Electronic case report form generation method and device, electronic equipment and storage medium

Abstract

The invention discloses an electronic case report form generation method and device, electronic equipment and a storage medium, and belongs to the technical field of electronic data processing.The electronic case report form generation method comprises the steps that key information of a clinical test is obtained; inputting the key information into a trained deep learning model to generate an electronic case report form; the trained deep learning model is obtained through memory caching and small-batch training, and the trained deep learning model adopts a multi-layer self-attention mechanism and an encoder-decoder structure. According to the method, key information of a clinical test is automatically analyzed by using a trained deep learning model, and a structured electronic case report form is generated. And researchers do not need to manually write and adjust the form content, so that repeated labor is reduced, and the working efficiency is improved. In addition, time and labor cost of manual design are remarkably reduced.

Description

Technical Field

[0001] The present application belongs to the technical field of electronic data processing, and specifically relates to a method, device, electronic device and storage medium for generating an electronic case report form. Background Art

[0002] Electronic case report forms (eCRFs) are an important tool for recording patient data in clinical trials. The traditional eCRF design process relies on manual writing and expert review, which is not only time-consuming and labor-intensive, but also prone to human errors. Designing an eCRF form requires a detailed understanding of the study purpose and research plan, and customizing the form structure and content according to the needs of the trial.

[0003] As the complexity and scale of clinical trials continue to increase, manually designing eCRF forms faces increasing challenges. In order to improve efficiency and reduce human errors, electronic data collection systems have gradually been widely used in clinical trials. However, existing electronic data collection systems still rely on manually designing eCRF forms, which requires a lot of time and manpower. Summary of the invention

[0004] The purpose of this application is to provide a method, device, electronic device and storage medium for generating an electronic case report form to realize automatic generation of an electronic medical record report form, thereby reducing time and labor costs.

[0005] According to a first aspect of an embodiment of the present application, a method for generating an electronic case report form is provided, and the method may include:

[0006] Obtain key information on clinical trials;

[0007] The key information is input into the trained deep learning model to generate an electronic case report form; the trained deep learning model is obtained through memory caching and small batch training, and the trained deep learning model adopts a multi-layer self-attention mechanism and an encoder-decoder structure.

[0008] In some optional embodiments of the present application, the trained deep learning model is obtained by the following training:

[0009] Get key training information for clinical trials;

[0010] Perform consistency check and enhancement processing on key training information to obtain preprocessed data;

[0011] Label the preprocessed data to obtain training data;

[0012] The deep learning model is trained using the training data to obtain a trained deep learning model.

[0013] In some optional embodiments of the present application, the key data includes:

[0014] Study objectives, patient types and characteristics, disease stage, interventions and treatment types, study endpoints.

[0015] In some optional embodiments of the present application, the consistency check includes:

[0016] Data logic error and inconsistency checking;

[0017] Check the consistency between actual information and clinical records.

[0018] In some optional embodiments of the present application, the enhancement process includes:

[0019] Use the trained language model to expand key training information;

[0020] Insert medical terms based on key training information.

[0021] In some optional embodiments of the present application, data annotation is performed on the preprocessed data to obtain training data, including:

[0022] Use the pre-trained language model for machine labeling to obtain preliminary labeling results;

[0023] Review and modify the preliminary annotation results to obtain the secondary annotation results;

[0024] Convert the secondary annotation results into a table format to obtain training data.

[0025] In some optional embodiments of the present application, the deep learning model is trained using the training data to obtain a trained deep learning model, including:

[0026] Use multi-process concurrent libraries and fast interfaces to load training data;

[0027] Unify the format of training data to obtain standard data;

[0028] Divide the standard data into small batches for deep learning model training, and use the small batch loader to load the data to obtain a trained deep learning model.

[0029] According to a second aspect of an embodiment of the present application, a device for generating an electronic case report form is provided, and the device may include:

[0030] Acquisition module, used to obtain key information of clinical trials;

[0031] The electronic report generation module is used to input key information into the trained deep learning model to generate an electronic case report form; the trained deep learning model is obtained through memory caching and small batch training, and the trained deep learning model adopts a multi-layer self-attention mechanism and an encoder-decoder structure.

[0032] According to a third aspect of an embodiment of the present application, an electronic device is provided, which may include:

[0033] processor;

[0034] a memory for storing processor-executable instructions;

[0035] The processor is configured to execute instructions to implement the method for generating an electronic case report form as shown in any one of the embodiments of the first aspect.

