Medical image report structuring method and related device

Abstract

The invention discloses a medical image report structuring method and a related device, and relates to the technical field of natural language processing, and the medical image report structuring method comprises the steps: obtaining a target medical image report; based on a pre-obtained medical image report structured model, structured processing is conducted on the target medical image report, structured information of the target medical image report is obtained, and the medical image report structured model is obtained by conducting instruction fine adjustment on a pre-trained large language model through training data; the training data at least comprises a first training medical image report marked with structured information. The medical image report structuring method disclosed by the invention is an end-to-end structuring method based on a large language model, the key information contained in the medical image report can be accurately, efficiently and normatively extracted by the method, and the method is simple in implementation process and relatively low in manual annotation cost.

Description

Technical Field

[0001] The present application relates to the field of natural language processing technology, and in particular to a medical imaging report structuring method and related devices. Background Art

[0002] Medical imaging reports carry rich and valuable clinical information, including imaging findings, detailed descriptions, and diagnostic results of various examinations. However, these reports are usually stored in a semi-structured or even unstructured form, with many scattered and non-standardized descriptive terms, which poses great challenges to the analysis, processing, and mining of clinical information.

[0003] Text structuring technology in natural language processing can standardize and structure the extraction of various information in medical imaging reports, and at the same time associate the main information with related subsidiary information, facilitating subsequent automated data mining and processing. With the booming development of artificial intelligence technology today, the structured information of medical imaging reports can also prepare high-quality, standardized, and information-focused corpus for the training of medical artificial intelligence models, and enhance the model's ability to learn medical knowledge, discover clinical events, and infer disease diagnosis from massive, complex, and redundant original report texts, which is of great significance for intelligent and automated medical decision-making.

[0004] Since there are a large number of image findings, diagnostic entities, and related clinical events in medical image reports, the current medical image report structuring scheme mainly breaks down the medical image report structuring task into multiple tasks such as noun recognition, relationship extraction, and terminology normalization, and trains different natural language processing models on different tasks, and then uses the trained natural language processing model to process the corresponding task, and finally integrates the processing results of each task into the final structured information. Although the current medical image report structuring method can obtain the structured information of medical image reports, the implementation process is relatively complicated. In addition, since each task requires a training model, and the training model requires manually labeled data, the manual labeling cost of the current medical image report structuring method is relatively high. Summary of the invention

[0005] In view of this, the present application provides a medical imaging report structuring method and related devices, which are used to solve the problems of the existing medical imaging report structuring method having a relatively complex implementation process and high manual annotation cost. The technical solution is as follows:

[0006] The first aspect of the present application provides a method for structuring a medical imaging report, comprising:

[0007] Obtain target medical imaging reports;

[0008] Based on the pre-acquired medical image report structured model, the target medical image report is structured to obtain structured information of the target medical image report;

[0009] The medical imaging report structured model is obtained by fine-tuning a pre-trained large language model using training data, and the training data at least includes a first training medical imaging report annotated with structured information.

[0010] In a possible implementation, the step of performing structural processing on the target medical image report based on the pre-acquired medical image report structural model to obtain structural information of the target medical image report includes:

[0011] Processing the preset instruction prompt information and the target medical image report into prompt instructions to obtain prompt instructions corresponding to the target medical image report, wherein the prompt instructions corresponding to the target medical image report are used to prompt the medical image report structured model to list key information in the target medical image report;

[0012] The prompt instruction corresponding to the target medical image report is input into the medical image report structured model to obtain the structured information of the target medical image report output by the medical image report structured model.

[0013] In a possible implementation, the structured information includes: an entity category label and an anatomical location label and an additional information label corresponding to the entity category label, wherein the entity category label is one of a normal entity label and an abnormal entity label.

[0014] In a possible implementation, the process of fine-tuning the pre-trained large language model using training data to obtain a medical imaging report structured model includes:

[0015] Using the first training medical imaging report annotated with structured information, fine-tuning the pre-trained large language model, and using the fine-tuned large language model as a medical imaging report structured model;

[0016] or,

[0017] The first training medical imaging report annotated with structured information is used to fine-tune the pre-trained large language model;

[0018] Based on the large language model fine-tuned by the instruction, a structural processing is performed on the unlabeled second training medical image report to obtain structured information of the second training medical image report as labeling information of the second training medical image report;

[0019] A first training medical imaging report annotated with structured information and a second training medical imaging report annotated with structured information are used to further fine-tune the large language model after instruction fine-tuning, and the large language model after further instruction fine-tuning is used as a medical imaging report structured model.

[0020] In a possible implementation, the first training medical imaging report annotated with structured information is used to fine-tune the pre-trained large language model, including:

[0021] Processing the preset instruction prompt information and the first training medical image report into prompt instructions, and obtaining prompt instructions corresponding to the first training medical image report, wherein the prompt instructions corresponding to the first training medical image report are used to prompt the pre-trained large language model to list key information in the first training medical image report;

[0022] Inputting the prompt instruction corresponding to the first training medical imaging report into the pre-trained large language model to obtain a structured information prediction result of the first training medical imaging report;

[0023] Determining a first prediction loss according to the structured data prediction result of the first training medical imaging report and the structured information annotated in the first training medical imaging report;

[0024] According to the first prediction loss, parameters of the pre-trained large language model are updated.

[0025] In a possible implementation, the method of fine-tuning the pre-trained large language model using the first training medical imaging report annotated with structured information further includes:

[0026] Performing data enhancement processing on the first training medical image report annotated with structured information to obtain an enhanced medical image report annotated with structured information;

[0027] Processing the preset instruction prompt information and the enhanced medical image report into prompt instructions, generating prompt instructions corresponding to the enhanced medical image report, wherein the prompt instructions corresponding to the enhanced medical image report are used to prompt the pre-trained large language model to list key information in the enhanced medical image report;

[0028] Inputting the prompt instruction corresponding to the enhanced medical imaging report into the pre-trained large language model to obtain the structured information prediction result of the enhanced medical imaging report;

[0029] Determining a second prediction loss according to the structured information prediction result of the enhanced medical imaging report and the structured information annotated in the enhanced medical imaging report;

[0030] The updating of parameters of the pre-trained large language model according to the first prediction loss includes:

[0031] Parameters of the pre-trained large language model are updated according to the first prediction loss and the second prediction loss.

