Medical record diagnosis and surgical operation code intelligent generation method and system based on AI

Abstract

The invention provides an AI-based medical record diagnosis and surgical operation code intelligent generation method and system, and relates to the technical field of medical information, and the method comprises the steps: obtaining multi-modal data from a hospital information system, including a text medical record, a medical image, a surgical video and a pathological report; entities and relationships are extracted through a multi-modal fusion model, and a hospital-level disease knowledge graph containing space-time dimensions is constructed; training a coding rule generation model and a verification model based on a reinforcement learning framework, and generating candidate codes and confidence scores thereof; and correcting the candidate codes according to manual feedback, and dynamically updating the knowledge graph and the rule base by using the graph neural network. Accurate disease diagnosis codes and surgical operation codes are generated in an intelligent mode, and the accuracy of the codes can be remarkably improved through automatic medical record code recommendation.

Description

Technical Field

[0001] The present invention relates to the field of medical information technology, and in particular to a method and system for intelligently generating medical record diagnosis and surgical operation codes based on AI (artificial intelligence). Background Art

[0002] With the continuous development of artificial intelligence technology, the application of AI in the medical industry has gradually become a trend, especially in the field of hospital medical record coding and surgical procedure code generation. Currently, traditional medical record coding relies heavily on manual operations or similarity calculations based on natural language processing (NLP) technology for matching. Traditional NLP-based coding methods typically require word segmentation and semantic similarity calculations to generate codes, but this method has low accuracy, especially in hospital-level applications. Its poor adaptability leads to frequent hospital coding errors and requires extensive manual correction, resulting in high labor costs.

[0003] Therefore, how to improve the accuracy of medical record coding and reduce the cost of manual intervention has become an urgent problem to be solved in the field of medical informationization. Summary of the Invention

[0004] In view of this, the purpose of this invention is to propose an AI-based method and system for intelligently generating medical record diagnosis and surgical operation codes. This system aims to intelligently generate diagnosis codes and surgical operation codes by combining historical hospital medical record data, medical knowledge graphs, and AI technology. This method can intelligently generate accurate disease diagnosis codes and surgical operation codes. Through automated medical record code recommendations, it can significantly improve coding accuracy, thereby increasing coding efficiency, reducing manual intervention and error rates, and lowering hospital labor costs.

[0005] To achieve the above object, the present invention provides the following technical solutions: Based on the above objectives, in a first aspect, the present invention provides an AI-based method for intelligently generating medical record diagnosis and surgical operation codes, comprising the following steps: Acquire multimodal data from hospital information systems, including text medical records, medical images, surgical videos, and pathology reports; Entities and relationships are extracted through a multimodal fusion model to construct a hospital-level disease knowledge graph that includes time and space dimensions; Based on the reinforcement learning framework, the coding rule generation model and verification model are trained to generate candidate codes and their confidence scores; Modify candidate codes based on manual feedback, and use graph neural networks to dynamically update knowledge graphs and rule bases.

[0006] As a further solution of the present invention, entities and relationships are extracted through a multimodal fusion model to construct a hospital-level disease knowledge graph with spatiotemporal dimensions, including the following steps: Clean multimodal data obtained from hospital information systems, uniformly encode text medical records, image labels, and laboratory time series data into heterogeneous nodes in the knowledge graph, and annotate disease code labels; Use the BERT-BiLSTM-CRF model to extract entities and relationships of symptoms, examinations, and medications from multimodal data; Through semantic version control, the hospital automatically detects differences between the terminology library and the standard coding system, generates incremental alignment rules, aligns entities with standard medical terminology, and constructs a hospital-level disease knowledge graph containing multi-dimensional diagnosis and treatment data; When a new term is detected, the knowledge graph node is automatically expanded and associated with a temporary encoding.

[0007] As a further solution of the present invention, constructing a hospital-level disease knowledge graph including time and space dimensions also includes constructing: Timeline: embeds diagnosis and treatment events into timestamp attributes of graph nodes according to their occurrence time; Spatial axis: Annotate the 3D spatial coordinates of anatomical structures based on medical image segmentation results and construct a sub-atlas of surgical sites.

[0008] As a further solution of the present invention, before training the coding rule generation model and the verification model based on the reinforcement learning framework, the following steps are included: Rule generation model: Generate candidate coding sequences through random walk sampling of knowledge graph; Rule verification model: A discriminant network trained with artificial historical coding data simulates the behavior of human coders, performs confidence scoring on candidate codes, and obtains a coding rationality score; Strategy optimization: Maximize the reward function based on the PPO algorithm and update the generation model parameters.

