Medical data processing method and device, computer program product and electronic equipment

Abstract

The embodiment of the invention relates to a medical data processing method and device, a computer program product and electronic equipment, and relates to the technical field of computers, and the medical data processing method comprises the following steps: obtaining medical data of a plurality of target objects; and according to various description information in the medical data, performing relation pair extraction on the medical data to obtain a plurality of medical data relation pairs with a causal relationship, so as to train a language model according to the medical data relation pairs. According to the technical scheme, the accuracy of the data relation pair can be improved, and the model training effect is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a medical data processing method, a medical data processing device, a computer program product, and an electronic device. Background Art

[0002] Large language models enable intelligent predictions in a variety of application scenarios. To improve accuracy, language models can be trained using electronic medical data to enhance their effectiveness in medical applications.

[0003] Related technologies typically streamline medical record models and use basic or important data from electronic medical records as the training corpus for the larger model. Alternatively, all information in the electronic medical record is directly used, organized according to the data of a single medical visit, and then the cleaned data is used directly as training data for the larger model.

[0004] In the above approach, information in the simplified medical record model may be lost and may contain certain errors, resulting in a less professional language model after training as corpus. Using the full medical record model, all medical record data is organized together, resulting in less accurate and stable language models.

[0005] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the Invention

[0006] The purpose of the present disclosure is to provide a medical data processing method and device, a computer program product, and an electronic device, thereby overcoming, at least to a certain extent, the problems of inaccurate data and low model accuracy caused by the limitations and defects of related technologies.

[0007] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by practice of the present disclosure.

[0008] According to a first aspect of the present disclosure, a medical data processing method is provided, comprising: acquiring medical data of multiple target objects; extracting relationship pairs from the medical data based on multiple descriptive information in the medical data to obtain multiple medical data relationship pairs with causal relationships, so as to train a language model based on the medical data relationship pairs.

[0009] In an exemplary embodiment of the present disclosure, the medical data is subjected to relationship pair extraction based on multiple descriptive information in the medical data to obtain multiple medical data relationship pairs with causal relationships, including: when the type of the descriptive information is physical examination information, data is composed according to the physical examination information and the physical examination conclusion in the medical data to obtain medical data relationship pairs related to the physical examination information; when the type of the descriptive information does not belong to physical examination information, the medical data is subjected to relationship pair extraction based on each descriptive information to obtain an extraction result, and the relationship pair extraction is continued in combination with the extraction result to generate medical data relationship pairs related to each descriptive information.

[0010] In an exemplary embodiment of the present disclosure, the descriptive information includes basic information; the medical data is subjected to relationship pair extraction based on each descriptive information to obtain an extraction result, and the relationship pair extraction is continued in combination with the extraction result to generate a medical data relationship pair related to each descriptive information, including: using the basic information as an independent variable, determining a first relationship pair between the basic information and the inspection and examination instruction; updating the independent variable based on the result of the inspection and examination instruction in the first relationship pair, continuing to extract relationship pairs based on the updated independent variable, and determining a second relationship pair between the updated independent variable and the diagnosis result; continuing to update the independent variable based on the diagnosis result in the second relationship pair, and determining a third relationship pair between the updated independent variable and the treatment method, so as to determine multiple medical data relationship pairs related to the basic information.

[0011] In an exemplary embodiment of the present disclosure, the descriptive information includes case characteristics; the relationship pair extraction of medical data is performed according to each descriptive information to obtain an extraction result, and the relationship pair extraction is continued in combination with the extraction result to generate medical data relationship pairs related to each descriptive information, including: using case characteristics as independent variables, determining a fourth relationship pair between the case characteristics and the preliminary diagnosis result; using the case characteristics and the fourth relationship pair as independent variables, determining the medical data relationship pairs between multiple diagnosis and treatment related information corresponding to the preliminary diagnosis result; wherein, the diagnosis and treatment related information includes at least one of a diagnosis basis, a diagnosis and treatment plan, and a differential diagnosis.

[0012] In an exemplary embodiment of the present disclosure, the descriptive information includes medical course information; the medical data is subjected to relationship pair extraction based on each descriptive information to obtain an extraction result, and the relationship pair extraction is continued in combination with the extraction result to generate a medical data relationship pair related to each descriptive information, including: using the medical course information as an independent variable to determine a fifth relationship pair between the medical course information and the test and examination instructions; using the medical course information and the results of the test and examination instructions in the fifth relationship pair as independent variables to determine the medical data relationship pair with the diagnosis and treatment plan.

[0013] In an exemplary embodiment of the present disclosure, the medical course information is used as an independent variable to determine the fifth relationship pair between the medical course information and the test and examination instructions, including: grouping the medical course information according to the recording date to obtain multiple medical course groups, and grouping the test and examination instructions according to the examination issuance date to obtain multiple test and examination groups; determining the recording time according to the recording date and the current date, and determining the examination issuance time according to the examination issuance date and the current date; for each test and examination group, combining the medical course information whose recording time is greater than the examination issuance time of the test and examination instruction with the test and examination instruction of the current date to obtain the fifth relationship pair.

[0014] In an exemplary embodiment of the present disclosure, the medical course information and the results of the inspection and examination instructions in the fifth relationship pair are used as independent variables to determine the medical data relationship pair between the medical course information and the treatment method, including: merging the results of the inspection and examination instructions in the fifth relationship pair into the medical course information to obtain updated medical course information; grouping the updated medical course information according to the recording date to obtain multiple medical course groups, and grouping the diagnosis and treatment plans according to the treatment start date to obtain multiple treatment method groups; determining the recording time according to the recording date and the current date, and determining the treatment start time according to the treatment start date and the current date; for each treatment method group, combining the updated medical course information whose recording time is greater than the treatment start time with the treatment method of the current date to determine the medical data relationship pair.

