Clinical examination result analysis method based on medical knowledge graph

Abstract

The invention relates to the technical field of diagnosis and treatment data processing, and particularly discloses a clinical examination result analysis method based on a medical knowledge graph, and the method comprises the steps: collecting historical clinical examination data of patients with different target diseases in an electronic medical record system; screening and extracting a disease keyword text sequence of the target disease according to the historical clinical examination data; constructing an ontology model based on the medical knowledge mapping relationship in the database, inputting the disease keyword text sequence into the ontology model to construct a medical knowledge graph model of the target disease, and outputting a plurality of disease keyword parameters; acquiring clinical manifestation information of target diseases of different patients in real time, extracting a knowledge graph of current clinical examination data of the patients and illness state trend values corresponding to the illness state keyword parameters according to the illness state keyword parameters, judging whether the illness states of the patients are abnormal or not, generating illness state evaluation parameters and outputting clinical examination scheme results of the patients; the accuracy of clinical examination results is improved; and the medical detection investment cost is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of diagnosis and treatment data processing, and particularly to a method for analyzing clinical test results based on a medical knowledge graph. Background Art

[0002] In the field of clinical medicine, the analysis of clinical test results is a key step in the diagnosis and treatment process. With the development of medical informatization and big data technology, traditional analysis methods can no longer meet the growing demand for data processing and analysis. As a structured knowledge representation method, the medical knowledge graph can reveal the complex relationships between different entities, providing a new perspective and solution for the analysis of clinical test results.

[0003] By integrating and organizing a large number of medical entities (such as diseases, symptoms, drugs, test items, etc.) and the relationships between them, the medical knowledge graph forms a huge semantic network; it can not only help doctors and researchers better understand and navigate complex medical information, but also support advanced data analysis and knowledge discovery.

[0004] In the existing process of using the medical knowledge graph for clinical disease test analysis, it usually stays mostly in the computable theoretical research of clinical disease test knowledge, lacking the actual transformation analysis process of actual clinical information and diagnosis and treatment status information digitization, and lacking the prediction of the results of disease treatment interventions. As a result, when faced with a large number of patients, it takes time to conduct empirical analysis on the historical medical record information of patients, relying on the experience of medical staff and the accuracy of diagnosis results, and lacking a unified standard. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for analyzing clinical test results based on a medical knowledge graph, and solve the following technical problems:

[0006] How to improve the accuracy of clinical test results, unify the disease diagnosis standard, and reduce the cost of medical detection input.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] A method for analyzing clinical test results based on a medical knowledge graph, the method includes:

[0009] S1. Collect the historical clinical test data of different patients with the target disease in the electronic medical record system; screen and extract the text sequence of disease symptom keywords according to the historical clinical test data; the text sequence of disease symptom keywords includes the text sequence of physiological entity keywords and the text sequence of pathological entity keywords;

[0010] S2. Construct an ontology model based on the medical knowledge mapping relationship in the database, input the text sequence of disease keyword into the ontology model to construct a medical knowledge graph model of the target disease, and output multiple disease keyword parameters;

[0011] S3. Obtain the clinical manifestation information of the target disease of different patients in real time, extract the disease keyword parameters and the corresponding disease trend values of the disease keyword parameters according to the knowledge graph of the current clinical test data of the patient, and judge whether there is an abnormality in the patient's condition:

[0012] If so, go to step S4;

[0013] If not, generate a disease evaluation parameter based on the preset disease evaluation model;

[0014] S4. Input the knowledge graph of the abnormal clinical test data of the patient and the corresponding abnormal disease trend value into the medical knowledge graph model for entity construction, and output the clinical test plan result of the patient.

[0015] Preferably, the screening and extraction method of the text sequence of disease keywords of the target disease in step S1 is:

[0016] S11. Perform preprocessing on the clinical test data text, remove irrelevant information, and correct spelling mistakes;

[0017] S12. Identify the target disease text through named entity recognition technology;

[0018] S13. Input the historical target disease feature association information and the target disease text into a machine learning model for training, and screen and output the disease keyword text;

[0019] S14. Construct a text sequence of disease keywords by arranging the keyword texts according to semantic logic, disease importance, and time sequence.