[0009] As a further solution of the present invention, a coding rule generation model and a verification model are trained based on a reinforcement learning framework to generate candidate codes and their confidence scores, including the following steps: Based on graph neural networks, we mine high-frequency co-occurring diagnosis and treatment patterns and coding mapping rules, and calculate the rule confidence, support, and hospital-level specific weight. By parsing the new medical record text, extracting key entities and matching them with the knowledge graph, activating association rules to generate candidate codes, and combining confidence and rule priority to generate a recommendation list.

[0010] As a further solution of the present invention, a graph neural network is used to dynamically update the knowledge graph and rule base, including: Node update: When adding a rare disease entity, similar case subgraphs are matched using locality sensitive hashing (LSH) and association rules are inherited; Edge weight update: adjust the weight coefficient of the "Disease→Surgery" edge according to the frequency of coding usage.

[0011] As a further solution of the present invention, the AI-based intelligent generation method for medical record diagnosis and surgical operation codes also includes cross-hospital medical record coding knowledge transfer, the steps of which are as follows: Aggregate local knowledge graph features of multiple hospitals through federated learning to generate a global encoding rule model; Use knowledge distillation to compress the global model into a lightweight transfer model for new hospitals to initialize their local systems; Based on differential privacy technology, hospital data privacy is protected and noise perturbation is added during the knowledge transfer process.

[0012] As a further solution of the present invention, the lightweight migration model is constructed in the following manner: Extract high-frequency diagnosis and treatment patterns from the global model as meta-rules; Encode the meta-rules as the initial weight matrix of the graph attention mechanism (GAT); Fine-tune the attention weights on the local data of the new hospital to adapt to the hospital-level terminology habits.

[0013] In a second aspect, the present invention provides an AI-based intelligent generation system for medical record diagnosis and surgical operation codes, comprising: Multimodal data parsing module for processing text, image and video data; Spatiotemporal knowledge graph database, which stores medical entities and relationships with spatiotemporal attributes; Reinforcement learning engine, including generator, discriminator and reward calculation unit; Incremental learning interface supports cross-institutional knowledge transfer and dynamic update of local rule base.

[0014] As a further solution of the present invention, the multimodal data parsing module integrates the following sub-models: Text entity extraction model: Named entity recognition network jointly trained with BioBERT and CRF; Image feature extraction model: surgical video key frame classifier based on 3D-ResNet; Temporal alignment model: Transformer-based diagnostic and treatment event sequence generator.

[0015] As a further aspect of the present invention, the incremental learning interface supports: Version control and rollback of local rule base; Blockchain-based rule update log storage ensures data traceability; Automate compliance checks to prevent coding rules from violating healthcare regulatory policies.

[0016] In another aspect of the present invention, a computer device is provided, comprising a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, any one of the above-mentioned methods for intelligent generation of medical record diagnosis and surgical operation codes based on AI according to the present invention is executed.

[0017] In another aspect of the present invention, a computer-readable storage medium is provided, which stores computer program instructions, and when the computer program instructions are executed, any of the above-mentioned AI-based medical record diagnosis and surgical operation code intelligent generation methods according to the present invention is implemented.

[0018] Compared with the existing technology, the AI-based medical record diagnosis and surgical operation code intelligent generation method and system proposed in this invention have the following beneficial effects: The present invention integrates data from medical record texts, medical images, surgical videos, and pathology reports to construct a hospital-level knowledge graph covering the entire diagnosis and treatment process, which helps to comprehensively capture the patient's condition and treatment process, improve the accuracy of coding, and construct a dynamic causal chain reasoning mechanism by embedding time and space dimensions. The spatiotemporal modeling method can help the system better understand the development process of the disease and related treatments, support more accurate coding reasoning, and use the generative adversarial network framework to train the coding rule generation model and verification model. The rule generation model generates candidate codes, and the verification model scores the candidate codes. By continuously optimizing the rule base, the coding system is getting closer and closer to the actual coding preferences of the hospital; and through federated learning and knowledge distillation, the problem of knowledge transfer across hospitals is solved, so that hospitals can quickly start and optimize the medical record coding system through a lightweight migration model without a large amount of local data.

[0019] Therefore, this invention fully realizes the efficient integration and reasoning of different types of data, dynamically adjusts node embedding representations through graph neural networks, and further optimizes the accuracy of the knowledge graph. Furthermore, the system supports cross-hospital knowledge transfer, using federated learning and differential privacy technologies to protect data privacy and enable knowledge sharing among multiple hospitals.