[0015] In an exemplary embodiment of the present disclosure, the method further includes: determining the priorities of multiple cleaning methods; screening the medical data relationship pairs according to the priorities to obtain cleaned medical data relationship pairs; wherein the multiple cleaning methods include data cleaning, data scoring, data deduplication and data desensitization.

[0016] According to a second aspect of the present disclosure, a medical data processing device is provided, comprising: a data acquisition module for acquiring medical data of multiple target objects; a relationship pair generation module for extracting relationship pairs from the medical data based on multiple descriptive information in the medical data to obtain multiple medical data relationship pairs with causal relationships, wherein the medical data relationship pairs are used to train a language model.

[0017] According to a third aspect of the present disclosure, a computer program product is provided, on which a computer program is stored. When the computer program is executed by a processor, the medical data processing method described above is implemented.

[0018] According to a fourth aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute any one of the above-mentioned medical data processing methods by executing the executable instructions.

[0019] The medical data processing method, medical data processing device, computer program product, and electronic device provided in the embodiments of the present disclosure can, on the one hand, extract relationship pairs from the medical data based on the various descriptive information in the medical data, thereby obtaining medical data relationship pairs with causal relationships. Since the medical data relationship pairs with causal relationships can be obtained, the problem of language model error prediction caused by the confusion of causal relationships in related technologies is avoided, and the accuracy and stability of the language model are improved, and the rationality and interpretability are improved. On the other hand, the medical data can be extracted from the dimensions of multiple descriptive information based on the various descriptive information in the medical data, thereby obtaining medical data relationship pairs with causal relationships, thereby avoiding the problem of structural confusion caused by the data containing unstructured data and structured data, increasing the comprehensiveness, completeness, and accuracy of the data, and enhancing the professionalism of the language model.

[0020] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0022] Figure 1 A schematic diagram schematically illustrates a medical data processing method in an embodiment of the present disclosure.

[0023] Figure 2 The following schematically illustrates a flow chart of determining medical data relationship pairs in an embodiment of the present disclosure.

[0024] Figure 3 A schematic diagram schematically illustrates the relationship between disease course information and test and examination instructions in an embodiment of the present disclosure.

[0025] Figure 4 A schematic diagram schematically illustrates the relationship between disease course information and a diagnosis and treatment plan in an embodiment of the present disclosure.

[0026] Figure 5The following schematically illustrates a method for generating medical data relationship pairs in an embodiment of the present disclosure.

[0027] Figure 6 A block diagram of a medical data processing device in an embodiment of the present disclosure is schematically shown.

[0028] Figure 7 A block diagram schematically illustrates an electronic device in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0029] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the present disclosure will be more comprehensive and complete and will fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or that other methods, components, devices, steps, etc. may be employed. In other cases, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0030] In addition, the accompanying drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0031] Training language models using electronic medical data can be achieved through the following methods: First, a streamlined medical record model uses essential and important data from electronic medical records. This data is cleansed and used as the training corpus for the larger model. For example, for hospitalized and discharged patients, only the presenting medical history and the diagnosis and treatment process description are used. The presenting medical history describes the chief complaint, symptom development, examinations, treatment, and disease progression. The diagnosis and treatment process description describes the examinations, symptom development, treatment, and disease progression during the hospitalization period. Together, they form a relatively complete closed-loop process for the patient's disease: symptoms => examinations => diagnosis => treatment. For hospitalized and non-discharged patients, case characteristics are used, which contain unique attributes regarding clinical presentation, disease progression, diagnosis, and treatment response. These two sets of data, after simple cleaning, serve as the training corpus for the larger model.

[0032] The second full medical record mode directly uses all the information in the electronic medical record, including the chief complaint, current medical history, personal history, marital history, past history, family history, physical examination, auxiliary examination, specialist examination, examination, test, medical history, surgery, disease, course of disease, medication, etc. These information are organized according to the data of a single medical visit, and after simple cleaning, they are directly used as training corpus for the large model.

[0033] The streamlined medical record model primarily uses the unstructured information in the medical record: the current medical history, the description of the treatment process, and the case characteristics. Information may be lost. The current medical history only provides a description and summary of the patient's past condition. For patients seeking treatment outside the hospital, most information is verbalized by the patient, which carries a certain probability of error. There is also a risk of misinterpretation of test results. The third treatment process description briefly describes the patient's treatment process, and the treatment methods or medications used may be relatively concise. For example, the treatment process description often mentions maintenance treatment, but without the doctor's instructions, the specific medications used for maintenance treatment and their dosages are unknown. Therefore, the language model obtained after training with the corpus may not reflect professionalism or personalization. The full medical record model organizes all medical record data, including both unstructured and structured data. Due to the way medical records are written, this full medical record model can lead to confusion in causal relationships. For example, a breast cancer patient with hypertension is prescribed both breast treatment drugs and antihypertensive drugs during hospitalization. The large model trained with such corpus can include antihypertensive drugs in consulting on what drugs should be used to treat breast cancer.

[0034] In order to solve the technical problems in the related technology, a medical data processing method is provided in the embodiment of the present disclosure, which can be executed by a server or a client to process medical data, generate data for training a language model, and then perform corresponding operations through the language model.

[0035] Next, refer to Figure 1 The figure shows a detailed description of the medical data processing method in the embodiment of the present disclosure.

[0036] In step S110 , medical data of a plurality of target objects are acquired.