[0020] Preferably, step S3 includes:

[0021] Obtain the disease coefficient Dis of the i-th patient of the target disease through the formula ; i ;

[0022] where M is the total number of items of disease keywords of the target disease, and j ∈ [1, M]; Dk j is the parameter conversion function of the j-th item of disease keyword; Tr ij (t) is the real-time disease trend value corresponding to the j-th item of disease keyword of the i-th patient; Δt is the length of the preset disease change time period; δ j is the preset weight coefficient of the j-th item of disease keyword.

[0023] Preferably, it also includes the disease coefficient Dis of the i-th patienti Compare the size with the preset disease condition coefficient threshold range [Dis 1 , Dis 2 :

[0024] If then it is determined that the patient's disease condition has changed abnormally, and clinical tests are performed;

[0025] If Dis i ∈ [Dis 1 , Dis 2 , then it is determined that the patient's disease condition has changed normally, and disease condition evaluation is performed.

[0026] Preferably, the generation process of the disease condition evaluation parameter is as follows:

[0027] SS1. Extract the positive example samples of the positive keyword texts and the negative example samples of the negative keyword texts after the transformation of the disease keyword text sequences from the semantic database;

[0028] SS2. Construct an initial disease condition evaluation model based on the positive example samples, negative example samples and historical disease condition evaluation indicators;

[0029] SS3. Input the current clinical manifestation information of the patient, the positive example samples and negative example samples of the current disease keywords into the disease condition evaluation model for training, and output the disease condition evaluation parameter.

[0030] Preferably, it further includes judging the prognosis of the patient according to the disease condition evaluation parameter:

[0031] If the disease condition evaluation parameter is greater than or equal to the set threshold, it is determined that the patient has a good prognosis;

[0032] If the disease condition evaluation parameter is less than the set threshold, it is determined that the patient has an average prognosis.

[0033] Preferably, the physiological entity keywords include: patient's blood pressure, patient's blood sugar, patient's disease-related sign information; the pathological entity keywords include: patient's disease treatment cycle, patient's disease symptoms, patient's disease treatment response.

[0034] Preferably, the method further includes:

[0035] S5. Input the results of different patients' clinical test plans and their historical clinical test data into a preset prediction model for updating to obtain a disease condition prediction model, and output prediction parameters.

[0036] Preferably, warning prompt information is generated according to the prediction parameters.

[0037] Advantages of the present invention: By screening and constructing a text sequence of disease keywords, the present invention deepens the understanding of the disease information of the current patient. By using methods such as statistical analysis and machine learning, the keyword text sequence is analyzed to identify disease characteristics and correlations, and the process and evolution pattern of the current disease are determined. An ontology model is constructed based on the medical knowledge mapping relationship in the database, and the text sequence of disease keywords is input into the ontology model to construct a medical knowledge graph model of the target disease and output multiple disease keyword parameters. By creating a visual and structured medical knowledge graph, the disease keyword parameters extracted from the knowledge graph corresponding to the real-time clinical test data obtained by analyzing the current clinical manifestation information of the patient and the disease trend shown in the knowledge graph are further reflected by the magnitude of the disease trend value. According to the magnitude of the obtained disease keyword parameters, it is judged whether the change of the current real-time disease trend value is abnormal. If so, a test plan can be automatically generated. Otherwise, if the current disease change does not require test adjustment, a disease evaluation is automatically generated to facilitate setting an intervention adjustment cycle later and reducing the investment in medical tests.