[0036] According to a fourth aspect of an embodiment of the present application, a storage medium is provided. When instructions in the storage medium are executed by a processor of an information processing device or a server, the information processing device or the server implements a method for generating an electronic case report form as shown in any one of the embodiments of the first aspect.

[0037] The above technical solution of the present application has the following beneficial technical effects:

[0038] The method of the embodiment of the present application automatically parses the key information of the clinical trial by using the trained deep learning model to generate a structured electronic case report form. Researchers no longer need to manually write and adjust the form content, which reduces repetitive work and improves work efficiency. In addition, the time and labor cost of manual design are significantly reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 is a flowchart of a method for generating an electronic case report form in an exemplary embodiment of the present application;

[0040] Figure 2 is a schematic structural diagram of an electronic case report form generating device in an exemplary embodiment of the present application;

[0041] Figure 3 is a schematic structural diagram of an electronic device in an exemplary embodiment of the present application;

[0042] Figure 4 It is a schematic diagram of the hardware structure of an electronic device in an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with specific implementations and with reference to the accompanying drawings. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present application. In addition, in the following description, the description of known structures and technologies is omitted to avoid unnecessary confusion of the concepts of the present application.

[0044] The accompanying drawings show schematic diagrams of layer structures according to embodiments of the present application. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clarity. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may additionally design regions / layers with different shapes, sizes, and relative positions according to actual needs.

[0045] Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0046] In the description of the present application, it should be noted that the terms "first", "second" and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0047] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0048] The following, in conjunction with the accompanying drawings, describes in detail the electronic case report form generation method, device, electronic device and storage medium provided in the embodiments of the present application through specific embodiments and their application scenarios.

[0049] like Figure 1 As shown, in a first aspect of an embodiment of the present application, a method for generating an electronic case report form is provided, and the method may include:

[0050] S110: Obtain key information of clinical trials;

[0051] S120: Input key information into the trained deep learning model to generate an electronic case report form; the trained deep learning model is obtained through memory caching and small batch training, and the trained deep learning model adopts a multi-layer self-attention mechanism and an encoder-decoder structure.

[0052] The method of this embodiment automatically parses the key information of clinical trials by using a trained deep learning model to generate a structured electronic case report form. Researchers no longer need to manually write and adjust the form content, which reduces repetitive work and improves work efficiency. In addition, the time and labor cost of manual design are significantly reduced.

[0053] In some embodiments, the trained deep learning model is obtained by the following training:

[0054] Get key training information for clinical trials;

[0055] Perform consistency check and enhancement processing on key training information to obtain preprocessed data;

[0056] Label the preprocessed data to obtain training data;

[0057] The deep learning model is trained using the training data to obtain a trained deep learning model.

[0058] In this embodiment, the data of the deep learning model is strictly screened and processed to ensure the quality and consistency of the input data, so that the data is suitable for further processing by the large language model.

[0059] In some embodiments, key data includes:

[0060] Study objectives, patient types and characteristics, disease stage, interventions and treatment types, study endpoints.

[0061] The information in this example is the basis for generating the eCRF form, which covers the main objectives and expected results of the trial.

[0062] In some embodiments, the consistency check includes:

[0063] Data logic error and inconsistency checking;

[0064] Check the consistency between actual information and clinical records.

[0065] This embodiment specifically provides the content of the consistency check, checking the logical errors and inconsistencies in the data, such as age less than 0 or gender inconsistent with the clinical record. The inspection process can define a series of inspection rules in the rule engine, for example, the age cannot be less than 0, the gender field must be consistent with the clinical record, etc. The data is checked, marked and corrected by the rule engine. The actual information can be the patient's gender information, and the patient's gender information is checked to see if it is consistent with the disease field. Specifically, SpaCy (NLP model) can be used to perform text parsing on clinical records to extract gender and disease field information. Define consistency check rules for gender and disease fields in the rule engine. Perform consistency check on the data through the rule engine and mark inconsistent data. Generate prompt information and correct data as needed.

[0066] In some embodiments, the enhancement process includes:

[0067] Use the trained language model to expand key training information;

[0068] Insert medical terms based on key training information.

[0069] This embodiment uses natural language processing technology to enhance data and improve data diversity and model robustness. Specifically, a pre-trained language model (such as Wenxinyiyan, Qianwenda model interface) can be used to generate and enhance text data. Use the pre-trained language model to generate new text data and expand it based on existing data. Then, the enhanced data is verified to ensure its quality and consistency.

