks, but the application of natural language processing technology still faces many challenges. First, the complexity, diversity and multimodal characteristics of surgical operation records increase the difficulty of automatic coding; second, the existing technology is not accurate enough when processing professional terms, semantic relationships and contextual information; in addition, the accuracy of the coding results directly affects medical insurance payment and hospital performance evaluation. Therefore, there is an urgent need for an automatic coding method and system that can efficiently process surgical records, support multimodal data fusion and introduce a manual review mechanism to improve the accuracy and compliance of coding. At present, there is no method and system for automatic coding of surgical operation records on the market that covers the above characteristics. Summary of the invention

[0004] In view of the above-mentioned defects of the prior art, the purpose of the present invention is to provide a method and system for automatic coding of surgical operation records, which can reduce the burden of manual coding and improve the efficiency of medical insurance settlement. At the same time, it optimizes model performance through continuous learning, aiming to solve the problem of weak interactivity between coders and automatic coding systems in the prior art.

[0005] The present invention solves the technical problem by adopting the following technical solution: In a first aspect, a method for automatically encoding surgical operation records comprises the following steps: Step S1: Access the hospital's medical record system through a standard interface, obtain and pre-process the patient's surgical operation records, and display the results in the record interaction module; Step S2: Select the AI ​​coding function in the toolbar module, generate the surgical coding results and their corresponding evidence fragments based on the surgical coding model, and display them synchronously in the coding module and the evidence interaction module; Step S3: reviewing the coding results in the coding module and the corresponding evidence fragments in the evidence interaction module, and correcting the errors of the coding results and evidence fragments that are over-coded, wrongly coded, or omitted, and marking the causes of the errors; Step S4: manually adjusting the coding results in the coding module and the order of the corresponding evidence fragments in the evidence interaction module; Step S5: Collect typical surgical operation records and describe the reasons for collection; Step S6: Save the coding results and evidence fragments given by the surgical coding model, the reviewed coding results and evidence fragments, the collected surgical operation records and the reasons for collection in the database module.

[0006] Furthermore, in step S1, if the modality of the surgical operation record is an image modality, image text recognition is performed; if the modality of the surgical operation record is a sound modality, voice text recognition is performed, and finally the record is displayed in the text modality in the record interaction module.

[0007] Furthermore, in step S2, the surgical coding model used to generate automatic coding results and their corresponding evidence fragments should be an open source or closed source large language model or other natural language processing model that has been fine-tuned and trained on the coding task; the output format of the surgical coding model should include the surgical coding results and their corresponding evidence fragments in the surgical operation records.

[0008] Furthermore, in step S2, the coding results output by the surgical coding model and their corresponding evidence fragments are displayed one by one in the coding module and the evidence interaction module; the evidence fragments in the evidence interaction module will be synchronously displayed on the record interaction module and highlighted.

[0009] Furthermore, in step S3, the error correction performed on the encoding result and the evidence fragment includes the following types: Multiple coding: The surgical coding model gives a coding result outside the correct coding result set and its corresponding evidence fragment. Select the delete coding function in the toolbar module to mark the coding result as not in the correct coding result set. At the same time, the evidence fragment corresponding to the coding result is removed from the record interaction module and the error cause is marked as multiple coding; Wrong coding: mainly includes three types: the coding result is correct but the corresponding evidence segment is wrong, the coding result is wrong but the corresponding evidence segment is correct, and there are errors in both the coding result and the corresponding evidence segment. For wrong coding, the correct coding result is selected from the coding dictionary in the coding module, and the corresponding evidence segment is modified in the evidence interaction module, and the error cause is marked as wrong coding; Omission: The surgical coding model does not give some coding results in the correct coding result set. Select the new coding function in the toolbar module, and select the evidence segment corresponding to the coding result in the record interaction module. The selected evidence segment will be synchronously displayed in the evidence interaction module and separated by special symbols. The surgical coding model will generate recommended codes based on the selected evidence segment, select the correct coding result from it, and mark the cause of the error as omission.