[0038] Of course, it is not necessary for any product implementing the present invention to achieve all the advantages described above simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0040] Figure 1 It is a flowchart of the steps of a clinical test result analysis method based on a medical knowledge graph of the present invention;

[0041] Figure 2 It is a flowchart of the method for screening and extracting the text sequence of disease keywords of the target disease of the present invention;

[0042] Figure 3 It is a flowchart of the generation steps of the disease evaluation parameters of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0044] Please refer to Figure 1As shown in the figure, the present invention is a method for analyzing clinical test results based on a medical knowledge graph, and the method includes:

[0045] S1. Collect historical clinical test data of different patients with the target disease from the electronic medical record system; screen and extract the disease keyword text sequence of the target disease according to the historical clinical test data; the disease keyword text sequence includes a physiological entity keyword text sequence and a pathological entity keyword text sequence;

[0046] S2. Construct an ontology model based on the medical knowledge mapping relationship in the database, input the disease keyword text sequence into the ontology model to construct a medical knowledge graph model of the target disease and output multiple disease keyword parameters;

[0047] S3. Obtain the clinical manifestation information of different patients with the target disease in real time, extract the disease keyword parameters and the corresponding disease trend values of the disease keyword parameters according to the knowledge graph of the patient's current clinical test data, and judge whether the patient's condition is abnormal:

[0048] If so, go to step S4;

[0049] If not, generate a condition evaluation parameter based on a preset condition evaluation model;

[0050] S4. Input the knowledge graph of the patient's abnormal clinical test data and its corresponding abnormal disease trend value into the medical knowledge graph model for entity construction, and output the clinical test plan result of the patient.

[0051] In the above technical solution, in this embodiment, by designing a method for analyzing clinical test results based on a medical knowledge graph, the accurate analysis of the patient's clinical test results and the prediction process of the prognosis state are realized, ensuring the improvement of the analysis accuracy of the clinical test results, the diagnosis efficiency and the comprehensive test result output efficiency.

[0052] Among them, the medical knowledge graph is a structured knowledge representation method, which represents entities (such as diseases, drugs, genes, etc.) in the medical field and their relationships through a graphical network structure. The purpose of the medical knowledge graph is to provide a comprehensive, consistent and queryable medical knowledge base, to facilitate the integration and standardization of medical information, to provide a structured knowledge representation, and to facilitate efficient knowledge retrieval and reasoning by computer algorithms through the medical knowledge graph model. It is mainly a process of building by using a machine model through a large amount of disease patient data information in the medical database, and outputting the visualization result of the medical knowledge graph model.

[0053] In the process of the structured display of the above medical knowledge graph, the screening of specific diseases, the integration of patient information, and the selection and data analysis of current disease symptoms are the main exploration ideas of this design, which are specifically implemented through the following steps:

[0054] First step, collect the historical clinical test data of different patients with the target disease through the electronic medical record system; retrieve the medical record materials of patients with specific diseases from the electronic medical record system of hospitals or medical institutions and ensure the compliance of the data, obtain the consent of the patients, and then clean and standardize the collected data, remove irrelevant information, such as patient personal identity information, and integrate the collected patient name, patient gender, patient age, and patient medication records into patient personal characteristics for marking, and unify the data format for subsequent analysis; and use natural language processing (NLP) technology to identify and extract keywords related to the disease. This design is divided into physiological entities (such as blood pressure, blood sugar, etc.) and pathological entities (symptom characteristics, treatment information); the above processing is a common data information integration and processing method in this field and will not be described in detail here.

[0055] In addition, it is necessary to screen and extract the symptom keyword text sequence of the target disease according to the historical clinical test data; the symptom keyword text sequence includes the physiological entity keyword text sequence and the pathological entity keyword text sequence; by screening and constructing the symptom keyword text sequence, the understanding of the current patient's disease information can be deepened, and by using methods such as statistical analysis and machine learning, analyze the keyword text sequence to identify disease characteristics and correlations to determine the process and evolution pattern of the current disease.

[0056] Among them, as an implementation manner of the present invention, the physiological entity keywords include: patient blood pressure, patient blood sugar, and patient disease-related physical sign information; the pathological entity keywords include: patient disease treatment cycle, patient disease symptoms, and patient disease treatment response.

