Medical data processing method, computing device, server, storage medium and product

Abstract

The invention provides a medical data processing method, computing equipment, a server, a storage medium and a product. The method comprises the following steps: acquiring text information of a medical detection sheet; extracting the detection result data, the content of the basic information item and the medical purpose information contained in the medical detection sheet from the text information, identifying an index entity and a detection item entity contained in the medical detection sheet, searching medical knowledge related to the index entity and the detection item entity, and obtaining the medical knowledge corresponding to the medical detection sheet; and inputting the detection result data contained in the medical detection sheet, the content of the basic information item, the medical purpose information and the medical knowledge corresponding to the medical detection sheet into an intelligent interpretation model, and generating an interpretation result of the medical detection sheet through the intelligent interpretation model. According to the method, rich background information can be provided for the intelligent interpretation model by identifying the index entity and the detection item entity and combining the searched medical knowledge, so that the interpretation model can generate a more accurate and personalized interpretation result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computers, and in particular to a medical data processing method, a computing device, a server, a storage medium and a product. BACKGROUND

[0002] With the continuous improvement of people's health awareness and the increasing demand for medical services worldwide, individuals pay more attention to their own health status. Medical test sheets, as an important part of health management, have become the main way for patients and their families to obtain health information.

[0003] However, medical test sheets generally have the problem of poor readability. Medical test sheets are filled with a large number of professional terms and complex test indicators. These information may be daily language for medical professionals, but it is difficult for ordinary people without a medical background to understand.

[0004] Therefore, patients or their families usually need to spend a lot of time searching the Internet to obtain explanations of the terms and indicators in the test sheet. However, the information obtained through search engines is of varying quality, which may result in inaccurate and incomplete information, even misleading patients, leading to poor efficiency and accuracy in interpreting medical test sheets. SUMMARY

[0005] The present application provides a medical data processing method, a computing device, a server, a storage medium and a product to solve the problem of poor efficiency and accuracy in interpreting medical test sheets.

[0006] In a first aspect, the present application provides a medical data processing method, comprising:

[0007] obtaining text information of a medical test sheet;

[0008] extracting, from the text information, test result data contained in the medical test sheet, content of at least one basic information item and medical use information;

[0009] identifying index entities and test item entities contained in the medical test sheet, searching for medical knowledge related to the index entities and test item entities, and obtaining medical knowledge corresponding to the medical test sheet;

[0010] inputting the test result data contained in the medical test sheet, the content of the at least one basic information item and the medical use information, and the medical knowledge corresponding to the medical test sheet into an intelligent interpretation model, and generating an interpretation result of the medical test sheet through the intelligent interpretation model.

[0011] In a second aspect, the present application provides a medical data processing method, comprising:

[0012] An image of a medical test sheet sent by a terminal-side device;

[0013] Performing OCR recognition on the image to obtain text information of the medical test sheet;

[0014] Extracting, from the text information, detection result data contained in the medical test sheet, content of at least one basic information item, and medical use information;

[0015] Recognizing index entities and detection item entities contained in the medical test sheet, searching medical knowledge related to the index entities and the detection item entities, and obtaining medical knowledge corresponding to the medical test sheet;

[0016] Inputting, into an intelligent interpretation model, the detection result data contained in the medical test sheet, the content of the at least one basic information item, the medical use information, and the medical knowledge corresponding to the medical test sheet, and generating, by the intelligent interpretation model, an interpretation result of the medical test sheet;

[0017] Outputting the interpretation result of the medical test sheet to the terminal-side device.

[0018] In a third aspect, the present application provides a computing device, comprising:

[0019] A memory and a processor;

[0020] The memory is configured to store computer programs / instructions, and the processor is configured to execute the computer programs / instructions, and the computer programs / instructions, when executed by the processor, implement the method of any one of the preceding aspects.

[0021] In a fourth aspect, the present application provides a server, comprising at least one processor and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions, when executed by the at least one processor, cause the server to perform the method provided in any one of the preceding aspects.

[0022] In a fifth aspect, the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer execution instructions, and when a processor executes the computer execution instructions, the method provided in any one of the preceding aspects is implemented.

[0023] In a sixth aspect, the present application provides a computer program product, comprising a computer program, and the computer program, when executed by a processor, implements the method provided in any one of the preceding aspects.

[0024] The medical data processing method, computing device, server, storage medium and product provided by the application help to store and analyze data in a structured manner, so that complex medical test single information becomes more organized and easier for model understanding, facilitating subsequent processing and analysis of intelligent interpretation models; by identifying index entities and test item entities and combining the searched medical knowledge, the intelligent interpretation model can be provided with rich medical field knowledge, so that the interpretation model can generate more accurate and personalized interpretation results, help users better understand their health status, and also provide personalized suggestions and information related to the specific health status of the user, and finally the automatic extraction of information reduces the time and possible errors of manual interpretation, improves the efficiency and accuracy of medical data processing, and reduces human errors and subjective bias. BRIEF DESCRIPTION OF DRAWINGS

[0025] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and serve to explain the principles of the application.

[0026] Figure 1 A system architecture diagram of the medical data processing system provided by the application;

[0027] Figure 2 A medical data processing method flowchart provided by an exemplary embodiment of the application;

[0028] Figure 3 A classification diagram of test single types provided by an exemplary embodiment of the application;

[0029] Figure 4 A flowchart of another medical data processing method provided by an exemplary embodiment of the application;

[0030] Figure 5 A flowchart of searching medical knowledge related to index entities and test item entities provided by an exemplary embodiment of the application;

[0031] Figure 6 A flowchart of generating an interpretation result of a medical test single by an intelligent interpretation model provided by an exemplary embodiment of the application;

[0032] Figure 7 A flowchart of training an intelligent interpretation model provided by an exemplary embodiment of the application;

[0033] Figure 8 A flowchart of another medical data processing method provided by an exemplary embodiment of the application;

[0034] Figure 9System architecture diagram of a medical data processing method according to an exemplary embodiment of the present application;

[0035] Figure 10 Structure block diagram of a computing device according to an embodiment of the present application;

[0036] Figure 11 Structure diagram of a server according to an embodiment of the present application.

[0037] The specific embodiments of the present application have been shown through the above-described drawings, and will be described in more detail hereinafter. These drawings and written descriptions are not intended to limit the scope of the present application concept in any way, but to illustrate the present application concept to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0038] The exemplary embodiments will be described in detail herein with reference to the accompanying drawings. In the following description, the same numbers refer to the same or similar elements throughout the drawings, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not meant to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present application as detailed in the appended claims.

[0039] It should be noted that the user information (including but not limited to user device information, user attribute information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards, and provide corresponding operation portal for user to choose authorization or refusal.

[0040] First, the terms involved in the present application are explained:

[0041] Test report: The medical test report is a formal document filled out by medical personnel after various medical tests on patients. This report records the specific test items, test methods, test dates, test results, and doctors' preliminary diagnoses or recommendations.

[0042] Test report: The test report is a formal document issued by the medical institution after completing the laboratory analysis of patient samples (such as blood, urine, etc.). It records specific test items, test methods, test dates, results data, and reference ranges, helping doctors make diagnoses or adjust treatment plans.

[0043] Report Interpretation: The process of analyzing and interpreting information from medical examination reports, test reports, and other medical documents by professional medical personnel or experts in related fields. This process aims to help patients understand their health status and how doctors make diagnoses, develop treatment plans, or provide health management recommendations based on this information.

[0044] OCR: Optical Character Recognition, a technology that converts printed or handwritten text from paper documents, images, or other media into electronic documents.

[0045] LLM (Large Language Model): A large-scale language model trained on massive amounts of text data, usually based on the Transformer architecture. It has the ability to generate, understand, reason, and translate natural language by learning statistical patterns and semantic associations in text.

[0046] Large Model: A deep learning model with a large number of model parameters, typically containing hundreds of millions, billions, or even tens of billions of model parameters. Large models can also be referred to as Foundation Models (FM). Through large-scale unlabeled corpus pre-training, pre-trained models with billions of parameters are produced. Such models can adapt to a wide range of downstream tasks and have good generalization ability. For example, large-scale language models, multi-modal pre-training models, etc.

[0047] In practical applications, large models only need a small amount of sample data to fine-tune pre-trained models and can be applied to different tasks. Large models can be widely used in natural language processing (NLP) and computer vision fields. Specifically, they can be applied to tasks such as visual question answering (VQA), image captioning (IC), image generation, sentiment classification based on text, text summarization generation, machine translation, etc. The main application scenarios of large models include digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc.

[0048] With the rise of health awareness and the increasing demand for medical services worldwide, individuals are paying more attention to their own health status. Medical test reports, as an important part of health management, have become the main way for patients and their families to obtain health information. However, these reports generally have poor readability. The reports are filled with a large number of professional terms and complex test indicators, which are difficult for ordinary people without a medical background to understand. This information asymmetry leads patients and their families to feel confused when facing health data, unable to accurately grasp their own or their family's health status, and may even trigger unnecessary anxiety or misunderstanding.