[0010] Furthermore, in step S4, the adjustment of the order of the coding results and their corresponding evidence fragments is mainly based on the surgery ranking model trained with the medical insurance settlement importance ranking data, which is used to screen out the main surgery in a series of surgeries and put it at the top of the coding results.

[0011] Furthermore, in step S6, the coding results and corresponding evidence fragments given by the surgical coding model, the coding results and corresponding evidence fragments after manual review, and the surgical operation records collected in step S5 and the reasons for collection are saved in the database module for storage; these wrong examples and typical cases are used to optimize the performance of the surgical coding model and the model results are corrected using retrieval enhanced generation.

[0012] In a second aspect, a surgical operation record automatic coding system is provided, which is used to implement any of the above-mentioned surgical operation record automatic coding methods, comprising: The record interaction module is used to display the patient surgical operation records read from the standard interface, and simultaneously highlight the evidence fragments corresponding to the encoding results in the evidence interaction module. After selecting a surgical operation record fragment, it is appended to the corresponding evidence fragment in the evidence interaction module; The evidence interaction module is used to record the evidence fragments corresponding to each coding result. When a coding result corresponds to multiple evidence fragments, special symbols are used to separate the evidence fragments. After adding or deleting evidence fragments, the changes will be synchronously displayed in the recording interaction module. When correcting errors, the causes of the errors will be recorded in the evidence interaction module. The coding module is used to record the coding result of the patient's surgical operation record. The coding result should be the most appropriate surgical operation noun selected from the mapping dictionary; Function bar module, used to provide automatic encoding function, delete encoding function, add encoding function and perform function expansion; The database module is used to save the coding results and corresponding evidence fragments given by the surgical coding model and the coding results and corresponding evidence fragments after manual review. The saved data is used for tuning the model performance.

[0013] Furthermore, the evidence interaction module is associated with the encoding module, and the evidence fragment corresponding to each encoding result in the encoding module should be in the same order in the evidence interaction module as the encoding result in the encoding module. When the coder rearranges the order, the encoding result and its corresponding evidence fragment should be moved synchronously; the record interaction module is synchronized with the evidence interaction module, and the evidence fragment recorded in the evidence interaction module should be synchronously highlighted in the record interaction module. For newly added or deleted encoding results, the highlighting of the corresponding evidence fragment should be synchronously updated or cancelled in the record interaction module.

[0014] In a third aspect, a terminal device comprises a memory, a processor, and a program of a method for automatic encoding of surgical operation records stored in the memory and executable on the processor, wherein when the processor executes the program of the method for automatic encoding of surgical operation records, the steps of any one of the above-mentioned methods for automatic encoding of surgical operation records are implemented. Beneficial Effects

[0015] The present invention provides a method and system for automatic coding of surgical operation records. The AI ​​model automatically generates coding results and combines them with a manual review mechanism, which significantly improves coding efficiency and accuracy, supports multimodal data processing, and provides manual error correction, sequence rearrangement, and model continuous optimization functions. The system design is user-friendly and has strong functional scalability. It effectively solves the problem of medical data standardization, provides high-quality data support for medical research, medical insurance payment, and hospital performance evaluation, and promotes the digital transformation of the medical industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a flow chart of the automatic encoding method of the present invention.

[0017] Figure 2 It is a relationship diagram of various system modules of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Example

[0019] refer to Figure 1 This embodiment provides a method for automatically encoding surgical operation records, comprising the following steps: Step S1: Access the hospital's medical record system through a standard interface, obtain and pre-process the patient's surgical operation records, and display the results in the record interaction module; In the pre-processing of the input surgical records, if the modality of the surgical operation record is an image modality, image text recognition is performed; if the modality of the surgical operation record is a sound modality, voice text recognition is performed, and finally displayed in the record interaction module in text mode. If a patient undergoes multiple surgeries during this hospitalization, the record interaction module will also display the surgical operation records of multiple surgeries for subsequent coding and correction.