[0020] These and other aspects of the present application will be more clearly understood in the following description of the embodiments. It should be understood that the above general description and the following detailed description are merely exemplary and explanatory and cannot limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or related technologies, the following briefly introduces the drawings required for the exemplary embodiments or related technical descriptions. The drawings are used to provide a further understanding of the present invention and constitute part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the drawings: Figure 1 This is a flowchart of an AI-based method for intelligently generating medical record diagnosis and surgical operation codes according to an embodiment of the present invention.

[0022] Figure 2 The present invention provides a flowchart for constructing a hospital-level disease knowledge graph including spatiotemporal dimensions in an AI-based method for intelligently generating medical record diagnosis and surgical operation codes in an embodiment of the present invention.

[0023] Figure 3 The present invention provides a flowchart for generating candidate codes and their confidence scores in an AI-based intelligent generation method for medical record diagnosis and surgical operation codes according to an embodiment of the present invention.

[0024] Figure 4 The present invention provides a flowchart of the cross-hospital medical record coding knowledge transfer in an AI-based intelligent generation method for medical record diagnosis and surgical operation codes in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] Below, the present application is further described in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0026] To make the purpose, technical solutions and advantages of the present invention more clearly understood, the following is a further detailed description of the embodiments of the present invention in conjunction with specific embodiments and with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0027] It should be noted that all expressions using "first" and "second" in the embodiments of the present invention are intended to distinguish two non-identical entities or non-identical parameters with the same name. Therefore, "first" and "second" are used for convenience of expression only and should not be understood as limitations on the embodiments of the present invention. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, other steps or units inherent to a process, method, system, product, or device that includes a series of steps or units.

[0028] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0029] The flowcharts shown in the accompanying drawings are for illustrative purposes only and do not necessarily include all contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may be decomposed, combined, or partially merged, so the actual execution order may vary depending on the actual situation.

[0030] The following describes some embodiments of the present application in detail with reference to the accompanying drawings. In the absence of conflict, the following embodiments and features therein may be combined with each other.

[0031] In order to address the problem that traditional medical record coding relies on manual operation or similarity calculation based on natural language processing for matching, and generates codes through word segmentation and semantic similarity calculation, which has low accuracy. The present invention proposes an AI-based intelligent generation method and system for medical record diagnosis and surgical operation codes, aiming to intelligently generate diagnosis codes and surgical operation codes by combining hospital historical medical record data, medical knowledge graphs and AI technology. This method can generate accurate disease diagnosis codes and surgical operation codes in an intelligent manner, and can significantly improve the accuracy of coding through automated medical record coding recommendations, thereby improving coding efficiency, reducing manual intervention and error rates, and reducing hospital labor costs.

[0032] See also Figure 1 As shown, an embodiment of the present invention provides an AI-based method for intelligently generating medical record diagnosis and surgical operation codes, the method comprising the following steps: Step S10: Acquire multimodal data from the hospital information system, including text medical records, medical images, surgical videos, and pathology reports.

[0033] In this step, the hospital information system (HIS) contains various data sources, such as electronic medical records (EMRs), medical images (such as CT and MRI), surgical videos, and pathology reports. This data is obtained through the hospital's API interface or data warehouse, and the integrity and accuracy of the data are ensured. The data can be in JSON, XML, or other structured formats. The purpose of this step is to extract and standardize the different types of data in the hospital information system into a format that can be used for subsequent processing. Multimodal data includes text data (such as medical records), imaging data, video data, and other medical laboratory data, providing comprehensive information about patient diagnosis and treatment.

[0034] Step S20: Extract entities and relationships through a multimodal fusion model to construct a hospital-level disease knowledge graph that includes time and space dimensions.

[0035] In this step, see Figure 2 As shown in the figure, entities and relationships are extracted through a multimodal fusion model to construct a hospital-level disease knowledge graph with spatiotemporal dimensions, which includes the following steps: Step S201: Clean the multimodal data obtained from the hospital information system, uniformly encode the text medical records, image labels and laboratory time series data into heterogeneous nodes in the knowledge graph, and mark the disease code labels.

[0036] In this step, after cleaning the multimodal data, the text information in the medical records is converted into a unified code. Image processing algorithms are used to extract important labels and annotations (such as tumor size and location) from the imaging data, and these data are mapped into nodes in the knowledge graph. Data cleaning includes removing redundant information, correcting errors, and standardizing the format. Information in the text medical records, such as symptoms, diagnoses, and medications, is mapped into nodes in the knowledge graph using specific rules. Image labels are uniformly encoded and associated with relevant diseases.