[0032] In a possible implementation, performing data enhancement processing on the first training medical image report annotated with structured information to obtain an enhanced medical image report annotated with structured information includes:

[0033] Using one or more of several data enhancement methods to perform data enhancement processing on the first training medical image report to obtain an enhanced medical image report, wherein the several data enhancement methods include inserting segmentation words, disrupting the order of sentences, and report splicing, and the report splicing refers to splicing another first training medical image report to the first training medical image report that needs data enhancement;

[0034] If data enhancement processing is performed on the first training medical image report by inserting segmentation words and / or disrupting the order of sentences, the annotation information of the first training medical image report is determined as the annotation information of the enhanced medical image report;

[0035] If at least one report splicing method is used to perform data enhancement processing on the first training medical image report, the labeling information of the two spliced ​​first training medical image reports is spliced ​​and de-redundant to obtain the labeling information of the enhanced medical image report.

[0036] A second aspect of the present application provides a medical imaging report structuring device, comprising: a medical imaging report acquisition module and a medical imaging report structuring module;

[0037] The medical imaging report acquisition module is used to acquire a target medical imaging report;

[0038] The medical image report structuring module is used to perform structural processing on the target medical image report based on a pre-acquired medical image report structuring model to obtain structural information of the target medical image report;

[0039] The medical imaging report structured model is obtained by fine-tuning a pre-trained large language model using training data, and the training data at least includes a first training medical imaging report annotated with structured information.

[0040] A third aspect of the present application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0041] The memory is used to store computer programs;

[0042] The processor is used to execute the computer program so that the electronic device can implement the steps of any one of the above-mentioned medical imaging report structuring methods.

[0043] A fourth aspect of the present application provides a computer storage medium, which carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the steps of any one of the medical imaging report structuring methods described above.

[0044] By means of the above technical scheme, the medical image report structuring method provided in the present application first obtains the target medical image report, and then performs structured processing on the target medical image report based on the pre-acquired medical image report structuring model to obtain the structured information of the target medical image report. Since the medical image report structuring model in the present application uses the training medical image report annotated with structured information to perform instruction fine-tuning on the pre-trained large language model, the target medical image report is processed based on the medical image report structuring model to directly obtain the structured information. That is, the medical image report structuring method provided in the present application is an end-to-end structuring method based on the large language model. The implementation process of the method is simple, and the key information contained in the medical image report can be accurately, efficiently and normatively extracted. In addition, since the medical image report structuring method provided in the present application only needs to train one model, and the strong generalization learning ability exhibited by the large language model enables fine-tuning of a small amount of data to achieve ideal performance, the amount of data that needs to be manually labeled is relatively small, and thus the cost of manual labeling is low. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0046] Figure 1 A schematic diagram of the system architecture involved in this application;

[0047] Figure 2 A schematic diagram of a hardware structure of a terminal provided in an embodiment of the present application;

[0048] Figure 3 A schematic diagram of a hardware structure of a server provided in an embodiment of the present application;

[0049] Figure 4A schematic diagram of a process for a medical imaging report structured method provided in an embodiment of the present application;

[0050] Figure 5 A schematic diagram of fine-tuning instructions on a pre-trained large language model using only a first training medical imaging report annotated with structured information provided in an embodiment of the present application;

[0051] Figure 6 A schematic diagram of fine-tuning instructions on a pre-trained large language model using a first training medical image report annotated with structured information and an enhanced medical image report annotated with structured information provided in an embodiment of the present application;

[0052] Figure 7 A schematic diagram of further fine-tuning a large language model after fine-tuning instructions using a first training medical image report annotated with structured information, an enhanced medical image report annotated with structured information, and a second training medical image report annotated with structured information provided in an embodiment of the present application;

[0053] Figure 8 A schematic diagram of structuring a target medical imaging report using a medical imaging report structuring model provided in an embodiment of the present application;

[0054] Fig. 9 A schematic diagram of the structure of a medical imaging report structuring device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0055] The following describes the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. The terms used in the implementation method section of the present application are only used to explain the specific embodiments of the present application, and are not intended to limit the present application.

[0056] The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0057] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, which is only to describe the distinction mode adopted by the objects of the same attributes when describing in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0058] Before introducing the medical imaging report structuring method provided by this application, the system architecture involved in this application is first explained.

[0059] like Figure 1 As shown, the system architecture involved in the present application may include a terminal 100 and a server 200, wherein the server 200 may include one or more servers ( Figure 1 A server is included as an example for explanation).

[0060] The user may use the terminal 100 to interact with the server 200 through a network (wired network or wireless network) to receive or send messages, etc. Various client applications may be installed on the terminal 100 .

[0061] Next, the product form of the terminal 100 will be described.

[0062] The terminal 100 mentioned above can be a mobile phone, a tablet computer, a wearable device, a vehicle-mounted device, an augmented reality (AR) / virtual reality (VR) device, a laptop computer, an ultra-mobile personal computer (UMPC), a netbook, a personal digital assistant (PDA), etc., and the embodiments of the present application do not impose any restrictions on this.

[0063] Figure 2 An optional hardware structure diagram of the terminal 100 is shown.

[0064] refer to Figure 2 As shown, the terminal 100 may include a radio frequency unit 110, a memory 120, an input unit 130, a display unit 140, a camera 150 (optional), an audio circuit 160 (optional), a speaker 161 (optional), a microphone 162 (optional), an earphone jack 163 (optional), a processor 170, an external interface 180, a power supply 190 and other components. Those skilled in the art will appreciate that Figure 2 The terminal 100 is merely an example and does not constitute a limitation on the terminal 100 , and may include more or less components than those shown in the figure, or may combine certain components, or may include different components.