[0020] Step S2: Select the AI ​​coding function in the toolbar module, generate surgical coding results based on the surgical coding model, save the coding results and their corresponding evidence fragments in the coding module and the evidence interaction module respectively, and synchronously display the evidence fragments in the record interaction module; Specifically, the surgical coding model used to generate automatic coding results and their corresponding evidence fragments should be an open source or closed source large language model or other natural language processing model that has been fine-tuned and trained on the coding task; the output format of the surgical coding model should include the surgical coding results and their corresponding evidence fragments in the surgical operation records.

[0021] For example, the format of the instruction data for constructing an encoding task is as follows: Input: "Which standard surgical terms are contained in the following text, and which fragments of the text do they correspond to? Output in json format\n{surgical record data in text mode}"; Output: "{"{standard surgical term 1}": ["{text fragment 1 corresponding to standard surgical term 1}",…],…}".

[0022] To further optimize the technical solution, the system first uses the vector database to retrieve the historical surgical operation records and their corresponding coding results that are most similar to the current surgical operation record text, so as to make full use of the experience of existing data to optimize the coding process of new data. In this way, the system can perform reference matching based on historical coding knowledge to improve the accuracy and consistency of coding. Subsequently, the system inputs the current surgical operation record text and the most similar historical coding results retrieved into the automatic coding module. On this basis, the system constructs specific prompt words and calls the large language model to generate new coding results to ensure the intelligence and automation of the coding process.

[0023] Specifically, the core data stored in the vector database includes the text of surgical operation records, their vectorized representation, and the correct coding results after review by the coder. The construction of the vector database follows the following steps: First, for each surgical operation record, an embedding model that supports multiple languages ​​and is optimized for sentence similarity (such as bge-m3) is used to encode the preprocessed text and generate a high-dimensional text vector representation to ensure the efficiency and accuracy of the retrieval process. Subsequently, a vector database is established based on the generated text vector representation, and the surgical operation record text, the corresponding text vector, and the correct coding results reviewed and confirmed by the coder are stored. At the same time, an index is built on the text vector field to support subsequent efficient and accurate similarity retrieval.

[0024] After retrieving similar historical surgical operation records and their corresponding encoding results, this step uses specific template assembly prompt words to guide the large language model to perform text classification. The structure of the template assembly prompt words is as follows: "{Role definition label}: {Role definition statement that sets the large language model to encode surgical operation records}\n{Example label}: \n{Example surgical operation records and their corresponding encoding results}\n{Input label}: {Sentence that guides the large model to encode surgical operation records}\n{Input surgical operation data}\n{Encoding result output label}:" Among them, {example text and classification results} can include a single example or multiple retrieved similar historical surgical operation records and their corresponding coding results.

[0025] In some specific implementations, the specific content of the template is determined as follows based on the structural composition of the template assembly prompt word: "You are an excellent surgical operation record coding expert. You are familiar with various surgical operations and their coding. Now you need to identify the surgical operation codes and their evidence from the input surgical operation records. You should output them in JSON format. You can refer to the examples provided in [Example] to complete your task.

[0026] [Example] Example 1: {Surgical operation record 1} Output 1: {Encoding result} Example 2: {Surgical operation record 2} Output 2: {Encoding result}

Data to be classified

[0027] {Enter surgical operation record} [Output]" After the surgical coding model is output, the system first parses the JSON format text output by the surgical coding model. Once it is found that the text cannot be directly converted into the standard JSON format, the system will automatically enable predefined rules to correct the text to ensure that a complete and standardized JSON text is finally obtained.