[0037] In the disclosed embodiments, the multiple target objects may be multiple patients, who may be patients suffering from any type of disease. The medical data may be all of a patient's medical records at one or more medical institutions, such as outpatient medical data or inpatient medical data. The medical data may be medical records stored in any medical institution's internal database, or publicly available in other platform databases. The internal database of a medical institution must be authorized by the medical institution, and the data in the database must meet reasonableness and legality requirements.

[0038] Medical data may contain multiple descriptive information. Descriptive information is used to describe the basic situation of the disease, such as the symptoms of the disease, the course of the disease, etc. In addition, medical data may also contain diagnosis and treatment information, which is used to indicate the test and inspection instructions that need to be executed, the name of the disease, and the diagnosis and treatment plan, treatment information, such as diabetes, medication, etc. For example, disease data may include chief complaint, current medical history, personal history, marital history, past history, family history, physical examination, auxiliary examination, specialist examination, examination, test, medical history, surgery, disease, course of disease, medicine and other information. This disclosure does not specifically limit the specific information contained in the disease data, which can be determined based on the actual medical records.

[0039] In step S120, relationship pairs are extracted from the medical data based on various descriptive information in the medical data to obtain a plurality of medical data relationship pairs with causal relationships, so as to train a language model based on the medical data relationship pairs.

[0040] In the disclosed embodiment, the descriptive information in the medical data may include any one or more of basic information, case characteristics, course of disease information, and physical examination information. Basic information is used to understand the basic situation of the patient. Basic information may include chief complaint, current medical history, past medical history, personal history, family history, marital history, auxiliary examinations done in the past, physical examinations, and specialist examinations. Case characteristics refer to the analysis of information with diagnostic and differential significance in the basic information. That is, the analysis of evidence with diagnostic and differential significance in the chief complaint, current medical history, personal history, physical examination, and auxiliary examinations. Course of disease information refers to information used to record changes in the patient's condition over time. Physical examination information can be a physical imaging examination prescribed for the patient's medical record data.

[0041] Relationship pair extraction refers to the extraction of data pairs from medical data according to causal relationships, thereby obtaining multiple medical data relationship pairs that conform to causal relationships.

[0042] In some embodiments, depending on the type of descriptive information in the medical data, a round of relationship pair extraction can be performed based on the descriptive information to obtain multiple medical data relationship pairs related to the descriptive information, or a round of relationship pair extraction can be performed based on the descriptive information, and relationship pair extraction can be continued based on the extraction results to obtain multiple medical data relationship pairs related to each descriptive information.

[0043] Figure 2 The flowchart for determining the medical data relationship pair is schematically shown in FIG. Figure 2 As shown in , it mainly includes the following steps:

[0044] In step S210 , it is determined whether the type of the description information is physical inspection information; if not, the process proceeds to step S220 ; if so, the process proceeds to step S230 .

[0045] In step S220, if the type of descriptive information does not correspond to physical examination information, a pair extraction is performed on the medical data based on each descriptive information to obtain an extraction result. The extraction results are then combined to further extract multiple medical data pairs. The descriptive information may include one or more of basic information, case characteristics, and course of disease information.

[0046] In step S230, when the type of the description information belongs to physical examination information, data is composed according to the physical examination information and the corresponding physical examination conclusion to obtain a medical data relationship pair.

[0047] Next, the specific process of generating medical data relationship pairs is explained.

[0048] First, if the descriptive information includes basic information, the basic information can be used as the independent variable to determine the first relationship pair between the basic information and the examination and inspection instructions, and this first relationship pair is used as the extraction result. Basic information can include consultation information and specialist examinations. Consultation information may include: a complaint and five histories, a physical examination, and auxiliary examinations. Examination and inspection instructions refer to medical examination and inspection instructions issued by a doctor based on the basic information, such as test orders, examination orders, etc. Based on this, medical data relationship pairs can be constructed between the basic information and the test and inspection orders.

[0049] For example, the basic information mentioned by the patient during the consultation process can be obtained, such as the chief complaint, current medical history, past medical history, personal history, family history, marital history, physical examination, auxiliary examination (examinations done), etc.; then the specialist examinations performed by the doctor on the patient can be obtained, such as specialist examinations for ophthalmic diseases. Further, the doctor's test and examination orders issued based on the corresponding consultation information and specialist examinations can be obtained. Based on this, the one complaint and five histories, auxiliary examinations, specialist examinations, test orders, and examination orders in the medical record data can be obtained to form a complete logical relationship pair as the first relationship pair to determine the extraction results.

[0050] After the first round of relationship pair extraction is performed to obtain the extraction results, the independent variables can be iteratively updated based on the results of the previous round of extraction, that is, the independent variables of the current round of relationship pair extraction process are all updated based on the results of the previous round of extraction. Based on this, the basic information and the extraction results can be used as independent variables, and the relationship pair extraction can be iteratively performed in combination with the extraction results, and the iterative cycle can be repeated to determine the basic information and the diagnosis result or treatment method. Specifically, the basic information and the results of the inspection and examination instructions in the first relationship pair can be used as the updated independent variables, and the relationship pair extraction can be continued to determine the second relationship pair between the updated independent variable and the diagnosis result. Furthermore, the basic information and the results of the inspection and examination instructions in the first relationship pair and the diagnosis result in the second relationship pair can be used as the updated independent variables, and the relationship pair extraction can be continued to determine the third relationship pair between the updated independent variable and the treatment method, thereby determining multiple medical data relationship pairs related to the basic information.