[0049] To address the above challenges, patients or their families usually take the following measures:

[0050] The first way is to search the Internet for explanations of terms and indicators in the test report to provide some basic information and background knowledge. However, the information obtained through search engines is of varying quality, and patients may receive inaccurate or not applicable explanations for their specific situation, which may lead to unnecessary anxiety due to over-interpretation or misinterpretation of online information. Moreover, online information often lacks context related to the patient's specific health condition, and it may not be possible to fully interpret the report based on this information alone.

[0051] The second way is to take the test report to a professional doctor for a follow-up visit to obtain professional explanations and further medical advice. However, scheduling a follow-up visit usually requires waiting, especially in situations where medical resources are scarce or doctors are busy with appointments. Patients may need to spend a long time to see a doctor, and the doctor's follow-up visit may be limited in time, unable to answer all of the patient's questions or provide sufficient health education.

[0052] To solve the above technical problems, the present application provides a medical data processing method. The server first extracts text information from a medical test sheet, extracts detection result data, content of at least one basic information item and medical use information contained in the medical test sheet from the text information, and can also identify index entities and detection item entities contained in the medical test sheet, and search medical knowledge related to the index entities and the detection item entities. The index entity refers to a specific, measurable health parameter or variable, which is usually used to evaluate a specific physiological or biochemical state. For example, the index entity can be the number of red blood cells, white blood cell count, hemoglobin concentration, etc. The detection item entity refers to a set of related tests, which is usually used to evaluate the health status of a certain aspect. The detection item entity contains multiple index entities. For example, the detection item entity can include blood routine, urine routine, etc. The present application does not limit the order of identifying the index entity (or the item entity) or generating the structured information. Finally, the detection result data, the content of the at least one basic information item and the medical use information contained in the medical test sheet, and the medical knowledge corresponding to the medical test sheet are input into an intelligent interpretation model, and the intelligent interpretation model can generate an interpretation result of the medical test sheet.

[0053] For example, the medical test sheet is a blood routine test sheet. The server first extracts text information from the blood routine test sheet, then extracts detection result data, content of at least one basic information item and medical use information contained in the medical test sheet from the text information, and the medical knowledge corresponding to the medical test sheet, and can also identify index entities and detection item entities in the blood routine test sheet. The detection item entity is blood routine, and the index entities include red blood cell count, white blood cell count, hemoglobin, hematocrit, mean corpuscular volume, etc. Then, the medical knowledge related to "blood routine", "red blood cell count", "white blood cell count", "hemoglobin", "hematocrit", "mean corpuscular volume" is searched respectively. Finally, the detection result data, the content of the at least one basic information item and the medical use information contained in the medical test sheet, and the searched related medical knowledge are input into an intelligent interpretation model, and the intelligent interpretation model generates an interpretation result of the blood routine test sheet.

[0054] In this way, by extracting the detection result data, the basic information items and the medical use information from the text information, it is helpful to store and analyze the data in a structured manner, so that the complex medical detection single information becomes more organized and easier for model understanding, facilitating the subsequent processing and analysis of the intelligent interpretation model; by identifying the index entities and the detection item entities and combining the searched medical knowledge, the intelligent interpretation model can be provided with rich medical field knowledge, so that the interpretation model can generate more accurate and personalized interpretation results, help users better understand their health status, and also provide personalized suggestions and information related to the specific health status of the user, and finally the automatic extraction of information reduces the time and possible errors of manual interpretation, improves the efficiency and accuracy of medical data processing, and reduces human errors and subjective bias.

[0055] Figure 1 The system architecture diagram of the medical data processing system provided in the present application is shown in FIG. 1, which includes a server and a client device. The server and the client device have a communicable communication link therebetween, and can realize the communication connection between the server and the client device. Figure 1

[0056] The server is a device with computing capability deployed in the cloud or locally, such as a cloud cluster. The server is a server device in the medical data processing system, which is responsible for generating the interpretation result of the medical detection single based on the text information of the medical detection single.

[0057] The client device can be an electronic device running the client of the medical data processing system, which can be specifically a hardware device with network communication function, operation function and information display function, including but not limited to smart phones, tablet computers, desktop computers, local servers, cloud servers, etc. The user interacts with the server through the client device used to obtain the interpretation result of the medical detection single.

[0058] ​In this embodiment, the user uploads the text information of the medical test sheet through the client, the client sends the text information of the medical test sheet uploaded by the user to the server, the server obtains the text information of the medical test sheet, and extracts the detection result data, the content of at least one basic information item and the medical use information contained in the medical test sheet from the text information. The detection result data, the content of at least one basic information item and the medical use information contained in the medical test sheet can be presented in the form of structured information. The index entity and the detection item entity contained in the medical test sheet are identified from the text information. For example, the identified entities are: detection item entity 1, index entity 1, index entity 2 and index entity 3. The first medical knowledge related to the detection item entity 1, the second medical knowledge related to the index entity 1, the third medical knowledge related to the index entity 2 and the fourth medical knowledge related to the index entity 3 are searched respectively. The detection result data, the content of at least one basic information item and the medical use information contained in the medical test sheet, the first medical knowledge, the second medical knowledge, the third medical knowledge and the fourth medical knowledge are input into the intelligent interpretation model, and the intelligent interpretation model generates an interpretation result of the medical test sheet.

[0059] Further, the client displays the interpretation result of the medical test sheet on the display interface after receiving the interpretation result of the medical test sheet.

[0060] The medical data processing method provided by the present application can provide rich background information for the intelligent interpretation model, so that the interpretation model can generate more accurate and personalized interpretation results, help users better understand their health status, and also provide personalized suggestions and information related to the specific health status of the user, and reduce the time and possible errors of manual interpretation, improve the efficiency and accuracy of medical data processing, and reduce human errors and subjective bias.

[0061] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail in the specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present application will be described below with reference to the accompanying drawings.

[0062] Figure 2 The medical data processing method flowchart provided by an exemplary embodiment of the present application. The execution subject of this embodiment is the server in the system architecture described above. As shown in the figure, the specific steps of the method are as follows: Figure 2

[0063] Step S201, obtaining the text information of the medical test sheet.

[0064] Among them, the medical test sheet includes test report and examination report.

[0065] ​A laboratory report typically involves laboratory test results and is used to document the results of analyzing patient samples (such as blood, urine, tissue samples, etc.) through chemical, physical, or biological methods.

[0066] A laboratory report typically includes the following:

[0067] Patient information: including name, age, gender, patient ID, etc.

[0068] Test items: such as blood routine, urine routine, liver function test, etc.

[0069] Test indicators: specific measurement parameters, such as hemoglobin concentration, red blood cell count, blood sugar level, etc.

[0070] Result values: specific measurement results for each indicator.

[0071] Reference range: used to determine whether the results are within the normal range.

[0072] Test date and time: the date and time of the test.

[0073] Notes: may include doctors' explanations or suggestions.

[0074] An examination report typically involves imaging examination or other non-laboratory medical examination results, and is used to document the results of direct observation or imaging of the patient's body through equipment or instruments.

[0075] An examination report typically includes the following:

[0076] Patient information: including name, age, gender, patient ID, etc.

[0077] Examination type: such as X-ray, CT (Computed Tomography), MRI (Magnetic Resonance Imaging), ultrasound examination, etc.

[0078] Examination results: describe the observed images or physical characteristics.

[0079] Image description: detailed description of the structures or abnormalities visible in the image.

[0080] Diagnostic opinion: doctors' professional interpretation of the examination results and possible diagnosis.

[0081] Examination date and time: the date and time of the examination.

[0082] Examination agency information: including agency name, address and contact information.

[0083] Note: May include follow-up recommendations or recommendations for further review.

[0084] The text information includes all the textual content recorded on the medical test sheet, as well as the format information corresponding to the text. The format information refers to the presentation and structural layout of the text content, i.e. how to organize and display the content. The format information helps to understand the structure and semantics of the text. For example, the format information can include:

[0085] Text position: the specific position coordinates of the text in the image.

[0086] Font attributes: including font type, size, color, boldness (such as bold), italic, etc.

[0087] Text alignment: the alignment of the text, such as left alignment, right alignment, center alignment, etc.

[0088] Paragraph and line spacing: spacing between paragraphs and line spacing information.

[0089] Table structure: if the text is located in a table, the OCR (Optical Character Recognition) can identify the row and column structure of the table, helping to understand the organization of the data.

[0090] Markers and symbols: such as bullets, numbers, asterisks, etc.

[0091] Specifically, the server can obtain the text information of the medical test sheet. The application does not limit the way the server obtains the text information of the medical test sheet.

[0092] In an optional implementation, the user inputs the text information of the medical test sheet through the client, the client sends the medical test sheet text information input by the user to the server, and the server obtains the text information of the medical test sheet.