[0028] Next, in the successfully parsed JSON content, for each recorded surgical term, the system will compare and verify it with the built-in standard surgical term library. If a surgical term fails to find a corresponding match in the standard term library, the system uses Jaccard Similarity as a measurement indicator to calculate the similarity between the term and all entries in the standard term library, and selects the closest one as a candidate match. The Jaccard similarity calculation formula is:

[0029] in A is the character set of surgical terms generated by the model, B It is a character set of standard surgical terms in the standard surgical terminology library. Representing a collection A With Collection B The number of elements in the intersection, that is, the number of characters they have in common; Representing a collection A With Collection B The number of elements in the union of , that is, the total number of different characters in the two sets.

[0030] In order to ensure the consistency and accuracy of the data, the system sets a similarity threshold for the above process. If, after comparative analysis, even the most similar standard surgical terminology fails to meet this threshold requirement, it means that the surgical record and its associated evidence fragments may lack sufficient accuracy or applicability, and the system will choose to ignore this record, thereby maintaining the quality and reliability of the entire data.

[0031] After processing the output results of the surgical coding model, the system will display the coding results and evidence fragments output by the surgical coding model one by one in the coding module and the evidence interaction module respectively; the evidence fragments in the evidence interaction module will be synchronously displayed on the record interaction module and highlighted.

[0032] Step S3: The coder reviews the coding results in the coding module and the evidence fragments corresponding to the coding results in the evidence interaction module, and can correct the errors of multiple coding, wrong coding and missing coding of the coding results and evidence fragments. For the coding results with errors, the coder can mark the reasons for the errors; Specifically, error correction covers the following categories: Overcoding: When the coding results generated by the surgical coding model exceed the correct set of coding results, that is, there are redundant coding results and their corresponding evidence fragments, the coder needs to manually delete these unnecessary codes. Such errors may be caused by AI's failure to correctly perform "merge coding", AI's errors in text annotation, and doctors' writing errors in surgical records. In addition, other unforeseen factors may also cause such problems. All such situations are collectively referred to as "overcoding".

[0033] Miscoding: If the surgical coding model provides coding results and corresponding evidence fragments that are similar to the correct coding result set but inaccurate, the coder needs to manually modify these coding results and adjust their evidence fragments accordingly. This type of error is usually caused by AI's inaccurate recognition of the surgical approach, AI's incorrect annotation of the text, or the doctor's inadequate writing of the surgical record. In addition, there may be other factors that lead to this type of error, and these situations are collectively classified as "miscoding."

[0034] Missed coding: If the surgical coding model fails to provide some necessary parts of the correct coding result set, the coder can add the missing coding results and their related evidence fragments through the Insert Coding function in the toolbar module. This may be caused by the AI's failure to identify the content that needs to be coded, the failure to fully cover the marked text, or the omissions in the surgical record content provided by the doctor. In addition, there are other possible reasons for this type of error, collectively referred to as "missed coding".

[0035] In addition to the above classifications, coders can also add detailed notes for each error type to more accurately describe the specific circumstances of the error and the basis for correction, thereby ensuring the professionalism and accuracy of the entire coding process. In this way, not only can the quality of the data be improved, but it can also provide reliable support for subsequent data analysis and medical decision-making.

[0036] During the error correction process, coders can select specific surgical record text in the record interaction module and append it to the evidence segment corresponding to the relevant coding result. This function ensures that all necessary information is accurately included in the evidence segment, which facilitates subsequent review and processing.

[0037] Step S4: the coder can manually adjust the coding results in the coding module and the order of the corresponding evidence fragments in the evidence interaction module; Specifically, after completing the correction of the output results of the surgical coding model, the coder can manually rearrange the surgical coding results according to the output of the surgical sorting model. The main purpose of this function is to optimize the sorting of the coding results generated by the surgical coding model according to the importance of medical insurance settlement. In specific operations, the coder can click on a coding result and drag it up and down to adjust its order to achieve the most reasonable arrangement. Usually, the coding of major surgeries will be placed at the top of all coding results. This is because major surgeries often have a key impact on the settlement of overall medical expenses.