[0065] The input unit 130 can be used to receive input digital or character information, and generate key signal input related to the user settings and function control of the terminal 100. Specifically, the input unit 130 may include a touch screen 131 (optional) and / or other input devices 132. The touch screen 131 can collect the user's touch operations on or near it (such as the user's operation on or near the touch screen using any suitable object such as fingers, joints, stylus, etc.), and drive the corresponding connection device according to a pre-set program. The touch screen can detect the user's touch action on the touch screen, convert the touch action into a touch signal and send it to the processor 170, and can receive and execute the command sent by the processor 170; the touch signal at least includes the touch point coordinate information. The touch screen 131 can provide an input interface and an output interface between the terminal 100 and the user. In addition, the touch screen can be implemented using multiple types such as resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch screen 131, the input unit 130 can also include other input devices. Specifically, other input devices 132 may include, but are not limited to, one or more of a physical keyboard, function keys (such as a volume control key, a switch key, etc.), a trackball, a mouse, a joystick, and the like.

[0066] The display unit 140 may be used to display information input by a user or information provided to a user, various menus of the terminal 100, an interactive interface, file display, and / or playback of any multimedia file.

[0067] The memory 120 can be used to store instructions and data. The memory 120 can mainly include an instruction storage area and a data storage area. The data storage area can store various data, such as multimedia files, texts, etc.; the instruction storage area can store software units such as operating systems, applications, instructions required for at least one function, or their subsets and extensions. It can also include a non-volatile random access memory; provide the processor 170 with hardware, software and data resources including management of computing and processing equipment, and support control software and applications. It is also used for the storage of multimedia files, and the storage of running programs and applications.

[0068] The processor 170 is the control center of the terminal 100. It uses various interfaces and lines to connect various parts of the entire terminal 100. By running or executing instructions stored in the memory 120 and calling data stored in the memory 120, it executes various functions of the terminal 100 and processes data, thereby controlling the terminal 100 as a whole. Optionally, the processor 170 may include one or more processing units; preferably, the processor 170 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface and application program, and the modem processor mainly processes wireless communication. It is understandable that the above-mentioned modem processor may not be integrated into the processor 170. In some embodiments, the processor and the memory may be implemented on a single chip, and in some embodiments, they may also be implemented separately on independent chips. The processor 170 may also be used to generate corresponding operation control signals, send them to corresponding components of the computing and processing device, read and process data in the software, especially read and process data and programs in the memory 120, so that each functional module therein performs corresponding functions, thereby controlling the corresponding components to act according to the requirements of the instructions.

[0069] The radio frequency unit 110 (optional) can be used to receive and send information or receive and send signals during a call. For example, after receiving the downlink information of the base station, it is sent to the processor 170 for processing; in addition, the designed uplink data is sent to the base station. Generally, the radio frequency unit 110 includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (LowNoiseAmplifier, LNA), a duplexer, etc. In addition, the radio frequency unit 110 can also communicate with network devices and other devices through wireless communication. The wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile communication (Global System of Mobile communication, GSM), General Packet Radio Service (General PacketRadio Service, GPRS), Code Division Multiple Access (Code Division Multiple Access, CDMA), Wideband Code Division Multiple Access (Wideband Code Division Multiple Access, WCDMA), Long Term Evolution (Long Term Evolution, LTE), email, Short Messaging Service (SMS), etc.

[0070] In the embodiment of the present application, the radio frequency unit 110 can send data to other devices, and can also receive data sent by other devices. It should be understood that the radio frequency unit 110 is optional, and can be replaced by other communication interfaces, such as a network port.

[0071] The terminal 100 also includes a power supply 190 (such as a battery) for supplying power to various components. Preferably, the power supply can be logically connected to the processor 170 through a power management system, so that the power management system can manage functions such as charging, discharging, and power consumption.

[0072] The terminal 100 further includes an external interface 180 , which may be a standard Micro USB interface or a multi-pin connector, and may be used to connect the terminal 100 to communicate with other devices, or to connect a charger to charge the terminal 100 .

[0073] Although not shown, the terminal 100 may further include a flashlight, a wireless fidelity (WiFi) module, a Bluetooth module, sensors with different functions, etc., which will not be described in detail herein.

[0074] Next, the product form of the server 200 will be described.

[0075] Figure 3 A structural diagram of the server 200 is provided. Figure 3 As shown, the server 200 includes a bus 201, a processor 202, a communication interface 203, and a memory 204. The processor 202, the memory 204, and the communication interface 203 communicate with each other via the bus 201.

[0076] The bus 201 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0077] The processor 202 may be any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

[0078] The memory 204 may include a volatile memory, such as a random access memory (RAM). The memory 204 may also include a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0079] It should be noted that the medical imaging report structuring method provided in the embodiment of the present application is generally executed by the server 200. In view of this, the memory 204 can be used to store software codes related to the medical imaging report structuring method, and the processor 202 can call the software code stored in the memory 204, and can also schedule other units to implement corresponding functions.

[0080] It should be understood that the above-mentioned terminal 100 and server 200 can be centralized or distributed devices, and the processors in the above-mentioned terminal 100 and server 200 (such as processor 170 and processor 202) can be hardware circuits (such as application specific integrated circuit (ASIC), field-programmable gate array (FPGA), general-purpose processor, digital signal processor (DSP), microprocessor or microcontroller, etc.), or a combination of these hardware circuits. For example, the processor can be a hardware system with an instruction execution function, such as a CPU, DSP, etc., or a hardware system without an instruction execution function, such as an ASIC, FPGA, etc., or a combination of the above-mentioned hardware systems without an instruction execution function and hardware systems with an instruction execution function.

[0081] Next, the medical imaging report structuring method provided by this application is introduced through the following embodiments.

[0082] See also Figure 4 , shows a schematic flow chart of a medical imaging report structuring method provided in an embodiment of the present application, and the medical imaging report structuring method may include:

[0083] Step S401: Obtain a target medical imaging report.