[0037] Step S202: Use the BERT-BiLSTM-CRF model to extract entities and relationships of symptoms, examinations, and medications from multimodal data.

[0038] In this step, entities such as disease names, examination items, and medications are extracted from the medical record text, and relationships between these entities, such as "diagnosis" and "treatment," are identified. In the BERT-BiLSTM-CRF model, BERT, as a pre-trained language model, encodes the medical record text by understanding its context; BiLSTM captures information about the preceding and following text sequences; and CRF annotates entity sequences. This combination of the three effectively identifies and extracts key entities in the medical field, such as terminology, symptoms, and medications, and also extracts relationships between them.

[0039] Step S203: Automatically detect the differences between the hospital terminology library and the standard coding system through semantic version control, generate incremental alignment rules, align entities with standard medical terms, and build a hospital-level disease knowledge graph containing multi-dimensional diagnosis and treatment data.

[0040] In this step, the system automatically detects and updates discrepancies between the hospital's terminology and standard medical coding systems (such as ICD-10 and CPT) through semantic versioning, generating incremental alignment rules and achieving standardization. Comparing the differences between the hospital's terminology and standard medical terminology and generating rules ensures that the hospital's terminology aligns with the standard coding system, which is crucial for the long-term maintenance and upgrade of the hospital system.

[0041] Step S204: When a new term is detected, the knowledge graph node is automatically expanded and associated with a temporary code.

[0042] In this step, when the system identifies new terms or diseases, it automatically adds them to the knowledge graph and assigns them a temporary code until further verification. If a new term or entity is identified by the model, it is automatically expanded to the corresponding location in the knowledge graph. The expansion process involves assigning a temporary code to the new node and marking it as "pending confirmation."

[0043] Among them, building a hospital-level disease knowledge graph with spatiotemporal dimensions also includes building: Timeline: embeds diagnosis and treatment events into timestamp attributes of graph nodes according to their occurrence time; Spatial axis: Annotate the 3D spatial coordinates of anatomical structures based on medical image segmentation results and construct a sub-atlas of surgical sites.

[0044] Step S30: Train the coding rule generation model and verification model based on the reinforcement learning framework to generate candidate codes and their confidence scores.

[0045] In this step, before training the coding rule generation model and verification model based on the reinforcement learning framework, the following steps are performed: Rule generation model: Generate candidate coding sequences through random walk sampling of knowledge graph; Rule verification model: A discriminant network trained with artificial historical coding data simulates the behavior of human coders, performs confidence scoring on candidate codes, and obtains a coding rationality score; Strategy optimization: Maximize the reward function based on the PPO algorithm and update the generation model parameters.

[0046] In this embodiment, see Figure 3 As shown in the figure, the coding rule generation model and verification model are trained based on the reinforcement learning framework to generate candidate codes and their confidence scores, including the following steps: Step S301: Based on the graph neural network, high-frequency co-occurring diagnosis and treatment patterns and coding mapping rules are mined, and the rule confidence, support and hospital-level specific weight are calculated.

[0047] In this step, the graph neural network (GNN) model is used to mine common diagnosis and treatment patterns from the encoded case data, such as the common combination of "diabetes and hypertension". By calculating the confidence and support of these patterns, the GNN learns the medical record data through the graph structure, mines the association patterns between symptoms, and evaluates their frequency (support) and accuracy (confidence).

[0048] Step S302: parse the new medical record text, extract key entities, match them with the knowledge graph, activate association rules to generate candidate codes, and generate a recommendation list based on confidence and rule priority.

[0049] In this step, when a new medical record is entered, the graph neural network searches for relevant diagnosis and treatment models, generates candidate codes, and then generates a recommendation list based on rule priority. The candidate codes are generated by the rule generation model, and the rationality of the candidate codes is verified by the discriminant network. The discriminant network is trained with historical manual coding data to simulate the judgment process of human coders and evaluate the rationality of the codes.

[0050] Step S40: Modify the candidate encoding based on manual feedback, and use the graph neural network to dynamically update the knowledge graph and rule base.

[0051] In this step, the knowledge graph and rule base are dynamically updated using the graph neural network, including: Node update: When adding a rare disease entity, similar case subgraphs are matched using locality sensitive hashing (LSH) and association rules are inherited; Edge weight update: adjust the weight coefficient of the "Disease→Surgery" edge according to the frequency of coding usage.