[0038] In this way, the system not only supports accurate correction of the original coding results, but also allows the order of coding results to be flexibly adjusted according to actual needs, so as to better meet the requirements of medical insurance settlement and other aspects. This design fully takes into account the professionalism and complexity of medical data processing, and helps to improve the efficiency and accuracy of the entire workflow. In addition, this flexible adjustment mechanism also facilitates the response to specific requirements that may exist between different regions or institutions, enhancing the adaptability and practicality of the system.

[0039] Step S6: The coder saves the coding results and evidence fragments given by the surgical coding model, the coding results and evidence fragments reviewed by the coder, and the surgical operation records collected by the coder and the reasons for collecting them in the database module.

[0040] To further optimize the technical solution, the coding results and corresponding evidence fragments given by the surgical coding model, the coding results and corresponding evidence fragments after manual review by the coder, and the surgical operation records collected by the coder and the reasons for collection are stored in the database module; these wrong examples and typical cases can be used to optimize the performance of the surgical coding model and use retrieval enhanced generation to correct the model results.

[0041] Specifically: The system will first write the coding results processed by the coder accurately into the hospital's medical record system database to ensure that all updated coding information can be reflected in the medical records in a timely manner, providing a reliable basis for clinical decision-making, medical insurance settlement and patient management.

[0042] At the same time, the system will also record the coding results initially generated by the surgical coding model, including the coding content and its corresponding evidence fragments, and record in detail all error types annotated by the coder and their specific causes. In this way, not only can we fully understand the performance of the surgical coding model, but we can also identify common error patterns, laying the foundation for subsequent data analysis and model optimization. These detailed records support the evaluation of the performance of the surgical coding model from multiple angles, which helps to make targeted adjustments and optimizations, thereby continuously improving the accuracy and efficiency of the surgical coding model.

[0043] In addition, detailed records also support the hospital's internal quality control processes and external audit requirements. They provide the necessary documentation support for checking the compliance, accuracy and consistency of coding work, which helps to improve the overall quality of medical services. By saving the original output of the surgical coding model and the coder's correction records, the system provides the possibility for subsequent data analysis, making it possible to continuously improve the surgical coding model based on actual operational feedback. This approach effectively combines the advantages of artificial intelligence and professional judgment, promotes the development of medical information management systems to a higher level, and not only improves work efficiency, but also enhances data reliability. Example

[0044] refer to Figure 2 The present invention also provides a surgical operation record automatic coding system, which is used to implement any of the above-mentioned surgical operation record automatic coding methods, including: The record interaction module is used to display the patient surgical operation records read from the standard interface, and simultaneously highlight the evidence fragments corresponding to the encoding results in the evidence interaction module. After selecting a surgical operation record fragment, it is appended to the corresponding evidence fragment in the evidence interaction module; The evidence interaction module is used to record the evidence fragments corresponding to each coding result. When a coding result corresponds to multiple evidence fragments, special symbols are used to separate the evidence fragments. After adding or deleting evidence fragments, the changes will be synchronously displayed in the recording interaction module. When correcting errors, coders can record the causes of errors in the evidence interaction module. A coding module is used to record the coding results of the patient's surgical operation records. The coding results should be the most appropriate surgical operation terms selected from a mapping dictionary. The mapping dictionary refers to standard surgical operation terms and their corresponding codes issued by national functional departments or professional associations; A function bar module is used to provide the automatic coding function, the coding deletion function, the new coding function and to expand the functions, such as a collection function that can mark special surgical operation records and their coding results, and an AI dialogue function that obtains coding information by communicating with a generative artificial intelligence model; The database module is used to store the coding results and corresponding evidence fragments given by the surgical coding model and the coding results and corresponding evidence fragments after manual review by the coder. The stored data is used to optimize the model performance.