[0051] For example, the type of disease determined by the doctor based on the patient's basic information and the results of the test instructions can be obtained to determine the patient's diagnosis result, forming a medical data relationship pair between the basic information, the test results and the diagnosis result.

[0052] Furthermore, the doctor can determine the diagnosis based on the patient's medical history and test results, and then prescribe a treatment based on the disease type and corresponding symptoms. Treatments can include medication orders and treatment orders. Medication orders include information such as the drug name, usage, and dosage; treatment orders include corresponding treatment orders, such as suture removal orders for patients who need suture removal after surgery. Based on this, a causal relationship can be established, which determines the treatment based on basic information, specialist examinations, auxiliary examinations, test results, and diagnosis results.

[0053] When the descriptive information includes case characteristics, the case characteristics can be used as independent variables to extract relationship pairs to obtain extraction results, and the independent variables can be further updated based on the extraction results to determine the various diagnosis and treatment related information corresponding to the extraction results according to the updated independent variables, and to construct medical data relationship pairs between the case characteristics and the various diagnosis and treatment related information. The diagnosis and treatment related information may include relevant information used to explain, process or identify the preliminary diagnosis results, and specifically may include any one or more of the diagnostic basis, diagnosis and treatment plan, and differential diagnosis. Specifically, the case characteristics can be used as independent variables to determine the fourth relationship pair between the case characteristics and the preliminary diagnosis results as the extraction result; further, the case characteristics and the preliminary diagnosis results in the fourth relationship pair are used as independent variables to determine the medical data relationship pairs between the various diagnosis and treatment related information corresponding to the preliminary diagnosis results.

[0054] For example, the preliminary diagnosis refers to the disease diagnosis information obtained based on the patient's basic information, such as the disease type or name. For hospitalized patients, the patient's age, gender, chief complaint, current medical history, personal history, past history, family history, physical examination, specialist examination, auxiliary examination and other related information can be obtained to preliminarily obtain the patient's disease and obtain a preliminary diagnosis. Case characteristics are a line-by-line analysis of evidence with diagnostic and differential significance in the chief complaint, current medical history, personal history, physical examination, auxiliary examination, etc. Based on this, the preliminary diagnosis can be inferred based on the case characteristics, and a self-contained causal relationship can be formed based on the case characteristics and the preliminary diagnosis.

[0055] Furthermore, the diagnostic basis for the disease diagnosis formed when the doctor obtains a preliminary diagnosis based on the patient's case characteristics can be obtained. Diagnostic basis is a set of standards or evidence used in medicine to confirm a disease or health condition, typically based on the patient's clinical manifestations, medical examinations, and laboratory test results. Specifically, the diagnostic basis generally includes clinical symptoms and signs, medical imaging examinations, laboratory tests, special diagnostic procedures, medical history and epidemiological information, biomarkers, genetic testing, etc. Based on this, a self-contained causal relationship pair can be formed between the case characteristics, the preliminary diagnosis result, and the diagnostic basis, namely the medical data relationship pair.

[0056] In addition, the doctor can obtain the diagnosis and treatment plan formulated based on the patient's case characteristics and preliminary diagnosis results. The diagnosis and treatment plan may include plans related to testing, examination, nursing, treatment, etc. Based on this, a causal relationship can be established between the case characteristics, preliminary diagnosis results, and diagnosis and treatment plan.

[0057] Similarly, a physician can obtain a differential diagnosis based on the patient's case characteristics and initial diagnosis results. A differential diagnosis is used to identify and distinguish between multiple diseases or causes that could explain a patient's symptoms or signs. Specifically, after considering the patient's clinical presentation, all possible diagnoses are listed and eliminated one by one through further medical examinations, laboratory tests, or imaging studies, ultimately identifying the most likely disease. This allows for the formation of a causal relationship between medical data pairs between case characteristics, initial diagnosis results, and differential diagnosis.

[0058] When the descriptive information includes medical course information, a medical data relationship pair between the medical course information, the test and examination instructions, and the treatment method can be obtained. Specifically, using the medical course information as the independent variable, a fifth relationship pair between the medical course information and the test and examination instructions is determined. Furthermore, using the medical course information and the test and examination instruction results in the fifth relationship pair as independent variables, a medical data relationship pair between the medical course information and the diagnosis and treatment plan is determined.

[0059] In some embodiments, medical history information is used to record the progression of a patient's condition during hospitalization, and may include initial medical history records, daily medical history records, ward rounds medical history records, postoperative medical history records, and medical treatment records. When a doctor reviews medical history information and identifies corresponding symptoms, they will issue corresponding laboratory test orders. Therefore, medical history information and laboratory test orders form a causal relationship. However, the number of medical history information and laboratory test orders during hospitalization is relatively large, so the medical data relationship between the two can be determined by date.

[0060] Figure 3 The schematic diagram of determining the relationship between the course of disease information and the test and examination instructions is shown schematically in FIG. Figure 3 As shown in , medical history information can be grouped by recording date to obtain multiple medical history groups, such as medical history group 1, medical history group 2, and medical history group 3. Test and examination instructions can also be grouped by examination issuance date to obtain multiple test and examination groups, such as test and examination group 1, test and examination group 2, and test and examination group 3. Based on this, the recording time can be determined by the difference between the recording date and the current date, and the examination issuance time can be determined by the difference between the examination issuance date and the current date.

[0061] For each test group, combine the medical records with a recording time greater than the test order's issuance time with the test order for the current date to obtain a fifth relationship pair. For example, for test group 1, if the recording time for both medical records in medical records group 1 and medical records group 2 is greater than the test order's issuance time, combine the medical records corresponding to medical records group 1 and medical records group 2 with the test order corresponding to medical records group 1 to obtain a causal relationship pair, i.e., the fifth relationship pair.