[0057] In the above technical solution, combining physiological entities and pathological entities can further reflect the overall situation of the patient's illness, and build a complete patient information through the knowledge graph.

[0058] Specifically, please refer to Figure 2 As shown, as an implementation manner of the present invention, the screening and extraction method of the symptom keyword text sequence of the target disease in step S1 is:

[0059] S11. Perform preprocessing on the clinical test data text, remove irrelevant information, and correct spelling mistakes;

[0060] S12. Identify the target disease condition text through named entity recognition technology;

[0061] S13. Input the historical target disease feature association information and the target condition text into a machine learning model for training, and screen and output the condition keyword text;

[0062] S14. Construct a sequence of disease keyword texts by arranging the keyword texts according to semantic logic, disease importance, and chronological order.

[0063] In the above technical solution, in this embodiment, first, the preprocessing of the clinical test data text is performed to exclude irrelevant information and improve the accurate range of the current disease clinical test data. Then, the named entity recognition technology in natural language processing is used to process the target condition text to identify entities such as diseases, symptoms, and test indicators in the text, and the entity types are marked to distinguish which are diseases, which are physiological indicators, and which are treatment measures in this embodiment; and they are marked one by one. Next, the screening of the condition keywords is carried out. Through the historical target disease feature association information and the target condition text, the machine model is used to learn these association information and texts to identify which keywords are more important for disease diagnosis. This process can be obtained through the existing neural network model and is screened through machine training to output a set of keyword texts strongly related to the target disease. Finally, the sorting is carried out according to semantic logic (such as symptoms preceding diagnosis), disease importance (severe diseases preceding mild diseases), and chronological order (early symptoms preceding late symptoms). Ensure that an ordered and logically clear sequence of disease keyword texts is formed for clinicians to understand and apply.

[0064] In the second step, an ontology model is constructed based on the medical knowledge mapping relationship in the database. This is a common way to construct an ontology model in the medical field based on a knowledge graph and will not be elaborated here. Then, the sequence of disease keyword texts is input into the ontology model to construct a medical knowledge graph model of the target disease and output multiple disease keyword parameters; by creating a visual and structured medical knowledge graph; using the ontology model as a skeleton, adding disease keyword parameters, and establishing multi-layer associations between entities; the machine training is carried out by inputting the sequence of disease keyword texts into the ontology model, and a complete graph including diseases, symptoms, test results, etc. is output to ensure the realization of supporting the intuitive understanding of the complexity of the disease. In addition, according to the training results, the disease keyword information, including the disease keyword parameters, can be extracted from the knowledge graph.

[0065] In the third step, by obtaining the clinical manifestation information of the target disease of different patients in real time, the disease keyword parameters and the disease trend values corresponding to the disease keyword parameters are extracted from the knowledge graph of the patient's current clinical test data to judge whether the patient's condition is abnormal. If it is judged to be yes, then enter step S4; if it is judged to be no, then generate a condition evaluation parameter based on a preset condition evaluation model.

[0066] Based on this, in this embodiment, the disease keyword parameters extracted from the knowledge graph corresponding to the real-time clinical test data obtained by analyzing the current clinical manifestation information of the patient and the disease condition change trend shown in the knowledge graph are further reflected by the magnitude of the disease condition trend value. According to the magnitude of the obtained disease keyword parameters, it is judged whether the change of the current real-time disease condition trend value is abnormal. If so, a test plan can be automatically generated. Otherwise, if the current disease condition change does not require test adjustment, a disease condition evaluation is automatically generated to facilitate setting an intervention adjustment cycle in the later stage and reducing the investment in medical tests.