[0084] The target medical imaging report is a medical imaging report to be structured.

[0085] Step S402: Based on the pre-acquired medical image report structured model, the target medical image report is structured to obtain structured information of the target medical image report.

[0086] Specifically, based on the pre-acquired medical image report structured model, the process of performing structured processing on the target medical image report to obtain structured information of the target medical image report may include:

[0087] Step S4021: Process the preset instruction prompt information and the target medical imaging report into a prompt instruction, and obtain a prompt instruction corresponding to the target medical imaging report.

[0088] Exemplarily, the preset instruction prompt information may be “Suppose you are a medical expert, and your task is to list all the key information mentioned in the medical imaging report.”.

[0089] In this embodiment, the preset instruction prompt information can be spliced ​​with the target medical imaging report to obtain the prompt instruction corresponding to the target medical imaging report. The prompt instruction corresponding to the target medical imaging report is used to prompt the medical imaging report structured model to list the key information in the target medical imaging report.

[0090] Step S4022: input the prompt instruction corresponding to the target medical image report into the medical image report structured model to obtain the structured information of the target medical image report output by the medical image report structured model.

[0091] The medical imaging report structured model in this embodiment is obtained by fine-tuning the pre-trained large language model using training data, wherein the training data at least includes a plurality of first training medical imaging reports annotated with structured information.

[0092] The annotation information of the first training medical imaging report can be obtained in the following manner: according to a pre-constructed medical imaging knowledge base (which may include multiple medical imaging knowledge information, each of which may but is not limited to include a medical term and the definition of the medical term), the first training medical imaging report is keyword matched according to set rules to obtain rough structured information. After obtaining the rough structured information, it can be manually verified and adjusted to obtain the final more accurate structured information.

[0093] It should be noted that the pre-trained large language model in this embodiment can be any open source large language model with any parameter amount that supports the instruction fine-tuning function.

[0094] Considering that large language models have powerful cross-context matching and recognition capabilities, this application proposes to use large language models to extract valuable clinical information from medical imaging reports and convert them into structured outputs.

[0095] Furthermore, the present application proposes that an instruction fine-tuning strategy can be used to fine-tune the pre-trained large language model to adapt it to the task of structuring medical imaging reports and achieve end-to-end one-time output of structured information.

[0096] In this embodiment, the structured information annotated in the first training medical imaging report may include entity category labels (normal / abnormal entity labels) and anatomical location labels and additional information labels corresponding to the entity category labels. The structured information obtained by structuring the target medical imaging report based on the medical imaging report structuring model also includes entity category labels and anatomical location labels and additional information labels corresponding to the entity category labels. The additional information may include information such as the uncertainty, severity, time progression, relative significant position, and quantitative description of the abnormality in the report. It should be noted that the additional information label corresponding to the normal entity label is empty.

[0097] For example, the target medical imaging report is "possible severe pneumonia in the left lung, and new solid nodules in the upper lobe of the right lung. No abnormality in the shape and size of the heart. Please follow up in combination with clinical practice." The following structured information can be obtained by using the medical imaging report structuring method provided in the embodiment of the present application to structure it:

[0098] - Abnormalities: Pneumonia

[0099] --Location: Left lung

[0100] ---Uncertainty: Uncertain

[0101] ---Severity: Severe

[0102] -Abnormalities: Solid nodules

[0103] --Location: right upper lobe

[0104] ---Time Progress: Added

[0105] -Normal: No abnormality found in the heart

[0106] --Location: Heart

[0107] It should be noted that the structured information can be in the form of a hierarchical JSON format, or a list of triples in the form of <normal / abnormal entity label, anatomical location label, additional information label>, or graph structured data converted from triples.

[0108] The medical image report structuring method provided in the embodiment of the present application first obtains a target medical image report, and then performs structured processing on the target medical image report based on a pre-acquired medical image report structuring model to obtain structured information of the target medical image report. Since the medical image report structuring model in the embodiment of the present application is obtained by fine-tuning the pre-trained large language model using at least a training medical image report annotated with structured information, the target medical image report is processed based on the medical image report structuring model to directly obtain structured information. That is, the medical image report structuring method provided in the embodiment of the present application is an end-to-end structuring method based on a large language model. The implementation process of the method is simple, and the key information contained in the medical image report can be accurately, efficiently and normatively extracted. In addition, since the medical image report structuring method provided in the embodiment of the present application only needs to train one model, and the strong generalization learning ability exhibited by the large language model enables fine-tuning of a small amount of data to achieve ideal performance, the amount of data that needs to be manually labeled is relatively small, and thus the cost of manual labeling is low.

[0109] The above-mentioned embodiment mentioned that the medical imaging report structuring model used for structuring the target medical imaging report is obtained by fine-tuning the instructions of a pre-trained large language model using training data. In another embodiment of the present application, the process of obtaining the medical imaging report structuring model by fine-tuning the instructions of a pre-trained large language model using training data is introduced.

[0110] There are many ways to implement the method of using training data to fine-tune the instructions of the pre-trained large language model to obtain a medical imaging report structured model. This embodiment provides the following two optional implementation methods.

[0111] The first implementation method of using training data (including the first training medical imaging report annotated with structured information) to fine-tune the pre-trained large language model to obtain a medical imaging report structured model:

[0112] Step a: using the first training medical imaging report annotated with structured information, and performing instruction fine-tuning on the pre-trained large language model, and using the instruction fine-tuned large language model as a medical imaging report structured model.

[0113] There are many ways to implement fine-tuning instructions on the pre-trained large language model using the first training medical imaging report annotated with structured information. This embodiment provides the following two optional implementation methods.

[0114] The first implementation method is to use the first training medical imaging report annotated with structured information to fine-tune the instructions of the pre-trained large language model:

[0115] Step a11: Process the preset instruction prompt information and the first training medical image report into prompt instructions, and obtain the prompt instructions corresponding to the first training medical image report.