[0062] It should be noted that, based on the above method, each test and examination group can be processed to obtain the fifth relationship pair between the course of disease information and the test and examination instructions as multiple medical data relationship pairs.

[0063] Furthermore, the medical data relationship pair between the course of disease information and the execution result of the fifth relationship pair is used as an independent variable to determine the medical data relationship pair between the treatment method. For example, the results of the inspection and examination instructions in the fifth relationship pair are merged into the course of disease information to obtain updated course of disease information; the updated course of disease information is grouped according to the recording date to obtain course of disease groups, and the treatment methods are grouped according to the treatment issuance date to obtain treatment method groups. Based on this, the recording time can be determined by the difference between the recording date and the current date, and the treatment issuance time can be determined by the difference between the treatment issuance date and the current date. The treatment issuance date refers to the date on which a treatment instruction is issued.

[0064] For each treatment method group, the updated course of disease information whose recording time is greater than the treatment prescription time is combined with the treatment method of the treatment method group to determine the medical data relationship pair.

[0065] Figure 4 A schematic diagram of determining the relationship between disease course information and treatment methods is shown in FIG. Figure 4 As shown in , the results of the test instructions in the fifth relationship can first be merged into the medical course information, that is, the test instruction results can be recorded in the medical course information to obtain updated medical course information. The updated medical course information can be grouped according to the recording date to obtain multiple medical course groups, such as medical course group 1, medical course group 2, and medical course group 3. The treatment methods can also be grouped according to the treatment prescription date to obtain multiple treatment method groups, such as treatment method group 1, treatment method group 2, and treatment method group 3.

[0066] For each treatment group, the updated medical course information with a recording time greater than the treatment prescription time is combined with the treatment method of the current date to determine a medical data relationship pair. For example, for treatment group 2, if the recording time of medical course groups 1 and 2 is greater than the treatment prescription time, the medical course information corresponding to medical course groups 1 and 2 can be combined with the treatment method corresponding to treatment group 2 to obtain a medical data relationship pair with a causal relationship.

[0067] If the descriptive information type is physical examination information, data is assembled based on the physical examination information and the physical examination conclusions in the medical data to form a medical data relationship pair. Physical examination information refers to the examination findings. For example, physical examination information can be imaging examinations. The imaging department can interpret the corresponding physical examination conclusions based on the physical imaging. Examination findings in medical data such as electronic medical records and physical examination conclusions have a causal relationship, forming a medical data relationship pair.

[0068] refer to Figure 5 As shown in , based on the above steps, a variety of medical data relationship pairs with causal relationships can be generated.

[0069] In the disclosed embodiments, to improve the accuracy of medical data relationship pairs, the medical data relationship pairs may be filtered after they are formed. Alternatively, the medical data may be filtered before they are formed, and relationship pairs may be extracted based on the filtered medical data. This description will take the filtering of medical data relationship pairs as an example.

[0070] In some embodiments, after obtaining multiple medical data relationship pairs, the medical data relationship pairs may be screened to obtain target medical data relationship pairs in order to improve the accuracy of the data pairs. The medical data relationship pairs may be screened based on various cleaning methods. These cleaning methods include data cleaning, data scoring, data deduplication, and data desensitization. Other cleaning methods may also be included and are not specifically limited herein.

[0071] Based on this, we can first determine the priorities of various cleaning methods. Then, we can filter the medical data relationship pairs in descending order of priority to obtain the target medical data relationship pairs. Data cleaning takes precedence over data deduplication, which takes precedence over data scoring, which takes precedence over data desensitization.

[0072] On the basis of determining the priorities of multiple cleaning methods, the medical data relationship pairs can be cleaned first, and then the data after cleaning can be deduplicated. The medical data relationship pairs after deduplication can be scored to obtain scores, and the medical data relationship pairs whose scores meet the score conditions can be desensitized to obtain the target medical data relationship pairs.

[0073] In some embodiments, data cleaning can be achieved through data filtering and data modification. For example, data filtering can be performed on medical data relationship pairs based on the content, content type, and content format contained in the medical data. Specifically, data cleaning can be performed based on whether the medical data contains a chief complaint, whether the chief complaint type is a preset type, whether the chief complaint contains a duration, whether the length of other information other than the chief complaint is less than a threshold, and whether the physical examination conclusion is a preset conclusion. For example, medical data relationship pairs without a chief complaint are discarded, medical data relationship pairs where the chief complaint does not contain a duration are discarded, medical data relationship pairs where the chief complaint is a preset type such as infectious pneumonia, consultation, physical examination, and physical examination are discarded, medical data relationship pairs where the length of current medical history, family history, past history, physical examination, etc. is less than a threshold are discarded, and medical data relationship pairs where the physical examination conclusion is a preset conclusion such as the same as above, reference, etc. are discarded. In addition, telephone numbers in the medical data can be cleaned, images in the medical data can be stored in base64 format to achieve data modification, and meaningless data that appears frequently in the medical data can be replaced.

[0074] Next, the cleaned medical data relationship pairs can be deduplicated. Specifically, deduplication can be performed using a complete match method, and the deduplication results can be further deduplicated using a fuzzy deduplication method to obtain deduplicated medical data relationship pairs. Specifically, if multiple medical data relationship pairs are completely matched, these multiple medical data relationship pairs are filtered out. Furthermore, the minhash method can be used to calculate the similarity between multiple medical data relationship pairs, and the data with a similarity greater than a threshold can be filtered to achieve fuzzy deduplication. The threshold here can be 0.7. By performing data deduplication, the problem of model overfitting caused by too much data of the same type can be avoided.