[0045] Specifically, the evidence interaction module is associated with the encoding module, and the evidence fragment corresponding to each encoding result in the encoding module should be in the same order in the evidence interaction module as the encoding result in the encoding module. When the coder rearranges the order, the encoding result and its corresponding evidence fragment should be moved synchronously; the record interaction module is synchronized with the evidence interaction module, and the evidence fragment recorded in the evidence interaction module should be synchronously highlighted in the record interaction module. For newly added or deleted encoding results, the highlighting of the corresponding evidence fragment should be synchronously updated or cancelled in the record interaction module. Example

[0046] This embodiment provides a terminal device, which includes a memory, a processor, and a program of a method for automatic encoding of surgical operation records stored in the memory and executable on the processor. When the processor executes the program of the method for automatic encoding of surgical operation records, the steps of any one of the above-mentioned methods for automatic encoding of surgical operation records are implemented.

[0047] The terminal device may include one or more processors, a memory, and a computer program stored in the memory and executable on one or more processors, for example, a surgical operation record automatic coding system. When one or more processors execute the computer program, the various steps in an embodiment of a surgical operation record automatic coding method can be implemented. Alternatively, when one or more processors execute the computer program, the functions of various modules in an embodiment of a surgical operation record automatic coding system can be implemented, which is not limited here.

[0048] In one embodiment, the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0049] In one embodiment, the memory may be an internal storage unit of an electronic device, such as a hard disk or memory of the electronic device. The memory may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device. Further, the memory may also include both an internal storage unit of the electronic device and an external storage device. The memory is used to store computer programs and other programs and data required by the terminal device. The memory may also be used to temporarily store data that has been output or is to be output.

[0050] Those skilled in the art will understand that the structure of the terminal device described above is only a partial structure related to the solution of the present invention, and does not constitute a limitation on the terminal device to which the solution of the present invention is applied. The specific terminal device may include more or fewer components than those described in the above requirements, or combine certain components, or have a different arrangement of components.

[0051] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0052] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for automatically encoding surgical operation records, characterized in that: The steps include: Step S1: Access the hospital's medical record system through a standard interface, obtain and pre-process the patient's surgical operation records, and display the results in the record interaction module; Step S2: Select the AI ​​coding function in the toolbar module, generate the surgical coding results and their corresponding evidence fragments based on the surgical coding model, and display them synchronously in the coding module and the evidence interaction module; Step S3: reviewing the coding results in the coding module and the corresponding evidence fragments in the evidence interaction module, and correcting the errors of the coding results and evidence fragments that are over-coded, wrongly coded, or omitted, and marking the causes of the errors; Step S4: manually adjusting the coding results in the coding module and the order of the corresponding evidence fragments in the evidence interaction module; Step S5: Collect typical surgical operation records and describe the reasons for collection; Step S6: Save the coding results and evidence fragments given by the surgical coding model, the reviewed coding results and evidence fragments, the collected surgical operation records and the reasons for collection in the database module.

2. The method for automatic coding of surgical operation records according to claim 1, characterized in that: In step S1, if the modality of the surgical operation record is an image modality, image text recognition is performed; if the modality of the surgical operation record is a sound modality, voice text recognition is performed, and finally the record is displayed in the text modality in the record interaction module.

3. The method for automatic coding of surgical operation records according to claim 2, characterized in that: In step S2, the surgical coding model used to generate automatic coding results and their corresponding evidence fragments should be an open source or closed source large language model or other natural language processing model that has been fine-tuned and trained on the coding task; the output format of the surgical coding model should include the surgical coding results and their corresponding evidence fragments in the surgical operation records.

4. The method for automatic coding of surgical operation records according to claim 1, characterized in that: In step S2, the coding results output by the surgical coding model and their corresponding evidence fragments are displayed one by one in the coding module and the evidence interaction module; the evidence fragments in the evidence interaction module will be synchronously displayed on the record interaction module and highlighted.