[0067] Specifically, as an implementation manner of the present invention, step S3 includes:

[0068] Through the formula Calculate to obtain the disease condition coefficient Dis of the i-th patient with the target disease i ;

[0069] Wherein, M is the total number of items of the disease keyword of the target disease, and j ∈ [1, M]; Dk j Is the parameter conversion function of the j-th disease keyword; Tr ij (t) is the real-time disease condition trend value corresponding to the j-th disease keyword of the i-th patient; Δt is the length of the preset disease condition change time period; δ j Is the preset weight coefficient of the j-th disease keyword.

[0070] In the above technical solution, in this embodiment, the analysis method in step S3 is analyzed data-wise, and through the formula Calculate to obtain the disease condition coefficient Dis of the i-th patient with the target disease i ; According to the magnitude of the disease condition coefficient, it can be judged the change of the target disease of the current patient, and further test and evaluation processing are carried out; among them, the time period accumulation of the real-time disease condition trend values of different disease keywords of the patient can reflect the periodic change situation after the patient's treatment intervention. Of course, this time period Δt is also pre-selected and set according to the treatment intervention cycle, which can be one cycle or multiple cycles. Specifically, it needs to be selected and set according to the current disease treatment situation of the patient.

[0071] It needs to be further explained that the parameter conversion function Dk j Of the j-th disease keyword is a function set according to the influence of the historical disease keyword parameters of the patients with this disease on the severity of the historical deterioration of this disease, which is different; this conversion function ensures that the influence of the obtained parameters of the disease keyword on the disease condition trend value is reasonably converted, so that the obtained disease condition coefficient value meets the coefficient analysis requirements; the preset weight coefficient δ jThe result obtained by quantitatively analyzing the influence of the disease keywords of the current j-th item on the magnitude of the disease coefficient. For example, assume that the disease being analyzed is diabetes, and the keywords include "blood sugar level", "weight", and "insulin resistance". Determine "blood sugar level", "weight", and "insulin resistance" as key factors; obtain the relevant numerical values and descriptions of these keywords from clinical data; according to the consensus of medical experts, initially set the weight coefficients of each keyword. For example: blood sugar level: 0.7 (highest importance), weight: 0.2, insulin resistance: 0.5. Further, conduct statistical analysis on a large number of cases to observe the performance of each keyword in the disease process; adjust the weights according to the analysis results. For example, if it is found that the actual influence of "insulin resistance" is greater than the initially set value, then increase its weight to 0.6.

[0072] In addition, during actual use, periodic review is also required: regularly check the effectiveness of the weight coefficients and make corrections based on new data and clinical feedback.

[0073] Example calculation:

[0074] Suppose in a case, the contribution of an abnormal blood sugar level (set as hyperglycemia) is 0.7, the contribution of overweight is 0.2, and the contribution of insulin resistance is 0.6; if these factors are aggregated, the disease coefficient (i.e., the severity of the disease) of this case can be roughly evaluated; this kind of quantitative analysis can help doctors more accurately judge the urgency of the disease and the treatment direction.

[0075] As an implementation manner of the present invention, it further includes comparing the disease coefficient Dis of the i-th patient i with the preset disease coefficient threshold interval [Dis 1 , Dis 2 to compare their magnitudes:

[0076] If then it is determined that the patient's disease condition has changed abnormally, and clinical tests are performed;

[0077] If Dis i ∈[Dis 1 , Dis 2 , then it is determined that the patient's disease condition has changed normally, and disease evaluation is performed.

[0078] In the above technical solution, in this embodiment, the disease coefficient Dis of the i-th patient is also compared with the preset disease coefficient threshold interval [Dis i , Dis 1 , Dis 2 to make a range comparison, and a preliminary judgment on the subsequent disease condition changes is made through the preset range interval. When it does not belong to the current interval [Dis 1 , Dis 2, it is determined that the patient's condition has changed abnormally, and specific determination is also made in combination with clinical tests; when it belongs to this range, it is determined that the patient's condition has changed normally, and then the condition is evaluated, and the test cycle is adjusted or the setting and reminder of the next test cycle are carried out, and the result is notified to the patient user terminal or mobile terminal device suffering from the disease.