[0052] In the embodiments of the present invention, see Figure 4 As shown, the AI-based intelligent generation method for medical record diagnosis and surgical operation codes also includes cross-hospital medical record coding knowledge transfer, and the steps are as follows: Step S501: Aggregate the local knowledge graph features of multiple hospitals through federated learning to generate a global coding rule model.

[0053] In this step, coding data from different hospitals is stored locally. Knowledge graph features from multiple hospitals are aggregated using a federated learning framework to generate a global coding rule model. Federated learning allows multiple hospitals to collaboratively optimize coding rules by updating and aggregating model parameters without sharing data. Each hospital's data is computed locally, ensuring privacy.

[0054] Step S502: Use knowledge distillation to compress the global model into a lightweight migration model for the new hospital to initialize the local system.

[0055] In this step, the global encoding model is compressed into a lightweight model through knowledge distillation technology, so that it can adapt to the local system of the new hospital and reduce computational overhead.

[0056] Step S503: Protect hospital data privacy based on differential privacy technology and add noise disturbance during the knowledge transfer process.

[0057] During the knowledge transfer process, differential privacy technology is used to add noise to the data, ensuring that hospital patient data is not leaked. By adding noise perturbations to hospital data, differential privacy ensures that patient privacy is protected while maintaining the effectiveness of the overall knowledge transfer process.

[0058] The lightweight migration model is constructed in the following way: Extract high-frequency diagnosis and treatment patterns from the global model as meta-rules; Encode the meta-rules as the initial weight matrix of the graph attention mechanism (GAT); Fine-tune the attention weights on the local data of the new hospital to adapt to the hospital-level terminology habits.

[0059] The AI-based intelligent generation method of medical record diagnosis and surgical operation codes of the present invention integrates data from medical record texts, medical images, surgical videos and pathology reports to construct a hospital-level knowledge graph covering the entire diagnosis and treatment process, which helps to comprehensively capture the patient's condition and treatment process, improve the accuracy of coding, and construct a dynamic causal chain reasoning mechanism by embedding time and space dimensions. The spatiotemporal modeling method can help the system better understand the development process of the disease and related treatments, support more accurate coding reasoning, and use the generative adversarial network framework to train the coding rule generation model and verification model. The rule generation model generates candidate codes, and the verification model scores the candidate codes. By continuously optimizing the rule base, the coding system is getting closer and closer to the actual coding preferences of the hospital; and through federated learning and knowledge distillation, the problem of knowledge transfer across hospitals is solved, so that hospitals can quickly start and optimize the medical record coding system through a lightweight migration model without a large amount of local data.

[0060] It should be noted that the above figures are merely illustrative of the processes included in the method according to exemplary embodiments of the present invention and are not intended to be limiting. It is readily understood that the processes illustrated in the above figures do not indicate or limit the temporal order of these processes. Furthermore, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0061] It should be understood that, although the above is described in a certain order, these steps are not necessarily performed in sequence according to the above order. Unless clearly stated herein, the execution of these steps does not have strict order restrictions, and these steps can be performed in other orders. Moreover, a part of the steps of the present embodiment may include multiple steps or multiple stages, and these steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of the steps or stages in other steps or other steps.

[0062] In a second aspect of the embodiments of the present invention, the present invention further provides an AI-based intelligent generation system for medical record diagnosis and surgical operation codes, comprising: Multimodal data parsing module for processing text, image and video data; Spatiotemporal knowledge graph database, which stores medical entities and relationships with spatiotemporal attributes; Reinforcement learning engine, including generator, discriminator and reward calculation unit; Incremental learning interface supports cross-institutional knowledge transfer and dynamic update of local rule base.

[0063] In the present invention, in the spatiotemporal knowledge graph database, the knowledge graph is a database of medical entities and their relationships containing spatiotemporal attributes, which is used to store medical knowledge extracted from different data sources. Among them, the spatiotemporal attributes of the knowledge graph are assigned time and space attributes to each knowledge node (such as disease, medicine, surgery, etc.). The time attribute indicates the order of occurrence of diagnosis and treatment events, and the spatial attribute is associated with the anatomical location in the imaging data, especially the surgical site. For example: in the database: a disease node may include the timestamp of the first diagnosis of the disease, and the surgical node includes the start time and end time of the surgery, as well as the three-dimensional spatial coordinates of the surgical site.