[0116] Among them, the prompt instruction corresponding to the first training medical imaging report is used to prompt the pre-trained large language model to list the key information in the first training medical imaging report.

[0117] Step a12: input the prompt instruction corresponding to the first training medical imaging report into the pre-trained large language model to obtain the structured information prediction result of the first training medical imaging report.

[0118] like Figure 5 As shown, after the preset instruction prompt information and the first training medical imaging report are processed into prompt instructions, the prompt instructions are input into the pre-trained large language model, and the pre-trained large language model predicts and outputs structured information for the first training medical imaging report based on the input prompt instructions.

[0119] Step a13: determining the prediction loss according to the structured data prediction result of the first training medical image report and the structured information annotated in the first training medical image report.

[0120] Among them, the prediction loss can use the standard autoregressive next token prediction loss.

[0121] Step a14: fine-tune the parameters of the pre-trained large language model according to the prediction loss.

[0122] In this embodiment, the parameter fine-tuning strategy during model training can be a full-parameter fine-tuning strategy, or it can be an efficient parameter fine-tuning strategy such as LORA, Q-LORA, p-tuning, etc.

[0123] The second implementation method is to use the first training medical imaging report annotated with structured information to fine-tune the pre-trained large language model:

[0124] Step a21: perform data enhancement processing on the first training medical image report annotated with structured information to obtain an enhanced medical image report annotated with structured information.

[0125] Specifically, the process of performing data enhancement processing on the first training medical image report annotated with structured information to obtain an enhanced medical image report annotated with structured information may include:

[0126] Step a211: Use one or more of several data enhancement methods to perform data enhancement processing on the first training medical image report to obtain an enhanced medical image report.

[0127] Among them, several data augmentation methods can include inserting segmentation words, disrupting sentence order, and report splicing.

[0128] Inserting a segmentation word refers to inserting a segmentation word at any position in the report. Specifically, a segmentation word can be randomly sampled from a pre-constructed invalid segmentation set, and the sampled segmentation word can be inserted at any position in the report, wherein the segmentation words in the invalid segmentation set are segmentations that do not involve key clinical information. Shuffling the order of sentences refers to shuffling the order of sentences in the report. Specifically, the report can be split into sentences, and then the order of the split sentences can be shuffled, and finally the shuffled sentences can be combined to obtain a new report. Report splicing refers to splicing a first training medical imaging report that needs data enhancement with another first training medical imaging report. Specifically, the two first training medical imaging reports are spliced ​​end to end, and redundant information is removed.

[0129] Step a212-a: If data enhancement processing is performed on the first training medical image report by inserting segmentation words and / or disrupting the order of sentences, the annotation information of the first training medical image report is determined as the annotation information of the enhanced medical image report.

[0130] If only the method of inserting participles is used to perform data enhancement processing on the first training medical image report that needs data enhancement, or only the method of scrambling the sentence order is used to perform data enhancement processing on the first training medical image report that needs data enhancement, or both the method of inserting participles and scrambling the sentences are used to perform data enhancement processing on the first training medical image report that needs data enhancement, then the annotation information of the first training medical image report is used as the annotation information of the enhanced medical image report.

[0131] Exemplarily, the first training medical image report that needs data enhancement is x, and the structured information annotated with x is y. If data enhancement processing is performed on x by inserting word segmentations and / or disrupting the order of sentences, y will be used as the annotation information of the enhanced medical image report x′, thereby obtaining an enhanced medical image report x′ annotated with y.

[0132] Step a212-b: If at least one report splicing method is used to perform data enhancement processing on the first training medical image report, the labeling information of the two spliced ​​first training medical image reports is spliced ​​and de-redundant to obtain the labeling information of the enhanced medical image report.

[0133] If only report splicing is used to perform data enhancement processing on the first training medical image report, or if report splicing is used in combination with inserting word segmentation and / or scrambling the sentence order to perform data enhancement processing on the first training medical image report, the annotation information of the two spliced ​​first training medical image reports is spliced ​​and de-redundant to obtain the annotation information of the enhanced medical image report.

[0134] Exemplarily, the first training medical image report that needs data enhancement is x1, and the structured information annotated with x1 is y1. If data enhancement processing is performed on x1 only by report splicing, then x1 is spliced ​​with another first training medical image report x2 (x1 and x2 are spliced ​​end to end). After splicing x1 and x2, redundant information can be removed to obtain an enhanced medical image report x′. The structured information y1 annotated with x1 and the structured information y2 annotated with x2 are spliced. After splicing y1 and y2, redundant information can be removed to obtain the annotated information y′ of the enhanced medical image report x′. Through the above process, an enhanced medical image report x′ annotated with the structured information y′ can be obtained.

[0135] Exemplarily, the first training medical image report that needs to be data enhanced is x1, and the structured information annotated with x1 is y1. If the first training medical image report is data enhanced by report splicing and sentence shuffling, the first training medical image report x1 and the first training medical image report x2 are sentence split respectively, the sentences of x1 are split and the sentences are shuffled and then recombined, the sentences of x2 are split and the sentences are shuffled and then recombined, the two recombined reports are spliced, and then redundant information is removed to obtain an enhanced medical image report x′, the structured information y1 annotated with x1 and the structured information y2 annotated with x2 are spliced, and after splicing y1 and y2, the redundant information can be removed to obtain the annotated information y′ of the enhanced medical image report x′. Through the above process, the enhanced medical image report x′ annotated with the structured information y′ can be obtained.

[0136] Step a22-a: Process the preset instruction prompt information and the first training medical image report into prompt instructions, and obtain the prompt instructions corresponding to the first training medical image report.

[0137] Among them, the prompt instruction corresponding to the first training medical imaging report is used to prompt the pre-trained large language model to list the key information in the first training medical imaging report.

[0138] Step a22-b: Process the preset instruction prompt information and the enhanced medical imaging report into prompt instructions, and obtain prompt instructions corresponding to the enhanced medical imaging report.