[0093] In another optional implementation, the user uploads the image of the medical test sheet through the client, the client sends the image of the medical test sheet to the server, the server performs OCR identification on the received image of the medical test sheet to obtain the OCR identification result of the medical test sheet, and the server takes the OCR identification result as the text information of the medical test sheet. It should be noted that OCR identification is one of the methods, and the application does not exclude other possible technical means as long as the purpose of converting the image into text information can be achieved.

[0094] In yet another optional implementation, the user uploads an image of the medical test sheet through the client, the client performs OCR recognition on the image of the medical test sheet to obtain an OCR recognition result, and sends the OCR recognition result as the text information of the medical test sheet to the server, and the server obtains the text information of the medical test sheet sent by the client.

[0095] In this way, the present application allows two different input approaches: image recognition through OCR and direct text input. This enables the system to adapt to different use scenarios and user needs. By automatically converting images to text information through OCR recognition, the time and labor intensity of manual input are reduced, and the efficiency of obtaining text information is improved. By allowing direct text input, character recognition errors that may occur in the OCR process are avoided, thereby improving the accuracy of the data.

[0096] Step S202, extracting the detection result data contained in the medical test sheet, the content of at least one basic information item, and the medical use information from the text information.

[0097] Among them, the plurality of basic information items are used to indicate the basic information of the user. For example, "gender", "age", "department", "report time", "audit time", "clinical diagnosis", "specimen type", "test sheet name", etc. all belong to basic information items. The content of the basic information item is the field value corresponding to the basic information item, for example, the field value corresponding to "gender" is "male" or "female". The medical use information of the basic information item indicates the medical use or clinical significance of the basic information item.

[0098] The detection result data is used to indicate the detection result of the user. The detection result can include detection result information of at least one index. For example, in a test report, at least one index corresponds to at least one test item. For example, in a blood routine test report, the at least one index includes red blood cell count, white blood cell count, hemoglobin concentration, etc. The detection result information of each index can include the Chinese name, English name, detection result (qualitative or quantitative), reference value (qualitative or quantitative), unit, detection method, etc. of the index.

[0099] In the examination report, at least one index corresponds to at least one examination item. For example, in a liver and spleen examination report, at least one index includes liver image and spleen image. The detection result of the liver image can include shape, size, density, lobe ratio, diameter, etc.

[0100] The detection result data, the content of the at least one basic information item, and the medical use information included in the medical detection sheet can be presented in the form of structured information. The structured information includes detection sheet information and detection result data. The detection sheet information includes the at least one basic information item of the medical detection sheet and the content and medical use information of each basic information item. The detection result data includes detection result information of at least one index.

[0101] Structured information refers to converting unstructured or semi-structured data into an organized format, making it easier to be processed and analyzed by computer programs. The structured information of the medical detection sheet includes converting text information into specific fields and numerical pairs.

[0102] Specifically, the present application does not limit the way the server extracts the detection result data, the content of the at least one basic information item, and the medical use information included in the medical detection sheet from the text information, i.e., generates the structured information of the medical detection sheet according to the text information.

[0103] In an optional implementation, a standardized detection sheet template can be created to define the position and format of each field, and a template matching algorithm can be used to map the text information into the standardized template.

[0104] In another optional implementation, the text information can be input into an information structure model, and the information structure model can generate the structured information of the medical detection sheet. The information structure model can be any large model, such as various LLMs.

[0105] It should be noted that the structured information is the information extracted from the text information by the information structure model or the template matching algorithm, and does not necessarily include all the information in the text information.

[0106] Table 1 is a structured information template of a test report sheet provided by an embodiment of the present application. It should be noted that when describing a structured information template of a test report sheet, the present application uses a table to visually display the organization of information. However, in actual applications, especially in digital storage and processing, data formats such as JSON, XML, or database table structures are usually used to express these information.

[0107] The test report sheet in Table 1 includes two first-level fields, namely, "basic information" and "test result", each of which includes a plurality of second-level fields. For example, the first-level field "basic information" includes the second-level fields of "gender", "age", "department", "report time", "audit time", "clinical diagnosis", "specimen type", and "test sheet name". Each of the second-level fields has a corresponding field value. The "example" column in Table 1 is the possible field value. The "format" column is used to indicate the format of the corresponding field value. The "description" column is a detailed explanation or clinical effect of the corresponding second-level field. The "whether necessary" column indicates whether the corresponding second-level field is necessary. For example, referring to Table 1, all the second-level fields corresponding to the first-level field "basic information", such as "gender", "age", "department", "report time", "audit time", "clinical diagnosis", "specimen type", and "test sheet name", are basic information items. The content of the basic information item corresponds to the field value of the corresponding second-level field. For example, the field value of "gender" is "male" or "female". The medical use information of the basic information item corresponds to the "description" column in Table 2.

[0108] Table 1

[0109]

[0110]

[0111] Table 2 is a structured information template of the test report sheet provided in the embodiment of the present application.

[0112] Table 2

[0113]

[0114] In addition to the above-mentioned manner in the embodiment, the present application can also use other manners when extracting the test result data, the content of at least one basic information item, and the medical use information of the medical test sheet from the text information, that is, generating the structured information of the medical test sheet according to the text information.

[0115] In an optional implementation manner, the server can obtain the test sheet name. The test sheet name can be obtained from the text information or extracted from the name of the medical test sheet by other manners, which is not limited in the present application. After obtaining the test sheet name, the test sheet name and the text information are input into the structured model. The structured model extracts the structured information of the medical test sheet from the text information according to the test sheet name.

[0116] In another optional implementation method, the server can perform text classification and recognition on the text information of the medical test form, determine the test form type of the medical test form, input the test form type and text information of the medical test form into the information structuring model, and extract the structured information of the medical test form from the text information according to the test form type through the information structuring model.

[0117] Figure 3 A schematic diagram of the classification of test order types provided for an exemplary embodiment of the present application, such as Figure 3 As shown in the figure, the test report types include target document categories, easily confused document categories, and non-target document categories. The target document category includes test reports and examination reports, the easily confused document category includes medical records, prescriptions, physical examination reports, drug instructions, and pet medical reports, and the non-target document category includes all non-medical text.

[0118] This application does not limit the way in which the server performs text classification and recognition on the text information of the medical test form and determines the test form type of the medical test form. For example, a text classification model can be trained, and the server inputs the text information of the medical test form into the text classification model, and the text classification model can output the test form type of the medical test form.

[0119] When the type of the test form is the target document classification, that is, when the test form is a test report or an examination report, the test form type and text information of the medical test form are input into the information structuring model. The information structuring model can extract the structured information of the medical test form from the text information according to the test form type.

[0120] When the type of the test order is not an inspection report or an examination report, a first preset wording will be output to the client. The first preset wording is used to explain that the type of the input test order exceeds the scope that this model can handle. For example, the output wording can be "Sorry, the type of the test order exceeds the processing range of this model."

[0121] Figure 4 A flowchart of another medical data processing method provided by an exemplary embodiment of the present application is shown as follows: Figure 4As shown, the user uploads a picture of a medical test sheet through the client, the client sends the picture of the medical test sheet to the server, the server performs OCR recognition on the picture of the medical test sheet, if the OCR recognition is successful, the medical test sheet is classified according to the OCR recognition result, the type of the medical test sheet is obtained, when the type of the medical test sheet is a target document classification, that is, the medical test sheet is a test report sheet or an examination report sheet, the type of the medical test sheet and the text information are input into an information structured model, the information structured model extracts structured information of the medical test sheet from the text according to the type of the test sheet. If the medical test sheet is not a test report sheet or an examination report sheet, a first preset dialogue is output to the client. If the OCR recognition fails, a second preset dialogue is output to the client, the second preset dialogue is used to indicate that the OCR recognition fails.

[0122] In this way, by using the text classification technology, the type of the test sheet can be accurately recognized, and the test sheet type and the text information are input into the structured information model, which can improve the accuracy of the structured information extracted by the structured information model.

[0123] In step S203, the index entity and the detection item entity contained in the medical test sheet are recognized, and the medical knowledge related to the index entity and the detection item entity is searched to obtain the medical knowledge corresponding to the medical test sheet.

[0124] The index entity refers to a specific, measurable health parameter or variable, which is usually used to evaluate a specific physiological or biochemical state. For example, the index entity can be the number of red blood cells, the white blood cell count, the hemoglobin concentration, etc.

[0125] The detection item entity refers to a set of related detections, which is usually used to evaluate the health status of a certain aspect. The detection item entity contains multiple index entities. For example, the detection item entity can include blood routine, urine routine, etc.

[0126] The related medical knowledge can include various enumerated medical data of intent, for example, the related medical knowledge can be the definition, normal range, clinical significance, treatment method, medical advice, preventive measures, etc. of the index entity or the detection item entity.

[0127] In the recognition of the index entity and the detection item entity contained in the medical test sheet, various ways can be used, which are not limited by the present application.