[0079] Further, please refer to Figure 3 As shown in, as an embodiment of the present invention, for patients within the normal range of condition changes, condition evaluation parameters are generated for the data of the patients. Specifically, the generation process of the condition evaluation parameters is as follows:

[0080] SS1. Extract the positive example samples of the positive keyword texts and the negative example samples of the negative keyword texts after the conversion of the disease keyword text sequences from the semantic database;

[0081] SS2. Construct an initial condition evaluation model based on the positive example samples, negative example samples and historical condition evaluation indicators;

[0082] SS3. Input the patient's current clinical manifestation information, positive example samples and negative example samples of the current disease keywords into the condition evaluation model for training, and output the condition evaluation parameters.

[0083] In the above technical solution, in this embodiment, positive keywords and negative keywords are selected from the disease keyword text sequences, and corresponding positive example samples and negative example samples are constructed respectively according to the positive keywords and negative keywords in all the disease keywords as the sample set for input. In addition, an evaluation model is selected according to the historical condition evaluation indicators, and the historical condition evaluation indicators of the disease and the sample set are input into the initial evaluation model together, and the condition evaluation parameters of the corresponding patients of the disease are output according to the positive example samples and negative example samples of the actual clinical manifestations, and the prognosis of the patient's disease is judged.

[0084] Specifically, as an embodiment of the present invention, it further includes judging the prognosis of the patient according to the condition evaluation parameters:

[0085] If the condition evaluation parameter is greater than or equal to the set threshold, it is judged that the patient has a good prognosis;

[0086] If the condition evaluation parameter is less than the set threshold, it is judged that the patient has a general prognosis.

[0087] In the fourth step, the knowledge graph of the patient's abnormal clinical test data and its corresponding abnormal condition trend value are input into the medical knowledge graph model for entity construction, and the clinical test plan result of the patient is output.

[0088] In the above technical solution, based on the real-time update of the clinical test information data of the target disease screened from the abnormal clinical test data of the patient, a clinical test plan after more accurate entity information fusion and a knowledge graph model for visual display of accurate test information are provided. By providing a detailed individualized clinical test plan, it is ensured that: first, according to the analysis results, a targeted clinical test plan is proposed, including suggestions for further examinations, treatment monitoring points, etc.; second, combined with factors such as the patient's medical history, age, and gender, a personalized clinical strategy is formulated.

[0089] As an implementation manner of the present invention, a method for analyzing clinical test results based on a medical knowledge graph further includes:

[0090] In the fifth step, the clinical test plan results of different patients and their historical clinical test data are input into a preset prediction model to update and obtain a disease prediction model, and prediction parameters are output; and warning prompt information is generated according to the prediction parameters.

[0091] In the above technical solution, in this embodiment, the disease prediction model is set to dynamically adapt to individual differences of patients and changes in the medical environment, providing prospective data support for clinical decision-making; and the establishment of the warning system strengthens disease management, timely reminds the medical team to take measures to avoid the deterioration of the condition, and improves the treatment effect and quality of life of patients.

[0092] Each embodiment in this specification is described in a progressive manner. The same or similar parts among the embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments. In particular, for the embodiments of the device, equipment, and non-volatile computer storage medium, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0093] The above specifically describes certain embodiments of this specification. Other embodiments are within the scope of the appended documents. In some cases, the actions or steps recorded in this application can be executed in a different order from that in the embodiments and still achieve the desired results. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In certain embodiments, multi-tasking and parallel processing are also possible or may be advantageous.

[0094] The above content is only an example and illustration of the concept of the present invention. Those skilled in the art of this technology can make various modifications, supplements, or use similar methods to replace the specific embodiments described, as long as they do not deviate from the concept of the invention or exceed the scope defined in this application, they should all belong to the protection scope of the present invention.