[0075] Furthermore, a data scoring strategy can be used to evaluate the scores of the deduplicated medical data relationship pairs to obtain corresponding scores. The data scoring strategy can be a linear regression method based on the variables and their corresponding weights. Specifically, each variable can be weighted and summed with its corresponding weight to obtain the corresponding score. A weight can be assigned to each variable, and the importance of the variable is positively correlated with the size of the weight. For example, the more important the variable, the greater the weight.

[0076] For example, variables could include current medical history, personal history, marital and reproductive history, past medical history, family history, physical examination, and auxiliary examinations. Each of these variables could be scored out of 5. If the variable is a diagnosis that is present in the case characteristics, a score of 0 is assigned. If the variable is a treatment that includes tests, examinations, treatment, nursing care, and general diet, 5 points are added for each variable. The score is determined by multiplying the similarity between the diagnostic basis and the case characteristics by 100. If a variable is an uncertain variable (e.g., "to be determined," "?," "uncertain," or "unknown") in the diagnosis, a lower score, such as 1, is assigned.

[0077] In addition, medical data relationships that meet the score conditions can also be desensitized. For example, LTP (Language Technology Platform) can be used for entity recognition to obtain the entity names contained therein, such as names of people and organizations. In addition, other entity recognition methods can also be used for entity recognition. Furthermore, the identified entity names can be replaced with special symbols to improve the privacy and confidentiality of the data.

[0078] Based on the above-mentioned data cleaning, data deduplication, data scoring, and data desensitization, multiple medical data relationship pairs with causal relationships are screened to obtain screened medical data relationship pairs. Furthermore, a language model can be trained based on the screened medical data relationship pairs to obtain a trained language model. The language model can be any type of language model, for example, a large language model such as ChatGPT (Chat Generative Pre-trained Transformer, a generative artificial intelligence large language model), or other types of language models. The ChatGPT large language model is used as an example for illustration. For example, the screened medical data relationship pairs can be converted into vectors, which are input into the language model. The language model is then fitted to obtain an output result. Based on the difference between the output result and the true result, the model parameters of the language model are adjusted until the difference is minimized, thereby obtaining a trained language model. After obtaining the trained language model, the medical data to be processed can be input into the trained language model to obtain a prediction result. The prediction result can be determined based on the function that the trained language model can achieve, for example, predicting disease symptoms, predicting disease type, etc., or predicting medication recommendations. Among them, the trained language model is trained based on multiple medical data relationship pairs with causal relationships obtained by extracting relationship pairs from medical data.

[0079] The technical solution in the disclosed embodiments extracts medical data relationship pairs from medical data. Since the medical data relationship pairs contain corresponding causal relationship pairs, the problem of causal relationship confusion in related technologies is avoided, which can improve the accuracy and comprehensiveness of training data, enhance the efficiency of model training, and also improve the model training effect, thereby improving the accuracy and stability of the model and thus improving the accuracy of the model output results. In addition, in the disclosed embodiments, medical data relationship pairs are screened through various data cleaning methods, and a language model is trained based on the screened medical data relationship pairs. Since the accuracy of the medical data relationship pairs can be improved, the accuracy of model training can be improved.

[0080] The present disclosure provides a medical data processing device, referring to Figure 6 As shown in , the medical data processing device 600 may include:

[0081] A data acquisition module 601 is used to acquire medical data of multiple target subjects;

[0082] The relationship pair generation module 602 is used to extract relationship pairs from the medical data based on various descriptive information in the medical data to obtain multiple medical data relationship pairs with causal relationships, so as to train a language model based on the medical data relationship pairs.

[0083] In an exemplary embodiment of the present disclosure, the medical data is subjected to relationship pair extraction based on multiple descriptive information in the medical data to obtain multiple medical data relationship pairs with causal relationships, including: when the type of the descriptive information is physical examination information, data is composed according to the physical examination information and the physical examination conclusion in the medical data to obtain medical data relationship pairs related to the physical examination information; when the type of the descriptive information does not belong to physical examination information, the medical data is subjected to relationship pair extraction based on each descriptive information to obtain an extraction result, and the relationship pair extraction is continued in combination with the extraction result to generate medical data relationship pairs related to each descriptive information.

[0084] In an exemplary embodiment of the present disclosure, the descriptive information includes basic information; the medical data is subjected to relationship pair extraction based on each descriptive information to obtain an extraction result, and the relationship pair extraction is continued in combination with the extraction result to generate a medical data relationship pair related to each descriptive information, including: using the basic information as an independent variable, determining a first relationship pair between the basic information and the inspection and examination instruction; updating the independent variable based on the result of the inspection and examination instruction in the first relationship pair, continuing to extract relationship pairs based on the updated independent variable, and determining a second relationship pair between the updated independent variable and the diagnosis result; continuing to update the independent variable based on the diagnosis result in the second relationship pair, and determining a third relationship pair between the updated independent variable and the treatment method, so as to determine multiple medical data relationship pairs related to the basic information.

[0085] In an exemplary embodiment of the present disclosure, the descriptive information includes case characteristics; the relationship pair extraction of medical data is performed according to each descriptive information to obtain an extraction result, and the relationship pair extraction is continued in combination with the extraction result to generate medical data relationship pairs related to each descriptive information, including: using case characteristics as independent variables, determining a fourth relationship pair between the case characteristics and the preliminary diagnosis result; using the case characteristics and the fourth relationship pair as independent variables, determining the medical data relationship pairs between multiple diagnosis and treatment related information corresponding to the preliminary diagnosis result; wherein, the diagnosis and treatment related information includes at least one of the diagnosis basis, treatment method and differential diagnosis.