[0128] In an optional implementation, the NER (Named Entity Recognition model) model can be used to perform named entity recognition on the text information of the medical test sheet to obtain the index entity and the detection item entity contained in the medical test sheet.

[0129] In another optional implementation, natural language processing tools can be used for word segmentation, and the text information of the medical test sheet is processed to obtain a plurality of word units. The plurality of word units are matched with the indexes and test items contained in the medical vocabulary library. When matching, a plurality of matching methods can be used, such as direct matching, fuzzy matching, and context matching, etc. The present application does not limit this.

[0130] Direct matching: the word units after word segmentation are directly matched with the terms in the vocabulary library. Hash table or dictionary data structure can be used to speed up the search process.

[0131] Fuzzy matching: for word units that may have spelling errors or variants, fuzzy matching techniques such as edit distance (Levenshtein distance) or Jaccard similarity can be used to identify possible matches.

[0132] Context matching: considering the context of the word units in the text to improve the accuracy of matching, context window or n-gram model can be used to capture relevant information.

[0133] For any word unit, if the word unit matches any index in the medical vocabulary library, the matched index is taken as the index entity, and if the word unit matches any test item in the medical vocabulary library, the matched test item is taken as the test item entity.

[0134] In yet another optional implementation, the two above-mentioned implementations can be used respectively to obtain the index entities and test item entities contained in the medical test sheet. Then, the index entities and test item entities obtained by the two methods are respectively taken to perform a set operation to obtain the index entities and test item entities finally contained in the medical test sheet.

[0135] In this way, the NER model usually needs a large amount of labeled data for training, and the vocabulary library matching depends on the term library, which reduces the dependence on labeled data. NER, vocabulary library matching or a combination of the two can be selected according to specific needs to flexibly adjust the identification strategy and improve the accuracy of identifying the index entities and test item entities contained in the medical test.

[0136] When searching for medical knowledge related to the index entities and test item entities, different methods can be used, which are not limited by the present application.

[0137] In an optional implementation, the server can directly search in the medical knowledge graph according to the index entities and test item entities to obtain medical knowledge related to the index entities and test item entities. The medical knowledge graph is a structured data model used to organize and represent medical knowledge.

[0138] In another alternative implementation, the indicator entity and the detection item entity identified from the text are first mapped to standardized names, in the following manner:

[0139] For any indicator entity, the indicator entity is matched with any indicator name in the standard medical database, and the indicator name that matches the indicator entity is the standard indicator name corresponding to the indicator entity.

[0140] The specific matching method can be complete matching, alias matching, fuzzy matching, etc., which are not limited in the present application. The complete matching means that the indicator entity is matched with the indicator name in the standard medical database word by word and symbol by symbol, which means that the indicator entity must be completely consistent with the indicator name in the standard medical database, including character order, spelling and format. Alias matching involves matching the indicator entity with the indicator name and its known alias in the standard vocabulary library. Alias refers to different names or expressions of the indicator entity, such as abbreviations, synonyms or common variants. And the standard medical database contains multiple indicator names and multiple detection item names, the indicator name is the standard indicator name in the medical field, and the detection item name is the standard detection item name in the medical field.

[0141] For any detection item entity, the detection item entity is matched with any detection item name in the standard medical database, and the detection item name that matches the detection item entity is the standard detection item name corresponding to the detection item entity. The matching process of the detection item entity is similar to that of the indicator entity, which will not be described here.

[0142] After obtaining the standard indicator name and the standard detection item name, the server searches the medical knowledge related to the standard indicator name and the medical knowledge related to the standard detection item name in the medical knowledge graph.

[0143] In this way, by mapping the entity to the standardized name, consistent queries in the medical knowledge graph can be ensured, and retrieval errors caused by term diversity can be reduced. The standardization process helps to eliminate the ambiguity caused by different expressions of the same entity, and improves the accuracy of the retrieval result.

[0144] Figure 5 The flowchart of searching the medical knowledge related to the indicator entity and the detection item entity provided by an exemplary embodiment of the present application is shown in Figure 5As shown, first, the text information of the medical test sheet is acquired, then the NER model can be used to perform named entity recognition on the text information of the medical test sheet, and the text information of the medical test sheet can also be processed to obtain a plurality of word units, the plurality of word units are matched with the indexes and detection items contained in the medical vocabulary library, the word units matched with any index in the medical vocabulary library are taken as index entities, the word units matched with any detection item in the medical vocabulary library are taken as detection item entities, the index entities and detection item entities obtained through the NER model and the dictionary matching logic are merged, standardized mapping is performed according to the type of the medical test sheet, the index entities are mapped into corresponding standard index names, and the detection item entities are mapped into corresponding standard detection item names; the mapping process can include complete matching, alias matching and fuzzy matching, in the medical knowledge graph, medical knowledge related to the standard index name and medical knowledge related to the standard detection item name are searched.

[0145] In step S204, the detection result data contained in the medical test sheet, the content of the at least one basic information item and the medical use information, and the medical knowledge corresponding to the medical test sheet are input into an intelligent interpretation model, and an interpretation result of the medical test sheet is generated through the intelligent interpretation model.

[0146] The intelligent interpretation model can be any kind of large language model or other deep learning model with similar capabilities, which is not limited here. The intelligent interpretation model can generate an interpretation result of the medical test sheet according to the detection result data contained in the medical test sheet, the content of the at least one basic information item and the medical use information, and the medical knowledge corresponding to the medical test sheet.

[0147] The interpretation result can include the following aspects:

[0148] Diagnosis information: the model can provide a diagnostic explanation of the medical test sheet. For example, if the test sheet involves blood tests, the model can point out some abnormal indicators and may give hints of potential health problems or diseases.

[0149] Health advice: based on the medical test sheet, the model can provide health management advice, which can include lifestyle changes, dietary advice, further medical examinations or advice to consult professional medical personnel.

[0150] Risk assessment: the model can assess the patient's health risks based on the medical test sheet and known medical knowledge, and point out health risk factors that may need attention.

[0151] Data interpretation: the model can convert complex medical terminology and data in the medical test sheet into easy-to-understand language, helping patients or non-professionals understand the meaning of the test results.

[0152] Specifically, after obtaining the detection result data contained in the medical detection sheet, the content of at least one basic information item, medical use information, and medical knowledge corresponding to the medical detection sheet, the server inputs the detection result data contained in the medical detection sheet, the content of at least one basic information item, medical use information, and medical knowledge corresponding to the medical detection sheet into the intelligent interpretation model, and generates an interpretation result of the medical detection sheet through the intelligent interpretation model.

[0153] For example, the server first obtains the prompt corresponding to the medical detection sheet, which can include task information, precautions, examples, and slots corresponding to embedded information, etc. The specific design and adjustment can be made according to the actual task requirements, which is not limited here. After obtaining the prompt, the detection result data contained in the medical detection sheet, the content of at least one basic information item, medical use information, and medical knowledge corresponding to the medical detection sheet are embedded into the corresponding slot of the prompt to obtain prompt information, which is input into the intelligent interpretation model to obtain the interpretation result of the medical detection sheet.

[0154] Optionally, the prompts corresponding to the examination report and the test report can be the same or different. When the prompts corresponding to the examination report and the test report are different, the corresponding prompt is first selected according to the type of the medical detection sheet, and then the detection result data contained in the medical detection sheet, the content of at least one basic information item, medical use information, and medical knowledge corresponding to the medical detection sheet are embedded into the corresponding prompt to obtain the corresponding prompt information, which is then input into the intelligent interpretation model to obtain the interpretation result of the medical detection sheet.

[0155] For example, the prompt of a medical detection sheet interpretation task is as follows:

[0156] "[Task information]:

[0157] You are a chief physician, please give scientific interpretation and suggestions according to the patient's medical detection sheet and reference relevant medical materials, according to the medical diagnosis logic.

[0158] [Symbol definition]:

[0159] 1. "" is a special mark indicating the knowledge content needed;

[0160] 2. <...> represents a text to be filled in, and the attribute specifies the requirements for the content of the text, and the `attribute` usually refers to the attributes or requirements related to the placeholder. These attributes can specify what content should be filled in the placeholder.

[0161] [Rules that the interpretation result must meet]:

[0162] … (omitted multiple rules here) [example]:

[0163] Input:

[0164] """[patient situation]

[0165] Age: 57,

[0166] Gender: female

[0167] """

[0168] """[medical test information]

[0169] {"department": "cardiology department",

[0170] "clinical diagnosis": "",

[0171] "test sheet": "chest CT test sheet",

[0172] "test results": "Both sides of the chest shape are symmetrical, the mediastinum and trachea are in the middle, the lung markings are increased, the right lung middle lobe (thin layer I arc 134) sees glass nodules shadow, size about 3m x 3mm, left lung sees point calcification shadow, both lungs see cystic transparent shadow and a little bit of cord shadow. Bilateral lung lobes, segmental bronchial morphology is normal, both sides of the lung hilum and mediastinal structure are normal. Part of the aortic wall calcification.",

[0173] "test diagnosis": "Right lung micro nodules, suggest review. Left lung calcification. Localized emphysema in both lungs. Cord shadow in both lungs. Part of the aortic wall calcification. Please combine with clinical, if symptoms persist, review in time.""