[0064] In the present invention, the reinforcement learning engine includes a generator, a discriminator, and a reward calculation unit, which are used to generate surgical operation codes and diagnostic codes and verify their accuracy. The generator is used to generate preliminary coding candidates (such as ICD-10 diagnosis codes, CPT surgical operation codes, etc.) based on input data such as medical records and images. The discriminator is used to use the discriminant network trained with historical coding data to determine whether the generated candidate codes are reasonable, simulating the judgment process of a human coder. The reward calculation unit is used to use the PPO (Proximal Policy Optimization) algorithm to continuously optimize the generator through the reinforcement learning engine to maximize the reward for generating accurate codes. For example: when a new medical record is input, the generator will combine the knowledge graph and learned rules to generate candidate codes (such as the ICD code corresponding to "diabetes", and the discriminator will verify these codes. If the code meets the standards, the reward calculation unit will strengthen the learning of the generator through a reward mechanism.

[0065] In this embodiment, the multimodal data parsing module integrates the following sub-models: Text Entity Extraction Model: A named entity recognition network jointly trained with BioBERT and CRF. This text entity extraction model uses BioBERT (a BERT model pre-trained in the medical field) and conditional random fields (CRF) for named entity recognition (NER). It can extract medical entities such as disease names, medications, treatments, and patient symptoms from electronic medical records, pathology reports, and other text. For example, if a medical record contains the sentence "The patient suffered a headache due to hypertension and was given antihypertensive medication," the system will use BioBERT to identify "hypertension" and "headache" as disease entities, and "antihypertensive medication" as a medication entity.

[0066] Image feature extraction model: A surgical video keyframe classifier based on 3D-ResNet. This image feature extraction model, based on a 3D-ResNet model, processes medical imaging data, particularly keyframes in surgical videos. By extracting and analyzing video frames, the system can identify lesions, anatomical structures, and their changes. For example, for example, a surgical video depicts a tumor removal procedure. By analyzing keyframes, the system can extract the tumor's location, size, and the extent of the removal process, providing data support for subsequent encoding.

[0067] Temporal Alignment Model: A Transformer-based diagnostic and treatment event sequence generator. This model, based on the Transformer architecture, processes a patient's diagnostic and treatment event sequence, including events extracted from timestamps in medical records and medical imaging and video data, to generate a sequential stream of diagnostic and treatment events. For example, if a patient undergoes different stages of treatment, from examination to diagnosis to surgery, the temporal alignment model can arrange these different events in chronological order and provide a foundation for subsequent encoding.

[0068] The incremental learning interface supports: Version control and rollback of local rule base; Blockchain-based rule update log storage ensures data traceability; Automate compliance checks to prevent coding rules from violating healthcare regulatory policies.

[0069] In the present invention, the incremental learning interface realizes cross-hospital knowledge migration and dynamic update of local rule bases, and provides version control and log evidence functions. Among them, it has local rule base version control and rollback functions, supports version control of the hospital's local rule base, and can roll back to the previous coding version as needed to cope with changes in coding rules. The incremental learning interface can also be based on blockchain rule update log evidence. All rule updates and modification records are recorded through blockchain technology to ensure that the data cannot be tampered with and enhance the transparency and traceability of the system. For example: when the hospital needs to update the coding rules, the system will record the time, content and responsible person of each update through the blockchain to ensure that the specific modification history can be traced back in the future.

[0070] The AI-based medical record diagnosis and surgical operation code intelligent generation system of the present invention realizes the intelligent processing of medical record data by combining multimodal data processing, knowledge graph construction, reinforcement learning and incremental learning. This system can not only improve the efficiency and accuracy of diagnosis and surgical coding, but also adapt to the needs of different hospitals, and ensure the timeliness and compliance of coding rules through cross-hospital knowledge migration and local rule base updates. In addition, the system of the present invention enhances the traceability and transparency of data by integrating blockchain technology, and ensures compliance with medical regulatory policies through automated compliance checks. Through continuous learning and optimization, the system can adapt to new medical knowledge and coding standards to ensure long-term effectiveness.

[0071] Through the above detailed steps, the AI-based medical record diagnosis and surgical operation code intelligent generation system of the present invention is used to execute the steps of the AI-based medical record diagnosis and surgical operation code intelligent generation method in the above embodiment, which will not be repeated here.

[0072] This AI-based system for intelligently generating medical record diagnoses and surgical operation codes fully implements efficient integration and reasoning of diverse data types. It uses graph neural networks to dynamically adjust node embedding representations, further optimizing the accuracy of the knowledge graph. Furthermore, the system supports cross-hospital knowledge transfer, using federated learning and differential privacy technologies to protect data privacy and enable knowledge sharing across multiple hospitals.

[0073] According to a third aspect of an embodiment of the present invention, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the method of any one of the above embodiments is implemented.