[0086] In an exemplary embodiment of the present disclosure, the descriptive information includes disease course information; the medical data is subjected to relationship pair extraction based on each descriptive information to obtain an extraction result, and the relationship pair extraction is continued in combination with the extraction result to generate a medical data relationship pair related to each descriptive information, including: using the disease course information as an independent variable to determine a fifth relationship pair between the disease course information and the test and examination instructions; using the disease course information and the result of the test and examination instructions in the fifth relationship pair as independent variables to determine the medical data relationship pair between the disease course information and the treatment method.

[0087] In an exemplary embodiment of the present disclosure, the medical course information is used as an independent variable to determine the fifth relationship pair between the medical course information and the test and examination instructions, including: grouping the medical course information according to the recording date to obtain multiple medical course groups, and grouping the test and examination instructions according to the examination issuance date to obtain multiple test and examination groups; determining the recording time according to the recording date and the current date, and determining the examination issuance time according to the examination issuance date and the current date; for each test and examination group, combining the medical course information whose recording time is greater than the examination issuance time of the test and examination instruction with the test and examination instruction of the current date to obtain the fifth relationship pair.

[0088] In an exemplary embodiment of the present disclosure, the medical course information and the results of the inspection and examination instructions in the fifth relationship pair are used as independent variables to determine the medical data relationship pair between the medical course information and the treatment method, including: merging the results of the inspection and examination instructions in the fifth relationship pair into the medical course information to obtain updated medical course information; grouping the updated medical course information according to the recording date to obtain multiple medical course groups, and grouping the treatment methods according to the treatment prescription date to obtain multiple treatment method groups; determining the recording time according to the recording date and the current date, and determining the treatment prescription time according to the treatment prescription date and the current date; for each treatment method group, combining the updated medical course information whose recording time is greater than the treatment prescription time with the treatment method of the current date to determine the medical data relationship pair.

[0089] In an exemplary embodiment of the present disclosure, the device further includes: determining the priorities of multiple cleaning methods; screening the medical data relationship pairs according to the priorities to obtain cleaned medical data relationship pairs; wherein the multiple cleaning methods include data cleaning, data scoring, data deduplication and data desensitization.

[0090] It should be noted that the specific details of each module in the above-mentioned medical data processing device have been described in detail in the corresponding medical data processing method, and therefore will not be repeated here.

[0091] The exemplary embodiments of the present disclosure further provide an electronic device. This electronic device may be the aforementioned client or a server. Generally, the electronic device may include a processor and a memory, the memory being configured to store executable instructions of the processor, and the processor being configured to execute the aforementioned medical data processing method by executing the executable instructions.

[0092] Refer to the following Figure 7 700 according to this embodiment of the present disclosure will be described. Figure 7 The electronic device 700 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0093] like Figure 7 As shown, electronic device 700 is implemented as a general-purpose computing device. Components of electronic device 700 may include, but are not limited to, the aforementioned at least one processing unit 710, the aforementioned at least one storage unit 720, a bus 730 connecting various system components (including storage unit 720 and processing unit 710), and a display unit 740.

[0094] The storage unit stores program codes, which can be executed by the processing unit 710, so that the processing unit 710 performs the steps described in the "Exemplary Method" section of the present disclosure according to various exemplary embodiments. For example, the processing unit 710 can perform the following steps: Figure 1 Follow the steps shown in .

[0095] The storage unit 720 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 7201 and / or a cache memory unit 7202 , and may further include a read-only memory unit (ROM) 7203 .

[0096] The storage unit 720 may also include a program / utility 7204 having a set (at least one) of program modules 7205, such program modules 7205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0097] Bus 730 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics interface, a processing unit, or a local bus using any of a variety of bus architectures.

[0098] The electronic device 700 can also communicate with one or more external devices 800 (e.g., a keyboard, a pointing device, a Bluetooth device, etc.), one or more devices that enable a user to interact with the electronic device 700, and / or any device that enables the electronic device 700 to communicate with one or more other computing devices (e.g., a router, a modem, etc.). Such communication can occur via an input / output (I / O) interface 750. Furthermore, the electronic device 700 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network such as the Internet) via a network adapter 760. As shown, the network adapter 760 communicates with other modules of the electronic device 700 via a bus 730. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the electronic device 700, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0099] It should be noted that some embodiments of the present disclosure further provide a computer program product, which includes a computer program, and the computer program implements the above method when executed by a processor.

[0100] In one embodiment, a computer program product may be a tangible product containing a computer program, such as a computer-readable storage medium storing the computer program. The computer-readable storage medium may be a storage medium based on electrical, magnetic, optical, electromagnetic, infrared, or other signals, including but not limited to random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory (Flash), mechanical hard disk drive (HDD), solid-state drive (SSD), and the like. Exemplarily, the computer program product may be implemented as a non-volatile storage medium storing the computer program, such as a read-only memory, NAND flash memory, and the like.

[0101] In one embodiment, the computer program product may be an intangible product containing a computer program. For example, the computer program product may be implemented as a virtual digital product, such as a digital file such as an executable file or installation package storing the computer program.

[0102] The code of the computer program can be written in one or more programming languages. Programming languages ​​include C, Java, C++, etc. The program code can be executed entirely on the user computing device, partially on the user computing device, or as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device via any type of network, such as a local area network (LAN), a wide area network (WAN), etc., or can be connected to an external computing device (e.g., via an Internet connection provided by a carrier).