[0174] """[related medical data]

[0175] {

[0176] "left lung calcification": {"title": "What does left lung calcification mean",

[0177] "Content":"Left lung calcifications usually indicate a past tuberculosis infection, which the patient may or may not be aware of. During tuberculosis infection, the primary repair process is calcium deposition, leaving calcified lesions in the left lung. Calcified lesions after tuberculosis can vary in size, regular or irregular, spherical, or have other morphologies. When left lung calcifications also extend to pleural calcifications, the calcified lesions may appear larger and have clear boundaries, but their morphology may be irregular. Normally, lung calcifications are mostly located in the upper lung fields and are mostly spherical. However, Calcifications in the left upper lung can be spherical or sheet-like. They are considered calcifications because they have clear boundaries, a high density, and can remain unchanged for one, two, ten, twenty, or thirty years. If a calcification in the left upper lung is dense and well-defined, its shape may be regular or irregular, and it may be spherical or sheet-like. Because it remains unchanged for ten, twenty, thirty, or even forever, even if the patient was previously unaware of their tuberculosis history, the calcification in the left upper lung is considered a calcification, a remnant, or a remnant of a past tuberculosis infection. "},……(Multiple medical knowledge omitted here)"""

[0178] Interpretation of the results:

[0179] What you provided is a routine chest CT scan report.

[0180] 1. Report Interpretation and Analysis

[0181] Right lung micronodule: A 3mm*3mm ground-glass nodule is present in the right middle lobe of the lung. This nodule is usually benign, but regular chest CT scans are required to monitor the growth of the nodule.

[0182] Left lung calcification: Dot-like calcifications are seen in the left lung. These are traces left behind by previous lung inflammation and infection and usually do not require special treatment.

[0183] Localized emphysema in both lungs: Cystic lucencies are seen in both lungs. This is caused by overexpansion and fusion of alveoli in localized areas of the lungs, leading to airway insufficiency. Localized emphysema is usually minor, but if symptoms such as chest tightness and shortness of breath occur, further evaluation with pulmonary function tests is indicated.

[0184] Linear shadows in both lungs: A few linear shadows in both lungs. This is usually related to factors such as lung inflammation and fibrosis. A small number of linear shadows is usually not serious and needs to be comprehensively judged based on the presence of cough and sputum symptoms and other test results such as blood routine.

[0185] Aortic partial wall calcification: aortic partial wall calcification. This is a manifestation of arterial vascular sclerosis, which may be related to factors such as age, high blood pressure, high blood sugar, high blood lipids, etc. You are advised to pay attention to blood pressure, blood sugar and blood lipids.

[0186] 2. Department of Medicine Recommendation

[0187] Respiratory medicine or cardiovascular medicine.

[0188] 3. Home life suggestion

[0189] Daily quit smoking, avoid contact with dust, second-hand smoke environment.

[0190] Daily attention to a balanced diet, reduce high oil and high fat food intake. Ensure 7-8 hours of sleep per day, and insist on appropriate exercise (such as 150 minutes of walking or jogging per week).

[0191] [Input]:

[0192] Detection single information: {medical detection single information}

[0193] Related medical information: {medical knowledge}

[0194] [Interpretation result]:

[0195] ”

[0196] Among them, {medical detection single information} represents a slot for embedding detection result data, content of at least one basic information item and medical use information contained in a medical detection single, and {medical knowledge} represents a slot for embedding medical knowledge. It should be noted that the prompt words can also include slots corresponding to other information to be input into the intelligent interpretation model, including but not limited to slots for embedding abnormal indicator information, slots for embedding abnormal processing prior knowledge. In this embodiment, by extracting detection result data, basic information item and medical use information from text information, it is helpful to store and analyze data in a structured manner, so that complex medical detection single information becomes more organized and easier for the model to understand, facilitating subsequent processing and analysis of the intelligent interpretation model; By identifying the indicator entity and the detection item entity, and combining the medical knowledge obtained by searching, the intelligent interpretation model can provide rich medical knowledge, so that the interpretation model can generate more accurate and personalized interpretation results, help users better understand their health status, and also provide personalized suggestions and information related to the user's specific health status. Finally, the automatic extraction of information reduces the time and possible errors of manual interpretation, improves the efficiency and accuracy of medical data processing, and reduces human errors and subjective bias.

[0197] Figure 6A flowchart of generating an interpretation result of a medical test sheet by an intelligent interpretation model is provided for an exemplary embodiment of the present application. The structured information of the medical test sheet includes test sheet information and test result data, wherein the test sheet information includes a plurality of basic information items of the medical test sheet and the content and medical use information of each basic information item, and the test result data includes detection result information of at least one index, such as Figure 6 The specific steps of the method are as follows:

[0198] Step S601, obtaining text information of a medical test sheet.

[0199] Step S602, generating structured information of the medical test sheet according to the text information.

[0200] Step S603, identifying index entities and detection item entities contained in the medical test sheet, and searching for medical knowledge related to the index entities and the detection item entities in a medical knowledge graph.

[0201] The specific implementation modes of steps S601-S603 are the same as those of steps S201-S203, which are not limited herein.

[0202] Step S604, determining abnormal index information according to the test result data.

[0203] Among the abnormal index information, there can be only a plurality of abnormal indexes, or both a plurality of abnormal indexes and a plurality of abnormal degrees corresponding to the abnormal indexes, wherein the abnormal degree is related to the degree of the test result exceeding or being less than the reference value or reference range. For example, the abnormal degree can be divided into slight abnormality, general abnormality and critical situation. The specific division logic of the abnormal degree can need to be configured according to actual needs, which is not limited herein. Optionally, the indexes related to severe cases can be divided by the above-mentioned method of dividing the abnormal degree, and the indexes not related to severe cases can not be divided or can be divided more simply, such as the difference between the test result and the reference value being greater than b belonging to severe abnormality, and the difference between the test result and the reference value being less than b belonging to mild abnormality.

[0204] The server can adopt different implementation modes when determining the abnormal index information according to the test result data, which is not limited herein. The following several implementation modes can be used alone or in combination, which is not limited herein.

[0205] In an optional implementation, it can be identified whether the detection result data contains special identifiers, which can be exemplarily "↑", "↓" and "*", wherein "↑" represents that the corresponding detection result value is high, "↓" represents that the corresponding detection result value is low, and "*" can have different meanings in different laboratories or medical systems, but common uses include that "*" is used to mark an abnormal value, indicating that the value of the detection result significantly deviates from the normal range and needs special attention. Or in some cases, "*" can be used to identify a critical value, prompting that the value of the detection result can have an urgent clinical significance and needs immediate attention or action.

[0206] When the server contains special identifiers in the detection result data, the abnormal index information is determined according to the contained special identifiers. For example, "↑" or "↓" can determine which indexes belong to abnormal indexes, and "*" can determine the abnormal degree of the abnormal indexes.

[0207] In another optional implementation, the server can extract the actual value and reference value of the detection result of each index in the detection result data. Exemplarily, the reference value can be a numerical range, which can be a normal value or an abnormal value of the detection result. If the reference value is a normal value, the index is an abnormal index when the detection result of the index is different from the reference value, that is, the actual value corresponding to the test result is not in the numerical range corresponding to the reference value. If the reference value is an abnormal value, the index is an abnormal index when the detection result of the index is the same as the reference value, that is, the actual value corresponding to the test result is in the numerical range corresponding to the reference value. And for any abnormal index, the abnormal degree information of the abnormal index can be determined according to the detection result and the reference value of the abnormal index.

[0208] Exemplarily, the abnormal degree of the abnormal index can include slight abnormality, general abnormality and critical situation, the value corresponding to the detection result of the abnormal index is c, the normal value range corresponding to the abnormal index is [d, e], and c < d < e. The difference f between c and d is calculated. If 0 < f < a, the abnormal degree of the abnormal index is slight abnormality. If a < f < b, the abnormal degree of the abnormal index is general abnormality. If f > b, the abnormal degree of the abnormal index is critical situation, wherein 0 < a < b.

[0209] Optionally, if the server does not extract the reference value of the detection result of a certain index from the detection result data, the reference value of the detection result of the index can be obtained from the medical knowledge base or the medical knowledge graph described above, and then the subsequent abnormality and abnormal degree judgment are performed.

[0210] In this way, by comparing the detection results with the reference values, it can be accurately identified which indicators are abnormal. This helps to quickly locate potential health problems and reduce the possibility of misdiagnosis or missed diagnosis. In addition to identifying abnormal indicators, it further quantifies the degree of abnormality, as different degrees of abnormality may correspond to different clinical significance and processing priorities, helping the subsequent intelligent interpretation model to generate more accurate interpretation results.