[0070] The inserted medical terms in this embodiment can be extracted through the API of SNOMED CT to extract medical terms related to the patient's condition or treatment. For example, common complications and treatment side effects of chronic hepatitis are extracted. Then, the patient description text is preprocessed using NLP tools (such as SpaCy), including word segmentation, part-of-speech tagging, and named entity recognition. The semantic structure of the patient description text is analyzed to determine the location of the inserted term to ensure the logical and contextual consistency of the inserted text. According to the results of the semantic analysis, the extracted medical terms are inserted into the appropriate location. For example, when describing the type and characteristics of the patient, "possible complications include cirrhosis and liver cancer" or "side effects that may be caused by treatment include gastrointestinal discomfort and allergic reactions" are inserted. The inserted text can be generated or adjusted using a large model (Wen Xin Yi Yan) to ensure that the sentences are smooth and logically consistent. Finally, a feedback mechanism can be established to manually review the inserted text, collect feedback, and optimize the term insertion process. Based on the feedback results, the term insertion strategy and technology are continuously optimized to improve data quality and consistency.

[0071] In some embodiments, data annotation is performed on the preprocessed data to obtain training data, including:

[0072] Use the pre-trained language model for machine labeling to obtain preliminary labeling results;

[0073] Review and modify the preliminary annotation results to obtain the secondary annotation results;

[0074] Convert the secondary annotation results into a table format to obtain training data.

[0075] In this embodiment, after the data processing is completed, the data needs to be labeled to ensure that the model can accurately understand and generate the content required by the eCRF form. First, define the labeling rules and standards according to the requirements of the eCRF form. For example, define the labeling standards for fields such as patient type, disease stage, and intervention measures. Use a pre-trained language model (Wenxin Yiyan large model) for preliminary labeling. Then, professionals review and correct the labeling results of the large model. Finally, the labeled data is verified to ensure the quality and consistency of the labeling. This labeling method combines large model-assisted labeling and manual review to improve the efficiency and accuracy of data labeling.

[0076] In addition, the annotated data is converted into Markdown table format to facilitate model training and data display. Specifically, a script can be used to convert the annotated data into a Markdown table. Then professionals will check the converted Markdown table. The combination of automated conversion and manual verification improves conversion efficiency and table quality.

[0077] For example, first define the format and structure of the Markdown table according to the requirements of the eCRF form. For example, define the field name, data type, validation rules, etc. of the table; then write a script to automatically convert the annotated data into the Markdown table format; then run the conversion script to convert the annotated data into the Markdown table in batches; finally, professionals verify the converted Markdown table to ensure that the table format is correct and the data is complete, and optimize and correct the conversion script and table format based on the verification results.

[0078] In some embodiments, using training data to train a deep learning model to obtain a trained deep learning model includes:

[0079] Use multi-process concurrent libraries and fast interfaces to load training data;

[0080] Unify the format of training data to obtain standard data;

[0081] Divide the standard data into small batches for deep learning model training, and use the small batch loader to load the data to obtain a trained deep learning model.

[0082] For example, you can use Python's multiprocessing library and TensorFlow's tf.data API to quickly load and process large-scale data sets and reduce data loading bottlenecks. Ensure that all text data is encoded in the same format, unified as UTF-8, to avoid errors caused by inconsistent data formats. Unify the format of numerical data to ensure data consistency and operability. Divide the data into small batches for training, and use batch gradient descent methods to improve training efficiency and model generalization. 32 or 64 data per batch helps balance memory usage and computational efficiency. Use memory caching technology (PyTorch's DataLoader) to preload the next batch of data while processing the current batch to reduce I / O bottlenecks.

[0083] The model in this embodiment is based on the Transformer architecture, using a multi-layer self-attention mechanism and an encoder-decoder structure. The encoder captures the contextual information of the input text, and the decoder generates the corresponding eCRF form to improve the semantic understanding and generation capabilities of the model. Through the multi-head attention mechanism, the model is allowed to focus on different parts of the input text and enhance the processing capabilities of long texts. Position encoding can be introduced to help the model understand the order of words and improve context understanding capabilities. Multiple GPUs are used for parallel training of large-scale models, and the DeepSpeed ​​distributed training framework is used to reduce the memory usage of a single GPU and improve training speed and efficiency through data parallelism and model parallelism.