[0139] Among them, the prompt instructions corresponding to the enhanced medical imaging report are used to prompt the pre-trained large language model to list the key information in the enhanced medical imaging report.

[0140] Step a23-a: input the prompt instruction corresponding to the first training medical imaging report into the pre-trained large language model to obtain the structured information prediction result of the first training medical imaging report.

[0141] Step a23-b: input the prompt instruction corresponding to the enhanced medical imaging report into the pre-trained large language model to obtain the structured information prediction result of the enhanced medical imaging report.

[0142] like Figure 6 As shown, after the preset instruction prompt information and the first training medical image report are processed into prompt instructions, the prompt instructions are input into the pre-trained large language model. At the same time, the preset instruction prompt information and the enhanced medical image report are processed into prompt instructions, and the prompt instructions are also input into the pre-trained large language model. The pre-trained large language model predicts and outputs structured information for the first training medical image report and the enhanced medical image report according to the input prompt instructions.

[0143] Step a24-a: determine a first prediction loss based on the structured data prediction result of the first training medical imaging report and the structured information annotated in the first training medical imaging report.

[0144] Step a24-b: Determine a second prediction loss based on the structured information prediction result of the enhanced medical imaging report and the structured information annotated in the enhanced medical imaging report.

[0145] The first prediction loss and the second prediction loss may adopt the standard autoregressive next token prediction loss.

[0146] Step a25: fine-tune the parameters of the pre-trained large language model according to the first prediction loss and the second prediction loss.

[0147] Specifically, the first prediction loss and the second prediction loss can be fused (such as direct summation, weighted summation, etc.), and the parameters of the pre-trained large language model can be fine-tuned according to the fused loss.

[0148] In this embodiment, the parameter fine-tuning strategy during model training can be a full-parameter fine-tuning strategy, or it can be an efficient parameter fine-tuning strategy such as LORA, Q-LORA, p-tuning, etc.

[0149] Next, a second implementation method is provided for fine-tuning the pre-trained large language model using training data (including first training medical images annotated with structured information and second training medical images without annotations) to obtain a medical image report structured model:

[0150] Step b1: Use the first training medical imaging report annotated with structured information to perform instruction fine-tuning on the pre-trained large language model.

[0151] The specific implementation process of step b1 can refer to steps a11 to 14, or steps a21 to 25.

[0152] Step b2: perform structural processing on the unlabeled second training medical image report based on the large language model fine-tuned by the instructions to obtain structured information of the second training medical image report as the labeling information of the second training medical image report.

[0153] In addition to the first training medical image annotated with structured information, the training data in this implementation method also includes an unlabeled second training medical image. This implementation method uses the first training medical image annotated with structured information to perform the first stage of instruction fine-tuning on the large language model. After the first stage of instruction fine-tuning is completed, the structured information of the unlabeled second training medical image report is obtained based on the large language model after the first stage of instruction fine-tuning. The obtained structured information is used as the annotation information of the second training medical image report. In this way, the second training medical image report annotated with structured information can be obtained.

[0154] Step b3: using the first training medical imaging report annotated with structured information and the second training medical imaging report annotated with structured information, further fine-tuning the large language model after instruction fine-tuning, and using the large language model after further instruction fine-tuning as the medical imaging report structured model.

[0155] After obtaining the second training medical image report annotated with structured information, the second training medical image report annotated with structured information and the first training medical image report annotated with structured information are used to perform second-stage instruction fine-tuning on the large language model. Specifically, the second training medical image report annotated with structured information and the first training medical image report annotated with structured information are used to further perform instruction fine-tuning on the large language model fine-tuned in step b1 to obtain a medical image report structured model.

[0156] Optionally, when further fine-tuning the large language model fine-tuned in step b1, a data enhancement strategy may be introduced, that is, data enhancement processing is performed on the first training medical image report annotated with structured information to obtain an enhanced medical image report annotated with structured information, and then, Figure 7 As shown, the second training medical image report annotated with structured information, the first training medical image report annotated with structured information, and the enhanced medical image report annotated with structured information are used to further fine-tune the large language model fine-tuned in step b1 to obtain a medical image report structured model.

[0157] It should be noted that when further fine-tuning the instructions of the large language model fine-tuned by step b1, the medical image report (the second training medical image report annotated with structured information, the first training medical image report annotated with structured information, and the enhanced medical image report annotated with structured information) and the preset prompt information are first processed into prompt instructions, and the prompt instructions are input into the large language model fine-tuned by step b1 to obtain the structured information prediction result, and then the prediction loss is determined according to the structured information prediction result and the structured information annotated in the medical image report, and the parameters of the large language model fine-tuned by step b1 are fine-tuned according to the prediction loss.

[0158] It should also be noted that the self-training process (i.e., using the large language model after instruction fine-tuning to annotate the unlabeled second training medical image report, and then using the second training medical image report annotated with structured information in combination with the first training medical image report annotated with structured information and the enhanced medical image report annotated with structured information to further fine-tune the large language model after instruction fine-tuning) can be carried out for multiple cycles, that is, the model in the previous self-training process can continue to annotate new unlabeled reports, and then be used in the next cycle of self-training process until the model converges.

[0159] After obtaining the structured model of medical imaging reports, Figure 8 As shown, the preset instruction prompt information and the target medical image report are processed into prompt instructions, and prompt instructions corresponding to the target medical image report are obtained. The prompt instructions corresponding to the target medical image report are input into the medical image report structured model to obtain the structured information of the target medical image report output by the medical image report structured model.

[0160] The above introduces the medical imaging report structuring method provided in the embodiment of the present application. The following will introduce the device corresponding to the above medical imaging report structuring method.

[0161] See also Fig. 9 , Fig. 9 A schematic diagram of the structure of a medical image report structuring device provided in an embodiment of the present application, the medical image report structuring device may include: a medical image report acquisition module 901 and a medical image report structuring module 902.

[0162] The medical imaging report acquisition module 901 is used to acquire a target medical imaging report.