[0211] Step S605, input the abnormal indicator information, the detection result data contained in the medical detection sheet, the content of at least one basic information item and the medical use information, and the medical knowledge corresponding to the medical detection sheet into the intelligent interpretation model.

[0212] Among them, the prompt words corresponding to the intelligent interpretation model also contain the slot corresponding to the abnormal indicator information.

[0213] Specifically, the server inputs the abnormal indicator information, the detection result data contained in the medical detection sheet, the content of at least one basic information item and the medical use information, and the medical knowledge corresponding to the medical detection sheet into the intelligent interpretation model. The intelligent interpretation model infers the detection result of the medical detection sheet according to the abnormal indicator information, the detection result data contained in the medical detection sheet, the content of at least one basic information item and the medical use information, and the medical knowledge corresponding to the medical detection sheet.

[0214] In this way, the abnormal indicator information provides specific abnormal data, while the detection result data contained in the medical detection sheet, the content of at least one basic information item and the medical use information provide background and basic information. Combined with medical knowledge, the model can more accurately judge the clinical significance of the abnormality, thereby improving the accuracy of interpretation.

[0215] Optionally, the medical data processing method provided by the present application, the server can also acquire abnormal processing prior knowledge first, wherein the abnormal processing prior knowledge refers to that if one or more abnormal indicators are obtained, if the disease information corresponding to the patient is inferred according to one or more abnormal indicators, for example, the abnormal processing prior knowledge can be: if indicators g and h are abnormal at the same time, the corresponding disease should be disease i.

[0216] After obtaining the abnormality processing prior knowledge, the server inputs the abnormality processing prior knowledge, the detection result data contained in the medical detection sheet, the content of the at least one basic information item and the medical use information, and the medical knowledge corresponding to the medical detection sheet into the intelligent interpretation model together with the medical knowledge, and the intelligent interpretation model performs reasoning according to the abnormality processing prior knowledge, the detection result data contained in the medical detection sheet, the content of the at least one basic information item and the medical use information, and the medical knowledge corresponding to the medical detection sheet and the medical knowledge to obtain an interpretation result of the medical detection sheet. The prompt words corresponding to the intelligent interpretation model further contain the slot corresponding to the abnormality processing prior knowledge.

[0217] In this way, the abnormality processing prior knowledge provides professional insights about the relationship between the abnormal index combination and the specific disease. This helps the model to more accurately infer the potential disease in the interpretation process and improves the accuracy of the interpretation result.

[0218] Figure 7 A flowchart of training of the intelligent interpretation model provided for an exemplary embodiment of the present application is shown in FIG. 7. As shown in FIG. 7, the specific steps of the method are as follows: Figure 7

[0219] Step S701, constructing an intelligent interpretation model.

[0220] Step S702, constructing a training data set, the training data set including: a medical detection sheet sample, text information of the medical detection sheet sample, structured information, and an annotated interpretation result, wherein the structured information includes: detection result data contained in the medical detection sheet sample, content of at least one basic information item, and medical use information.

[0221] Step S703, identifying index entities and detection item entities contained in the medical detection sheet sample, and searching for medical knowledge related to the index entities and the detection item entities in the medical knowledge graph to obtain medical knowledge related to the medical detection sheet sample.

[0222] Step S704, inputting the structured information of the medical detection sheet sample and the medical knowledge related to the medical detection sheet sample into the intelligent interpretation model to generate a predicted interpretation result of the medical detection sheet sample.

[0223] Step S705, adjusting parameters of the intelligent interpretation model according to the predicted interpretation result of the medical detection sheet sample and the annotated interpretation result.

[0224] The medical detection sheet sample can include an image of a medical detection sheet uploaded by a user or text information of a medical detection sheet input by a user.

[0225] ​Specifically, the server first constructs an intelligent interpretation model, which is used to interpret the medical test sheet and obtain an interpretation result. Then, a training data set is constructed, which includes a medical test sheet sample, text information and structured information of the medical test sheet sample, and a labeled interpretation result of the medical test sheet sample. Then, the index entity and the detection item entity contained in the medical test sheet sample are identified. The index entity and the detection item entity can be extracted from the text information of the medical test sheet sample or directly extracted from the medical test sheet sample.

[0226] After the index entity and the detection item entity are extracted, the medical knowledge related to the index entity and the detection item entity can be searched in the medical knowledge graph. Then, the structured information of the medical test sample and the extracted medical knowledge related to the index entity and the detection item entity are input into the intelligent interpretation model. The intelligent interpretation model can generate a predicted interpretation result of the medical test sheet sample. According to the predicted interpretation result of the medical test sheet sample and the labeled interpretation result, the loss value of the intelligent interpretation model can be determined. The parameters of the intelligent interpretation model are adjusted according to the determined loss value.

[0227] In this way, by comparing the labeled interpretation result with the predicted interpretation result, the model can identify the deviation in its interpretation and correct these deviations through parameter adjustment. Moreover, by searching for related knowledge in the medical knowledge graph, the model can utilize rich medical background information for interpretation, thereby improving the accuracy of the interpretation result generated by the trained intelligent interpretation model.

[0228] Optionally, the medical data processing method provided in the present application further comprises: obtaining a bad sample of the interpretation result generated by the intelligent interpretation model, the bad sample including detection result data contained in the medical test sheet, content of at least one basic information item and medical use information, related medical knowledge and interpretation result; correcting the interpretation result of the bad sample to obtain a corrected result of the bad sample; and optimizing the parameters of the intelligent interpretation model according to the bad sample and the corrected result of the bad sample.

[0229] The bad sample refers to a medical test sheet whose interpretation result generated by the intelligent interpretation model is inaccurate or does not meet the expectation. The bad sample includes detection result data contained in the medical test sheet, content of at least one basic information item and medical use information, related medical knowledge and the interpretation result generated by the model. The bad sample can be manually screened or automatically screened by an algorithm or a model. The present application does not limit the way of obtaining the bad sample.

[0230] For each bad sample, analyze the errors in its interpretation result and make corrections. The process of analyzing the errors in the interpretation result of the bad sample and making corrections to obtain the corrected result can be manual, such as a professional (e.g., a doctor or data analyst) manually checking the interpretation result of the bad sample, identifying errors therein, and making corrections according to professional knowledge and experience, which can improve the accuracy and reliability of the corrected result.

[0231] In an optional implementation, an interpretation correction model can be trained based on the LLM, and the errors in the interpretation result of the bad sample are automatically analyzed and corrected by the interpretation correction model to obtain the corrected result.

[0232] In another optional implementation, the process of analyzing the errors in the interpretation result of the bad sample and making corrections to obtain the corrected result can also be to develop and apply specific algorithms to automatically identify and correct errors in the interpretation result, which can be based on rules, statistical models or machine learning techniques, and use historical data and pattern recognition to make corrections, which can improve the processing efficiency of the bad sample.

[0233] After obtaining the corrected result of the bad sample, the server uses the bad sample and its corrected result to adjust and optimize the parameters of the intelligent interpretation model.

[0234] In this way, by identifying and correcting errors in bad samples, the model can better learn and understand the shortcomings in its interpretation, thereby improving overall accuracy, and processing bad samples helps the model to perform more stably in the face of abnormal or complex detection singles, reducing the occurrence of erroneous interpretation. The present embodiment provides a continuous feedback and learning mechanism, enabling the model to continuously adapt to new data and knowledge, and maintaining its performance for continuous improvement.

[0235] Optionally, the medical data processing method provided by the present application further comprises:

[0236] Using the medical detection data set, the intelligent interpretation model is continuously pre-trained and / or instruction fine-tuned, and the medical detection single related medical field knowledge is injected into the intelligent interpretation model, wherein the medical detection data set includes the medical field knowledge of the detection items and the medical field knowledge of the indicators in the medical detection single.

[0237] Specifically, the server first constructs a medical detection data set containing rich medical field knowledge, which includes detailed medical field knowledge of each detection item and indicator in the medical detection single. These knowledge may cover the normal range of indicators, the clinical significance of abnormal conditions, the diagnostic criteria of related diseases, etc.

[0238] After that, the intelligent interpretation model is continuously pre-trained using medical detection data sets. The purpose of pre-training is to enable the model to learn extensive medical knowledge and language patterns through a large amount of medical field knowledge, thereby improving its understanding ability of medical text.

[0239] And on the basis of pre-training, instruction fine-tuning training is performed, that is, by providing instructions and examples of specific tasks, the performance of the model is further optimized to enable it to more accurately perform interpretation tasks. Fine-tuning training can help the model better adapt to the specific interpretation task requirements.

[0240] During pre-training and fine-tuning, the model is injected with relevant medical field knowledge of medical detection forms, so that the intelligent interpretation model not only learns language patterns, but also masters professional knowledge in a specific field, which can improve its professionalism and accuracy in interpreting detection forms.