[0084] In addition, FP16 and FP32 precision can be combined to reduce memory usage and increase computing speed. Using NVIDIA's Apex tool for mixed precision training not only improves computing efficiency, but also reduces the requirements for hardware resources. Use a dynamic learning rate scheduling method (learning rate cosine annealing) to optimize the adjustment of the learning rate during training. A smaller learning rate is used in the initial stage, which is gradually increased and then slowly decayed to avoid shocks in the early stages of training and premature convergence in the later stages.

[0085] Finally, the model performance can be evaluated using indicators such as the accuracy of the generated form, grammatical correctness, and logical correctness. Specific indicators include BLEU score, ROUGE score, and accuracy of form fields. Compare and analyze with the benchmark model or other latest models to verify the superiority of the model.

[0086] In some embodiments, a knowledge base containing different research purposes can be established, and the model can understand and generate content related to specific research purposes through embedding technology. Different patient characteristics and their related data are stored in the knowledge base, and the model can better handle records of different types of patients through knowledge graph embedding. Incorporating research endpoint information into the knowledge base helps the model more accurately reflect the goals and expected results of the study when generating eCRF forms.

[0087] In addition, we can also introduce a domain-specific medical knowledge graph (SNOMED CT) and embed the nodes and edges in the medical knowledge graph into the vector space through the knowledge graph embedding technology to improve the model's ability to understand and generate clinical trial data. The knowledge graph can provide structured medical knowledge to help the model better understand and generate medical-related content.

[0088] This embodiment combines the knowledge base containing different research purposes, patient characteristics and research endpoints with the medical knowledge graph (SNOMED CT) through the RAG (Retrieval-Augmented Generation) method to enhance the GLM4 model's ability to understand and generate clinical trial data. The RAG method improves the accuracy and consistency of generated content by retrieving relevant knowledge and combining it with the generative model. Combining the knowledge base and the medical knowledge graph, the RAG method improves the model's ability to understand and generate medical knowledge. Provide structured medical knowledge to enhance the model's medical background and contextual understanding.

[0089] Specifically, the purpose of the study, patient characteristics, and study endpoint information are collected from clinical trial records, literature, and databases. The collected information is structured and stored in a database for quick retrieval.

[0090] Example:

[0091] Research purpose: To improve liver function in patients with chronic hepatitis B.

[0092] Patient characteristics: Age range 18-65 years, HBsAg positive, ALT levels between 2 and 10 times the upper limit of normal.

[0093] Study endpoints: ALT levels decreased, HBV DNA levels became undetectable, and liver function improved.

[0094] Acquiring and processing the SNOMED CT knowledge graph

[0095] Processing content: Obtain SNOMED CT knowledge graph and perform preprocessing.

[0096] Specific steps:

[0097] Download and import: Obtain the SNOMED CT knowledge graph from official channels and import it into the database.

[0098] Data cleaning: remove redundant information and retain key nodes (such as diseases, symptoms, treatment methods) and their relationships (such as the association between diseases and symptoms).

[0099] Example:

[0100] Nodes: Disease (e.g., Chronic Hepatitis B), Symptom (e.g., Liver Cirrhosis), Treatment (e.g., DrugX).

[0101] Side: The relationship between the disease and symptoms (such as Chronic Hepatitis B->Liver Cirrhosis).

[0102] This embodiment also requires integrating the knowledge base and knowledge graph into the RAG model. The knowledge base and knowledge graph are combined into the RAG model to enhance the model's medical knowledge understanding and generation capabilities.

[0103] Specifically, first, the retrieval module is used to retrieve the research purpose, patient characteristics and research endpoint information related to the query text from the knowledge base. For example, the query "45-year-old male patient, HBsAg positive, ALT level 4 times the upper limit of normal", the relevant information retrieved is "Study on patients with chronic hepatitis B, the research endpoint is a decrease in ALT levels". Secondly, the retrieval module is used to retrieve medical knowledge related to the query text from the SNOMED CT knowledge graph. For example, retrieve the relevant symptoms and treatment methods of "chronic hepatitis B". Then, the retrieved knowledge base information and knowledge graph information are fused with the query text. Next, the fused representation is input into the generation module (GLM4) for further processing and generation. Finally, the fields, data types, validation rules, etc. of the eCRF form are generated.