[0103] Computer programs can be carried or transmitted via electrical, magnetic, optical, electromagnetic, infrared, or other signals. Electronic devices can convert signals carrying computer programs into digital signals to run the computer programs. When the computer program is run on an electronic device, its code causes the electronic device to execute (more specifically, the processor of the electronic device to execute) the method steps of various exemplary embodiments of the present disclosure.

[0104] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the embodiments of the present disclosure.

[0105] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing what is disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the claims.

[0106] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A medical data processing method, characterized in that: include: Acquire medical data of multiple target subjects; Relationship pairs are extracted from the medical data based on various descriptive information in the medical data to obtain multiple medical data relationship pairs with causal relationships, so as to train a language model based on the medical data relationship pairs.

2. The medical data processing method according to claim 1, characterized in that: The extracting of relationship pairs from the medical data based on the multiple descriptive information in the medical data to obtain multiple medical data relationship pairs with causal relationships includes: In a case where the type of the description information is physical examination information, data composition is performed according to the physical examination information and the physical examination conclusion in the medical data to obtain a medical data relationship pair related to the physical examination information; When the type of the descriptive information does not belong to physical examination information, the medical data is subjected to relationship pair extraction according to each descriptive information to obtain extraction results, and the relationship pair extraction is continued in combination with the extraction results to generate medical data relationship pairs related to each descriptive information.

3. The medical data processing method according to claim 2, characterized in that: The description information includes basic information; extracting the medical data relationship pairs according to each description information to obtain extraction results, and continuing to extract the relationship pairs in combination with the extraction results to generate medical data relationship pairs related to each description information, including: Taking the basic information as an independent variable, determining a first relationship pair between the basic information and the inspection instruction; updating the independent variable based on the result of the inspection instruction in the first relation pair, continuing to extract relation pairs according to the updated independent variable, and determining a second relation pair between the updated independent variable and the diagnosis result; The independent variable is continuously updated based on the diagnosis result in the second relationship pair, and a third relationship pair between the updated independent variable and the treatment method is determined to determine a plurality of medical data relationship pairs related to the basic information.

4. The medical data processing method according to claim 2, wherein: The descriptive information includes case characteristics; performing relationship pair extraction on the medical data according to each descriptive information to obtain an extraction result, and continuing to perform relationship pair extraction in combination with the extraction result to generate medical data relationship pairs related to each descriptive information, including: Using case characteristics as independent variables, determining a fourth relationship pair between the case characteristics and the preliminary diagnosis results; Using the case characteristics and the preliminary diagnosis result in the fourth relationship pair as independent variables, determining a medical data relationship pair between a plurality of diagnosis and treatment related information corresponding to the preliminary diagnosis result; The diagnosis and treatment related information includes at least one of a diagnosis basis, a diagnosis and treatment plan, and a differential diagnosis.

5. The medical data processing method according to claim 2, characterized in that: The descriptive information includes disease course information; extracting the medical data relationship pairs according to each descriptive information to obtain extraction results, and continuing to extract the relationship pairs in combination with the extraction results to generate medical data relationship pairs related to each descriptive information, including: Using the disease course information as an independent variable, determining a fifth relationship pair between the disease course information and the test and examination instructions; The disease course information and the result of the inspection and examination instruction in the fifth relationship pair are used as independent variables to determine the medical data relationship pair between the treatment method.

6. The medical data processing method according to claim 5, characterized in that: The determining of a fifth relationship between the course of disease information and the test and examination instructions using the course of disease information as an independent variable includes: Grouping the medical course information according to the recording date to obtain a plurality of medical course groups, and grouping the test and examination instructions according to the examination issuance date to obtain a plurality of test and examination groups; Determining the recording time based on the recording date and the current date, and determining the inspection opening time based on the inspection opening date and the current date; For each test and examination group, the course of disease information whose recording time is greater than the examination issuance time of the test and examination instruction is combined with the test and examination instruction of the current date to obtain a fifth relationship pair.

7. The medical data processing method according to claim 5, characterized in that: The step of using the disease course information and the result of the inspection and examination instruction in the fifth relationship pair as independent variables to determine the medical data relationship pair between the medical course information and the treatment method includes: Merging the result of the fifth relationship alignment inspection instruction into the disease course information to obtain updated disease course information; Grouping the updated course of disease information according to the recording date to obtain a plurality of course of disease groups, and grouping the treatment methods according to the treatment prescription date to obtain a plurality of treatment method groups; Determining the recording time based on the recording date and the current date, and determining the treatment prescribing time based on the treatment prescribing date and the current date; For each treatment mode group, the updated course of disease information with a recording time greater than the treatment prescription time is combined with the treatment mode of the current date to determine the medical data relationship pair.

8. The medical data processing method according to claim 1, wherein: The method further comprises: Prioritize various cleaning methods; screening the medical data relationship pairs according to the priority to obtain cleansed medical data relationship pairs; Among them, the multiple cleaning methods include data cleaning, data scoring, data deduplication and data desensitization.

9. A medical data processing device, characterized in that: include: A data acquisition module, used to acquire medical data of multiple target objects; The relationship pair generation module is used to extract relationship pairs from the medical data based on multiple descriptive information in the medical data to obtain multiple medical data relationship pairs with causal relationships, so as to train a language model based on the medical data relationship pairs.

10. A computer program product having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the medical data processing method according to any one of claims 1 to 8 is implemented.

11. An electronic device, characterized in that: include: processor; as well as a memory for storing executable instructions of the processor; Wherein, the processor is configured to perform the medical data processing method according to any one of claims 1 to 8 by executing the executable instructions.

Images