[0241] In this way, by injecting professional medical field knowledge, the model can more accurately understand and interpret the information in the medical detection form, improving the accuracy of the interpretation result. And continuous pre-training and fine-tuning training enable the model to master deeper medical knowledge, enhancing its professionalism and reliability in the medical field. Through fine-tuning training, the model can better adapt to different types of interpretation tasks, improving its application ability in diversified medical scenarios.

[0242] Figure 8 Another flowchart of a medical data processing method provided for an exemplary embodiment of the present application. The execution subject of this embodiment is the server in the aforementioned system architecture. As shown in the figure, the specific steps of this method are as follows: Figure 8

[0243] Step S801, receiving the image of the medical detection form sent by the terminal side device.

[0244] Step S802, performing OCR recognition on the image to obtain the text information of the medical detection form.

[0245] Step S803, extracting the detection result data contained in the medical detection form, the content of at least one basic information item, and the medical use information from the text information.

[0246] Step S804, recognizing the index entity and detection item entity contained in the medical detection form, searching for medical knowledge related to the index entity and detection item entity, and obtaining the medical knowledge corresponding to the medical detection form.

[0247] Step S805, inputting the detection result data contained in the medical detection form, the content of at least one basic information item, and the medical use information, and the medical knowledge corresponding to the medical detection form into the intelligent interpretation model, and generating an interpretation result of the medical detection form through the intelligent interpretation model. ​

[0248] Step S806, output the interpretation result of the medical test sheet to the end-side device.

[0249] The implementation principle and technical effects of the embodiment are described above, and will not be repeated here.

[0250] Optionally, the medical data processing method provided by the embodiment of the application can also realize the functions of offline log analysis and online stability monitoring.

[0251] The offline log analysis is to perform offline processing and analysis on the log data generated in the running process, to mine potential problems and performance bottlenecks, and to provide a basis for system optimization and function improvement.

[0252] The online stability monitoring is to monitor the running state of the system in real time, including server performance, network status, application response time, etc., to ensure the high availability and stability of the system.

[0253] Figure 9 The system architecture diagram of the medical data processing method provided by an exemplary embodiment of the application is shown in FIG. 1. Figure 9 As shown in FIG. 1, the server obtains the image of the medical test sheet uploaded by the user, performs OCR identification on the image, obtains the OCR identification result of the medical test sheet, takes the OCR identification result as the text information of the medical test sheet, or receives the text information of the medical test sheet input by the user, generates the structured information of the medical test sheet according to the text information, performs named entity recognition on the text information of the medical test sheet, obtains the index entity and the detection item entity contained in the medical test sheet, maps the index entity into the corresponding standard index name, maps the detection item entity into the corresponding standard detection item name, searches for the medical knowledge related to the standard index name and the medical knowledge related to the standard detection item name in the medical knowledge graph, the structured information includes the detection result information, judges whether the detection result of each index is abnormal according to the detection result in the detection result information of each index and the reference value of each index, and determines the abnormal index, for any abnormal index, determines the abnormal degree information of the abnormal index according to the detection result and the reference value of the abnormal index, the abnormal index information includes the abnormal index and the abnormal degree information of the abnormal index, wherein the abnormal degree includes slight abnormality, general abnormality and critical situation, obtains the abnormal handling priori knowledge, and finally inputs the abnormal index information, the structured information, the abnormal handling priori knowledge and the medical knowledge into the intelligent interpretation model, so that the intelligent interpretation model generates the interpretation result of the medical test sheet.

[0254] Table 3 shows the capabilities of the intelligent interpretation model in the embodiment of the present application after training. The terms in Table 3 are explained below: SFT (Supervised Fine-Tuning); IFT (COT) is a training method that combines instruction fine-tuning and thought chain technology. CPT (Continual Pre-Training) can be pre-trained under specific conditions or contexts to better perform specific tasks. IFT (Instruction Fine-Tuning) is a method of fine-tuning a model by providing clear instructions or task descriptions to improve its performance on specific tasks.

[0255] Table 3

[0256]

[0257] Figure 10 This is a structural block diagram of a computing device according to an embodiment of the present application. Figure 10 As shown, the computing device may include: one or more (only one is shown in the figure) processors 1001 and a memory 1002. The memory 1002 is used to store computer programs / instructions, and the processor 1001 is used to execute the computer programs / instructions. When the computer programs / instructions are executed by the processor 1001, the technical solutions provided by any of the aforementioned method embodiments are implemented. The specific functions and technical effects that can be achieved are similar and will not be repeated here.

[0258] The above-mentioned computing device can be understood as an integrated intelligent terminal, including but not limited to a server, a desktop computer, a PC (Personal Computer), an all-in-one model machine, a mobile phone, a tablet computer or other portable intelligent terminal, etc., and the computing device can be pre-installed with the model in the above-mentioned embodiment of the present application.

[0259] Specifically, the computing device can pre-set multiple types of models, including but not limited to models in the fields of natural language processing, visual processing, speech processing, code processing, multimodal task processing, etc., so as to provide a variety of model choices. In different product forms, the computing device can support one or more model usage methods, including but not limited to model training, model calling, model fine-tuning, model deployment, model reasoning and application, etc. In some product forms, the computing device also supports model management, including but not limited to multi-type model management (supporting the management of multiple types of models such as discriminants and genesis), model version control (supporting the control of different model versions), model evaluation (based on model evaluation tools to evaluate the performance and effect of the model), etc. In other product forms, the computing device can also create applications based on the model, provide API (Application Programming Interface, application programming interface) calling capabilities, and can call the model into the created application through the API interface, while providing application management tools to achieve management and monitoring of the application.

[0260] Furthermore, the computing device can also include data management (supporting the creation and management of model tuning data sets), a training center (providing rich training resources to help users learn and master AI technology), and basic management and control capabilities (providing enterprise-level basic management and control capabilities to ensure the security and efficient operation of the system). Through the above functions, a comprehensive, integrated AI development, training, deployment and application device is provided.

[0261] Figure 11 This is a schematic diagram of the structure of a server provided in an embodiment of the present application. Figure 11 As shown, the server includes: a memory 1101 and a processor 1102. Memory 1101 is used to store computer-executable instructions and can be configured to store various other data to support operations on the server. Processor 1102 is communicatively connected to memory 1101 and is used to execute the computer-executable instructions stored in memory 1101 to implement the technical solutions provided by any of the above-mentioned method embodiments. The specific functions and technical effects achieved are similar and will not be further described here.

[0262] Optional, such as Figure 11 As shown, the server also includes: a firewall 1103, a load balancer 1104, a communication component 1105, a power supply component 1106 and other components. Figure 11 Only some components are shown schematically, which does not mean that the server only includes Figure 11 Components shown. Figure 11 The server is only taken as a cloud server deployed in the cloud as an example for exemplary description. The server can also be deployed locally, and this embodiment is not specifically limited here.

[0263] The embodiment of the present application further provides a computer readable storage medium, and the computer readable storage medium stores computer execution instructions. When a processor executes the computer execution instructions, the method of any of the foregoing embodiments is implemented, and the specific functions and the technical effects that can be achieved are not repeated here.

[0264] The embodiment of the present application further provides a computer program product, which comprises a computer program. When the computer program is executed by a processor, the method of any of the foregoing embodiments is implemented. The computer program is stored in a readable storage medium, and at least one processor of a server can read the computer program from the readable storage medium. The at least one processor executes the computer program, so that the server executes the technical solutions provided by any of the method embodiments, and the specific functions and the technical effects that can be achieved are not repeated here.

[0265] The embodiment of the present application provides a chip, which comprises a processing module and a communication interface. The processing module can execute the technical solutions of the server in the foregoing method embodiments. Optionally, the chip further comprises a storage module (for example, a memory). The storage module is used for storing instructions. The processing module is used for executing the instructions stored in the storage module. The execution of the instructions stored in the storage module causes the processing module to execute the technical solutions provided by any of the foregoing method embodiments.

[0266] The integrated modules in the form of software function modules described above can be stored in a computer readable storage medium. The software function modules described above are stored in a storage medium, and comprise a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to execute part of the steps of the method of each embodiment of the present application.

[0267] It should be understood that the processor described above can be a central processing unit (CPU), a graphics processing unit (GPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the application can be directly embodied as execution completed by a hardware processor, or executed by a combination of hardware and software modules in at least one processor.

[0268] The memory can include a high-speed random access memory (RAM) and can also include a non-volatile storage, such as at least one disk memory, and can also be a U disk, a mobile hard disk, a read-only memory, a magnetic disk or an optical disk, etc.

[0269] The memory described above can be an object storage (OSS).