[0104] In some embodiments, after the model training is completed, the text will be passed to the fine-tuned large language model (GLM-4) to generate the corresponding eCRF form. GLM-4 is a powerful large-scale language generation model with a high degree of natural language understanding and generation capabilities, which can generate detailed eCRF forms that meet the needs of clinical trials based on the input data. The generated eCRF form not only includes the field name and type, but also defines the validation rules, default values, and logical relationships of each field in detail. This information is output in Markdown format to ensure that the form structure is clear, easy to read and integrate.

[0105] The GLM-4 model is called for generation. The model processes the input data through a multi-layer Transformer architecture, understands its semantics and structure, and generates high-quality eCRF form text. The architectural design of GLM-4 enables it to handle complex contextual relationships and generate text output with logical consistency and accuracy. The eCRF form generated by the model contains detailed field definitions and validation rules, such as the mandatory nature of the field, value range, data type, etc., to ensure that the form meets the standards and requirements of clinical trials. During the generation process, the model uses pre-defined templates and rules to ensure that the generated form is not only clearly structured, but also logically rigorous and in line with the specifications of medical research.

[0106] The GLM-4 large language model is used to automatically generate eCRF forms through natural language processing (NLP) and fine-tuning of large-scale pre-trained models. Through pre-training and fine-tuning, the model can automatically parse information such as research objectives, patient characteristics, and research endpoints, and generate structured eCRF forms. This significantly reduces the time and labor costs of manual design. Researchers no longer need to manually write and adjust form content, reducing repetitive work and improving work efficiency.

[0107] In some embodiments, the generated eCRF form needs to be strictly verified and optimized to ensure that it meets the requirements and specifications of clinical trials. This step includes syntax checking, logic verification and optimization adjustments to ensure the completeness, accuracy and practicality of the form. Perform comprehensive syntax and logic checks on the generated form to ensure the correctness of the field definition and the rationality of the validation rules. It also includes the optimization of the form structure to ensure that it meets the best practices and standards of clinical trials. In addition, automated tools and scripts can be used to check whether the syntax of the generated Markdown form is correct to ensure that the form format is unified and error-free. Syntax checking includes the correctness of Markdown syntax, consistency of table format, etc. Verify whether the logical rules and validation conditions in the form are reasonable, including the mandatory nature of the field, value range and data type, etc., to ensure the logical correctness of each field definition. Logical validation is performed through a predefined rule base and logic checking algorithm to ensure that the logical relationship and validation rules of each field meet the standards of clinical trials. Make necessary adjustments and optimizations to the form based on the validation results, correct the problems found, improve the form structure and logic, and improve its use effect and user experience. The optimization process includes rearranging fields, adjusting validation rules, setting default values, etc., to ensure that the generated form is both practical and compliant with the specifications.

[0108] It should be noted that the execution subject of the method for generating an electronic case report form provided in the embodiment of the present application may be an electronic case report form generating device, or a control module in the electronic case report form generating device for executing the method for generating an electronic case report form. In the embodiment of the present application, the method for generating an electronic case report form performed by the electronic case report form generating device is taken as an example to illustrate the device for generating an electronic case report form provided in the embodiment of the present application.

[0109] like Figure 2 As shown, in a second aspect of an embodiment of the present application, a device for generating an electronic case report form is provided, and the device may include:

[0110] An acquisition module 210 is used to acquire key information of clinical trials;

[0111] The electronic report generation module 220 is used to input key information into the trained deep learning model to generate an electronic case report form; the trained deep learning model is obtained through memory caching and small batch training, and the trained deep learning model adopts a multi-layer self-attention mechanism and an encoder-decoder structure.

[0112] The electronic case report form generating device in the embodiment of the present application can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, an in-vehicle electronic device, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc., and the non-mobile electronic device can be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., which is not specifically limited in the embodiment of the present application.

[0113] The electronic case report form generating device in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0114] The electronic case report form generation device provided in the embodiment of the present application can achieve Figure 1 To avoid repetition, the various processes implemented by the method embodiment are not described here.

[0115] Alternatively, if Figure 3 As shown, an embodiment of the present application also provides an electronic device 300, including a first processor 301, a first memory 302, and a program or instruction stored in the first memory 302 and executable on the first processor 301. When the program or instruction is executed by the first processor 301, each process of the above-mentioned electronic case report form generation method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0116] It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and non-mobile electronic devices mentioned above.