Claims

1. A method for analyzing clinical test results based on a medical knowledge graph, characterized in that: The method comprises: S1. Collect historical clinical test data of different patients with target diseases in the electronic medical record system; screen and extract the symptom keyword text sequence of the target disease based on the historical clinical test data; the symptom keyword text sequence includes a physiological entity keyword text sequence and a pathological entity keyword text sequence; S2. Construct an ontology model based on the medical knowledge mapping relationship in the database, input a disease keyword text sequence into the ontology model to construct a medical knowledge graph model of the target disease and output multiple disease keyword parameters; S3. Obtain clinical manifestation information of target diseases of different patients in real time, extract disease keyword parameters and disease trend values ​​corresponding to disease keyword parameters based on the knowledge graph of the patient's current clinical test data, and determine whether the patient's condition is abnormal: If yes, proceed to step S4; If not, then generate condition evaluation parameters based on the preset condition evaluation model; S4. Input the knowledge graph of the patient's abnormal clinical test data and its corresponding abnormal disease trend value into the medical knowledge graph model for entity construction, and output the patient's clinical test plan results.

2. A clinical test result analysis method based on medical knowledge graph according to claim 1, characterized in that: The method for screening and extracting the text sequence of the target disease's symptom keywords in step S1 is: S11. Perform text preprocessing of clinical test data, remove irrelevant information, and correct spelling errors; S12, identifying the target disease text by using named entity recognition technology; S13, inputting historical target disease feature association information and target condition text into a machine learning model for training, and filtering and outputting condition keyword text; S14. Construct a disease keyword text sequence by arranging the keyword text according to semantic logic, disease importance, and time sequence.

3. According to claim 1, a clinical test result analysis method based on medical knowledge graph is characterized in that: Step S3 includes: By formula Calculate the disease coefficient Dis of the i-th patient of the target disease i ; Where M is the total number of keywords for the target disease, and j∈[1,M]; Dk j Tr is the parameter conversion function of the jth disease keyword; ij (t) is the real-time disease trend value corresponding to the jth disease keyword of the i-th patient; Δt is the length of the preset disease change time period; δ j is the preset weight coefficient of the ,th disease keyword.

4. A method for analyzing clinical test results based on a medical knowledge graph according to claim 3, characterized in that: It also includes the condition coefficient Dis of the i-th patient. i Compare the size with the preset disease coefficient threshold interval [Dis1, Dis2]: like If the patient's condition changes abnormally, clinical examinations are performed; If Dis i ∈[Dis1, Dis2], the patient's condition is judged to be normal and the condition is evaluated.

5. The method for analyzing clinical test results based on medical knowledge graph according to claim 1, characterized in that: The generation process of the condition evaluation parameters is as follows: SS1, extracting positive examples of keyword text and negative examples of keyword text converted from the disease keyword text sequence according to the semantic database; SS2: Construct an initial disease evaluation model based on positive samples, negative samples and historical disease evaluation indicators; SS3: Input the patient's current clinical manifestation information and the positive and negative samples of the current disease keywords into the disease evaluation model for training, and output the disease evaluation parameters.

6. A method for analyzing clinical test results based on medical knowledge graph according to claim 5, characterized in that: It also includes judging the patient's prognosis based on the disease evaluation parameters: If the condition evaluation parameter is greater than or equal to the set threshold, the patient is judged to have a good prognosis; If the condition evaluation parameter is less than the set threshold, the patient's prognosis is judged to be average.

7. A method for analyzing clinical test results based on medical knowledge graph according to claim 1, characterized in that: Physiological entity keywords include: patient blood pressure, patient blood sugar, and patient disease-related physical signs information; the pathological entity keywords include: patient disease treatment cycle, patient disease symptoms, and patient disease treatment response.

8. The method for analyzing clinical test results based on medical knowledge graph according to claim 1, characterized in that: The method further comprises: S5. Input the clinical test results of different patients and their historical clinical test data into the preset prediction model to update the disease prediction model and output the prediction parameters.

9. A method for analyzing clinical test results based on a medical knowledge graph according to claim 8, characterized in that: Generate early warning information based on prediction parameters.

Images