[0270] The memory described above can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0271] The communication component is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as a mobile hotspot (WiFi), a second generation mobile communication system (2G), a third generation mobile communication system (3G), a fourth generation mobile communication system (4G) / long term evolution (LTE), a fifth generation mobile communication system (5G), or other mobile communication networks, or a combination thereof. In an example embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an example embodiment, the communication component further includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared technology, ultra wide band (UWB) technology, Bluetooth technology, and other technologies.

[0272] The power component provides power to various components of the device in which the power component is located. The power component can include a power management system, one or more power sources, and other components associated with generating, managing, and distributing power to the device in which the power component is located.

[0273] The storage medium can be implemented by any type of volatile or non-volatile storage devices or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read only memory (EPROM), programmable read only memory (PROM), read only memory (ROM), magnetic storage, flash memory, magnetic or optical disk. The storage medium can be any available medium that can be accessed by a general or special purpose computer.

[0274] An exemplary storage medium is coupled to the processor so that the processor can read information from, and write information to, the storage medium. Of course, the storage medium can be a part of the processor. Consistent with the teachings provided herein, the processor and the storage medium can be located in a special purpose computing device. The processor and the storage medium can also be located in a general purpose computing device.

[0275] It is to be appreciated that the term "include," "comprise," or variations thereof, as used in this document, is intended to cover the presence of one or more elements, components, steps, actions, etc. but does not exclude the presence of one or more other elements, components, steps, actions, etc. Unless otherwise indicated, the use of the term "or" in this document is used to mean "and / or," that is, the term "or" is not exclusive, unless otherwise indicated. In addition, it is to be appreciated that the use of any form of "present" or "comprise" herein, such as "comprises", "comprising", "includes", "including", "contains", "containing" or variants thereof, is intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises, includes, contains elements or steps do not include only those elements or steps but can also include other elements or steps not expressly listed or inherent to such process, method, article, or apparatus.

[0276] The sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. In addition, in some of the processes described in the above embodiments and the accompanying drawings, a plurality of operations appear in a specific order, but it should be clearly understood that these operations can be executed in the order they appear in this document or in parallel, and only to distinguish different operations. The sequence itself does not represent any execution order. In addition, these processes can include more or fewer operations, and the operations can be executed in sequence or in parallel. It should be noted that the "first", "second", etc. described herein are used to distinguish different messages, devices, modules, etc. and do not represent the order. "First" and "second" are different types. The meaning of "multiple" is more than two, unless otherwise explicitly specified.

[0277] Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and the necessary general hardware platform, of course, can also be through hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of software products, and the computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a plurality of instructions to make a terminal device (may be a mobile phone, computer, server, air conditioner, or network equipment, etc.) execute the method described in various embodiments of the present application.

[0278] Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the application disclosed herein. It is intended that the present application cover any and all variations of the application that come within the scope of the present application along with its general principles and features. This application is intended to cover any adaptations or variations of the application.

[0279] The above is only the preferred embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation using the content of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A medical data processing method, characterized in that: include: Get the text information of the medical examination form; Extracting the test result data, the content of at least one basic information item, and the medical use information contained in the medical test form from the text information; Identify the indicator entity and the test item entity contained in the medical test form, search for medical knowledge related to the indicator entity and the test item entity, and obtain the medical knowledge corresponding to the medical test form; The test result data, the content of at least one basic information item and the medical use information contained in the medical test form, and the medical knowledge corresponding to the medical test form are input into the intelligent interpretation model, and the interpretation result of the medical test form is generated by the intelligent interpretation model.

2. The method according to claim 1, characterized in that The identification of the indicator entity and the test item entity contained in the medical test form includes: Performing named entity recognition on the text information of the medical examination form to obtain the indicator entities and test item entities contained in the medical examination form; and / or, The text information of the medical test form is segmented to obtain multiple word units, and the multiple word units are matched with the indicators and test items contained in the medical vocabulary library. The word unit that matches any indicator in the medical vocabulary library is used as the indicator entity, and the word unit that matches any test item in the medical vocabulary library is used as the test item entity.

3. The method according to claim 1, characterized in that The searching for medical knowledge related to the indicator entity and the detection item entity includes: Map the indicator entity to the corresponding standard indicator name, and map the test item entity to the corresponding standard test item name; In the medical knowledge graph, medical knowledge related to the standard indicator name and medical knowledge related to the standard test item name are searched.

4. The method according to claim 1, wherein The extracting the test result data, the content of at least one basic information item, and the medical use information contained in the medical test form from the text information includes: Performing text classification and recognition on the text information of the medical examination form to determine the examination form type of the medical examination form; The test order type of the medical test order and the text information are input into an information structuring model, and the structured information of the medical test order is extracted from the text information according to the test order type through the information structuring model, wherein the structured information includes test order information and test result data, the test order information includes at least one basic information item of the medical test order and the content and medical use information of each basic information item, and the test result data includes the test result information of at least one indicator.

5. The method according to claim 1, wherein The test result data includes test result information of at least one indicator, and the method further includes: Determining abnormal indicator information based on the test result data; The abnormal indicator information, the test result data contained in the medical test form, the content of at least one basic information item and medical use information, and the medical knowledge corresponding to the medical test form are input into the intelligent interpretation model.

6. The method according to claim 5, characterized in that Determining abnormal indicator information based on the detection result data includes: Determine whether the test results of each indicator are abnormal based on the test results in the test result information of each indicator and the reference value of each indicator, and determine the abnormal indicator; For any of the abnormal indicators, abnormality degree information of the abnormal indicator is determined according to the detection result and reference value of the abnormal indicator, and the abnormal indicator information includes the abnormal indicator and the abnormality degree information of the abnormal indicator.

7. The method according to any one of claims 1 to 6, characterized in that Also includes: Acquiring prior knowledge of abnormality processing, wherein the prior knowledge of abnormality processing includes disease inference result information corresponding to at least one abnormal indicator combination; The exception handling prior knowledge, the test result data contained in the medical test form, the content of at least one basic information item and medical use information, and the medical knowledge corresponding to the medical test form are input into the intelligent interpretation model together.

8. The method according to any one of claims 1 to 6, characterized in that The text information of the medical examination form is obtained, including: Obtaining the uploaded image of the medical test form, performing OCR recognition on the image, obtaining an OCR recognition result of the medical test form, and using the OCR recognition result as text information of the medical test form; or, Receive the input text information of the medical examination form.

9. The method according to any one of claims 1 to 6, characterized in that Also includes: Obtaining a bad sample of the interpretation result generated by the intelligent interpretation model, the bad sample including the test result data contained in the medical test form, the content of at least one basic information item and medical use information, relevant medical knowledge and the interpretation result; Correcting the interpretation result of the bad sample to obtain a corrected result of the bad sample; Optimizing parameters of the intelligent interpretation model according to the bad sample and the correction result of the bad sample.

10. The method according to any one of claims 1 to 6, characterized in that Also includes: Build intelligent interpretation models; Constructing a training data set, the training data set comprising: a medical test form sample, text information, structured information, and annotated interpretation results of the medical test form sample, wherein the structured information comprises: test result data contained in the medical test form sample, content of at least one basic information item, and medical use information; Identify the indicator entity and the test item entity contained in the medical test order sample, and search the medical knowledge related to the indicator entity and the test item entity in the medical knowledge graph to obtain the medical knowledge related to the medical test order sample; Inputting the structured information of the medical test sample and the medical knowledge related to the medical test sample into the intelligent interpretation model to generate a predicted interpretation result for the medical test sample; Adjust the parameters of the intelligent interpretation model according to the predicted interpretation results and the annotated interpretation results of the medical test sample.

11. The method according to claim 10, characterized in that Also includes: Using a medical test data set, the intelligent interpretation model is continuously pre-trained and / or instruction fine-tuned, and relevant medical field knowledge of the medical test order is injected into the intelligent interpretation model, wherein the medical test data set includes the medical field knowledge of the test items and the medical field knowledge of the indicators in the medical test order.

12. A medical data processing method, characterized in that: include: An image of the medical examination form sent by the receiving device; Performing OCR recognition on the image to obtain text information of the medical examination form; Extracting the test result data, the content of at least one basic information item, and the medical use information contained in the medical test form from the text information; Identify the indicator entity and the test item entity contained in the medical test form, search for medical knowledge related to the indicator entity and the test item entity, and obtain the medical knowledge corresponding to the medical test form; Inputting the test result data, the content of at least one basic information item and the medical use information contained in the medical test form, and the medical knowledge corresponding to the medical test form into an intelligent interpretation model, and generating an interpretation result of the medical test form through the intelligent interpretation model; Output the interpretation result of the medical examination form to the terminal side device.

13. A computing device, characterized in that include: memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer programs / instructions are executed by the processor, the method according to any one of claims 1 to 12 is implemented.

14. A server, characterized in that: include: at least one processor; as well as a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the server to execute the method according to any one of claims 1 to 12.

15. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-executable instructions, and when a processor executes the computer-executable instructions, the method according to any one of claims 1 to 12 is implemented.

16. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 12 is implemented.

Images