[0117] Figure 4 A schematic diagram of the hardware structure of an electronic device to implement an embodiment of the present application.

[0118] The hardware structure 400 of the electronic device includes but is not limited to: a radio frequency unit 401, a network module 402, an audio output unit 403, an input unit 404, a sensor 405, a display unit 406, a user input unit 407, an interface unit 408, a second memory 409, and a second processor 410 and other components.

[0119] Those skilled in the art will appreciate that the hardware structure 400 of the electronic device may also include a power source (such as a battery) for supplying power to each component, and the power source may be logically connected to the second processor 410 through a power management system, thereby implementing functions such as managing charging, discharging, and power consumption management through the power management system. Figure 4 The electronic device structure shown in the figure does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently, which will not be described in detail here.

[0120] It should be understood that in the embodiment of the present application, the input unit 404 may include a graphics processor (Graphics Processing Unit, GPU) 4041 and a microphone 4042, and the graphics processor 4041 processes the image data of the static picture or video obtained by the image capture device (such as a camera) in the video capture mode or the image capture mode. The display unit 406 may include a display panel 4061, and the display panel 4061 may be configured in the form of a liquid crystal display, an organic light emitting diode, etc. The user input unit 407 includes a touch panel 4071 and other input devices 4072. The touch panel 4071 is also called a touch screen. The touch panel 4071 may include two parts, a touch detection device and a touch controller. Other input devices 4072 may include but are not limited to a physical keyboard, a function key (such as a volume control button, a switch button, etc.), a trackball, a mouse, and a joystick, which will not be repeated here. The second memory 409 can be used to store software programs and various data, including but not limited to applications and operating systems. The second processor 410 can integrate an application processor and a modem processor, wherein the application processor mainly processes an operating system, a user interface, and applications, etc., and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the second processor 410 .

[0121] An embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, each process of the above-mentioned electronic case report form generation method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0122] The processor is a processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0123] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned electronic case report form generation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0124] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0125] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0126] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0127] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A method for generating an electronic case report form, characterized in that: include: Obtain key information on clinical trials; Inputting the key information into the trained deep learning model to generate an electronic case report form; The trained deep learning model is obtained through memory caching and small batch training, and the trained deep learning model adopts a multi-layer self-attention mechanism and an encoder-decoder structure.

2. The method for generating an electronic case report form according to claim 1, characterized in that: The trained deep learning model is obtained through the following training: Get key training information for clinical trials; Performing consistency check and enhancement processing on the key training information to obtain preprocessed data; Performing data labeling on the preprocessed data to obtain training data; The deep learning model is trained using the training data to obtain the trained deep learning model.

3. The method for generating an electronic case report form according to claim 2, characterized in that: The key data include: Study objectives, patient types and characteristics, disease stage, interventions and treatment types, study endpoints.

4. The method for generating an electronic case report form according to claim 2, characterized in that: The consistency check includes: Data logic error and inconsistency checking; Check the consistency between actual information and clinical records.

5. The method for generating an electronic case report form according to claim 2, characterized in that: The enhancement process includes: Expanding the key training information using a trained language model; Medical terms are inserted according to the key training information.

6. The method for generating an electronic case report form according to claim 2, characterized in that: The step of labeling the preprocessed data to obtain training data includes: Use the pre-trained language model for machine labeling to obtain preliminary labeling results; Reviewing and modifying the preliminary annotation results to obtain secondary annotation results; The secondary annotation results are converted into a table format to obtain training data.

7. The method for generating an electronic case report form according to claim 2, characterized in that: The step of training the deep learning model using the training data to obtain the trained deep learning model includes: Loading the training data using a multi-process concurrent library and a fast interface; Unifying the format of the training data to obtain standard data; The standard data is divided into small batches for deep learning model training, and the small batch loader is used to load the data to obtain the trained deep learning model.

8. An electronic case report form generating device, characterized in that: include: Acquisition module, used to obtain key information of clinical trials; The electronic report generation module is used to input the key information into a trained deep learning model to generate an electronic case report form; the trained deep learning model is obtained through memory caching and small batch training, and the trained deep learning model adopts a multi-layer self-attention mechanism and an encoder-decoder structure.

9. An electronic device, characterized in that: include: A processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the method for generating an electronic case report form as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by the processor, the steps of the method for generating an electronic case report form according to any one of claims 1 to 7 are implemented.

Images