5. The method for automatic coding of surgical operation records according to claim 1, characterized in that: In step S3, the error correction performed on the encoding result and the evidence fragment includes the following types: Multiple coding: The surgical coding model gives a coding result outside the correct coding result set and its corresponding evidence fragment. Select the delete coding function in the toolbar module to mark the coding result as not in the correct coding result set. At the same time, the evidence fragment corresponding to the coding result is removed from the record interaction module and the error cause is marked as multiple coding; Wrong coding: mainly includes three types: the coding result is correct but the corresponding evidence segment is wrong, the coding result is wrong but the corresponding evidence segment is correct, and there are errors in both the coding result and the corresponding evidence segment. For wrong coding, the correct coding result is selected from the coding dictionary in the coding module, and the corresponding evidence segment is modified in the evidence interaction module, and the error cause is marked as wrong coding; Omission: The surgical coding model does not give some coding results in the correct coding result set. Select the new coding function in the toolbar module, and select the evidence segment corresponding to the coding result in the record interaction module. The selected evidence segment will be synchronously displayed in the evidence interaction module and separated by special symbols. The surgical coding model will generate recommended codes based on the selected evidence segment, select the correct coding result from it, and mark the cause of the error as omission.

6. The method for automatic coding of surgical operation records according to claim 1, characterized in that: In step S4, the adjustment of the order of the coding results and their corresponding evidence fragments is mainly based on the surgery ranking model trained with the medical insurance settlement importance ranking data, which is used to screen out the main surgery in a series of surgeries and put it at the top of the coding results.

7. The method for automatic coding of surgical operation records according to claim 1, characterized in that: In step S6, the coding results and corresponding evidence fragments given by the surgical coding model, the coding results and corresponding evidence fragments after manual review, and the surgical operation records and reasons for collection collected in step S5 are saved in the database module for storage; These wrong examples and typical cases are used to optimize the performance of the surgical coding model and the model results are corrected using retrieval-enhanced generation.

8. An automatic coding system for surgical operation records, characterized in that: A method for automatically encoding surgical operation records according to any one of claims 1 to 7, comprising: The record interaction module is used to display the patient surgical operation records read from the standard interface, and simultaneously highlight the evidence fragments corresponding to the encoding results in the evidence interaction module. After selecting a surgical operation record fragment, it is appended to the corresponding evidence fragment in the evidence interaction module; The evidence interaction module is used to record the evidence fragments corresponding to each coding result. When a coding result corresponds to multiple evidence fragments, special symbols are used to separate the evidence fragments. After adding or deleting evidence fragments, the changes will be synchronously displayed in the recording interaction module. When correcting errors, the causes of the errors will be recorded in the evidence interaction module. The coding module is used to record the coding result of the patient's surgical operation record. The coding result should be the most appropriate surgical operation noun selected from the mapping dictionary; Function bar module, used to provide automatic encoding function, delete encoding function, add encoding function and perform function expansion; The database module is used to save the coding results and corresponding evidence fragments given by the surgical coding model and the coding results and corresponding evidence fragments after manual review. The saved data is used for tuning the model performance.

9. The automatic coding system for surgical operation records according to claim 8, characterized in that: The evidence interaction module is associated with the encoding module, and the evidence fragment corresponding to each encoding result in the encoding module should be in the same order in the evidence interaction module as the encoding result in the encoding module. When the coder rearranges the order, the encoding result and its corresponding evidence fragment should be moved synchronously; the record interaction module is synchronized with the evidence interaction module, and the evidence fragment recorded in the evidence interaction module should be synchronously highlighted in the record interaction module. For newly added or deleted encoding results, the highlighting of the corresponding evidence fragment should be synchronously updated or cancelled in the record interaction module.

10. A terminal device, characterized in that: The terminal device includes a memory, a processor, and a program of a method for automatic encoding of surgical operation records stored in the memory and executable on the processor. When the processor executes the program of the method for automatic encoding of surgical operation records, the steps of the method for automatic encoding of surgical operation records as described in any one of claims 1 to 7 are implemented.

Images