[0074] The computer device includes a processor and a memory, and may also include an input system and an output system. The processor, memory, input system, and output system may be connected via a bus or other means. The input system may receive digital or character input and generate signal input related to the intelligent generation and migration of AI-based medical diagnosis and surgical operation codes. The output system may include a display device such as a display screen.

[0075] As a non-volatile computer-readable storage medium, the memory can be used to store non-volatile software programs, non-volatile computer executable programs, and modules, such as the program instructions / modules corresponding to the AI-based medical record diagnosis and surgical operation code intelligent generation method in the embodiments of the present application. The memory may include a program storage area and a data storage area, wherein the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created by using the AI-based medical record diagnosis and surgical operation code intelligent generation method, etc. In addition, the memory may include high-speed random access memory and may also include non-volatile memory, such as at least one disk storage device, flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory may optionally include a memory remotely located relative to the processor, and these remote memories can be connected to the local module via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, corporate intranet, local area network, mobile communication network, and combinations thereof.

[0076] In some embodiments, the processor may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor is generally used to control the overall operation of the computer device. In this embodiment, the processor is used to run the program code stored in the memory or process data. The processors of the multiple computer devices of the computer device of this embodiment execute various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory, that is, implementing the steps of the intelligent generation method of medical record diagnosis and surgical operation codes based on AI in the above-mentioned method embodiment.

[0077] It should be understood that, to the extent that they do not conflict with each other, all the embodiments, features and advantages described above for the method for intelligently generating AI-based medical record diagnosis and surgical operation codes according to the present invention are also applicable to the intelligent generation and storage medium for AI-based medical record diagnosis and surgical operation codes according to the present invention.

[0078] It will also be appreciated by those skilled in the art that the various exemplary logic blocks, modules, circuits and algorithmic steps described in conjunction with the disclosure herein can be implemented as electronic hardware, computer software or a combination of the two. In order to clearly illustrate this interchangeability of hardware and software, a general description has been given of the functions of various schematic components, blocks, modules, circuits and steps. Whether this function is implemented as software or hardware depends on specific applications and the design constraints imposed on the entire system. Those skilled in the art can implement the function in various ways for each specific application, but this implementation decision should not be interpreted as causing a departure from the disclosed scope of the embodiments of the present invention.

[0079] Finally, it should be noted that the computer-readable storage medium (e.g., memory) herein may be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. By way of example and not limitation, non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which may act as external cache memory. By way of example and not limitation, RAM is available in various forms, such as synchronous RAM (DRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct Rambus RAM (DRRAM). The memory devices of the disclosed aspects are intended to include, but are not limited to, these and other suitable types of memory.

[0080] The various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure herein may be implemented or executed using a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination of these components, designed to perform the functions herein. A general-purpose processor may be a microprocessor, but alternatively, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, for example, a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP, and / or any other such configuration.

[0081] The above are exemplary embodiments disclosed in the present invention, but it should be noted that various changes and modifications may be made without departing from the scope of the embodiments disclosed in the claims. The functions, steps and / or actions of the method claims according to the disclosed embodiments described herein do not need to be performed in any particular order. In addition, although the elements disclosed in the embodiments of the present invention may be described or required in individual form, they may also be understood as multiple unless expressly limited to the singular.

[0082] It should be understood that, as used herein, the singular form "a" or "an" is intended to include the plural form as well, unless the context clearly supports an exception. It should also be understood that, as used herein, "and / or" refers to any and all possible combinations of one or more of the items listed in association. The serial numbers of the embodiments disclosed in the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0083] Those skilled in the art should understand that the discussion of any of the above embodiments is merely illustrative and is not intended to limit the scope of the disclosure of the present invention (including the claims) to these examples. Within the spirit of the present invention, the technical features of the above embodiments or different embodiments may be combined, and many other variations exist in different aspects of the above embodiments, which are not provided in detail for the sake of clarity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An AI-based method for intelligently generating medical record diagnosis and surgical operation codes, characterized in that: The method comprises the following steps: Acquire multimodal data from hospital information systems, including text medical records, medical images, surgical videos, and pathology reports; Through the multimodal fusion model, entities and relationships are extracted to build a hospital-level disease knowledge graph with spatiotemporal dimensions. Based on the reinforcement learning framework, the coding rule generation model and verification model are trained to generate candidate codes and their confidence scores; Modify candidate codes based on manual feedback, and use graph neural networks to dynamically update knowledge graphs and rule bases.