[0163] The medical image report structuring module 902 is used to perform structural processing on the target medical image report based on the pre-acquired medical image report structuring model to obtain the structural information of the target medical image report.

[0164] The medical imaging report structured model is obtained by fine-tuning the pre-trained large language model using training data, and the training data at least includes a first training medical imaging report annotated with structured information.

[0165] In a possible implementation, the medical image report structuring module 902 is specifically used to:

[0166] Processing the preset instruction prompt information and the target medical image report into prompt instructions to obtain prompt instructions corresponding to the target medical image report, wherein the prompt instructions corresponding to the target medical image report are used to prompt the medical image report structured model to list key information in the target medical image report;

[0167] The prompt instruction corresponding to the target medical image report is input into the medical image report structured model to obtain the structured information of the target medical image report output by the medical image report structured model.

[0168] In a possible implementation, the structured information includes: an entity category label and an anatomical location label and an additional information label corresponding to the entity category label, wherein the entity category label is one of a normal entity label and an abnormal entity label.

[0169] The medical imaging report structuring device provided in the embodiment of the present application may further include: a model fine-tuning module.

[0170] The model fine-tuning module is used to use the training data to fine-tune the pre-trained large language model to obtain a structured model of medical imaging reports.

[0171] The model fine-tuning module uses training data to fine-tune the pre-trained large language model to obtain a medical imaging report structured model. Specifically, it is used to:

[0172] Using the first training medical imaging report annotated with structured information, fine-tuning the pre-trained large language model, and using the fine-tuned large language model as a medical imaging report structured model;

[0173] Alternatively, a first training medical image report annotated with structured information is used to perform instruction fine-tuning on a pre-trained large language model; a second unlabeled training medical image report is structured based on the fine-tuned large language model to obtain structured information of the second training medical image report as the annotation information of the second training medical image report; the first training medical image report annotated with structured information and the second training medical image report annotated with structured information are used to further fine-tune the fine-tuned large language model, and the fine-tuned large language model is used as a medical image report structured model.

[0174] In a possible implementation, when the model fine-tuning module uses the first training medical imaging report annotated with structured information to perform instruction fine-tuning on the pre-trained large language model, it is specifically used to:

[0175] Processing the preset instruction prompt information and the first training medical image report into prompt instructions, and obtaining prompt instructions corresponding to the first training medical image report, wherein the prompt instructions corresponding to the first training medical image report are used to prompt the pre-trained large language model to list key information in the first training medical image report;

[0176] Inputting the prompt instruction corresponding to the first training medical imaging report into the pre-trained large language model to obtain a structured information prediction result of the first training medical imaging report;

[0177] Determining a first prediction loss according to a structured data prediction result of the first training medical imaging report and structured information annotated in the first training medical imaging report;

[0178] According to the first prediction loss, the parameters of the pre-trained large language model are updated.

[0179] In a possible implementation, when the model fine-tuning module uses the first training medical imaging report annotated with structured information to perform instruction fine-tuning on the pre-trained large language model, it is further used to:

[0180] Performing data enhancement processing on the first training medical image report annotated with structured information to obtain an enhanced medical image report annotated with structured information;

[0181] Processing the preset instruction prompt information and the enhanced medical image report into prompt instructions, generating prompt instructions corresponding to the enhanced medical image report, wherein the prompt instructions corresponding to the enhanced medical image report are used to prompt the pre-trained large language model to list key information in the enhanced medical image report;

[0182] Inputting the prompt instructions corresponding to the enhanced medical imaging report into the pre-trained large language model to obtain the structured information prediction results of the enhanced medical imaging report;

[0183] Determining a second prediction loss according to the structured information prediction result of the enhanced medical imaging report and the structured information annotated in the enhanced medical imaging report;

[0184] When the model fine-tuning module updates the parameters of the pre-trained large language model according to the first prediction loss, it is specifically used to:

[0185] According to the first prediction loss and the second prediction loss, the parameters of the pre-trained large language model are updated.

[0186] In a possible implementation, when the model fine-tuning module performs data enhancement processing on the first training medical image report annotated with structured information to obtain an enhanced medical image report annotated with structured information, it is specifically used to:

[0187] Using one or more of several data enhancement methods to perform data enhancement processing on the first training medical image report to obtain an enhanced medical image report, wherein the several data enhancement methods include inserting segmentation words, disrupting the order of sentences, and report splicing. Report splicing refers to splicing another first training medical image report to the first training medical image report that needs data enhancement;

[0188] If data enhancement processing is performed on the first training medical image report by inserting segmentation words and / or disrupting the order of sentences, the annotation information of the first training medical image report is determined as the annotation information of the enhanced medical image report;

[0189] If at least one report splicing method is used to perform data enhancement processing on the first training medical image report, the labeling information of the two spliced ​​first training medical image reports is spliced ​​and de-redundant to obtain the labeling information of the enhanced medical image report.

[0190] The medical image report structuring device provided in the embodiment of the present application adopts an end-to-end structuring method based on a large language model to obtain the structured information of the medical image report. The method is simple to implement and can accurately, efficiently and normatively extract the key information contained in the medical image report. In addition, since only one model needs to be trained and the strong generalization learning ability exhibited by the large language model enables fine-tuning of a small amount of data to achieve ideal performance, the amount of data that needs to be manually labeled is relatively small, and thus the cost of manual labeling is low.

[0191] An embodiment of the present application also provides an electronic device, which may include: at least one processor, at least one communication interface, at least one memory and at least one communication bus.

[0192] In the embodiment of the present application, the number of the processor, the communication interface, the memory, and the communication bus is at least one, and the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0193] The processor may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application;

[0194] The memory may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory;

[0195] The memory stores a program, and the processor can call the program stored in the memory, and the program is used to implement the steps of the medical imaging report structuring method provided in the above embodiment.

[0196] An embodiment of the present application also provides a computer storage medium, which carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the steps of the medical imaging report structuring method provided in the above embodiment.