2. The AI-based intelligent generation method for medical diagnosis and surgical operation codes according to claim 1, characterized in that: The multimodal fusion model is used to extract entities and relationships and construct a hospital-level disease knowledge graph that includes spatiotemporal dimensions. The following steps are involved: Clean multimodal data obtained from hospital information systems, uniformly encode text medical records, image labels, and laboratory time series data into heterogeneous nodes in the knowledge graph, and annotate disease code labels; Use the BERT-BiLSTM-CRF model to extract entities and relationships of symptoms, examinations, and medications from multimodal data; Through semantic version control, the hospital automatically detects differences between the terminology library and the standard coding system, generates incremental alignment rules, aligns entities with standard medical terminology, and constructs a hospital-level disease knowledge graph containing multi-dimensional diagnosis and treatment data; When a new term is detected, the knowledge graph node is automatically expanded and associated with a temporary encoding.

3. The AI-based intelligent generation method for medical diagnosis and surgical operation codes according to claim 2, characterized in that: Constructing a hospital-level disease knowledge graph with spatiotemporal dimensions also includes: Timeline: embeds diagnosis and treatment events into timestamp attributes of graph nodes according to their occurrence time; Spatial axis: Annotate the 3D spatial coordinates of anatomical structures based on medical image segmentation results and construct a sub-atlas of surgical sites.

4. The AI-based intelligent generation method for medical diagnosis and surgical operation codes according to claim 3, characterized in that: Before training the coding rule generation model and verification model based on the reinforcement learning framework, it includes: Rule generation model: Generate candidate coding sequences through random walk sampling of knowledge graph; Rule verification model: A discriminant network trained with artificial historical coding data simulates the behavior of human coders, performs confidence scoring on candidate codes, and obtains a coding rationality score; Strategy optimization: Maximize the reward function based on the PPO algorithm and update the generation model parameters.

5. The AI-based intelligent generation method for medical diagnosis and surgical operation codes according to claim 1, characterized in that: The coding rule generation model and verification model are trained based on the reinforcement learning framework to generate candidate codes and their confidence scores. The following steps are included: Based on graph neural networks, we mine high-frequency co-occurring diagnosis and treatment patterns and coding mapping rules, and calculate the rule confidence, support, and hospital-level specific weight. By parsing the new medical record text, extracting key entities and matching them with the knowledge graph, activating association rules to generate candidate codes, and combining confidence and rule priority to generate a recommendation list.

6. The AI-based intelligent generation method for medical diagnosis and surgical operation codes according to claim 5, characterized in that: Use graph neural networks to dynamically update knowledge graphs and rule bases, including: Node update: When adding a rare disease entity, similar case subgraphs are matched through local sensitive hashing and association rules are inherited; Edge weight update: Adjust the weight coefficient of the "Disease→Surgery" edge based on the frequency of coding usage.

7. The AI-based intelligent generation method for medical diagnosis and surgical operation codes according to claim 1, characterized in that: The AI-based intelligent generation method for medical record diagnosis and surgical operation codes also includes cross-hospital medical record coding knowledge transfer, and the steps are as follows: Aggregate local knowledge graph features of multiple hospitals through federated learning to generate a global encoding rule model; Use knowledge distillation to compress the global model into a lightweight transfer model for new hospitals to initialize their local systems; Based on differential privacy technology, hospital data privacy is protected and noise perturbation is added during the knowledge transfer process.

8. The AI-based intelligent generation method for medical diagnosis and surgical operation codes according to claim 7, characterized in that: The lightweight migration model is constructed in the following way: Extract high-frequency diagnosis and treatment patterns from the global model as meta-rules; Encode the meta-rules as the initial weight matrix of the graph attention mechanism; Fine-tune the attention weights on the local data of the new hospital to adapt to the hospital-level terminology habits.

9. An AI-based medical record diagnosis and surgical operation code intelligent generation system, characterized by: A method for intelligently generating medical record diagnosis and surgical operation codes based on AI according to any one of claims 1 to 8, the system comprising: Multimodal data parsing module for processing text, image and video data; Spatiotemporal knowledge graph database, which stores medical entities and relationships with spatiotemporal attributes; Reinforcement learning engine, including generator, discriminator and reward calculation unit; Incremental learning interface supports cross-institutional knowledge transfer and dynamic update of local rule base.

10. The AI-based medical record diagnosis and surgical operation code intelligent generation system according to claim 9, characterized in that: The multimodal data parsing module integrates the following sub-models: Text entity extraction model: Named entity recognition network jointly trained with BioBERT and CRF; Image feature extraction model: surgical video key frame classifier based on 3D-ResNet; Temporal alignment model: Transformer-based diagnostic and treatment event sequence generator.

Images