[0197] An embodiment of the present application also provides a computer program product, including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements the steps of the medical imaging report structuring method provided in the above embodiment.

[0198] It should also be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the device embodiments provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines.

[0199] Through the description of the above implementation mode, the technicians in the field can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. In general, all functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be various, such as analog circuits, digital circuits or special circuits. However, for the present application, software program implementation is a better implementation mode in more cases. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, a U disk, a mobile hard disk, a ROM, a RAM, a disk or an optical disk, etc., including a number of instructions to enable a computer device (which can be a personal computer, a training device, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0200] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0201] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a training device, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, training device, or data center. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

Claims

1. A method for structuring medical imaging reports, characterized in that: include: Obtain target medical imaging reports; Based on the pre-acquired medical image report structured model, the target medical image report is structured to obtain structured information of the target medical image report; The medical imaging report structured model is obtained by fine-tuning a pre-trained large language model using training data, and the training data at least includes a first training medical imaging report annotated with structured information.

2. The medical imaging report structuring method according to claim 1, characterized in that: The step of performing structural processing on the target medical image report based on the pre-acquired medical image report structural model to obtain structural information of the target medical image report includes: Processing the preset instruction prompt information and the target medical image report into prompt instructions to obtain prompt instructions corresponding to the target medical image report, wherein the prompt instructions corresponding to the target medical image report are used to prompt the medical image report structured model to list key information in the target medical image report; The prompt instruction corresponding to the target medical image report is input into the medical image report structured model to obtain the structured information of the target medical image report output by the medical image report structured model.

3. The medical imaging report structuring method according to claim 1, characterized in that: The structured information includes: an entity category label and an anatomical position label and an additional information label corresponding to the entity category label, wherein the entity category label is one of a normal entity label and an abnormal entity label.

4. The medical imaging report structuring method according to claim 1, characterized in that: The process of fine-tuning the pre-trained large language model using training data to obtain a medical imaging report structured model includes: Using the first training medical imaging report annotated with structured information, fine-tuning the pre-trained large language model, and using the fine-tuned large language model as a medical imaging report structured model; or, The first training medical imaging report annotated with structured information is used to fine-tune the pre-trained large language model; Based on the large language model fine-tuned by the instruction, a structural processing is performed on the unlabeled second training medical image report to obtain structured information of the second training medical image report as labeling information of the second training medical image report; A first training medical imaging report annotated with structured information and a second training medical imaging report annotated with structured information are used to further fine-tune the large language model after instruction fine-tuning, and the large language model after further instruction fine-tuning is used as a medical imaging report structured model.

5. The medical imaging report structuring method according to claim 4, characterized in that: The method of using the first training medical imaging report annotated with structured information to fine-tune the pre-trained large language model includes: Processing the preset instruction prompt information and the first training medical image report into prompt instructions, and obtaining prompt instructions corresponding to the first training medical image report, wherein the prompt instructions corresponding to the first training medical image report are used to prompt the pre-trained large language model to list key information in the first training medical image report; Inputting the prompt instruction corresponding to the first training medical imaging report into the pre-trained large language model to obtain a structured information prediction result of the first training medical imaging report; Determining a first prediction loss according to the structured data prediction result of the first training medical imaging report and the structured information annotated in the first training medical imaging report; According to the first prediction loss, parameters of the pre-trained large language model are updated.

6. The medical imaging report structuring method according to claim 5, characterized in that: The method of using the first training medical imaging report annotated with structured information to fine-tune the pre-trained large language model includes: Performing data enhancement processing on the first training medical image report annotated with structured information to obtain an enhanced medical image report annotated with structured information; Processing the preset instruction prompt information and the enhanced medical image report into prompt instructions, generating prompt instructions corresponding to the enhanced medical image report, wherein the prompt instructions corresponding to the enhanced medical image report are used to prompt the pre-trained large language model to list key information in the enhanced medical image report; Inputting the prompt instruction corresponding to the enhanced medical imaging report into the pre-trained large language model to obtain the structured information prediction result of the enhanced medical imaging report; Determining a second prediction loss according to the structured information prediction result of the enhanced medical imaging report and the structured information annotated in the enhanced medical imaging report; The updating of parameters of the pre-trained large language model according to the first prediction loss includes: Parameters of the pre-trained large language model are updated according to the first prediction loss and the second prediction loss.

7. The medical imaging report structuring method according to claim 6, characterized in that: The step of performing data enhancement processing on the first training medical image report annotated with structured information to obtain an enhanced medical image report annotated with structured information includes: Using one or more of several data enhancement methods to perform data enhancement processing on the first training medical image report to obtain an enhanced medical image report, wherein the several data enhancement methods include inserting segmentation words, disrupting the order of sentences, and report splicing, and the report splicing refers to splicing another first training medical image report to the first training medical image report that needs data enhancement; If data enhancement processing is performed on the first training medical image report by inserting segmentation words and / or disrupting the order of sentences, the annotation information of the first training medical image report is determined as the annotation information of the enhanced medical image report; If at least one report splicing method is used to perform data enhancement processing on the first training medical image report, the labeling information of the two spliced ​​first training medical image reports is spliced ​​and de-redundant to obtain the labeling information of the enhanced medical image report.

8. A medical imaging report structuring device, characterized in that: include: Medical imaging report acquisition module and medical imaging report structuring module; The medical imaging report acquisition module is used to acquire a target medical imaging report; The medical image report structuring module is used to perform structural processing on the target medical image report based on the pre-acquired medical image report structuring model to obtain the structural information of the target medical image report; The medical imaging report structured model is obtained by fine-tuning a pre-trained large language model using training data, and the training data at least includes a first training medical imaging report annotated with structured information.

9. An electronic device, characterized in that: The method comprises at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program so that the electronic device can implement the steps of the medical imaging report structuring method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: The storage medium carries one or more computer programs, and when the one or more computer programs are executed by an electronic device, the electronic device can implement the steps of the medical imaging report structuring method as described in any one of claims 1 to 7.

Images